Achenbach 1974

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

J. Fluid Mech. (1974),vol. 65, part 1, p p .

113-125 113
Printed in Great Britain

The effects of surface roughness and tunnel blockage


on the flow past spheres
By ELMAR ACHENBACH
Institut fur Reaktorbauelemente der Kernforschungsanlage Julich GmbH,
Julich, Germany

(Received 7 December 1973)

The effect of surface roughness on the flow past spheres has been investigated
over the Reynolds number range 5 x lo4 < Re < 6 x lo6.The drag coefficient has
been determined as a function of the Reynolds number for five surface rough-
nesses. With increasing roughness parameter the critical Reynolds number
decreases. At the same time the transcritical drag coefficient rises, having a
maximum value of 0-4.
The vortex shedding frequency has been measured under subcritical flow
conditions. It was found that the Strouhal number for each of the various
roughness conditions was equal to its value for a smooth sphere. Beyond the
critical Reynolds number no prevailing shedding frequency could be detected
by the measurement techniques employed.
The drag coefficient of a sphere under the blockage conditions 0.5 < ds/dt c 0.92
has been determined over the Reynolds number range 3 x lo4 < Re < 2 x 106.
Increasing blockage causes an increase in both the drag coefficient and the
critical Reynolds number. The characteristic quantities were referred to the flow
conditions in the smallest cross-sectionbetween sphere and tube. I n addition the
effect of the turbulence level on the flow past a sphere under various blockage
conditions was studied.

1. Introduction
I n a previous paper by Achenbach ( 1972) the effect of the Reynolds number on
the flow past smooth spheres was studied up to very high Reynolds numbers
(Re= 6 x lo6).It was expected that the flow might be influenced by the surface
conditions of the sphere and, therefore, great importance was attached to pro-
ducing very smooth surfaces. This paper describes the second part of those
experiments, in which the effects of surface roughness were examined. This topic
is not dealt with in the available literature.
The effect of tunnel blockage has been studied in connexion with the problem
of the pneumatic transport of a single sphere in a tube. Only high blockage ratios
0.5 < d,/d, < 0.92, where d, is the diameter of the sphere and d, the diameter of
the tube, were considered.
It is well known that the drag coefficient of a sphere in an infinite flow drops
considerably when the critical flow range is reached. The problem was to deter-
mine whether, for high blockage ratios, such a critical Reynolds number exists
8 PLM 65
114 E . Achenbach

I+- -4
FIGURE
1. Experimental arrangement for blockage tests; d, = 0.191 m.

and, if i t does, how its value depends on the blockage ratio. From the practical
point of view it must be established that in any flow state the drag forces are
strong enough to cause safe transport of the sphere.
The experimental results indicated the existence of a critical Reynolds number
range even for the highest blockage ratios. The tests were effected by measuring
the local static pressure and skin friction under the blockage condition
d,/d, = 0.916.

2. Experimental apparatus and measurement techniques


For the determination of the drag coefficient as a function of the Reynolds
number and roughness parameter k/ds, where k is the height of the roughness
elements, the same experimental techniques were applied as those described in
the paper on the flow past smooth spheres (Achenbach 1972).The high Reynolds
numbers were realized using a high-pressure wind tunnel. The drag forces were
measured by means of a balance incorporating a strain-gauge bridge. This balance
was mounted within the hollow test sphere. The sphere had a diameter of about
0.2 m and was supported from the rear by a sting which was 0.75 m long.
I n four of the tests, the rough surface was obtained by pasting glass spheres of
diameter k onto the surface; the sizes chosen were k = 2-5 x 1 x 10-3,
0.5 x and 0-3 x lOVm. In a further test, a smaller roughness height was
produced by abrading the surface with coarse emery paper: an irregular pattern
of grooves about 5 x low5, deep was formed. The roughness height k is divided
by the sphere diameter d, to obtain the roughness parameter kld,. With the
roughness heights mentioned above the following values of the roughness para-
meter result: kid, = 1250 x 500 x 250 x 150 x lO-5and 25 x
A sketch of the apparatus in which the effect of tunnel blockage was investi-
Surface roughness and tunnel blockage eflects 115
gatedis illustratedin figure 1. The sphere was suspended from a balance by means
of a wire. Thus the total drag force B could be measured directly. The mass flow m
was calculated from the pressure drop a t the entrance of the tube. The corre-
sponding inlet coefficient was determined to be a = 0.97. The local static pressure
and the wall shear stresses were measured using a test sphere from previous
investigations which was equipped with a calibrated skin-friction probe
(Achenbach 1972). The probes were inserted in a rotatable ring-shaped element
of the sphere and could be placed a t any circumferential position. The vortex
shedding frequency was indicated by hot wires which were mounted flush with
the surface of the sphere. This technique had been employed in previous experi-
ments (Achenbach 1974). It was found then that the optimum angular position
for installing the probe is q5 = 75", q5 being measured from the front stagnation
point. The hot-wire signals were relayed to a frequency analyser.
During the investigation of the blockage effect, the turbulence level of the
incident flow was varied by mounting screens of different mesh size in the inlet
section. The turbulence level T, which is defined as
T = (F)&/U,
where U is the mean velocity in the entrance, was measured by means of a hot
wire.

3. Results
3.1. The effect of surface roughness on the drag on spheres
Roughness elements distributed on the surface of a body in a fluid stream cause
transition from laminar to turbulent flow if the disturbances generated by them
are amplified. For bluff bodies without salient edges to fix the separation points,
the premature onset of turbulence in the boundary layer causes a shift of the
separation point and hence a change in the drag forces. This effect has been
demonstrated for a circular cylinder in a cross-flow.A similar result was expected
for the sphere.
I n figure 2 the drag coefficient cd of the sphere is plotted as a function of the
Reynolds number for various roughness parameters Icld,. With increasing rough-
ness parameter the critical Reynolds number Re,, defined as the value where cd is
a minimum, decreases (see also figure 3). For subcritical conditions the curves
corresponding to the particular roughnesses collapse. At a certain value of the
Reynolds number which is dependent on the roughness parameter, the drag
coefficient decreases rapidly. This is the critical flow range where small variations
in the Reynolds number cause considerable changes in the drag coefficient.
Beyond the critical Reynolds number the drag coefficient increases again. This
range is defined to be the supercritical flow range. It is succeeded by the trans-
critical regime, where the drag coefficient is nearly independent of the Reynolds
number. The boundary between the supercritical and transcritical flow regimes is
floating. This has already been mentioned in the paper on the flow past smooth
spheres. The transcritical flow is characterized by a more or less constant value
oft,hedrag coefficient for each roughness. It appears that there exists a maximum
8-2
116 E.Achenbach

0.5

0.4

0.3
cd

0.2

0.1

0
4 6 8 lo! 2 4 6 8 lo6 2 4 6x106
Re
FIGURE 2. Drag coefficient cd vs. Reynolds number for a sphere. Parameter: surface
roughness. -, smooth (Achenbach 1972); x , k/ds = 25 x v,
k/d8 = 1 5 0 x 10-6;
0, k/ds = 250 x A,kid, = 500 x 0, k/d, = 1 2 5 0 10-5.
~

1400

1200

1000

800

600

400

200

0
4 6 8 lo5 2 4 x lo5
Re
FIGURE
3. Roughness parameter us. critical Reynolds number.
x , ( k / 4 e f fsee
, §4.
Surface roughness and tunnel blockage eflects 117
0.4

0.3

0.1

0 200 400 600 800 1000 1200 1400


( k p x~105
FIGURE
4. Transcritical drag coefficient vs. roughness parameter for a sphere.
Re = 5 x 106.

value of the drag coefficient which is not exceeded though the roughness para-
meter increases. I n particular this evidence is illustrated in figure 4,which shows
the transcritical drag coefficient as a function of the roughness parameter for
Re = 5 x lo6. A similar trend is obvious on considering the results for a rough
circular cylinder in a cross-flow (Achenbach 1971). Here the transcritical drag
coefficients of the cylinders covered with the two highest roughnesses tested come
close together. Recent unpublished results on this point confirm that observation.
Comparison of the flow resistance curves for the rough sphere and a rough
circular cylinder in a cross-flow indicates a significant similarity. It therefore
appears justifiable to draw similar conclusions concerning the behaviour of the
boundary layer on the basis of the drag curves. For subcritical flow conditions
the boundary layer separates laminarly. I n the critical flow regime the separation
point of the boundary layer begins to shift downstream, but laminar separation
still occurs. With a further increase in the Reynolds number the free shear layer
becomes turbulent and reattaches to the wall. The boundary layer receives energy
from outside through turbulent fluctuations. Thus it is able to develop further
downstream in spite of a positive pressure gradient. Finally it separates turbu-
lently. The resulting reduction in the width of the wake causes an increase in the
static pressure a t the rear of the sphere, which leads to the low values of the drag
coefficient, I n the supercritical flow range the transition from laminar to turbu-
lent flow occursimmediately. The point of transition shifts to the front stagnation
point with increasing Reynolds number. At the same time the point of boundary-
layer separation moves upstream and causes higher drag coefficients. Under
transcritical flow conditions nearly the whole boundary layer is turbulent.
During the tests under atmospheric conditions it was possible to determine the
boundary-layer separation by means of the wool-tuft technique. Figure 5 illu-
strates the dependency of the separation point 4, on the Reynolds number for
roughness parameters kid, = 250 x and 1250 x The values referring to
I18 E . Achenbach
120

I10

I00

9 8

90

70
4 6 8 lo5 2 4 6 8 lo6 2 I x 106

Re
FIGURE
5 . Angle #8 of boundary-layer separation 2)s. Reynolds number for a sphere.
-, smooth (Achenbach 1972); 0,kid, = 250x 0, k/ds = 1 2 5 0 ~

the smooth sphere are taken from Achenbach (1972). The accuracy of the method
is not very good. However, the characteristic shift of the separation point a t
critical and supercritical flow conditions described above is evident. The curves
have nearly the same shapes as those showing the cd vs. R e relationship but
inverted.
3.2. Vortex shedding from rough spheres
Initially the tests on vortex shedding from rough spheres were intended to study
the vortex shedding mechanism in the critical flow range, employing velocities
as low as possible. The aim of this was to eliminate the effects of vibration of the
sphere which was produced by the powerful fluctuating aerodynamic forces.
However, it was found that immediately before the critical Reynolds number
was reached the periodic hot-wire signal vanished; this was also observed for the
smooth sphere as mentioned by Achenbach (1974). I n particular this effect
occurred at the Reynolds number where the drag coefficient starts dropping.
This fact becomes evident from figure 6 taken together with figure 2. I n addition
figure 6 indicates that for subcritical flow conditions the Strouhal number based
on the vortex shedding frequency f,sphere diameter d, and free-stream velocity
U, is almost identical for rough and smooth spheres.

3.3. Effect of tunnel blockage


If the cross-section of the flow is partially blocked by the sphere, the equatorial
velocity increases with increasing blockage. The static pressure at the rear of the
sphere is dependent on this equatorial velocity since the boundary layer separates
around the equator. Therefore it determines the drag. From this point of view it
seems to be reasonable to regard the mean velocity U, in the smallest cross-
section as the characteristic velocity. The length scale is the sphere diameter Cl,.
Since the tube diameter d, appears as a second length scale, the similarity laws of
Surface roughness and tunnel blockage effects 119

0.20
0.18

0.20
0.18

<11
0.20
0.18
w2
0.20
0.18

0.20
0.18

104 2 4 6 8 lo5 2 4x lo5


Re
FIGURE
6. Strouhal number I J ~ .Reynolds
number for a sphere under various roughness
conditions. -, smooth (Achenbach 1974) ; symbols as in figure 2.

the flow yield an additional parameter, viz. the blockage ratio B = ds/dt. I n
these terms the characteristics are defined as follows:
cd = 4.??/+pU:nd: (drag coefficient),
Re = U,d,p/q (Reynolds number),
B = ds/dt (blockage ratio).
p and J,? are the fluid density and fluid viscosity, respectively.
It is obvious that for B+ 0 the velocity V, is equal to the velocity U, of an
infinite flow. Thus far the new definitions of the characteristics are compatible
with those for an infbite flow.
I n figure 7 the drag coefficient cd is plotted as a function of the Reynolds
number. The turbulence level was about T = 0.3 yo.The parameter varied is the
blockage ratio B = d&. It is remarkable that even a t the highest tunnel block-
ages the decrease in the drag coefficient a t critical flow conditions is apparent.
With increasing blockage ratio the subcritical drag coefficient as well as the
critical Reynolds number increase. The lower limiting line represents the drag
coefficient of a sphere in an ‘infinite’ flow. These results, which have been pub-
lished previously by Achenbach (1972), were used as reference data. The drag
coefiicient measured for the highest blockage ratio under subcritical flow condi-
tions is about twice the reference drag coefficient. I n conventional terms and
using the velocity of the oncoming flow as a reference value, this factor would be
about 56. The triangles represent values obtained from integration of the local
static pressure and skin-friction distribution around the circumference of the
sphere at dsld, = 0.916. I n this case the maximum velocities a t the highest
Reynolds number were of the order of half the velocity of sound and, thus, Mach
number effects may be important.
120 E , Achenbach
The tests under the various blockage conditions have been carried out in an
atmospheric air stream. For this reason the maximum Reynolds numbers are
lower than those referring to infinite flows in the high-pressure wind tunnel.
At subcritical flow conditions the drag Coefficient is nearly independent of
the Reynolds number in the range tested. Thus the drag coefficient can be given
as a function of the blockage ratio B. I n figure 8 the experimental results are
plotted in terms of c d / c d a , us. B for Re = 2 x lo5. c d , denotes the corresponding
drag coefficient for an infinite flow; its value is 0.5 1. The curve providing the best
fit to the experimental results is

Equation (1)is experimentally confirmed for the blockage range 0.5 < B < 0.92.
As mentioned above the character of the flow past spheres is unchanged even
at the highest blockage ratios. This is illustrated in figure 9, where the local skin
friction and static wall pressure are plotted as functions of the circumferential
angle (6 for the blockage ratio 0.916. The reference quantities are again the
velocity U,, the static pressure pc, the density p and the viscosity T , all quantities
referred to the smallest cross-section. I n the immediate vicinity of the front
stagnation point, the variation of the static pressure is small. The wall shear
stress slowly increases with increasing distance from the front stagnation point.
When > 50" the flow becomes more and more accelerated and causes an
intensive pressure drop in the flow direction. At the same time the skin friction
increases considerably and reaches a maximum value a t # = 80". The static
pressure drops below the theoretical value calculated for the equatorial position
(4 = 90") because of the effect of displacement by the tube boundary layer and
the sphere boundary layer. The wall shear stress has already decreased for
(6 > SO", which means that the boundary layer on the sphere grows rapidly. At
subcritical flow conditions the boundary layer separates laminarly at about
4 = loo", such that the static pressure a t the rear of the sphere is nearly equal to
the theoretical value in the gap.
The curves representing the results for higher Reynolds numbers indicate by
the resurgence of the skin friction the phenomena of laminar intermediate
separation and turbulent reattachment of the boundary layer, as observed for
the circular cylinder and sphere in an infinite flow. The downstream shift of the
separation point causes a recovery of the static pressure a t the rear of the sphere
and hence a decrease in the drag coefficient.
The effect on the flow of an eccentric arrangement of the sphere in the tube is
demonstrated in figure 10. The pressure and skin-friction distribution in a plane
parallel to the main flow direction through the point of sphere-wall contact are
plotted against the circumferential angle 9. The asymmetry of the distribution
is obvious. The stagnation point is displaced about 15" towards the wall. For both
Reynolds numbers the separation point near the wall is a t about q5s = loo", while
in the opposite position the separation occurs a t about 45, = 125".It is remarkable
that the boundary layer shows a supercritical behaviour a t the large gap position
(right side of figure 10) for a mass flow a t which a subcritical Reynolds number is
observed for the symmetric sphere arrangement. It is surprising that the pressure
Surface roughness and tunnel blockage efj'ects 121

Re = U C ~ 8 P l 7
FIGURE 7. Drag coefficient vs. Reynolds number for a sphere a t various blockage ratios.
-, d,/dt = 0 (Achenbach 1972); 0, dJd, = 0.5; Q , dJdt = 0.6; 0,dJdt = 0.7; x ,
d,/d, = 0.8;U, d8/dt= 0.4;A,from integration, dJd, = 0.916.
2.5

2.0

-
0

0
8
a

1.5

I .o
0 0.2 0.4 0.6 0.8 1.0
cl,ldt
FIGURE
8. Effect of tunnel blockage on the drag for a sphere under subcritical flow
conditions (Re = 2 x lo5).-, cd/cdm = 1 $- 1.45(d8/dt)4'6.

at the rear of the sphere and hence the drag coefficient are considerably different
for both Reynolds numbers, though the skin-friction distributions are rather
similar. This can be explained by the fact that the local quantities were measured
only in one particular plane of the sphere. However, over the entire circum-
ference the flow will be predominantly subcritical for the low and supercritical
for the high Reynolds number.
122 E. AchenbacA

N U 0.6
aQ
-la
0.4
F;:
-
F4
I
0.2

-0.2

1.5

N U
1;3 1.0
9
4

G
' 0.5

FIGURE 9. ( a ) Local static pressure and ( b ) skin-friction distribution for a sphere at the
blockage ratio dJd, = 0.916. 0, Re = 2 x loe; x , Re = 5.7 x 106; A, Re = 1.4 x loE.

3.4. Ejfect of turbulence level under conditions of high blockage


The investigations of blockage effects have been carried out under the turbulence
conditions of an entrance flow from quiescent air. A turbulence level of 0.3 %
was determined from hot-wire measurements. Since a turbulence level of
T > 0-3yowas expected for the practical application to the pneumatic transport
of spheres in a tube, T was increased in two steps without investigating the
structure of the turbulence. In the first run a screen of wire diameter d, = 1 mm
and a 5mm mesh size was employed. The turbulence level was determined to
be 1.3 yo.The second run was performed using a grid of flat wires 2.2 mm broad
and a 7-5mm mesh size. This arrangement yielded a turbulence level of 2.2 %. In
Surface roughness and tunnel blockage effects 123

1.o
N U

2.
-d 0.5
4-
\

I
v
a
0
0 135 I80 2 2 3 - w 315 360

2.0

1.5

1.O

N U 0.5
l l
b
3
Gtp o

-0.5

- 1.0 i-
FIGURE
-1.5

-2.0
0 45
-k
-
1 135 180

9"
225 270 315 360

10. (a)Local static pressure and ( b ) skin-friction distribution for a smooth sphere
arranged eccentrically a t the blockage ratio &/d, = 0.916. 0,Re = 2 x lo5; A , Re =
1.4 x 10'.

figure 11 the effect of different turbulence levels on the flow past spheres under
the blockage condition d J d , = 0.9 is demonstrated. Increasing the turbulence
level causes a premature transition from laminar to turbulent flow and hence
a. drop in the drag coefficient a t lower Reynolds numbers.
124 E . Achenbach

11. Effect of turbulence level T on the drag for a smooth sphere at the blockage
FIGURE
ratio d,/d, = 0.9. 0, T = 0.3 %; 0, T = 1.3 %; A,T = 2.2 %.

4. Final remarks
The surface of the sphere was roughened by being covered with small glass
spheres as compactly as possible. I n previous experiments on cross-flow past
circular cylinders (Achenbach 197 1) the same technique was employed. Com-
parison of these latter results with those obtained by Fage & Warsap (1930), who
used emery paper to roughen the surface, indicated that the effective height keii
of spherical roughness is not equal to the diameter k of the sphere but to 0.55k.
This result is plausible as the lower hemisphere facing the test body is hardly
wetted by the fluid flow. If this result is taken into account the present effective
roughness parameters have the values (kfd),,, = 690 x 275 x 138 x
and 83 x 10-5. I n figure 3 the second curve represents the dependency of the
critical Reynolds number on the effective roughness parameter (k/d),ff. As a
further consequence of the experimental technique, the effective diameter of the
test body is equal to the true diameter plus 0.9k. This effective diameter has
been employed in the present evaluation of the results.
The determination of the mass flow for the experiments on the blockage effect
has been a problem. The intention to put the sphere into the entrance region of
the flow so that i t might experience a plane velocity profile conflicted with the
need to measure the static pressure of the undisturbed flow upstream of the
sphere a t a position as far from the sphere as possible. I n addition very small
pressure differences occurred, particularly for the tests a t high blockage condi-
tions. I n preliminary tests the arrangement problems were optimized experi-
mentally by minimizing the distances according to figure 1. The small pressure
differences were measured by means of a high-precision bourdon-tube pressure
gauge (Texas Instruments, Houston).
J'urface roughness and tunnel blockage effects 125
Finally, an unexpected effect should be mentioned. At critical flow conditions
the sphere was pushed by lift forces towards the wall of the tube. This movement
grew more and more intense as the critical Reynolds number was approached.
Occasionally the sphere rolled on its equator along the inner circumference of the
tube like the inner wheel of an epicyclic gear. The tension in the mounting wire
became very small. It appeared that the drag force more or less vanished. This
flow state could not be terminated by increasing the mass flow. The phenomenon
described could not be studied in detail. However, an attempt will be made to
explain its initiation. I n the critical flow range a small variation in the Reynolds
number causes a drastic change in the drag coefficient. This change is due to
a downstream shift of the boundary-layer separation point accompanied by a
recovery of the static pressure a t the rear of the sphere. The flow state is rather
unstable and therefore any slight asymmetry of the geometry due to eccentricity
or individual surface roughness would cause a local premature transition of the
boundary layer. The resulting three-dimensional pressure distribution yields
a force perpendicular to the main flow direction which drives the sphere towards
the wall. Since the sphere is in contact with the wall of the tube, it tends to stay
there. A pressure distribution as illustrated in figure 10 builds up and supplies
a transverse force which holds the sphere against the wall. The force parallel to
the main flow direction is compensated for at the wall by static friction forces
between the sphere and the tube. Thus the drag appears to have vanished. Usually
the transverse force does not act in a plane containing the point of contact and
the centre of the sphere. I n this case the sphere would experience a moment of
torque which would cause i t to roll. That this motion, once started by a random
event, continues is not easy to see. Possibly effects quite different from those
investigated here become relevant. I n any case, the asymmetric boundary-layer
separation will be sustained. It would move in the sense opposite to the rotation
of the sphere like the transverse force, which supports the rotation and whose
direction and magnitude remain steady relative to the point of contact.

The present results were carried out in the laboratories of the Institut fur
Reaktorbauelemente, Kernforschungsanlage Julich GmbH. The director of the
Institute, Dr C. B. von der Decken, supported this work with great interest. The
author wishes to thank him very much. He also wishes to express his gratitude to
his assistants H. Cillessen, I?. Hoffmanns, H. Reger, R. Rommerskirchen and
W. Schmidt for their valuable help during the preparation and execution of the
tests.

REFERENCES
ACHENBACR,
E. 1971 Influence of surface roughness on the cross-flow around a circular
cylinder. J . Fluid Mech. 46, 321-335.
ACHENBACE, E. 1972 Experiments on the flow past spheres a t very high Reynolds
numbers. J . Fluid Mech. 54, 565-575.
ACHENBACH, E. 1974 Vortex shedding from spheres. J . Fluid Mech. 62, 209-221.
FAUE,A. & WARSAP,J. H. 1930 The effect of turbulence and surface roughness on the
drag of a circular cylinder. Aero. Res. Counc. R. & M . no. 1283.

You might also like