SURFACE ROUGHNESS EFFECTS
ON THE MEAN FLOW
PAST CIRCULAR CYLINDERS
by
Oktay Guven, V. C. Patel, and Cesar Farell
Sponsored by
National Science Foundation
Grant No. GK-35795
uHR Report No. 175
Iowa Institute of Hydraulic Research The University of Iowa
Iowa City, Iowa
May 1975
ABSTRACT
The effeöts of surface roughness on the mean pressure distribu-tian and the boundary-layer development on a circular cylinder in a uniform stream have been investigated experimentally. Five different sizes %of
uniformly-distributed sand-paper roughness and several confïgurations of rectangular meridional ribs have been tested over a Reynolds-number range
4 5 -.
7x10 < Re < 5.5x10 . Measurements were also made on a smooth cylïnder
for comparison.
The experimental results .show a large influence of roughness
size and geometry on the mean pressure distribution as well as on the boundary-layer development. In general, the effects of rib roughness are lar to those of distributed roughness. In the case of large
ribs, howevér, rather strong local effects have been..óbserved. Some of
the important results have been examined in the light of boundary-layer
theory and previous data, nd supported by simple theoretical analysis..
In the analysis of .the data, special attentiOn has been gIven to the variations in the drag coefficient and important
pressure-distri-bution parameters with surface roughness at large Reynolds numbers. When
the Reynolds number exceeds a. certain value, which is determined by the
roughness, the pressure distribution becomes independent of Reyñolds numbèr and is dictated Only by the roughness geometry.
The present.study indicatés that su.bstantia]. reductions in the
magnitude of the ininimwn pressure coefficient on large cylindrical structures,
such as hyperbolic cooling towers, can be obtained by roughening the external surface with meridionàl ribs provided the ribs are sufficiently large and.
spaced in an optimum manner. The results also indicate that pressure
distributions On prototypes can be reproduced on scaled models by employing a proper combination of Reynolds number and surface roughness. 'A modelling
A series of experiments was also performed with a rough-walled cylinder and movable side-walls in the wind tunnel in order to study the influence of wind-tunnel blockage on the pressure distribution. The results, described in an Appendix, verify a blockage-correction procedure proposed
previously.
ACKNOWLEDGMENTS
ThIs study was sponsored by the National Science Foundation,
under Grant NO. GK-35795. Support for computer time was provided by the
Graduate College of the' University of Iowa. Proféssor John R. Glover
provided advice and .assistánce in the development of the compúterized mean
pressure data-acquisition system and the associated áomputer programs. Mr. Federico E. Maisch assisted with some of the ekperiments. Dr. Elmar Achenbach
kindly, provided some of his unpublished experimental data. The experimental
equipment was constructed at the Institute shop ünder Mr. Dale Harris's
supervision. All these are gratefully acknowledged.
TABLE OF CONTENTS
Page
LIST op TABLES V
LIST OF FIGURES
Vii
LIST OF SYMBOLs xli
INTRODUCTION i
BRIEF LITERATURE REVIEW 3
EXPERIMENTAL EQUIPMENT AND PROCEDURES 8
3.1 Wind tunnel 8
3.2 Circular cylinder models
3.3 Approach flow and Referenöè velocity 15
3.4 Mean pressure dataacqüisition H 18
3.5 Boundary-layer traversing mechanism 23
3.6 Surface rougInesses 23
3.6.1 Distributed roughness 23
3.6.2 Rib roughness 27
REDUCTION AND PBESENTATIOÑ OF DATA 32
4.1 Mean pressure distributions 32
4.1.1 Smooth cylinder pressure distributions 33
4.1.2 Pressure d-istributions with distributed roughness 33
4.1.3 Pressure distributioñs with rib roughness 40
4.1.4 Analysis and suuunary of mean pressure distributión data 46
4.2 Boundary-layer data 67
4.2.1 Cylinder with distributed roughness 67
4.2.2 Cylinders with rib roughiìess 74
4.2.3 Sununary ¿f boundary layer data 108
V. DISCUSSIOÑ OF RESULTS 112
5.1 Effècts of distributed roughness 112
5.1.1 Dräg coefficient 112
5.1.2 Pressure distribution 114
TABLE OF CONTENTS CONT.
Page
5.2 Effects of rib roughness 126
5.2.1 Drag coefficient 127
5.2.2 Pressure distribution 131
5.2.3 Local effects of ribs 145
5.2.4 Boundary-layer characteristics 152
5.3 Effects of roughness at high Reynölds number 158
5.3.1 Mean pressure distributions 158
5.3.2 Surface roughness and pressure rise to separation - 160
comparison with cooling tower results
5.4 Simulation of high Reynolds-number flows in wind tunnels 162
5.4.1 Reyrioids.number independence 163
5.4.2 Simulation by employing models with larger relative 166
roughness
5.5 Use of externàl ribs, on cooling toser shells 171
VI. SUMMARY 2ND CONCLUSIONS 173
REFERENCES 173
APPENDIX 1. Effects of wind-tunnel blockage 180
APPENDIX 2. Mean-pressure-distr-ibution plots and tables 184
(under separate .cover)
LIST OF TABLES
able
i?age3.1.
Coimnercial naines of sand paper and roughness characterIstics
253.2
Geometrical chracteristics of rib roughnesses
27Rib roughness confïgurations
284.1
Dêterùtinatïon of overall pressure distribution for
45different rib configurations testéd
4.2
Summary of nìean pressure distribution data;
48nooth cylinder
4.3
Sununary öf mean pressure distribution data;
49cylinders with distributed roughness
4.4
Suriry of mean pressure ditibution data;
54cylinders with ribs
4.5
Cylinder with distributedroughness.No:24 boundary-layer data.
69Re = 154000.
(Traverse at 1/8 in. above midsection)
4.6
Cylinder with distributed rougbnes.s No:24 boundary-layer
71data.
Re = 154000.
4.7
Cylinder with distributed roughness No:24 boundary-layer
73data.
Re = 304000.
4.8
Cylinder with ribs Rl bou dary-layer data.
Re = 152,000.
774.9
Cylinder with ribs Rl boundary-layer data.
Re287,000
794 10
Cylinder with ribs R2 boundary-layer data
Re = 118,000
814.11
Cylinder with ribs R2 boundary-layer data.
Re = 295,000.
834.12
Ciìinders with ribs RB-05 boundary-layer data.
Re = 295,000.
. 854.13
Cylinder with ribs RB-b boundary-layer däta.
Re = 295,000.
874.14
Cylinder with ribs RA-05 boundary-layer data.
Re118,000.
894.15
Cylinder with ribs RA-05 boundary-layer data.
Re200,000.
914.16
Cylindér with ribs RA-05 boundary-layer data.
Re = 295,000.
.934.17
Cylinder with ribs RA-10 boundary-layer data.
Re = 152,000.
964.18
Cylinder with ribs RA-10 boundary-layer data.
Re = 295,000.
97LIST OF TABLES CONT.
Table Page
4.19 ylinder with ribs RA-20 boundary-layer data. Re =152,000.
99
4.20 Cylinder with ribs RA-20 boundary-layer data. Re = 295,000. 101 4.21 Cylinder with ribs RC-05 boundary-layer datá. Re = 295,000. 103
4.22 Cylinder with ribs RC-lO boundary-layer data. Re =295,000. 105
4.23 :Cylinder with ribs RC-20 boundary-layer data. Re = 295,000. 107
4.24 Summary of boundary-layer data. Distributed 109
roughness (k/d = 2.66x103).
4.25 Sunmary of boundary-layer data. Cylinders with ribs. 110
5.1 Use of external ribs on a cooling tower shell. s/k = 20. 172
(Weisweiler tower: mean diameter d = 52o5 m, diáxnter at
waist = 44.6 m, height 105.1 m, shell thickness = t = 10 cm).
A.l Effectsof wind-tunnel blockage. Summary of results. 181
LIST OF FIGURES
Figure . Page
.2.1 Drag coefficient (corrected for blockage) of cylinders 4
with distributed roughness Fage and Warsap (1929)
(thin lines), and Achenbach (1971) (thick lines) results.
After Achenbaóh (1974).
2.2 Mean-pressure and skin-friction6distr-jbut.jorj on rough-walled 6
circular cylinders, Re= 3,0x10. (AfterAchenbach (1971).
10
3.1 Wind tunnel and cylinder
(Side view - Vertical Centerplane section).
3.2 Test Section and circular cylinder . 11
3.3 Cylinder in the test sectIon - viéw from upstream 13
3 4 Definition sketch and angular distribution of pressure 14 taps at midsection .
3.5 Distributi9n of normalized dynamic pressure. of the approach. 16
flow (y/y) , (V = 51 fps, x/r = -7.91)
3.6 Longitudinal velocity distribution along tuine1 axis 17
(End of contraction is at x/r = -8.80)
37 . Longitudinal velocity distribution along a line r/r -'3.84 17
Z/r = -1.74. (End of contraction is at x/r = -8.80)
3.8 Säheme of datà acquisition system for the mean pressure 19 distributions
3.9 Arrangement for calibration f mean-pressure measurement 21
system
3.10 Photographs of mean pressuré data-acquisItion equipment 22
3.11 Boundary-layer traversing mechanism, and cylinder, top view
24
3.12 Photographs of sand pa.pers (Flow is from left th right) 26
3.13 Location of ribs relative to pressure taps fQr the rib 29
conf-igurations R2, RA-05, RA-10 and RA-20
3.14 Sectional view of rIbs for configurations RB-lU, RA-10 and .30
RC-lO
3.15 Close-up view of cylinder with ribs RB-05 31
LIST OF FIGURES CONT.
viii
Figure
4.1 Smooth cylinder
pressure distributions in the subcritical
Reynolds number rañge.
4.2 Smooth cylinder pressure
distributions in the critical range
of Reynolds numbers..
4.3
Smooth cylinder pressure distribution, Re= 4.1x105. - 36 (Spanwise variations in pressure coefficient)
4.4 Pressure distributions ón
cylinder with distributed roughness; 37
k/d
l59xlQ3.
4.5 Pressure distributions on cylinder with distributed
roughness; 38
k/d = 6.2lxlO3.
4.6 Pressure distributions on cylinders with distributed
roughness. 39
4.7
Pressure distributjon..on cylinder
3.38x103, e lO°,first rib at 4.8 Pressure distributjön on òylinder 3.38x103, O 5°, first rib at 4.9 Pressure distribution on cylinder
3.38xl03, O 1O, first rib at
4.10 Pressure distribution on cylinder
3.38x103, M
200, first rib at 4.11 Boundary-layer velocity profiles. k/a 2.66xl03. Traverse at 1/84.12 Boundary-layer velocity profiles.
k/d = 2.66x103. Re = 154,000.
4.13 Boundary-layer velocity profiles.
k/d 2.66x103. Re =304,000.
82
4.18 Boundary-layer velocity profiles. Ribs
RB-05.. Re = 295,00Ö. 84
4.19 Boundary-layer velocity pofiles. Ribs RB-lb.
Re = 295,000. 86
4.14 Boundary-layer velocity pröfiles. Ribs
Rl. = 152,000.
4.15 Boundary-layer velocity profiles. Ribs
Rl. = 187,000.
4.16 Boundary-layer velOcity profiles.. R.ths
R2. Re. = 118,000.
4.17 Boundary-layer velocity profiles. Ribs.
R2. Re = 295,000.
with ribs R2 (k/d 41 e = 0) Re =476,000.
with ribs RA-05 (k/d = 42
O ±2.5), Re =-216,000.
with. ribs. RA-10 (k/d = 43
O = ±2.5) Re = 179,000.
with ribs RA-20 (k/d = 44 O = ±12.5) Re = 180,000.
Distributed roughness, 68
in. above midsection. Re=154,000.
Distributed roughness, 70 Distributed roughness, 72 Page 34 35 76 78 80
LIST 0F FIGURES coÑT.
5.1 Drag coeficient of cylinders with distributed roughness.
(Vàlues corrected for blockage)
5.2 Variation of C ànd C with Re,. and kid as pàrameter.
pb pm
5.3 Variation of O with Ré, and k/d as parameter. Cylinders with
distributed roughness.
5.4 Variation of C - C with Re, and k/d as parameter. Cylinders
with distribut ''
roughness. (Symbols same as in Fig. 5.2) 5.5 Boundary layer on a cylinder with distributed rouqhness (k/d =
2.66x103) at two Reynolds numbers in the supercr-itica:l znqe.
5.6 Effect of surfaöe roughness and Reynolds fluer on the boundary- 121
layer velocity profile at or near the location of minimum pressure
coefficient.
5.7 Boundary-layer separatiOn òriterion for à rough-walled circular
cylinder.
5.8 Variation of C with Re and kid for angular rib spacing of 5°.
5.9 Variation of Cd with Re and k/d for angular rib spacing of 10°, 129
5f 10 Variation of Cd with Re and k/d for angula rib spacing of 20°. 130
5.11 Variation of c and C with Re and k/d for angular rib spacing 132
.of59. p lLLL
Boundary-layer velocity profiles Ribs RA-05 Re = 118,000
Boundary-layer velocity profiles. Ribs RA-05. Re = 200,000.
Bourdary-layer velocity profïles. Ribs RA05. = 295,000.
Boundary-layer veloöity profiles Ribs RA-10. Re 152,000.
Boundary-layer velocity profiles Ribs RA-10. Re 295,000.
Boundary-layer velocity profiles. Ribs RA-20. Re 152,000.
Boundary-layer velocity profiles. Ribs RA-20. 295, 000
Boundary-layer velocity profiles. Ribs RC-05. = 95,OOÓ.
Boundary-layer velocity profiles. Ribs RC-lÖ. = 295,000.
Boundary-layer velocity profiles. Ribs -20. Ra = 295,000. Figure 4 20 4.21 4.22 4.23 4.24 4.25 4.26 4.27 4.28 I: 4.29 Page 88 90 92 94 96 98 100 102 104 106 113 115 117 118 120 125 128
5.12 Variation of.0 b
spacing of
5°f
5.17 EffeOt of angular
= l.97x103.
LIST 0F F-IGtiRES' CONT.
Figure
'Page pm'
5.13 Variation of Cpb and C
pm.
with Re and kid för angular rib 134'pacing of 10°.
5.14 Variation of. - C with Re and k/d for angular rib 135
pb pm
spacing of 100.
.5.15 Variation of C and C with Re and k/d for angular rib
136
- pb pm
spacing of 20°.:
5.16 Variation of C C with' Re and. kid for angular rib,
137
spacing of 200pb . pm
5.18 Effect of angular rib spácing
= 3.38x.I03.
5.19 Effect of angular rib spacing on
= 6.47x103.
with Re änd k/d för ang:u.lar rib 133
.rib spacing on C ' ànd C '-. C for k/d 139
d pb pm
on Cd and C = C ....for k/d 140
pb pm
C ' and Cd - C 'for' k/d 141 pb pm
5.20' Effect'of rib spacing on C
b - in the range of Reynolds 143
number independence. p . pm
5.21 Local influence of rIbs (adapted from Fig. 4.2 of Liu, 146 Kline, and Johnston (l966))
5.22 Local influence of ribs. RB-b (k/d = 1.97x103, s/k.= 148
44.2)
5.23 Local influence of ribs. Comparison of résults for R2 and
149
RA-10 (k/d 3.38x103, s/k = 25.8), Re = 4.33xl05.
5.24 Local influence of ribs. RA-20 (k/d = '3.38x10, .s/k =
150
51.6) Re = 304,000 (Results obtained manually during böundary layer measurements, Re = 295,Ó00, are also shown)
5.25 Local influence of rïbs. RC-lO, RC-20 and RC-40.
(k/d = 6.47x103)' .
151
5.26 Boundary layer on cylinders with'ribs. (Only one rib' is
'155
shown in (c), and only one rib for each configuration is
shown in Cd)). .' .
LIST OF FIGURES COT.
Figure Page.
5.27 Boundary-layer velocity profiles near the location 156
of minimum pressure coefficient.
5.27 (ôontinued) 157
5.28 Variation of .0 , C
b and C with k/d at .arge Re. (The 159
value of s/k ishon next
o each. point for cylinders with ribs.)5.29 Pressure rise to separation, C - C , as a function of . 161 pm
relative roughness, k/d, at. large Reynolds number.
Circular cylinders and hyperbolic cooling towers (The value of s/k is shown next to each point for cylinders
and towers with ribs.)
5.30 Reynolds number independence. . 164
5.31 Drag coefficient and minimum pressuré coefficient as a 1.68 function of roughness Reyncids number V k/v.
A.l Effect of wind-tunnel blockage on C-., C C and
c
-c
-- d pm pb pb pm
LIST OF SYQtS
A constant in Equation 5.1
B constant in Eqjzation 5.1
b rib width
drag coefficient C = force! (area x 1/2pv2))
Cf loàal friction coefficient C =
CdR drag coefficient of a rib in a turbulent boundary layer CfR "friction" coeffiôient due to a rib.
Cf nooth-wal1 friction coefficient C pressure coefficient ( = p-p/pv2)
Cb
average base-pressure .cpefficientC minimum pressure coefficient
pm -.
C pressure coéffïcient at separation C pressure coefficjet:at O = 180°.
p,l80
d diameter of cylinder, mean diameter of cooling tower
E subscript denoting edge .of boundary layer
f force per unit length shape factor (*/)
Ht total pressure in the boundary layer relative to stätic pressure of uniform stream
H dynamic. pressure of uniform stream
k rib height, roughness height
i span length of cylinder
p static pressure
p static pressure of uniform stream
P.LOC location of boundary-layer traverse plane
relative to.a rib
Ap differential pressure
LIST OF SYMBOLS CONT.
Reynolds number ( = V d/v)
o
Reynolds number at which mean presure distiibütjon becomes
independeñt of Re
r cylinder radius
s circumferential center-to-center distance between ribs
s1 distance Of first pressure tap downstream from a rib T subscript denoting laminar-turbulent transition t thiOkness of cooling tower shell
u velocity in boundary layer
velocity at edge of boundary layer
velocity at a póint in a plane upstream öf cylinder average velocity in a plane upstream of cylindér
VA velocity along the tunnel axis upstréam of cylinder at x/r= -6.78
V longitudinal cononent of a poténtlal-f].ow velocity upstream of
p cylinder
V longituj component of potential-flow velocity upstream of
p
cylinder at x/r = -7 V00 velocity at infinity
V approach velocity Of uniform stream
w width of wind tunnel test section
x,y,z right-handed Cartesian cOordinate system, x along tunnel axis
positive in downstream direction, z along dylinder axis normal distance from smooth surface of cylinder
nOrmal distánce from top of rouqhness elements
shear stress radient normal to the wáll at separation
AC variation of local pressure coefficient from overall pressure
p
coefficient
Re
r
LIST OF SYMBOLS CONT.
I(u/uE) error in U/UE due to local pressure variation angular spacing of ribs
boundary-layer thickness
boundary-layer displacement thickness
e meridional angle rnesured fröm the stagnation point boundary-layer momentum thiákness
angular location of pressüre minimum
O angular location of separation
approximate angle of beginning of wake region
K Karman constant
V kinematic" viscosity
p mass density wall sheár stress
wäll shear stress at locátion of pressure minimum parameter in Equation 5.1 '
W parameter in Equation 5.2
INTRODUCTION
The work reported here is part of, a wider research program under-taken at the Iowa Institúté of. Hydraulic Research to study the influence
of external surface roughness elements on the charácteritics of méà.n flow past circular cylinders and hyperbolic cooling tower shells. Resülts of experimental investigations of surface roughness effects on the mean pressure distributions on hyperbolic cooling tower models as well as an extensive
literature survey have been reported by Farell and Maisch (1974). Since
there are many similarities between the characteristics of flow past cooling
tower models and circular. cylinders, a more detailed experimental . and
theore-tical study of surface roughness effécts on the flow around circular
cylinders, including mean pressure distrìbutions and boundary-layer
devél-opment., was undertaken as part of this resèarôh program. The experiments
with circular cylinders are described in this report. In addition, a comparison of the essential features of the mean flow past circular oylinders ad cooling towers is made with special atténtion given to the application of present findings for circular cylinders to elücidàte the effects f surface roughness
on cooling-tower pressure distributions. . . .
These stud-ies were prompted largely by a controversy in thé
cooling tower in4ustr' concerning the influence of artificial surface roughness on the mean and fluctuating wind loads on hyperbolic cooling
tower shells. European manufacturers, notably in Germany, claim that thé
wind loads are substantially reduced if the prototypes are roughened extern-ally with vertical ribs or strakes. Their codes of building practice
reflect this claim and result in considerable savings in steel and cOncrete. Some American designers (see, for example, Rogérs and Cohen 1970) favor the use of ribs while others have remained unconvinced about the favorable effects of surface roughness due mainly to a lack of undisputed experimental
or theoretical evidence. Much of the present knowledge on the aerodynamics
of cooling towers has been obtained by means of model tests in wind tunnels at Reynolds ntbers generally two to three orders of maqnitude smaller than prototype Reynolds nÚbers. These model tests have disclosed that external
2
roughness elements, in the form of either uniformly-distributed random-shaped elements or geometrically regular
configurations of ribs or strakes, signif
i-cantly reduce the magnitudes of the negative pressures on the sides of the models (Niemann 1971, Fare].]. and Maisch 1974).
Such elements are therefore favorable if they can be shown to have a similar effect on prototype
structures. Furthermore, experiments with models
fitted with external roughness elements have produced mean pressure distributions representative of the
much greater Reynolds number flows past smoother-walled prototypes. There
is therefore a possibility of simulating prototype loading conditions in wind tunnel tests for the purposes of experimentally investigating the static and dynamic response of cooling tower shells.
In view of the foregoing considerations, the present research
program was undertaken with three main objectives: (a) to ôlarify the influence of surface roughness, especially at large Reynolds numbers, and investigate
the feasibility of simulating prototype conditions in wind tunnel experiments;
(b) to determine by means of systematic experiments the influence of different
types, sizes and configurations of external roughness elements on the mean pressure distributions so as to ascertain their relative merits for use on prototype structures; and (c) to identify the physical mechanisms responsible
for the observed roughness effects and elucidate the various observations by
theoretical analysis. The experiments on cooling tower models reported by
Farell and Maisch (1970) have verified that there is indeed a strong
favorable effect of surface roughness on the mean pressure distributions, and that wind-tunnel tests can be used to simulate the static wind loading of
prototype structures. Experiments with simple circular cylinders described
in this report, supplemented by theoretical boundary-layer and potential-flow analyses, to be described in greater detail in a separate report, also confirm these f ìndings. These experiments consisted of measurements of mean pressure distributions, as well as mean velocity profiles in the boundary layers on circular cylinders in a uniform stream in a large low-turbulence wind-tunnel
4 5 .
over a Reynolds number range: 7x10 to 5.5x10 . Several sizes of
distributed-type roughness, provided by commercial sand papers, and several sizes and
configurations of ribs, modelled by means of flat wires of rectangular section,
3
The results of the present study, as well as those of sorne other
recent investigations, also indicate, a close connection between the
character-istics of the bouhdary layer and the mean pressure distributions. Although
the overall effects of distributed and rib roughnesses on the mean pressure distributions are qualitatively similar, there are a number of important differences in the details of the flow due to the local disturbances caused
by the ribs. These differences warrant a careful interpretation
of the
experimental data and reqùiré modification Of the.usuaÏ theoretical treatment of rough-wall boundary layers in order to cOnsider rib roughness. Atx attempt is made here to elucidate the basic differences betweeñ the two types of
roughness.
It must be noted that the, mean pressure distributions on cylindrical
structures and hyperbolic cooling towers (as well as the statistical
proper-ties of the. pressure fluctuations) depend not Only on the Reynolds rnmiber and
surface roughness, but also on such factors as the mean velocity distribution and turbulence characteristics of the free stream, the presence of other large structures in the vicinity, and wind-tunnel blockage in the case
model tests (see, .g. Farell 1971), and even the span-to-diameter ratio (see,
e.g. Acheribach 1968). Since wind tunnel blockage is of particular
importance,
a series of experiments wïth a rough-walled circular cylinder was also made in the same wind tunnel to study the influence of the proximity of wind tunnel side walls on the.mêan pressure istributions. Thése experiments are described in Appendix 1 and verify the correction procedure proposed by Psh]ço (1961)
on the basis of the method of 'Allen and Vincenti (1944).
II. BRIEF LITERATURE REVIEW
The first study on thé effects of surface roughness on the flow past circular cylinders was made by Fage and Warsap (1929). In this well
known work, they measured the drag. coefficieñts' of cylinders covered with
roughness of the distìibuted type over the critical and supercritical
(see Achenbach (1971)) range Of Reynolds numbers. They also studied thé
effects of a pair of generator wires on the pressure distributions and the
effects. of grid-generated turbulence on the drag coefficients.
Partial
accounts of their wOrk caîi also be found in Goldstein (1938), and Schlichting
(1968). Fage 'and Warsap found a systematic effect of surface
roughness on
1.2 1.0 0.8 Cd 0.6 0.4 0.2 2x104 I
II
0. ReFigure 2.1: Drag coefficient (corrected for blockage)
of cylinders with distributed roughness.
Fage and Warsap (1929) (thin lines), and
Achenbach (1971) (thick lines) results.
After Achenbach (1971).
I I t I
5
particular, they attributed the increase of drag coefficient with roughness in the supercritical Reynolds nttÍer range to retardation of the boundary-layer flow by roughness and, hence, earlier separation. They mentioned also
that "It appears,1. .when the surface is very rough the flow around the
relatively large excrescences, and so around the cylinder, is unaffected by a change in a large valué of the Reynolds number."
It was not unt-il recently that another systematic study was
published on the effécts of roughness on circular cylinders. Aòhenbach (1971)
reported measurements of pressure and skin-friction over a Reynolds number range which extended up to Re = 3x106. His measurements shöwed, among
'i other things, that in the trancritical range the drag côeficient is
inde-pendent of the Reynolds numbér, as suggested earlier by Fage and Warsap, and only a function of the relative roughness kid. This can be seeñ frôm hi results which are reprOduced in F-ig. 21. Furthermore, although detailed
boundary-layer developments were not. measured, his Skin-friction results
showed the close connection between the pressure distributions and the boundary
layer behavioro Fig. 2.2 shows the skin friction and pressure distributions
for k/d = 1 1x103 and 4 5x103 at Re = 3x106 It will be seen that the larger roughness results in greàter'retardation of the boundary layer (hiqher skin friction), earlier separation and a larger magnitude of thé base-pressure
coefficient. It is also of intérest to note here that the pressure
distri-button is affected not only by the locatioñ of separation but also by the boundary-layer development ahead of separation. For example, separation was found to occur at O = 110° for kid = l.1xl03 at Re 4.3xl05 and for kid =
4.5xlO3at Re = 3.0x106 but the pressure distributions,and consequently the drag coefficient,were fOund to be cönsideràbly different. This can be seen from Fig. 10 of Achenbach's original paper and can be attributed to the differences in the boundary-layer development ahead of separation.
A careful examination of Fig. 2.1 shows that there is a remarkable
difference between the results of Acheithach and the eaxlier ones due tO
Fage and Warsap. It would be see that the valués of thé drag coefficient
measured by Fage and Warsap under nearly similar roughness and Reynolds nber conditions are considerably lowér than those of Acherthach in the
lOOxT
2 w/pV01.0
0.0
-1..
0 -2. 6 Static pressure 30 60 90 120 150 180 eFigure 2.. 2: Mean-pressure and skin-friction distribution on
rough-walled circular cylinders, Re 3x106. (Af.ter Achenbach (1971)).
7
C values for both k/dxlO3 = 4 and 7 are much lower than Achenbaôh's
3 ''t'
.5
values for k/dxlO = 4.5, at. Re 2.8x10 . As will be shown later on in
this report, however, the Cd values are. expected to be quite similar for
these roughnesses at such high Reynolds numbers. One pOssible reason for
this discrepancy can be fotind in the expeiirnental arrangement of Fage and
Warsap. In their study they used a 40-in.-löng cylinder suspended from a
drag balance in à 48-in.-wide test section. Two extension pieces of saine
diameter filled the remaining port-ion of thespan but 1/8-in. gaps were left between the test cylinder and these extension pieces. Furthermore the
span-» to-diameter ratio was 2Q2 or 7.88, depending on the diameter of the two
cylinders they used, as compared to 3.33 in the expetiinents o Achenbach.
Fage and Warsap point out that their results may have been affected by the
gapso Indeed,, with such gaps, the wake of the cylinder is supplied with high presüre flu-id from the front and as a result smaller values of Cd are expected mce the base. pressure is increased over the value it would
other-wise obtain. In addition, it is generally observéd that. values of Cd are
smaller for cylinders with larger span -to-diameter ratio. Both the presence
of the gaps and the larger value of l/d could therefore have resulted in the
lower drag coefficients. n the subcritical range of Re, however, these
effects appear to be negligible. Indeed, Morsbadh (1967) found that in the subcritical range there is no effect Of san-to-dïameter ratio. Due to these uncertainties concerning the experiments of Fage and Warsap further comparisons .ith their results are avoided in this report.
More recently, Batham (1973) has 'reporte6 experiments on the
effects of surface roughness of thé distributed type (k/a. 2.l7xlO3') and
free-stream turbulence on the mean and fluctuating pressure distributions on
circular cylinders at t Reynolds numbers (Re l.11xlO5 and Re = 2. 35x105),
and Szechenzi (1974) has made a study in which he measured steady drag coefficients and unsteady lift coeffiçients of rough walled c'1inders over a range of Reynolds numbers up to Re = 6.5xl06. Both investigators were interested in simulating the pressure distributions at high Reynolds numbers. in particular, Szechenyi (1974) plotted thé drag coefficient against roughness
Reynolds number Vk/v, and suggested that, in the supercritical flow regime,
values of kid = 1.6x104 to 2x103. (Incidentally, this roughness Reynolds number was also suggested by Armïtt (1968)). As will be discussed more fully later on, however, this observation is at variance with the previous as well
as present findings.
In the foregoing, we have mentioned briefly those studies dïrectly related to the problem at hand; and emphasized the effects of surface rough-ness on the mean flow past circular cylinders. A more extensive review of roughness and other effects on the flow past circular cylinders can be found,
for example, in Farell (1971),. and in the E.S.D.U. (1970) data item.
While the effects of roughness of the distributed type on circular cylinders and the effects of the rib-type roughness on cooling towers and cylinders of finite length have been studied in some detail there is very
little information at'ailable. on the effects of rib-type roughness on long
cylinders (i.e., cylinders without a free end). A comprehensive study of
rib roughness, therefore, forms an important part of the present invest-igation.
III. . EXPERIMENTAL EQUIPMENT PND PROCEDURES
3.1 Wind Tunnel
The experiments were conduôted in thé largest low-turbulence wind tunnel of the Iowa Institute ôf Hydraulic Research. The original 24 ft.-long, 5 ft.-octagönal test section of the tunnel was modified for the present.study, as described below, in order to achieve two-dimensional
flow. The turbulence intensity Of the approach flow after the tunnel modification. was 0..2 percent.
Tests made in the initial, phases of the study with ä smooth
ôylinder mounted vertically in the original octagonal section revealed a rather complex three-dimensional flow pattern on and àround the cylinder. These tests consisted of measurements of the mean pressure distribution on the cylinder, measurements of velocity profiles in the wake at three different elevations, and flow visualization by means of wool tufts. Strong
cross flow ina direction away from the midsection were observed in the
boundary layer of the cylinder. The velocity profiles in the wake also
exhi-bited strong three-dimensionality. For example, at a free-stream velocity öf 70 fps, the velOcity at the tunnel axis 5.07 cylinder diameters behind the cylinder was 51 fps, whereas the velocities at 0.66 diameter above and
below this point weré both 63fps. Some asynétry was also obsered in the pressure distributions on the cylinder.
In an attempt to eliminate the boundary-layer cross flows,
fences were placed around the cylinder at. levels about i. i cylinder diameters
above and bélow the mi.dsection. Meañ pressure distributions obtained with
i these fences did not show any substantial improvement.
The use of base plates was then attempted. Although these reduced the three-dimensionality, they seemed to affect the approach flow conditions in a complicated manner It appeared that the velocity of the plow between the plates was higher than the velocity above and below In order to achieve two-dimensionality and to eliminate uncertainties about the reference velocity and approaòh flow
conditiozis, it was finally dec±ded to implement a major modification of
the wind-tunnel test section.
The originalj 24 ft. long test section was modified as shown in
Fig. 3.1. There is now a 6 ft. long contraction leading
to a 95 ft. long rectangular test section, followed by a 8.5'ft. long diffuser. The present test section has a width of 5 ft. and a height of 32.855 in. The floor and ceiling o the test section intersect the inclined faces of the original octagonal section as shown in Fig. 3.2.
As a result of the modification, the th±ee-dimensionaiity induced by the original octagonal section was removed and at the same time the maximum velocity in the test section was increased to about 120 fps from
the original 90 fps. Tests carried out after the modification showed that
the approach flow was uniform across the test section. These tests are
described in section 3.3.
3.2 Circular Cylinder ModelS.
Two circular cylinders, each with diameter d = 10.65 in., have
been used in this study. Two süch cylinders were constructed so as to
mini-mize delays in data collection while the surface roughness on one of the cylinders ;was.being replaced. The cylinders were turned on: a lathe from an aluminum pipe, 10.75 in. nominal diameter. The resulting surface texture was smooth to the touch and further tests indicated that the surface was hydrody-nainically smooth. Fifty-three pressure taps were drilled at the midsection
of
Contraction 72.00 in. 46.875 in. Cylinder Flow Direction Axis Test Section Diffuser 114.00 in. 102.00 in. 24 ft. Figure 3.1:
Wind Tunnel and Cylinder (Side View - Vertical Centerplane Section)
-Cylinder Support Coupling 4.', Il 60.0" 24.855" 10.65" Midsection
-4"
Level -8" LévelF:igure 3.2: Test Section and Circular Cylinder
Floor 1/8 in. I.D. TYGON Ttthing C4 Lfl ir, 'o '-1 JOint Ceiling 14.5" 4.'9 +8" Level +4" Level -. p.
12
a total of four levels above and below the midsection at ± 4 in. and ± 8 in. in order to assess the two-dimensionàlity of the flow. The circumferential distribution of the holes at the midsection is given in Fig. 3.4. All
pressure taps had a diameter of 0.040 in., and 1/8 in. inside diameter plastic Tygon tubing was used to transmit the p±éssures to the mean-pressure
measure-ment system described in section 3.4 'below. ..
The cylinders were built in two sections to facilitate the
construc-tion of the pressure taps. The joïnt was 14.5 in. above the midsection and
was sealed with silicone grease. Care was taken to ensure that there was no offset or misalignment öf the two sectjons at the joint. A sketch of
the cylinder and test sectioñ including only the important 'features and
dimen-sions i,s shown in Fig. 3.2. A photograph of the cylinder in the test section
taken from upstream is given in Fig. 3.3.
The blockage ratio, d/w, where d is the cylinder diameter and w
is the wïdth of the. test.section was d/w = 0.178. In the cylinder experiments
of Achenbaöh (1968) and in some of the eperiments'of Fage and Falkner (1931),
the blockage, ratios were 0.166 and 0.185, respectively.
The cylinder axis was located. 46.875 in. from'the end' of the contraction
as shown in Fig. 3.1. The midsection of the cylinder was set at about 1/8 in.
below the horizontal centerplane of.the tunnel. The cylinder was supported at the bottom by a board underneath the working section of the wind tunnel and it could be rotated on this board around its axis. Additional supports were.
provided outside the tunnel floor and ceiling to securely fasten thecylinder after its orientation relative to the oncoming flow was adjusted. During the early phases of this investigation the cylinder was oriented relative to the oncoming flow b' first roughly aligning the O = 00 generator (9 is defined in Fig. 3.4) with the vérticàl centerplane of the tunnel and then rotating the cylinder until the pressure reading at 9 = 0° was maximum. As revealed later by the pressure distribution resüits, this procedure resulted in an
error of the order of ±3°. This is primarily due to the fact that the pressure
distribution close to the stagnation point is not very sensitive to angular
position. (The correct angular positions at the pressure holes relative to
the flow direction are considered and reported in this work.) ' A better
procedure was. followed to orient the cylinder in the later phases of this
study during which' the data with rib roúghnesses were Obtained. The cylinder
13
Flow
14
Pressure taps át 50 intervals Pressure tars at 10° intervals
Figure 3.4: Definition Sketch, and Angular Distribution
15
3.3 Approach Flow and Reference Velocity
Velocity measurements were made with the cylinder in place to check the uniformity of the oncoming flow after the new test seôtion was
installed. Velocity traverses were taken at a section 42.125 in. (3.95d)
upstream from the cylinder axis and the normalized dynamic pressure distri'
bútion (V/)'2 is depicted in Fig. 3.5. Here, denotes the average velocity
at the section, which was 51 fps. Similar measurements with. V 105.64 fps were made by Maisch (1974) 'with a hyperbolic coolïng tower model 'at a section 42.250 in. upstream of the model, axis, and similar results were
obtained. The data show á sufficiently uniform approach velocity distribution.
Iñ addition to the méasurement of approach velocity distribution, two sets of velocity measurements were made in the longitudinal direction to
detérmine the position where the referenòe velocity and pressure. should be measured,: one along the axis of the tunnel (y=0, .z=0), the other along a
line where y/r = -3 84, z/r = -1 74, where (x,y,z) is a right-handed Cartesian
coordinate system, with x along the tunnel axis ïn the flow direction, 'z
along the cylinder axis upwards, and the origin at the horizontal centerplane of the tunnel; and r is the radius of the cylinder. These experiments were also made at an approaôh velocity'of VA = 51 fps wheré VA is the velocity along the t,,,el axis at distance 36.12,5 in. (x/r = -6.78) upstream of
the áylinder axis. The normalized dynamic pressure (V/VA)2 distributions are
shown in Figs. 3.6 and 3.7.. Also included in these figures are the
longi-tudinal velocity' variations (V/V and (V/V72 corresponding to potential
flow, where V is the free-stream velocity 'at infinity, V is the
longitud-inal component of p0tential_f low velocïty, and V7 is the value of V at
x/r = -7. The measurements were ta,ke starting at a point 4.75 in. upstream
from the end of the contraction (x/r. -9.69) and both distributions show
therefore an increase in velocity due to the contraction, f011owed by a plateau from about x/r = -8 0 to x/r = -6 0 A corresponding decrease
or 'increase in velocity is then seen due to the. presence of the model. On
the basis of these results and a comparison with the poténtial-flow velocity distrIbution, it was decided to measure the reference velocity and pressure at a point 7.75 in. downstream from the end of the contraction (x/r -7.34) and 6.825 in. above the test section floor in the tunnel centerplane. This reference velocity will henceforth be denoted by V0, and the, reference
16
\\
0985
1.0J4.0.998 ,1.009 1.1.006
.1.006
+1.007i.1.006
.. 1.007 +0.992
+ 1.019
/
+1.013+1.002
1.009 + 0.990
+ 1.007
4.l.0I01.006
1.023 + 0.994
+ 0.995
+ 0.993 +0.999
0.992
+ 0.975
0.995
4.004+0998
+ 1.014 +1.007
1.010 0.999
+ 1.001
0.991
+0.999
+0.991. 0.983
0.987
41.003.1.006 + 1.002 +1.004
+ 1.000
+1.011
+ 0.994 +1.011
+0.993
.+0.9941.1.016
+1.004 +1O15
+ 1.001
+1.000
+1.001 +1.008.
+1.007
/
1.05 1.00 0.90 0.80 1.05 '1.00 0.90 0.80 li -10 -6
x/r e.
Figure 3.7: Longitudinal velocity distribution along a line
y/r = -3.84, z/r = -1.74. (End of contraction
is àt x/r = -8.80.)
-10 -9 -7 -6
x/r
-Figure 3.6: Lôngitudinal velocity distribution 'along tunnel axis
18
pressure by Po. The reference velocity and pressure were measured by means
of a Prandtl type Pitot-static tube of 0.125 in. oütside diameter in con-junction with a micro-manometer with a precision of 0.001 in. alcohol. The velocity of the approach, flow was constantly monitored for steadiness during
the course of.each experiment. The air temperature in the tunnel and the
temperature in the vïc-inity of the alcohol manometer were also monitored.
These ' together with the barometric pressure and. dry- and wet-bulb
tempera-tures in the laboratory, were. used to determine the approach velocity
and kinematic viscosity V and hence the Reynolds number Re. = V0d/v, in the manner described by Naudascher (1964).
3.4 Mean Pressure Data Acquisition
The. Institute's IBM 1801 Data Acquisition and Control System
was used to obtain the mean pressure dàta. A schematic representation of the overall arrangement is shown in Fig. 3.8. The pressure tubing from the cylinder and the wind-tunnel reference Pitot-static probe were connected
to the terminals of a 48-terminal scanning valve. : The scanning valve was
driven, by a solenoid drive controlled by a solenoid controller. The solnoid
controller can be operated manually or- automatically through the IBM 1801
System. In automatic operation it steps the scanivalve at prescribed time
intervals so that each terminal is scanned in succession and the pressure at
each terminal is fed -to à pressuré transdúcer. The signals from the pressure
transducer are passed through a Model 2850 v-2 DANA a.ntplifïer with a low-- pass filter set at 0.010 kHz bandwidth,and monitored, averaged, and recorded
by the IBM 1801 System. During the experiments, an averaging time of 5
seconds was employed at each terminal and, a. waiting timè of 0.6 seconds was
used to allow for the damping out of transients due to the switching before
thé averaging began. The waiting time was based on- the response of the
set-up to a step input of pressure at a scanivalve termiña.. The time- of rise was found to be about O.3'seôonds. The 5-seconds averaging time was found to be sufficient for the detçrmination of the mean pressures based on
preliminary experiments.
One -of two Stathain PM5TC differential pressure transducers, with ranges of ±0.15 psi and ±0.30 psi, was' used depending on the magnitude of
the approach velocity: the former.giving accurate resúlts for velocities less
- --Pressure
tubes
from
modeli Scanivalve and
position
¿ transmitter
Scónco
48D3-1/STM
19TQ pressure transducer
ScancoDS3-48-24vdc
Solenoiddrive
Scanco
ctLR2/s3
solenoidcontroller
IBM 1801
Data Acquisition and
Contröl System
Figure 3.8: Scheme of Data Acqu-is-ition System for the Mean
Pressure Distributions.
Statham
PM5TC
differential
pressure
transducer
i
Amplifier and
bridge-balancing
circuit
20
The reference lead of the pressure trandider was connected to the wind tunnel statïc pressure through a pressure chamber. Such a chamber was necessary since the dynamic pressure of the approach flow was monitored
constantly during eaòh experiment. A similar chamber was used for the total
pressure, which was connected to one of the scanïvalve terminals. W-ith
this arrangement all the pressures were measured relative to the wind tunnel
reference static pressure p0. The pressure data were finally obtained in
the form of punched cards for subsequent ana1yis on the IBM 360/65 computer. The pressure taps on the cylinder were scanned in sequence in a counter-clockwise manner (see Fig. 3.4), starting from the tap at 9 0, and the
last two terminals of the scanñing välve. were used fOr the .reference total
and static pressures. In most of the experiments, the maximum variation
of the reference dynamic pressure during each test, periOd (about 4.5 minutes)
was less thin ±2%. In the few cases where 'a drift in the wind tunnel speed
was observed"only 'the pressure measurements obtained for the west 'sideof
the,cylinder (negative angles) were considered in subsequent data 'analyses.
If the drift was more than 2%, the ecperiment was discarded altogether. The
cause of the drift was traced to a defective circuit in the servo-control mechanïsm of the wind tunnel drive and periodic maintenance work was necessary
to correct the situation.
Before each series of experiments a static calibration of the
system was obtained by applying known pressures to a scanivalve terminal.
and examining the (typed) output from the IBM 1801. The calibration curve was linear In order to provide the desired calibration pressures, a simple
apparatus was designed which essentially consisted of a flexible U-tube partially filled with water and a small-volume pressure chamber connected
to one end of the U-tube. By moving the' U-tube up or down the desired
pressures were generated in the chamber due to small volume changes of the
trapped air. The overall calibration arrangement is shown in Fig. 3.9. 'A
photograph showing, some. components of the measurement system and the:
calibra-tion apparatus is given in Fig. 3.10. (The boundary-layer traversiíig
21
Flexible J-tube and Stand
r Alcohol Manòxnetèr
LAPr
Pressure Chamber.I-r
.aIBM 1801
Scanivalve P. TransducerFigure 3.9: Arrangement for Calibratioñ. of mean-pressure
=
A
22
Figure 3.10: Photographs of mean-pressure data-acquisition
23
3.5 Boundary- Layer Traversing Mechanism
The boundary-layer tötal-pressure and mean-velocity profiles
reported in section 4.2 were obtained by mean s of stagnation tithes made from flattened hypcdermic needles and supported or the traversIng meòhaÁiiSm
described by Patel et al. (1973). The essential features of the traversiñg
mechanism are shown in Pig. 3.11. The mechanism consists of,a rigid rod
mounted on a slide sïtuated outside the tunnel, arid provides the rod with
three different modes of motion in the horizontal àenterplane of the tunnel: -motion along the length of the rod, motion along the slide situated outside
the tunnel, and rotation about a pivot oñ. this slide. The rod enters the
tunnel through a narrow slit out Of the tunnel wáll. The qrtión of the slit
not occupied by the rod. is sealed by a rubber sheet to prevent air leakage.
With this arrangement it is possible tO thke traverses in the direction normal
to the cylinder surface at any desired station between.O 65° and O 120°..
The normal distance from the cylinder Surface can be adjusted and measured from outside the tunnel with a esolution of 0.001 ft. The boundary-layer probe supported by the traversing mechanii was so òonstriicted that méasure-ments were madé in a plane 1 in. above the centerplanê for reasons explained in
sectiOn 4.2. The pressure dist±ibútion on the cylinder was not affected by
the preseñce of the probe. The total pressure from thé probe ws measured
by means of an alcohOl micrO-manometer. -
-3.6 Surface roughriesses
3.6.1 Distribúted roughness:
The distributed surface.rouqhnesses used in this. study were
commercially available s paper purohased f:rom the Norton Ço. and the 3M Co. The commercial names of the various kinds of sandpaper used and the average particle sizes k, as quoted by the manufacturers, are summatized
in Table 3.1.1 Also included in Table 3.1 are the relative roughnesses kid
based On the ooth cylinder diameter d (= 10.65 in.). Closeup photographs of these sandpapers are given in Fig. 3.12.
It should be noted that the value of k are reported differently in (18)
where they were estimated on the basis of the grit numbers Information
24
Figure 3.11: Boundary-layer traversing mechanism, and cylinder;
25
The sand paper was carefully wrapped around the cylinder in two
pieces,. leaving a gap of 1/16 in. above and below the center of the pressure
holes at the main measuring section. Double-sticking tape was used to stick the paper with the seam located at the rear of the cylinder. Care was taken to ensure that the paper fit snugly around the cylinder. The.thickness of the various papers, together with the double-sticking tape, varied from about
0.03 in. to about 0.065 in. On account of these small thicknesses, Reynolds
numbers were calculated on the basis of smooth cylinder diameter.
Table 3.1: Commercial names of sand paper and roughness characteristics
Commercial Name Roughness size k (mm) Relative roughness 3 k/dxl0
NOR'ION-ReSiflall, Adalox Paper, 0.430 1.59
Closekote Aluminum Oxide, Grit 40-E
NORTON-Resinall, Adalox Paper, 0.535 1.98
Closekote Aluminum Oxide Grit 36-E
NORTON-Resinall, Durite Cloth, Type 0.720 2.66
3, Closekote Aluminum Oxide, Grit 24-S
3M-Resinite, Floor Surfacing Paper 0.960 3.55
Type F Sheets, Open Coat, Grit 20-3½
3M-Resinall, Floor Surfacing Paper 1.680 6.21
i i i :1 Norton Co.
#36
3M Co. #20.
26 i 21 NortonCo. #40
NortonCo. #24
3M Co. #12
Figure
3.12:
Photographs of sand papers.(Flow is from left to right. Scale is in inches)
3.6.2 Rib roughness
The geometrical characteristics of rectangular wires used to obtain the various rib roughnesses are summarized in Table 3.2. These wires were purchased from the New England Wire Co..
Table 3.2 Geometrical Characteristics of Rib Roughnesses 27
* Height, k, includes thickness of the layer of adhesive of approximately
0.003 in.
The wires were glued along the cylinder (by means of Eastman 910 Adhesive) symmetrically about the leading generator (0=0) at equal angular
spacing 0. Several different configurations thus obtained are summarized in
Table 3.3. Included in Table 3.3 is a rib-roughness configuration code,
together with the corresponding rib type number, relative roughness height, location of the first rib in relation to the leading generator, angular
spacing 0, total number of ribs, and, also circumferencial spacing s,
and spacing ratio s/k. In all cases the ribs spanned the whole length of
a generator, except for the cases with the configuration codes R]., Ril
and P2. In the cases of Rl and R2, a gap of 1/16 in. (total 1/8 in.)
was left above and below the main measuring section, and in the case of Ru the gaps at the angular locations O = ±60, ±70'& ±80°were closed by glueing
additional pieces of wire. The relative locations of the ribs with respect
b
to the pressure taps along the circumference (between 0=-70 and 0=-95), for the rib configurations P2, RA-05, RA-10 and RA-20 are shown in Figure 3.13. Sectional drawings are presented in Fig. 3.14 for further illustration for the rib configurations RA-10, PB-lO and RC-lO. A close-up photograph of the cylinder with the RB-05 rib configuration is given in Fig. 3.15
i 0.021 0.036 1.97 1.71
2 0.036. 0.066 3.38 1.83
3 0.069 0.132 6.47 1.91
Rib Type Height*,k Width, b k/dxl b/k
Config-. uration Code Rib-type Number
.
Table 33 Rib Roughness Configurations
Circum-.
ferent:ial Spacing s (in)
Spacing Ratio s/k
Relätive Roughness k/dxlO3 First Rib at Angular Spacing 1O Total Number . of Ribs Ri I 1.97 0.0' 10 36 0.929 44.2 Ru R2 RA-05 i 2 2 1.07 3.. 38 . 3.38 0.0, .0.0 ±2.5' . , iO 10 5 36 36 72 0.929 0.929 0.465 44.2 25.8 12.9 RA-10 2 3.38 ±2.5 lO 36 0.929 25.8 RA-20 2 3.38. ±12.5 20 18 1.859 51.6 RB-OS i 1.97 ±2..5 5 72 0.465 22.1 'w-10 i 1.97 ±2.5 10 36 0.929 44.2 RB-20 i 1.97. ±12.5 20 18
1859
88.5 RC-05 3 6.47 ±2.5 5 72 0.465 6.7 RC-lO 3 6.47 ±2.5 10 36 . 0.929 13.5 RC-20 3 6.47 '±12.5 20 18 1.859 26.9 RC-4o 3 6.47 ±12.5 . 40 . 10 3.718 53.9fti
Pressure Tap R2°U°'U°Ü°
flo.fl
RA-05oflo
o
(C) RA-10o
o
o
o
Cd).. RA-20Figure 3.13: Location of ribs relative to pressuré taps for
rib configurations R2, RA-05, RA-10, and RA-20. 29 .929 in. O66 in. Scale 2:1
fl:
0 1/8
in. .0o
o
e= _750o
o
-85°
D
.30 RB-10
r77!?!77
RA-10RC-lo
-0.0
0.5 inòh 1.0 I scaleFigure 3.14: Sectional view of ribs for confIgurations RB-b,
31
32
IV. REDUCTION AND PRESENTATION OF DATA 4.1 Mean Pressure Distribution
Mean pressure distributions were obtained over a Reynolds number range of 7x104 'to 5.5x105 with the uooth cylinder, with each of the five
different sand papers listed in. Table 3.1, and with each of the rib conf
1g-urations listed in Table 3.3. The detai1edresults of the experiments have
been compiled in a. rather lengthy append-ix (Appendix 2)*to this report. This contains the computer plots and. tables of the variation of the pressure
coefficient C with the angular position O. Iñ theeplots, the data pöints
which belong to east and west sides of the cylinder (positive and negative
angles, respectively) are plotted with différent symbols in order to
illus-trate the symmetry of the mean flow. The pressure coéfficient C is defined
2 p
in the usual Ùanner:. C p - p/½pV, where p is. the pressure on the cylinder
at thé angular position O and p, Po and V are the mass density, static prssure and velocity óf the approach flow respectively. The data reported in Appendix 2 have not been corrected for blockage. The computer plots
were obtained by means of Simplotter, a high level plotting system ('Scranton
'and Màndhester, 1973). Ìi interpolation, mode, whïch made use of a
second-degree Lagrangian interpolation polynomial., wa selected as best suited to
draw curves through the data points for the cases of the smooth cylinder,
cylinders with distributéd roughness. and cylinders with the rib configurations
Rl, Ru and R2. - Due to the nature of the data (see Section 4.1.3), curves
were not drawn for the remainïng rib configurations, and only the data points
were plotted. Data points which were considered to be "bad" were disregarded
in the constructiön of the curves but are shown in the plots and given in
the tables. A "bad" point is one belonging to a pressure tap which consistently
g-ives a result removed from the other points, due. to, for example, a clog in
the measurement system, as revealed-by later ecamination. There was, at most,
one such point in some experiments. For the curves constructed for the
cases of' rib configurations Rl, Ru and R2,. the data points affected by the
local influence of the ribs, discusedtiength later on, were also
disregarded.
Owing to its length, Appendix 2 is produced under separate cover and can be
33
4.1.1 Smooth cylinder pressure distributions
Typical pressure distributions for the smooth cylinder are presented
in Figures 4.1, 4.2 and 4.3-. Some of the results Of Achenbàch (1968),
Batham (1973), añd Fage and Fàl-knèr (1931) are included in Figures 4.1 and
4.2 fOr òomparison. Except for the results of Batham, the data shown in
these Figures have not been corrected for blockage. There is of coúrse -a
large amount of data reported in the literature for smooth cylinders within the Reynolds-numbers range of this study. The available data are, however,
not cônsistènt, especially in the critiòal and supercritical Reyno1dsnu±nber ranges, due to the differences in the surface texture Of the different
cylinders and alo to the differences -in the free-stream turbulence chàrac-teristics and blockage ratios of differenttunnels. In the present study, it was therefore considered necessary to obtain the smooth cylinder data so. as to establish a useful reference for the effects of roughhèss. At the
saxnè -time, these experiments served to assess the degree of twö-dimensiona-lity
and to verify the experimental set-up and procedures. It òan be seen-from -Fig. 4..3 that the flow over the middle half (8 in, above -axd below -midsection)
of the cylinder is reasonably two-dimeñsional insofar as the pressure
-coefficient is sibstantial1y constant along the span. As ±nd-jcate earlier, -the two diÈensionality of -the mean flow was also verified -by making measurements
of velocity profiles in the wake at severäl spanwise statiOns. -.
4.1.2 Pressure Distributions with Distributed Roughness .Tyicai pressure distributions with distributed roughness,
uncorrected for blockage effects, ae presented in Figurés 4.4, 4.5 and
-4.6 for purposes of a general comparison. Included in Fig. 4.6, are some
results of Achenbac-h (1971) and Bàtham (1973) (the latter inclúde blockage
corrections). -A preliminary examination of these f-igures reveals the influence
of both the surface roughness and Reynolds number on the mean pressure
distri-butions. Detailed discussion of-thése effects and comparison with the
Figure 4.1:
Smooth Cylinder pressure distributions in the subcritical Reynolds-number range.
-1
-1 S .e 120 80 40 i'yO
120 160 180 80.1
II
/
-1 .' II III»ORe
2.07 ORe = lO5 Re = 1.11 x 10 (Achenbach, x l0 (present 1968) (Batham, expt.) 1973)O Re=3. 57x105 Re=3; 35x105 Re=2,. 39x105 180 160 I
(present experiment) (Fage & FaÏkner,1931:) (Batham, 1973)
Figure 4.2:
Smooth cylinder pressure distributions in the critical range of Reynolds numbers.
f
f
j
80 40 1 -3 C £ +8 in. level +4 in. level O Midsection
V
-4 in. level 40 80 I 120 ! I o-*
Figure 4.3:Smooth Cylinder Pressure distribution, Re = 4.1 X lOs. (Spanwise Variations in Pressure Coefficient)
160 l8 1.80 )60 120. I I
Figure 4.4:
Pressure distributions ori cylinder with distributed
roughness
1 I I I k/d 6.21
D
Re = 0.86 x 10 O Re = 5.16 x lO5 1ÒO 160 120 80 40 Figure 4.5.: Pressure. distributions oncylinder with distributed roughness
k/d =
O
k/d = 2.66 x 1O, Re = 2.14 x 10
(present expt)
-kid = 4.5 x 1O, Re = 1.7 x 10 (Achenbach,1971
A kid = 2.17 x iO, Re
= 2.35 x
(Batham, 1973)
Figurè 4.6:
Pressure distributions on circular cylinders with
diétributed
40
4.1.3 Pressure Distributions with Rib Roughness
The mean pressure distributions obtained with ribs are different in detail from the pressure distribùtions.with distributed roughness due to the local effects of the ribs Typical computer plots of pressure
distri-butions with ribs are. presented in Figures 4.7, 4.8, 4.9 and 4.10. (In
these figures, EAST POINTS and WEST POINTS, belong, respectively, to poitive
and negative values of O).
Figure 4.7 shows the pressure distribution with the rib
configura-tion R2. Recall that in this case the ribs were located at 10-degree
intervals starting from O = 0, and that the ribs had a discontinuity (or gap) of 1/8 in. at the midsection of the cylinder. .Therefore t1e .pressure
readings at the taps located at angular positions.corresponding to integral
multiples of 10 degrees were influenced by the presence of. the discontinuity.
This influence is rather large in the forward portion of the cylinder, but
littlé influence is observed in the wake. region. This particular rib conf
1g-uratiori was chosen with the objectives of determining the pressure doefficients midway between the ribs and observing the srnmletry of the presure distriu-tion at the same time, since most of the pressure taps on the east side of the cylinder (positive angles) were located at 10-degree intervals. No
definite influence of the gaps on the readings of the pressure taps midway
between ribs was detected, however, when the results were compared with the
results of the tests with the rib configuration RA-10, as discussed at length
later on in section 5.2.4. The results of the tests with the rib conf
Igura-tions Rl and Rl]. displayed asimilar influence of the gaps, as can be seen frOm the plots presented in Appendix 2. The influence was smaller in these cases than that observed with the rib configuration R2,due to the smaller
dimensions of the ribs in cases à]. and Ril. It may be noted that the
angular distribution of the ribs was the same in all three cases, but iii the
case of P.11 the gaps at the angular locations ±60°, ±70° and ±80° were closed.
Comparison of the resu.lts of the R]. tests with those of P.11, also showed no
systematic difference in the values of the pressure coefficients midway
between the ribs. .
Figurés 4.8, 4.9 and 4.10 show the pressure distributions for.configura-tions RA-05, RA-10, and RA-20 in which the ribs spanned the entire length of
the cylinder. (i.e. without gaps at the midsectiòn). In configuration RA-05
0.00
A
D
-e
3.00
6.00
9.00
ANGLE. (DEGREES/lW
CYL.IP«H W RIBSz2
(T. PUlPITS
b(ST POINTS
12.00
15.00
1x10 j18.00
Figure 4.7:Pressure distribution on cylinder with ribs R2 (k/d
=
3.38x103,
0=l0°,
first rib at 0=00).
.
o
. C,,0.0
I
£
I
L.
30.0
60.0
90.0
ANGLE (DEGREES/lO)
74101S01 ERST POINTS LP(ST POINTS
£
120.0
150.0
L
¡80.0
Figure 4.8:Pressure distribution on cylinder with ribs RA-05 (k/d =
3.38x103,
O=5°, first rib at O=±2.50).
Re =
N103102-L
ET POINTS
IST POINTS
£
£
0.0
30.0
60.0
90.0
ANGLE (DEGREES/lO)
A
Figure 4.9:Pressure distribution on cylinder with ribs RA-10 (k/a
3.38xl03,
O=l00, first rib a-t O=±2.5°).
Re =
-l.79x105.
120.0
150.0
£
N110701 ERST POINTS.£
l(ST POINTS0.0
30.060.0
90.0
120.0 150.0 180.0ANGLE (DEGREES/tO)
Figure 4.10:Pressure distribution on cylinder with ribs RA-20
(k/d =
3.38x103,
AO=20°, first rib at O=±25°).
Re =