4 Longitudinal Ribs
4 Longitudinal Ribs
4 Longitudinal Ribs
105-129 105
Printed in Great Britain
1. Introduction
The flow on surfaces with fine longitudinal ribs has become a subject of scientific
research because these non-planar surfaces are capable of reducing the shear stress
of a turbulent boundary layer t o below that of a smooth surface.
Historically, there is no doubt that nature has a clear priority for the development
of drag-reducing surfaces. The skin of fast sharks is covered with tiny scales which
have little longitudinal ribs on their surface. Shark scales with this structure have
been found in fossils more than 100 million years old (Reif 1985). In spite of the fact
that the particular shape of shark scales has been known to biologists for more than
a century, the first authors to speculate about their fluid dynamical significance were
Russian scientists in the late sixties (Burdak 1969; Chernyshov & Zayets 1970).
Unfortunately, at that time fluid dynamics was not sufficiently advanced t o provide
sensible explanations for the particular scale shapes visible under the microscope. I n
the meantime, however, the body of knowledge on shark scales has been increased
considerably (Reif 1982, 1985; Reif & Dinkelacker 1982; Rashi & Musick 1984). In
figure 1 the scale pattern of a fast shark can be seen, compared to that of a slow shark
in figure 2. The photographs in these figures are taken from the work of Reif (1985)
and the drawings are taken from the book on sharks by Steuben & Krefft (1978).
106 D . W . Bechert and M . Bartenwerfer
Until the late seventies, this issue had escaped the scrutiny of the community of
fluid dynamicists. This was partly due to the very fine structure of the shark scales
which can only be seen using a microscope. Also, it was considered impossible to
devise a surface having less friction drag in a turbulent boundary layer than that
produced by a smooth surface. Eventually, the situation changed when details of the
flow pattern of the viscous sublayer of a turbulent boundary layer became clearer.
The first documentation of this particular flow pattern was provided by Hama (see
Corrsin 1956, figure 15, p. 394). More detailed investigations were carried out in the
sixties by Kline and his coworkers (see e.g. Kline et al. 1967), and later by Smith and
his coworkers (see e.g. Smith & Metzler 1983). It has been shown by these authors
that the viscous sublayer exhibits a streaky structure with local regions of low
velocity, extending in the streamwise direction. One hypothesis to explain these
‘low-speed streaks ’ is that they are produced by slowly rotating longitudinal
vortices. There is some evidence to support this idea, for example the flow
visualization photographs by Cantwell (in Coles 1978, see also Van Dyke 1982, p.
93) and the theoretical investigations by Jang, Benney & Gran (1986). On the other
hand, it has been suggested that this streaky structure belongs to a ‘hairpin ’ vortex.
The viscous $ow on surfaces with longitudinal ribs 107
2. Basic considerations
One of the unsettled questions concerning riblet surfaces was where the average
origin of the velocity profile might be located. This question was raised in the paper
by Hooshmand, Youngs & Wallace (1983) and we suggested in a previous paper
(Bechert et al. 1985) that this problem could be solved by a viscous flow theory.
Figure 3 shows the apparent (or average) origin of the velocity profile which lies
below the tip of the rib and above the bottom of the valley between the ribs. We call
the distance between apparent origin and rib tip the protrusion height, because it
determines how far the rib tips protrude into the boundary layer.
We consider the average mean flow of the viscous sublayer in which the riblet
surface is immersed. I n the following equations we choose 2 as the direction of the
mean flow, y as the direction normal to the surface and z as the lateral direction. We
start with the first Navier-Stokes equation
au au au
-+u-+v-+w-+--
au 1 ap
vv2u.
pax
=
at ax ay a2
The viscous $ow on surfaces with longitudinal ribs 109
Mean velocity profiles
Apparent origin
3. Apparent origin of a riblet surface.
FIQURE
Obviously, for a mean flow calculation, the first term of the equation can be omitted.
If the riblet surface is completely immersed in the viscous sublayer, neglecting the
convective terms on the left-hand side of the equation is certainly a good
approximation. In fact, previous experiments do show that the relevant parameter
regime where drag reduction occurs corresponds to the situation where this condition
is just fulfilled. We shall return to this issue more explicitly in the discussion ( 5 10)
a t the end of this paper. We assume a flat-plate boundary-layer flow without mean
flow pressure gradient, so aplax = 0. For ‘two-dimensional ’ riblets which do not
vary in cross-section in the streamwise direction, we are left with the equation
which is a Laplace equation for the velocity u. To solve this equation in two
dimensions, various established techniques can be used. One particularly suitable
one is conformal mapping. This technique has two striking advantages: (i) Sharp
edges which produce singularities in the flow field do not lead to problems or
inaccuracies with this method whereas, for example, purely numerical methods
would encounter such problems. (ii) For a variety of configurations, solutions in
closed form can be worked out. In particular, simple formulae for the protrusion
height can be produced.
Conformal mapping utilizes the fact that for the Laplace equation (2) arbitrary
functions f(z+iy) are solutions. The function f, the mapping function, has to be
selected so that the boundary condition on the riblet surface is u = 0. The mean flow
u-distribution will be a uniform Couette shear flow in order to emulate the viscous
sublayer. What we shall do by conformal mapping is to transform a uniform shear
flow above a smooth plane surface into that above the particular riblet surface. So
the task is to do mathematically what is shown in figure 4. By the way, the solid lines
in figure 4 are not streamlines, but lines of constant velocity u (isotaches). In
conformal mapping we usually plot the real and imaginary part of a function, so we
obtain a netyvork with lines intersecting each other perpendicularly. If the imaginary
\
110 D . W . Bechert and M . Bartenwerfer
Velocity profile Smooth surface Riblet Reference grid
-
w,-plane
21
w,-plane w,-plane
FIGURE
4. Conformal transformations leading to the viscous flow on a sawtooth riblet surface.
part of our function is u, what is then the significance of the real part, i.e. the broken
lines in figure 4 1 The answer is simple : between two broken lines a shear force
AF = IT, AX (3)
is transmitted. In (3), 1 is the length in the streamwise (5) direction and Az is the
spacing between two broken lines of the undisturbed grid far above the riblet surface.
T~ is the shear stress r0 = ,udu/dy, where duldy is the velocity gradient above the
smooth surface or far away from the riblet surface. Thus, we call the broken vertical
lines in figure 4 force lines (isodynes).
C, is a constant and the integration of (4)will produce another constant, C,. These
two constants have to be determined later. a is the corner angle of the polygon (see
figure 4).I n order to carry out the integration, it is useful to introduce the following
substitution :
6=( w -1 L ii )-
with '
l =-a
w,+l n R' (5)
11 (7)
with 5 = (wz-l/w,+ 1);. For this case, the constants had been determined as
n 3 4 6 8
a 60" 45" 30' 22.5"
ridge angle 120' 90" 60' 45"
The formulae of the mapping functions are given in the Appendix. The grid pattern,
i.e. the velocity distributions and the force lines are given in figure 5. These drawings
are computer plots of our equations. The reference grid can be seen on the right-hand
side of each plot in figure 5. The apparent origin, i.e. the location of an equivalent
smooth surface, corresponds to the lowest horizontal line with zero velocity of this
reference grid. Several things become obvious from these plots: (i) The velocity
perturbations caused by the riblet structure are vanishing very rapidly with
increasing distance from the riblet surface. (ii) There is very little fluid motion and
extremely low shear stress a t the bottom of the grooves, in particular for the smaller
riblet angles. (iii) The protrusion height h, does not seem to increase above a certain
fraction of the riblet spacing s.
This latter observation can be quantified further. We find through the coordinate
transformations where the different points A and B move for a riblet surface, in
comparison t o the undisturbed reference pattern (represented by a = 90" or n = 2).
In this way, we can derive a general formula for the protrusion height h, for arbitrary
riblet angles. The details of this calculation are given in a previous conference paper
(Bechert et aZ. 1986). We find
y+21n2+--- (9)
s 2K tana a
I n this equation, y is the Euler constant y = 0.5772 and @ is the Digamma function,
as defined and tabulated in the tables of Abramovitz & Stegun (1972). I n figure 6 we
have plotted (9) for the protrusion height versus the height of the riblets. Both
quantit,ies have been made dimensionless with the riblet spacing s. For low values
of the relative riblet height his, the protrusion height is half the riblet height, i.e.
h, !z ih. For high riblets, however, there is a saturation of the protrusion height a t
hPmex= 0.2206s. This saturation limit (0.2206 = In 2/z) is an important and unex-
pected finding of this paper. We shall see in the next section whether this limit value
is also valid for other configurations.
112 D . W . Bechert and M . Bartenwerfer
FIGURE
5. u-velocity and fluid shear force distribution of the viscous flow on a sawtooth riblet
surface for various riblet angles.
The viscous flow on surfaces with longitudinal ribs 113
equation (19)
In 2
-
x ‘
0.2
P! ! equation (9)
S
0.1
ks4 ’
with s, being the width of the foot of the rib (see figure 7). For h / s > 1 we have
another approximate formula, which is essentially an interpolation between the
asymptote of the blade riblet and the curve of the sawtooth riblet:
%
s - (1--1)+k[y+21n8+----
s In2
n+4+;)].
tana a
114 D.W . Bechert and M . Bartenwerfer
0’25 :
FIGURE
7. Protrusion height h, as a function of geometrical height h for riblets with trapezoidal
valleys. The riblet foot breadth s, is kept constant for each curve.
Another probably more useful way to plot the data in figure 7 can be seen in figure
8. Here, for each individual curve, the rib tip angle is kept constant and the depth
of the grooves is varied. For small groove depth all curves behave like blade riblets.
On the other hand, the asymptotic value of h,/s for deep grooves is given by the
data of the sawtooth riblet. The asymptotic protrusion height for deep grooves is
reached a t a groove depth h/s between 0.4 and 0.6. Therefore, a groove deeper than
h / s x 0.6 has no influence whatsoever.
We have also worked out an approximate equation for the asymptotic value of
h,/s with the rib tip angle being kept constant :
In 2
- 0.2206
R
0.2094
0.20 0.1974
0.1845
0.1707
0.1558
0.15
0.1397
+!!
S
6;/ a=
n/24(7.5")
0.10 3 ~ 1 2 (22.5")
4
4n/24 (30.0')
5n/24 (37.5')
6 ~ 1 2 4(45.0')
0.05
iSJ
0 0.5 1 .o 1.5
hls
FIGURE
8. Riblets with trapezoidal valleys as in figure 7. The riblet tip angle is kept constant
for each dashed curve.
5. Blade-shaped riblets
Two-dimensional blade-shaped riblets have been introduced by Wilkinson & Lazos
(1987) as a drag-reducing surface. Therefore, we also carry out the viscous flow
calculation for this configuration. As in the preceding section, the viscous flow on
blade-shaped riblets can be calculated using the Schwarz4hristoffel transformation.
In fact, a transformation with n = 2 (i.e. rectangular corners) and a shift in
coordinates between the w2- and the w,-planes would do the job. However, the
following procedure to solve the problem is different and resembles somewhat the
procedure to calculate the flow in blade rows of turbomachines (Betz 1964).Its basic
advantage is that it can be modified to deal with other configurations such as
scalloped riblet shapes, which are found on shark scales.
If we use methods of turbomachinery flows, we may draw the reader's attention
to an obvious fact : the grids representing lines of u = const and F = const look like
streamlines !P = const and potential lines @ = const of a potential flow field. In spite
of the fact that we do not really consider a stream in the (x,y)-plane we may,
however, utilize this analogy. This has the advantage that, as fluid dynamicists, we
are used to thinking in terms of streamlines. It is, for example, easy to see what a
116 D . W . Bechert and M . Bartenwerfer
I--+ 22
FIGURE
9. Conformal transformations leading to the viscous flow on a blade riblet surface.
11
@ + i Y= 41n(z,+iy2).
2n
In order to fulfil the no-penetration condition on the surface of the cylinder, we need
a mirror-image vortex of identical strength 4, but with negative sign, inside the
cylinder a t an eccentric location E. Having only these two vortices would produce a
field with circles as streamlines everywhere. Also, as we would see later, the
streamlines in the final w,-plane would not be these of a horizontal parallel flow. This
latter condition is fulfilled if we install an additional positive vortex of strength 4
in the centre of the cylinder.
We shall not show all the details of the mathematics here. Since there are no
unusual steps involved, we shall show only a few equations and outline the
essential steps of the transformations. The stream function of the flow around the
cylinder in figure 9, which is the stream function of the sum of the three vortices, is
found to be
+
where w 2= z, iy,. The radius of the cylinder is a and the distance between the origin
a t A (where the inducing vortex is located) and the centre of the cylinder a t M has
been set equal t o one.
The streamlines can be calculated by inverting (14) to find w2 a t a given w l . Since
(14) is a quadratic equation for w,, this is a straightforward procedure. By following
one Y = const. line through the w,-plane, one can decide which one of the two
solutions of the quadratic equation is to be chosen.
The next step is t o 'smash' the cylinder in figure 9 t o obtain a single blade of finite
The viscous flow on surfaces with longitudinal ribs 117
length. The transformation capable of doing that is well known from airfoil theory,
i.e. the Kutta-Joukowsky transformation
a2
w3 = w,+-
W2
Before we apply this transformation, however, we have to shift the origin of the
coordinate system from A to the centre of the cylinder, M. Thus, with the appropriate
adjustment of constants, we have instead of (15)
w g= (w2-1)+---
a2
(5932-1)
+ 1 +a2.
After having applied the Kutta-Joukowsky transformation, the point A has moved
a bit to the left and, interestingly, the stagnation streamline Y = 0 still remains a
circle.
Before performing the next step, we move the origin back to the (shifted) centre
of the inducing vortex, A. By applying the transformation In w we obtain the desired
periodical flow pattern on the blade row. With the proper choice of constants, the
transformation equation is
is
where s is the riblet spacing. The result of this transformation can be seen in figure
9. The broken line is the stagnating streamline Y = 0 and is equivalent to the
location of the wall with u = 0. The height of the riblets can be controlled by
changing the radius a of the circle in the w,-plane (see figure 9). After some
intermediate calculations, one can find the relation that determines the riblet height
h:
h 1
- = - arctanh a.
s x
With the above outlined procedure, a set of flow patterns has been calculated for
various riblet heights, see figure 10.
The protrusion height hp is the height of the riblet h minus the upwards shift of the
streamlines (lines of constant u) in the w,-plane of figure 9. Its magnitude can be
determined by carefully determining the locations of the points in the different
planes after the different transformations. We find after some intermediate
calculations
s = i nl n ( l + t a n h e ) ) .
%
From (19), one can see, that h, x h for small h/s. This means that, for very small
blade riblets, the protrusion height is equal to the real height of the blade riblets.
On the other hand, for large blade heights, the blade riblets behave like the
sawtooth riblets, see figure 6. This is because for very deep grooves only the top of
the blade takes the viscous force. Obviously, the top of the blade is identical for deep
grooves on sawtooth riblets and on blade riblets. Thus, the limit value for the
protrusion height is again h,/s -+ In 2/n for large h/s.
118 D . W .Bechert and M . Bartenwerfer
i
h
i
T T
__-
I-
L
' I I I I I I I
FIGURE
10. u-velocity and fluid shear force distribution of the viscous flow on blade riblet
surfaces for various riblet heights: (a)h / s = 0.25; (6) 0.5; (c) 1.0.
6. Scalloped riblets
Riblets with scalloped cross-section are found on the scales of fast sharks. In
laboratory experiments, it has been shown that scalloped riblets with semicircular
shape match the best sawtooth riblets, with a drag reduction of 7 % (Walsh 1982 ;
Bechert et al. 1985). The viscous flow calculation on these scalloped riblets can be
carried out with a modification of the transformations that we used for blade riblets.
The Kutta-Joukowsky transformation in the preceding section (equation (15))
had a constant a equal to the cylinder radius a. We obtain scalloped riblets if this
radius of the mapping circle, which we call now b, is larger than the cylinder radius
a, see figure 11. In addition, the mapping circle is no longer coaxial with the cylinder.
The position of the mapping circle has to be chosen so that it touches two streamlines
with the same value (Y,,), but opposite sign, at the horizontal axis, yz = 0. The
+
streamlines between - Ycand Y, penetrate into the blade in the w3- and w,-planes.
I n the final plane, w4,the Ycstreamlines become the scalloped wall contour of the
The viscous flow on surfaces with longitudinal ribs 119
FIGURE
12. u-velocity and fluid shear distribution of the viscous flow on a scalloped riblet surface.
( a )Riblet height h/s = 0.25, parameter a / b = 1.155, protrusion height h,/s = 0.155. (b) Scalloped
riblet with nearly semicircular cross-section. Riblet height h/a = 0.50, parameter alb = 1.192,
protrusion height h,/s = 0.192. ( c ) Scalloped riblet; height h/s = 1.00, parameter a / b = 1.0325,
protrusion height hp/s = 0.216.
120 D . W . Bechert and M . Bartenwerfer
riblets. By numerical trial, cylinder radius and mapping circle radius can be adjusted
so that the riblet contour is very close to a semicircle. This particular configuration
and its flow distribution is shown in figure 12 ( b ) .Of course, the velocity value on the
surface has to be readjusted so that u = 0 a t the contour !Pc.I n this particular case,
the protrusion height is h,/s = 0.192hls. Other scalloped surfaces and their flow
distributions are given in figure 12(a, c). In general, the protrusion height of
scalloped riblets lies between sawtooth riblets and blade-shaped riblets. But, as one
can see from figure 6, they can indeed be very close to blade riblets. Of course,
mechanically, scalloped riblets are much more durable than blade riblets.
7. Rounded ridges
In technological applications of riblet surfaces, the wedge at the tip of the riblet
will never be sharp with a radius of curvature equal to zero, as assumed in the
preceding examples. Thus, it is important to know what the influence of afinite
radius of curvature may be.
Similarly to the preceding modification of the mapping function which produced
scalloped riblets, one can modify the Kutta-Joukowsky transformation into the
other direction, which is more familiar from airfoil theory. If one chooses a mapping
circle radius smaller than the cylinder radius, one obtains 'club '-like blade riblets
with rounded edges, see figure 13. Also the modified flow pattern can be seen in the
example calculated in figure 13.
The calculations, which are given in more detail in Bartenwerfer & Bechert (1987),
clearly show that there is a dramatic decrease of the protrusion height h, caused by
a finite radius of curvature R, at the rib wedge. A sample of our calculated examples
of club-like riblets is given in figure 14. For small R,/s we found an approximate
equation, which shows the decrease of h,/s (as compared t o the protrusion height for
infinitely small radius of curvature)
h h
4=41
Rk-0 2n (4.
- tanht(nh/s) ZR, 4
As one sees from this equation or from figure 14, one loses about 10% of the
(20)
protrusion height if the radius of curvature on the rib tip is only 0.5 % of the lateral
rib spacing. If one relates the protrusion height to the maximum thickness t of the
club-like wedges (see figure 13), one can compile another approximate equation from
The value of (h,/s) I R k=O can be taken from (19). Equation (21) is valid for t / s < 0.1
and 0.5 < h / s < 1.0.
One may consider our club-like riblets as too artificial and restricted in their shape.
Of course, there are also other ways to generate rib shapes with rounded tips. One
such possibility is this : we take the flow distribution above thin blade-like riblets as
considered in the preceding section, for example, those in figure 10. We choose a line
u = const. above the actual surface and consider this as the new surface. Obviously,
this produces a variety of riblet surfaces with rounded edges. Our numerical data
t This equation is identical to equation (23) in our Bechert el al. 1986, but it deviates by the
coefficient 2 from (76) in our previous report (Bartenwerfer & Bechert 1987) where this latter
equation is mistyped.
The viscous flow on surfaces with longitudinal ribs 121
FIGURE 13. Blade riblet with rounded edge: configuration and flow distribution. Riblet height
h/s = 0.50, parameter a / b = 0.773, protrusion height hp/a = 0.179, maximum thickness
t / s = 0.0740.
0.23
0.22 -
In 2
x
0.21
0.20
0.19
2!!!
S
0.18
0.17
0.16
0.15 I I I I
indicate a law very similar to that of (20).However, these data can be collapsed best
if one multiplies the second term on the right-hand side of (20) with the coefficient
1.3.
This additional enhancing coefficient reflects the fact that, with increasing radius
of curvature, the wedge angle of the riblets is also increased, which further decreases
the protrusion height (see $ 4 on trapezoidally grooved surfaces).
122 D . W . Bechert and M . Bartenwerfer
FIQURE 15. Convex riblet, u-velocity and shear force distribution. (a) Riblet height h / s = 0.50,
parameter a / x = 0.1305, protrusion height h,/s = 0.0724. (b) Riblet height h/a = 1.00, parameter
a / x = 0.0275, protrusion height h,/s = 0.0767.
8. Convex riblets
The mathematical methods in the preceding section are unable to produce convex
riblets with, say, semicircular shape. However, experiments with these configurations
have been carried out (Walsh 1982) and thus we were also interested in this
configuration.
Owing to a programming error in our data plotting program for sawtooth riblets,
we ran inadvertently into the solution of this problem. Consider the mapping
function, (7)in $3. By omitting the second term we obtain
Instead of the originally expected pattern, we get the viscous flow above two convex
riblets per division s. By varying the parameter n = n/ain the function 6 of (22),we
can vary the depth of the grooves. Figures 15 ( a )and 15 ( b ) show the viscous flow on
convex riblets, calculated with the mapping function (22), the grid size being
normalized to obtain one riblet per division s.
The protrusion height can be calculated for each pattern. It is h,/s = 0.0724 for
figure 15(a) and 0.0767 for figure 15(b). For convex riblets, the protrusion height is
always very small.
The viscous Jlow on surfaces with longitudinal ribs 123
Plexiglas
a s . , 2 V, 10 kHz
.. .. . . . .
I
/
Riblet electrode
FIUURE
16. Schematic diagram of the electrolytic experiment.
9. Electrolytic experiments
It appeared desirable to aIso be able to determine the protrusion height for riblet
configurations with shapes different from the ones already considered. For this
purpose we devised a simple experimental technique. I n addition, this will provide
an independent proof for our theoretical calculations. It is well known that the
electric field in a conductor, for example, an electrolyte, obeys the same Laplace
equation as a potential flow does. Obviously, this method is also applicable to our
viscous flow problem. I n the model experiment, lines of constant velocity u would
correspond to lines of constant voltage V in an electrolytic tank. However, besides
obtaining velocity distributions using a point probe, we can also easily measure the
protrusion height. For this latter purpose we have established a measuring set-up
which is shown schematically in figure 16. We measure the resistance R of thc
rectangular electrolytic tank without and with the riblet electrode via a voltage
measurement. The resistance of the tank without the riblet electrode is proportional
to the length L of the tank. The dccrease in the resistance caused by the presence of
the riblet electrode is equivalent to a decreased length L’ of the electrolyte. We have
L - L ’ = h-h, (23)
and h, = h-(L-L’), (24)
where h, is the protrusion height and h is the geometrical height of the riblet, as
before. The shape of the riblet electrode has to be chosen so that the walls of the
rectangular tank are symmetry planes.
The electrolytic measurements of the protrusion height h,/s plotted versus the
geometrical height h / s can be seen in figure 17. The agreement clearly shows that our
calculations are correct. There is, however, a systematic deviation :the measurements
on the blade riblets show a protrusion height about 5% lower than the theoretical
values. We assume that this minor discrepancy is caused by the finite thickness of the
blades of the blade riblet electrodes. The blade thickness is t = 0.025s. Then our
5 P L M 206
124 D . W . Bechert and M . Bartenwerfer
0.2
approximate formula for the decrease of the protrusion height, (201, would prcdict a
decrease of 5%. Thus, t h e discrepancy is explained.
The reader who is interest4edin the technical details of the experiment is referred
to our previous conference paper on this issue (Bechert et al. 1986).
10. Discussion
In this final section we shall try to establish a connection between our viscous flow
calculations and the issue of turbulent shear stress reduction by riblet surfaces. Since
that, problem as a whole cannot be solved analytically a t the present time, this part
of the paper will contain both proven and other still hypothetical material. Having
carried out numerous experiments on turbulent drag reduction, we are modest
enough to admit that many of our previous plausible-looking ideas have failed. Thus,
we are aware of the possibility that some of our present opinions may change in the
future. However, what we conclude below is compatible with the data available to
us a t the present time.
(i) Our theoretical calculations are certainly valid under purely viscous flow
conditions. Howcvcr, it took us some time to convince ourselvcs that such a theory
is also useful for a prediction of the viscous sublayer flow in a turbulent boundary
layer. In fact, for the parameter regime where drag reduction occurs, the riblet
surface is actually almost completely imbedded in the viscous sublayer. To give an
example: we consider sawtooth riblets with riblet height equal to lateral riblet
spacing (Walsh 1982). There, the optimal lateral spacing is about s+ = 15. s+ is given
in ‘wall units’, i t . s is made dimensionless with the shear stress velocity [ ~ ~ / p ]l
and
the kinematic viscosity u. Thus sf = ~ [ 7 ~ / p ] with
~ / u 70 being the shear stress of the
smooth reference plate under otherwise identical flow conditions. The protrusion
height for this geometry is h, = 0.18s (see figure 6) or hi = 2.7. Thus, the rib tips
protrude only 2.7 wall units above the origin of the velocity profile. By all available
definitions of the thickness of the viscous sublayer (y’ = 5 , or y+ = 3 for the ‘linear
sublayer’) this means that the riblets are imbedded in the viscous sublayer. Also,
hot-wire data of mean flow quantities on riblet surfaces (Wallace 1987) support our
approach. Clearly, our calculations can be valid only within a regime very close to the
riblet surface.
(ii) We would like to stress that our calculation cannot predict quantitatively any
turbulent shear stress reduction by riblet surfaces. It can only predict the velocity
The viscous flow on surfaces with longitudinal ribs 125
and shear stress distribution once the average shear stress is known. An integration
of the shear stress distribution on a riblet surface using our theory is possible, but will
lead to the trivial result that the average shear stress is given by the slope of the
Couette flow above the riblets.
(iii) The merit of our calculations is, however, that the origin of the velocity profile
can be predicted. The origin lies below the rib tips, usually a t a distance of 10-20%
of the rib spacing. This distance, which we call the protrusion height h,, cannot exceed
22% of the lateral rib spacing (0.22 = n-lln2).
(iv) The above-mentioned limit of the protrusion height is valid only for that class
of configurations where the tip of the rib looks, with increasing groove depth, more
and more like a thin blade. Thus, this limit is valid for sawtooth riblets, blade riblets,
riblets with trapezoidal grooves and riblets with scalloped cross-section. Con-
figurations that are altogether different, for example, thin wires stretched above a
smooth plate (suggested by Kramer 1939) are likely to have also a different limit for
the protrusion height.
(v) The determination of the origin of the velocity profile is not only of academic
interest. For instance, for experiments with riblet surfaces inside tubes, it is
important to know the effective inner diameter of the tubes. Without knowledge of
this effective inner diameter, it would be hard to make any statements about friction
losses in tubes with riblets.
(vi) We think that the method of conformal mapping to incorporate the boundary
conditions of riblet surfaces may be valuable for further work in this field. For
instance, the viscous sublayer model of Pearson (1988) utilizes this approach.
Furthermore, for numerical computations with the full Navier-Stokes equations, it
is desirable to have an orthogonal grid with a high resolution in the critical regions,
i.e. close to the rib tips. Our conformal mapping grids do have exactly that property
and may thus turn out to be useful for numerical computations.
We feel, however, that there should be a connection between our calculation and
the ability of riblet surfaces to decrease turbulent shear stress. Our basic hypothesis
is that sharp ribs do impede the instantaneous crossflow in the viscous sublayer,
which is generated by the turbulent motion. In this way the whole turbulent
momentum exchange in the boundary layer is also reduced, which is equivalent to
a shear stress reduction. This should occur if the rib tips protrude sufficiently above
the origin of the velocity profile. More insight into the mechanism can be obtained
from a discussion of the experiments by Wilkinson & Lazos (1987). A schematic
survey of their experiments can be seen in figure 18. Blade-shaped riblets with
different ratios h/s, i.e. height to lateral spacing, have been tested. On the vertical
axis 7 / 7 0 is plotted, the ratio of the turbulent shear stress of the riblet surface to that
of the smooth surface. On the horizontal axis, the dimensionless riblet spacing s+ is
plotted. The protrusion height is calculated from (Zl),taking the finite thickness of
the ribs into account. The data in figure 18 show an increase of drag if the blades
protrude too far into the boundary layer. For lower values of the dimensionless
protrusion height hp', drag reduction occurs. For sparse blades (see curve with h/5 =
0.153), little effect on the flow is exerted. The more blade there are, the higher is the
effect on the flow (see curve with h / s = 0.797). Our viscous calculation, however, tells
us that the protrusion height, and with it the effect on the cross-flow, finally
decreases if the blades are too close to each other. From these considerations, it
emerges that the protrusion height seems to have some relation to the ability of
riblets to impede the cross-flow. Roughly speaking, this ability should depend on how
far the ribs stick out from the surface, which indeed points to the protrusion height.
5-2
126 D . W . Rechert and M . Bartenwerfer
1.3 I
_ _'__ _- _ _ _ _- -
I I I
L-T h / s = 0.797
t l s = 0.078
S
' h,/s = 0.184
1.2
7
-
To 1.1
1.o
0.9
50 100 150 200 250
S' = S r o / p / v
FIGURE
18. Turbulent shear stress on blade-shaped riblets, according to
Wilkinson & Lazos (1987).
At first glance, there seems to be a contradiction between this suggestion and the
observation (see figure 18) that the ribs should not protrude too far into the turbulent
regime. However, given a maximum admissible protrusion height in terms of h;, it
is easy to see that we can accommodate more ribs in the lateral direction if the ratio
h,/s is maximal. This, again, points to a high h J s . At the present time, there is no
piece of experimental evidence available to us that contradicts this concept. In fact,
there are several previous experiments that support our point of view: (i) convex
riblets do not produce drag reduction (Walsh 1982); (ii) riblets with a very shallow
groove exhibit only a very small drag reduction (Walsh 1980); (iii) from Walsh's
(1982) and our own experimental data we know that riblets that do not have really
sharp wedges are inferior in their drag reducing ability.
Given this guideline, and using our viscous flow calculations, an optimal riblet
configuration should have :
(i) a sharp wedge for the rib, preferably with a radius of curvature smaller than
0.5-1 % of the lateral rib spacing;
(ii) a rib wedge angle as small as possible; and
(iii) a valley depth of about 60% of the lateral rib spacing.
At the present time, mainly sawtooth riblets with an h / s ratio (height to rib
spacing) of unity are used. Also, the wedge angle of these sawtooth shapes of about
55" seems perhaps too high. The riblet configuration which we found on the scales of
the Silky Shark (Carcharhinus falciformis) seems to us optimal in the light of the
above considerations. It has a scalloped cross-section with a groove depth h / s of
0 . 5 4 . 6 and very sharp ridges with a wedge angle of about 30". On the other hand,
we are aware of the fact that our suggested optimal riblet configuration would
promise only a modest improvement of the protrusion height and, with it, hopefully,
the drag-reducing performance.
[ I
w 3 = is l+-ln-
+1/3arctan--
[ ;(
w 3 = + I + - 1n*+2arctang
26- 1
+ 1/3 arctan - + d 3 arctan
d 3
~
'++
(v) n = 8 ; a = 22.5O; ridge angle = 45"
6'- d2 2 arctan 6
+2/2arctan(1/2 6-1)+.\/2arctan(2/26+1
REFERENCES
ABRAMOVITZ, M. & STEGUN, I. A. 1972 Handbook of Mathematical Functions. Dover.
ACARLAR, M. S. & SMITH, C. R. 1987 A study of hairpin vortices in a laminar boundary layer. Part
1 . Hairpin vortices generated by a hemisphere protuberance. J . Fluid Mech. 175, 1 4 1 .
BARTENWERFER, M. 6 BECHERT, D. W. 1987 Die viskose Stromung iiber Obedachen mit
Langsrippen. DFVLR-FB 87-21. (English trans]. The viscous flow on surfaces with streamwise
aligned riblets. ESA-TT-1091 1988.)
BECHERT, D. W., BARTENWERFER, M., HOPPE,G. & REIF,W.-E. 1986 Drag reduction mechanisms
derived from shark skin. 15th Congr. of the Int. Counc. of the Aeronautical Sciences, Sept. 7-12,
1986, London, Paper ICAS-86-1.8.3.AIAA.
BECHERT, D. W., HOPPE,G. & REIF,W.-E. 1985 On the drag reduction of the shark skin. AIAA-
Paper 85-0546.
BETZ,A. 1964 Konforme Abbildung. Springer.
BURDAK, V. D. 1969 Function of the ctenoid apparatus of fish in the presence of a turbulent
boundary layer. Zoologichevskiy Zhurnal48, 1053-1055 (in Russian).
CHERNYSHOV, 0. B. 6 ZAYETS,V. A. 1970 Some peculiarities of the structure of the skin of sharks.
In Hydrodynamic Problems of Bionics. Bionica Nr. 4, pp. 77-83 (in Russian).
CHOI, K. S. 1987 Test of drag reducing riblets on a one-third scale racing yacht. Proc. Turbulent
Drag Reduction by Passive Means, Royal Aeronautical Society, 15-1 7 Sept., London.
COLES,D. 1978 A model for flow in the viscous sublayer. Proc. Coherent Structures of Turbulent
Boundary Layers. AFOSRjLehigh Workshop 1978, pp. 462-475.
128 D. W . Bechert and M . Bartenwerfer
CORRSIN,S. 1956 Some current problems in turbulent shear flows. Symp. on Naval Hydrodynamics,
Washington, DC, 24-28 Sept., 1956. Publ. 525, National Academy of Sciences-National
Research Council, 1957.
DINKELACKER, A,, XITSCHKE-KOWSKY, P. & REIB, W.-E. 1988 On the possibility of drag
reduction with the help of longitudinal ridges in the walls. IUTAM ~Symp.on Turbulence
Management and Relaminarization, Bangalore, India, Jan. 1987 (ed. H. W . Liepmann & R.
Xarasimha). Springer.
GRADSTEIN, I. S. & RYSHIK,I. M. 1981 Tables of Series, Products, and Integrals. (Transl. from
Russian.) Frankfurt : Harri Deutsch.
HOOSHMAND, D., YOUNGS, R. & WALLACE, J. M. 1983 An experimental study of changes in the
structure of a turbulent layer due to surface geometry changes. AIAA-paper 83-0230.
JANG, P. S., BENNEY, 1).J. & GRAN, R. L. 1986 On the origin of streamwise vortices in a turbulent
boundary layer. J . Fluid Mech. 169, 109-123.
KLINE,S. J., REYNOLDS, W. C., SCHRAUB, F. A. & RUNSTADLER, P. W. 1967 The structure of
turbulent boundary layers. J . Fluid Mech. 30, 741-773.
KOBER,H. 1957 Dictionary of Conformal Representations. Dover.
KRAMER, M. 1939 Einrichtung zur Verminderung des Reibungswiderstandes. Reichspatentamt,
Patentschrift Nr. 669897, Klasse 62b, Gruppe 408. Patentiert vom 17.Marz 1937 an.
LIU, C. K., KLINE,S. J . & JOHNSTON, J . P. 1966 An experimental study of turbulent boundary
layer on rough walls. Stanford University, Dep. Mech. Eng.,Rep. MD-15, July 1966.
MCLEAN,J. D., GEORGE-FALVY, D. N. & SULLIVAN, P. P. 1987 Flight-test of turbulent skin-
friction reduction by riblets. Proc. Turbulent Drag Reduction by Passive Means, the Royal
Aermutical Society, 16-17 Sept., London.
NITSCHKE,P. 1983 Experimentelle Untersuchung der turbulenten Stromung in glatten und
liingsgerillten Rohren. Max-Planck-Institut fur Stromungsforschung, Gottingen. Bericht 3/ 1983.
(Trans. : Experimental investigation of the turbulent flow in smooth and longitudinally
grooved tubes. N A S A TM 77 480, 1984.)
PEARSON, C. F. 1988 Interaction of streamwise vorticity with a grooved boundary. Bull. A m .
Phys. SOC.,33, 2233.
RASCHI,W. G. & MUSICK, J. A. 1984 Hydrodynamic aspects of shark scales. Virginia Institute of
Marine Science, Gloucester Point, Va. 22062, Rep. 272, Jan. 1984.
REIF, W.-E. 1982 Morphogenesis and function of the squamation in sharks. Neues Jahrbuch
f u r Geologie und Palaentologie, Abhandlungen Band 164, pp. 172-183. Stuttgart :
E. Schweizerbart’sche Verlagsbuchhandlung.
REIF, W.-E. 1985 Squamation and ecology of sharks. Courier Forschungsinstitut Senckenberg,
FrankfurtlM., Nr. 78, 255 p.
REIF, W.-E. & DINKELACKER, A. 1982 Hydrodynamics of the squamation in fast swimming
sharks. Neues Jahrbuch fur Geologie und Palaentologie, Abhandlungen Band 164, pp. 184-187.
Stuttgart : E. Schweizerbart’sche Verlagsbuchhandlung.
SAWYER, W. G. & WINTER,K. G. 1987 An investigation of the effect on turbulent skin friction of
surfaces with streamwise grooves. Proc. Turbulent Drag Reduction by Passive Means, Royal
Aeronautical Society, London 15-1 7 Sept.
SMITH,C. R. & METZLER, S. P. 1983 The characteristics of low-speed streaks in the near-wall
region of a turbulent boundary layer. J . Fluid Mech. 129, 27-54.
STEUBEN, K. S. & KREFFT,G. 1978 Die Haie der Sieben Meere (The sharks of the seven seas).
Hamburg : Verlag Paul Parey.
VAN DAM,W. P. 1986 Weerstandsvermindering en een turbulente grenslag als gevolg van
microgroeven. Technische Hogeschool Delft, Vakgroep Stromingsleer, Rep. June 1986.
VANDYKE,M. 1982 An Album of Fluid Motion. Stanford : Parabolic.
WALLACE, J . M. 1987 On the mechanism of viscous drag reduction using streamwise aligned
riblets : a review with some new results. Proc. Turbulent Drag Reduction by Passive Means, Royal
Aeronautical Society, 15-17 Sept., London.
WALSH,M. J . 1980 Drag characteristics of V-groove and transverse curvature riblets. I n : Viscous
Flow Drag Reduction (ed. G. R. Hough), Progress in Astronautics and Aeronautics, vol. 72.
AIAA.
The viscous $ow on surfaces with longitudinal ribs 129
WALSH,M. J. 1982 Turbulent boundary layer drag reduction using riblets. AIAA-paper 82-0169.
WALSH,M. J. 1983 Riblets as a viscous drag reduction technique. AIAA J . 21, 485486.
WALSH,M. J. 1984 Optimization and application of riblets for turbulent drag reduction. AIAA-
Paper 84-0347.
WILKINSON, S. P. & LAZOS,B. S. 1987 Direct drag and hot-wire measurements on thin-element
riblet arrays. IUTAM Symp. on Turbulence Management and Relaminarization. Bangalore,
India, Jan. 1987 (ed. H. W. Liepmann & R. Narasimha). Springer.