The Aerodynamics of an Isolated Wheel

Rotating in Contact with the ground

by

John Edward Fackrell

Thesis submitted for the degree of Doctor of

Philosophy in the 'aculty of Engineering,

University of London

(.
1O\l,t\
AcknowledgementS

I should like to express my deepest gratitude to

my supervisor, Dr. John Harvey, for all his help and encoura-
gement during the course of this work. I should also like to
thank all the other members of the department of Aeronautics
for the cheerful assistance provided whenever required, in
part'icular Peter Gasson, who did the technical drawing, John

O"Leary, who took the photographs, Mac, John Churchward,John

Coles and the rest of the workshop crew.(Not exactly angels

but occasionally heaven sent). Financial assistance with the

apparatus was provided by the Dunlop Tyre Co. and financial

assistance with existence was provided by the Science Research
Council and my wife, not necessarly in that order. For all
of which,I am deeply indebted.
 Summary

The problems associated with modelling the wheel.

flow in the wind tunnel are discussed and it is shown that
rotation of the wheel and contact between the wheel and floor
are both necessary if the simulation is to include the impor-.
tant features of' the flow.

Six wheels (three different widths times two different

edge profiles) were tested in the wind tunnel, using a moving
floor rig • The drag and lift of the wheels, obtained by
integrating surface pressure measurements, were found to
increase with width, but a drag coefficient based on an area
given by multiplying the tread width in contact with the ground
times the diameter, remains fairly constant. Of the two edge
shapes, the sharper edged wheels had the better aerodynamic
characteristics.

The rotation of the wheel is shown to produce high

pressures near the ground and a forward movement of separation
from the top of the wheel. Tyre tread patterns, simulated by
grooves in the wheel's surface, were found to decrease the
- lift of' the wheel,but to have little effect on the drag,because
of the ].ocalieed nature of the changes they produced.
 The special features associated with the rotation

of the wheel and the presence of the ground are discussed in

detail, and some theoretical models are used to examine these

features. A free streamline solution for a cylinder on the

ground is derived and compared with experimental result8.

Co NTENTS

(i) INTRODUCTION 1

(i-i) Features of Bluff Body Flows 2

(1-2) Flow about Circular Cylinders 4

(1-3) Lower Aspect Ratio Bodies 6

(i-4) Rotating Bodies 9
(1-5) Separation from Rotating Bodies 13

(1-6) The Effect of the Ground 17
(1-7) Effect of Ground Clearance on
Wheel Flow Simulation 21
(1-8) Conclusions 28

2 SUMMARY OF EXPERIMENTAL APPARATUS

AND PROCEDURE 30

3 RESULTS 34

34
Note on Errors

(3-1) Lift and Drag Coefficents 35

(3-2) Comparison with Previous Work 36

(3-3) Variation of CL and CD across the Wheels 38

(3....4) Pressure Distributions for A-Wheels 40

(3-s) Pressure Distributions for other Wheels 46

(3-6) Effects of Rotation 48
(3-7) The Effect of Grooves in the Wheel
Surface 55

(3-8) Variation in Wind Speed 58

(3-9) Two Dimensional Flow 61

(3-i O)Conclus ions 66

(4) DISCUSSIONOF FLOW FEATURES ASSOCIATED

WITH THE WHEEL 69

(4-i) Features Associated with Rotation 70
(4-i-i) The boundary layer on the wheel

near separation 70
(4-1-2) The two-dimensional. boundary

layer on a rotating cylinder 74
(4-i-3) The movement of separation
84
position
(4-i-4) Flow near the stagnation point

on a moving surface 87
91
(4-1-5) Conclusion
(4-2) Features Associated with the Ground
92
and Rotation
(4-2-1) Potential flow solutions for
cylinder 92
(4-2-2) Viscous flow very near the ground 101
(4-3) Free Streamline Theory for Cylinder on

Ground 109
109
(4-3-i) Free streamline theory
(4-3-2) Transform for cylinder on ground 125

(4-3-3) Comparison of results 137
(4-4) Conclusions 141
 143
References

Foot Note 151

Appendix - Details of Experimental Apparatus 152

(A-i) The Wind Tunnel 152

(A-2) The Moving Floor 153

(A-3) The Wheel 155

(A-k) The Pressure Measuring System 157
164
(A-5) Data Reduction
171
(A-6) Additional Apparatus

3 pages
Tables

95 pages
Figures

Total number of pages - 277

 1

(i) INTRODUCTION

Although there have been many investigations into

the aerodynamics of cars in recent years, both of specific
vehicles and of more generalized shapes, '(e.g. see City
University (1969)),the flow field associated with the wheels
has received very little attention; , mainly due to the difficult-
ies in obtaining an adequate simulation of this complex flow.
Yet, in many cases where aerodynamic design is important,the
wheels can play an important role in the behaviour of the
vehicle. The extreme example is the "Grand Prix" racing car,
which travels at high speeds with totally exposed wheels and,
because the body is relatively well "streamlined", these pro-
duce the dominant effects on the drag force and also greatly
influence the lift characteristics. It is therefore desirable
to have some understanding of the wheel's flow field,in order
to develop realistic simulation techniques for wind tunnel
testing of vehicles, where having rotating wheels in contact
with the ground is not always feasible. This understanding may
also help in dealing with the entrainment and dispersal of
spray by wheels on a wet road, which can form a hazard to
following vehicles.

We can expect the aerodynamic effects on an auto-

mobile wheel to be influenced by the following factors; the
 2

wheel is a bluff body and, in particular, is basically a

circular cylinder of low aspect ratio; it is rotating; and
it is on the ground. We shall therefore need to consider all
these factors in attempting to understand the aerodynamics of
the wheel. In the following sections a review of re'evant work
on the above effects will be presented. First, by considering
aspects of flow common to all bluff bodies, then by looking in
detail at the two dimensional flow past. circular cylinders ,
followed by a consideration of the effects of lowering aspect
ratio, of rotation, and of the ground on these flows.

(i.i) Features of bluff body flows

A bluff body can be defined as a body of any shape,

which experiences complete boundary layer separation before
the trailing edge due to the large adverse pressure gradient
set up over that part of the body behind the position of ma-
ximum thickness. This pressure gradient decelerates the slow
moving fluid within the boundary layer near to the surface
and eventually causes a reversed flow and hence separation
(Fig.(1-1)). Bearman (1972) has given a description of the pre-
sent knowledge regarding the behaviour of the separated flow.
After separating, the shear layers entrain fluid from within
the separated region, lowering the pressure within this re-
gion and hence causing themselves to curve inwards. When the
shear layers meet, fluid is returned to the base region to
balance that removed by entrainment. The downstream position
 3

of this meeting point is stabilised by, and depends on, the

entrainment of fluid from the base region. Thus, in the mean,
we have the flow configuration shown in Fig.(1-1).

The vast difference between this flow and the fully

attached potential model of the flow leads to corresponding
changes in the pressure acting on the body. In particular,the
pressure over the rear of the body does not attain the stagna-
tion value but is usually constant and negative within the se-
parated region, thus leading to a large pressure drag on the
body. Typical results fbrbluff body flows, including the re
circulating bubble, can be found in the work of Fail,Lawford
and Eyre (1959), who studied the wakes of flat plates normal
to the free stream.

So far we have described only the mean flow about

the body. In practice, except at very low Reynolds numbers,
the detached shear layers are unstable, create highly un-
steady flow downstream of the body and have a tendency to
roll up into discrete vortices. The latter is especially true
of high aspect ratio bodies. In fact, in two dimensional flow,
at most Reynolds nvmbers, it is the entrainment of fluid from
the base region by the formation of discrete vortices, that
dominates the wake mechanics (see Gerrard (1966)), whereas in
three dimensional flow, although discrete vortices are formed
(e.g. see Fail, Lawford and Eyre (1959)), they are relatively
weak and entrainment is mainly by the shear layer8.
 1

However, on a bluff body with finite curvature the

attached boundary layer can have as important an effect on
the wake mechanics as the free shear layer interaction, in
particular , in establishing the Reynolds number dependence
of the flow. Thus a study of the two dimensional flow about
a circular cylinder can be expected to yield information y e-
levant to the three dimensional flow about a wheel.

(1.2) Flow about circular cylinders

Morkovin (1964) has given an excellent review of

the flow features of circular cylinders, giving many re-
ferences on the work carried out in this field up to 1964.
For our purposes, the important point to emerge from this
work is that the flow exhibits distinct flow regimes depend-
ent on Reynolds number, Fig. (1-2). It is the higher Reynolds
number which are of interest in the present study; for auto-
mobile wheels,it will normally be around io-io6 , spanning
the subcritical, critical and postcritical regions and at
very high speeds entering the transcritical region.

The factor governing the flow alterations at these

higher Reynolds numbers, is the variation in the position of
transition to turbulence within the shear layers. The Von
Karinan vortc street at Re. <150-300 is set up by a purely
laminar instability of the two shear layers due to mutual
interaction (see Abernathy and Kronauer (1962)). At Reynolds
numbers above 300, transition occurs in the separated shear
layers which then roll up to form turbulent vortices (Sub-.
critical Regime). However, this does not alter the basic vrtex
wake structure, the cylinder still suffering fluctuating lift,
due to the alternate shedding of vortices,and the Strouhal
number (f d/iJ ) remains fairly constant throughout the
range.

The point of transition to turbulence moves up-

stream with increasing Reynolds number and at the same time
the individual shear layers can undergo local lamirta?instabil-
ities ahead of the turbulence. When these instabilities reach
the surface of the cylinder, they can radically alter the
flow producing a closed separation bubble (Critical Regime).

As the Reynolds number increases further, the separation

bubbles disappear and the boundary layers become fully tur-
bulent before separation (Supercritical Regime).

In the critical and supercritica]. regimes, the

separation line is highly three dimensional, being affected
by local transitions to turbulence. This random three-dimen-
sionality destroys the regular shedding of vortices bygrtly
decreasing any epanwise correlation of the vortex shedding
(i.e. reduces correlation length). However, once the transitLon
 6

to turbulence within the boundary layer occurs sufficiently

far upstream of the separation position to prevent local fran-

sitions affecting the separation line, regular vortex ahe&Iing

recommences. (Transcritical Regime). The boundary layers then

separate in a fully turbulent state, but the separation position

occurs at about the same angular position across the span of
the cylinder, restoring the two dimensionality of the flow.

The effect of the above changes in the flow on the

mean pressure distribution around the cylinder is shown in
Fig. (1-3). In those regimes, with regular vortex shedding,
the pressures fluctuate producing an oscillating liftforce.
The affect on the drag coefficient is shownin Fig.
- j--
Note that the flow around sharp edged objects is not soJs_
-
ceptibile to Reynolds number changes, becauseepha-redges
maintain straight separation lines, which are little affected
by transitions to turbulence.

(1.3) Lower Aspect Ratio Bodies

In deciding how much of the above is relevant to

the flow around the wheel, we must take into account that
the wheel has an aspect ratio (width/diameter) of less than
one. The effect of aspect ratio on the drag coefficients of
the circular cylinder and flat plate is given in Fig.(i-5).
The variation of the three curves between AR= 1 and 10 is
very similar, tho ii'ag ratio increasing fairly slowly. Above
 7

10 the drag of the flat plate appeaiBto increase more rapidly

than that of the cylinder. Fail, Lawford and Eyre (1959) give
detailed results for the effects of aspect ratio on flat plates
and in view of the above similarity in drag behaviour, these
may have some relevance to the circular cylinder. It was
found that the base pressure became more negative with in-
creasing aspect ratio, as might be expected since there i
always a close correspondence in the behaviour of the nega-
tive base pressure and the drag of bluff bodies. The recir-
culating bubble length divided by the square root of the
plate area decreases with aspect ratio, although the same
length divided by the chord increases to a maximum atASR=10
before decreasing. Sharply defined shedding frequencies were
detected down to an aspect ratio of .5 and the Strouhal num-
ber based on the chord varied slightly, having a minimum at
AR = 10. This vortex shedding will not play as dominant a
part in the wake mechanics as in two dimensional flow, how-
ever, as mentioned previously.

The above results cannot tell us whether the Reynolds

number dependence of the circular cylinder, associated with
transition to turbulence, still continues at lower aspect
ratios. We might suspect that it does since this is basically
a curved surface phenomenon; the downstream movement of the
separation position with transition is due to the turbulent
boundary layer being more efficient at obtaining energy from
the free stream and so able to sustain a greater adverse
pressure gradient. To examine this we can turn to results
for the flow pastspree (which do bear some resemblance to
 8

our round edged wheels). Fig.(1-k) shows the variation of

drag coefficient with Reynolds number given by Schlichting
(19 68 ), and Fig. (1-6) the variation in pressure distribution

found by Fage (1937). The great similarity between these fi-

gures and the corresponding ones for the circular cylinder

suggests that the same processes are taking place, including
the formation of local separation bubbles and the transition
to turbulence.

Both cylinder and sphere are very susceptible to

small disturbances when in the critical regime. Such agents
as free stream turbulence, 5upport interference and trip wires
can promote an earlier transition to turbulence, with a corres-
ponding change in the pressure distribution and drag. (For
example see the photographs of Weisselberger (191k) in
Schlichting (1968) for the effect of a trip wire on the flow
around a sphere). Because of this there is much scatter in
the data from different experiments in this range of Reynolds
number depending on how adequately the above agents have
- been eliminated.

The effect of surface roughness on both is also si-

milar (ref. Goldstein (193 8 ),vo l.2) . With increase in rough-
ness the fall in drag occurs at a lower Reynolds number and
is of smaller extent. When the roughness is very coarse, the
critical drop in CD tends to disappear and the drag to become
independent of Reynolds number. From our discussion of the
circular cylinder flow we see that, because the coarse sur-
face is provoking an early transition, the critical and
 9

supercritical regimes are being supressed and the transcritical

regime encouraged. These effects may be significant when
considering the flow over a treaded te, which may exhibit
characteristics of the transcritica]. regime at Reynolds num-
bers lower than would be necessary for a smooth wheel.

However, the wheels used in the experiments described

herein are smooth, as are the tyres used on racing cars, so
it becomes very necessary to choose the Reynolds number to
at least fall within the same flow regime as that of the real
flow that we wish to model. An average racing car wheel of 20"
dia. has a Reynolds number of 5 x 105 at 30 m.p.h. and 2.5 x
io6 at 150 m.p.h. both of which are in the supercritical range.

A choice of Reynolds number within this range would therefore

seem to cover most real situations of interest.

(i. Zi) Rotating Bodies

Apart from any direct applicability to the wheel, it

is desirable to have an understanding of the flow around cir-
cular cylinder and spheres because these shapes have featured
in all experiments on the effects of rotating bluff bodies.
Simple potential theory yields a solution for flow past a
rotating cylinder, which indicates a ]ift force on the cylin-
der proportional to the circulation about the cylinder, Nost
of the early experimental work on rotating cylinderswas res-
tricted to measuring the lift and drag forces in order to com-
pare the performance with that of aerofoil sections. When it
was found that the aerofoils were generally more efficient for
 10

aircraft application, (having less drag for a given lift)

interest in rotating cylinderswaned. There is, therefore, a
fair amount of data on the forces experienced by the cylinder
but little regarding those details of the flow field which
might be of use in understanding the wheel flow. In fact the
only work containing detailed measurements in e external
flow field is that of Thom (1931), whilstPrandtl and Tietjen8
(i93Z) have photographed the flow at various values of v/Uoo,
the ratio of the circumferential velocity to the free stream
velocity.

Thom's work was mainly at a Reynolds number of

2xi0Ll and that of Prandtl would appear to be at a much lower
number. That the flow is still sensitive to Reynolds number
is shown in Figs. (1-7) and (1-8), where the lift and drag
coefficients are plotted against v/U 0
, and the variation with
Reynolds number indicated. The lift curves incorporate the
results of several experiments and yetthe results agree
fairly well overthe xange of v/U 0, shown. However, only one
set of results has been used for the drag curves, as the
value of the drag of even the stationary cylinder varies
considerably between investigators,seeming to depend on the
aspect ratio used. The variation of the drag with v/U 0
, is
similar in all cases,however,although the absolute values
are different, At higher values of v/U 0
, than are shown there
is considerable difference in the results of different expe-
riments in both lift and drag. The range of aspect ratio
covered by these rosults is quite oxtensive and it is doubt-
ful if oven approximate two dimensional flow would have been
 11

achieved in many cases so that end effects would be important.

For example the results of Griffiths and Ma (1969) given in
Fig (1-8) were obtained with a cylinder of aspect ratio of
only 2 and although end plates were fitted the CD of 8 for
the stationary case is lower than would be expected for two
dimensional flow at the Reynolds number indicated. Because
of these discrepancies and also because the detailed results
near the critical range will be influenced by the conditions
of the various experiments only broad trends with Reynolds
number are indicated in Figs. (1-7) and (1-8).

To see what effect a lower aspect ratio will have

on the above results we can compare tham with some results
obtained for rotating spheres. The lift and drag coefficiits
for sphere8 are presented on Fig. (1-9), together with the
results of Reid (192 L ) for a cylinder. The results of Davies
(19 L 9) are for smooth and dimpled golf balls, which were set
spinning at a constant rotational speed and then allowed to
fall through an air stream, the forces being obtained from
measurements of the trajectory. The results of Maccoil (1928)
were obtained using a conventional balance arrangement.

Davies' method avoided complications due to support

interference,but was subject to other errors, principally the
fact that the ball has an appreciable velocity downwind near
the end of its trajectory and was also influenced by the
walls of the tunnel. Despite these differences the variation
in the coefficients with v/U is of a similar form for all
smooth ephores, although the drag coefficient found by Maccoil
 12

for v/U 0 seems a little high. The cylinder also 8h0w5

a similar but more extreme variation.

An interesting feature of both cylinder and sphere

flows is the negative lift,(opposite to that indicated by
potential theory),which is obtained at Reynolds numbers
approaching the critical number and at small values of
Griffiths and l'la (1969) suggest that for the cylinder this
is due to the upstream moving surface provoking transition
and hence a downstream movement of the separation point on
that side of the cylinder. This gives rise to a pressure
distribution resembling A in Fig. (1-3) on the downstream
moving side of the cylinder and one resembling Cm Fig.(1-3)
on the upstream moving side. The net result is a negative
lift. For subcritical Reynolds number, PrandtL's photographs
(Prandt]. and Tietjens (193 4 )) and Thorn's pressure distribulions
(Fig.(1-10)) show that the downstream moving surface suppresses
separation, whilst the flow separates closer to the stagna-
tion point on the other side. Apparently, for v/TJ greater
than .5, this influence of rotation overcomes the effects of
transition since a positive lift is found for all Reynolds
number up to 4,6x10 5 , the maximum at which data is available.

The similar results for the lift force on smooth

spheres suggest that the same transition effects are occuring
in the three dimensional flow.
(i.5) Separation From Rotating Bodies

The concept of separation over a rotating body

needs some consideration. For the present, we shall define
the separation point in terms of the pressure distribution
as the position at which the mean static pressure on the stir-
face becomes constant. On a stationary body this definition
would give a position only just downstream of that at which
a ui8 I
becomes zero, the normal definition of separation
point. On a rotating body, the distinction between these two
definitions becomes more necessary because au/ayt can be-
come zero even in a favourable pressure gradient. To see this,
consider Fig.(1-11) which presents Thom's results for the
pressure distribution at v/U,,= 2, together with his measure-
ments of velocity profiles and static pressure at selected
points around the cylinder. The velocity profile at 9 = 200
has au,ayl negative, whilst at 8 = ZOO it is positive,
therefore, between these two, au/ayl must be zero, without
having any noticeable effect on the measured pressures.
Similarly 8U/ay must be zero between 9 =60-120 but between
points where the pressure distribution shows significant be-
haviour,i.e, e = 160°-28o°, aU/ay is always non zero.
However, a13 , the static pressure gradient at the wall,
L
does become zero and even negative within this region. These
profiles were obtained withPitot and static tubes, held tan-
gential to the surface, which are likely to have errors near
a rapidly rotating curvod body, so we must beware of drawing
 14

too many conclusions from them. However, a aP/a <0 does

violate the boundary layer assumptions which require aP/ay
= Kpu2 (K = curvature, positive for a convex body). The obvious
conclusion is that the normal velocity is no longer small in
the neighbourhood of points where ap,ayI is negative.

Some idea of the errors involved in the use of stat

tubes can be gained from Fig.(1-11), where the results from
the static tube readings are compared with the pressures ob-
tained from a static hole in the surface of the cylinder. The
pressures were measured through this hole by opening a valve
at a preset angle on each revolution of the cylinder, the
pressure being taken from the cylinder through a mercury seal.
This method was not without error since the valve was open
for a finite angle and tended to have a pumping action which
raised the measured pressure. However, the error is likely to
be less than that associated with the static tube, The major
difference between the two occurs at 6 = 1600 where reading
of static pressurecicsest to the surface (at.5 mm.) seems to
indicate an increasing pressure inwards. If this were so the
appreciable normal velocities would probably make the reactns
invalid, so the results for this position must be viewed with
suspicion. Agreement over the rest of the cylinder is much
better,suggesting that the profiles at the other points may
be more accurate.

Anothor roniarkablo foature of Fig. (i-il) is that

the pressure distribution given is the difference in the
 15

pressures measured with the wind on and with it off. That is,
in making the correction for centrifugal effects, no account
has been taken of the thin layer of air which is set in rota-
tion by the cylinder in still air and so lowers the centri-
fugal pressure below the theoretical value, If allowance is
made for this layer, then the values of c , given in Fig.
(i-li) are too high and the maximum value would be less than
one, The conclusion Thom draws is that the cylinder induces
the same rotation in the layer of air next to it independent
of any external flow, The pressures shown therefore are those
over the surface,not of the cylinder, but of an annulus
corresponding to this thin layer of air.

From the above measurements of Thom, the correspond-

ing photographs of Prandt]. and our own measurements and
theoretical work in section(4) we can conjecture that the
mean flow about a rotating cylinder with v/U = 2 is of the
form shown in Fig. (1-12). Behind the cylinder is the usual
recirculating bubble associated with bluff bodies, although
now offset to one side and asynietric due to the rotation. Next
to the cylinder is a layer of air which remains attached to
it and rotates in the same direction as the cylinder's surfa-
ce. This means that separation must occur from this layer of
air as shown not from the cylinder's surface and that each
separation point will involve the meeting of four streamlines,
The separation from the upstream moving 8urface is that rele-
vant to the wheel,as indicated in Fig. (1-19). Here the profi-
les of tangential velocity are basically similar before and
 i6

after separation, although after separation there would be a

greater normal velocity. (c.f. Fig.(1-1) and velocity pro-
files at 0 = 280 and 300° Fig. (1-11).)

Separation in three dimensional flow is more complex

than in two dimensions so one must be careful of applying two
dimensional conclusions • Maskell (1955) has identified two
basic forms of separation from a general curved surface (Fig.
(1-13)), He uses the concept of limiting streamlines in the
surface to indicate the direction of the streamlines as the
normal distance tends to zero, These streamlines can be obser-
ved by painting paraffin oil onto the body. In the bubble type

of separation the fluid enclosed by the separation surface is

the recirculating bubble already mentioned. When the 'free
shear layer' is formed the fluid either side is main stream
fluid. Separation over the top of the wheel will be of the
first type. In the formation of the bubble there is a singular
point S, at which the behaviour of the flow is the same as
in two dimensional flow. At all other points of separation
the limiting streamlines run tangentially into the line of
separation, from which a surface of separation originates.
The behaviour in the neighbourhood of separation on a rotating
body is obviously going to be still more complex. However, if
we can apply our two dimensional ideas near the singular
point,we might expect that the line of separation lies in a
surface above the cylinder and that there is a thin layer
of fluid next to the cylinder,moving with the cylinder's sur-
face.
 17

(1-6) The Effect of the Ground

The last of the features of the flow around the

wheel mentioned at the beginning of the introduction is the
effect of the ground.

The real situation which we wish to model in the

wind tunnel is that of a wheel rolling forward in contact with
the ground through still air. In the wind tunne].,we fix the
position of the wheel and move the air past it, but we must
also move the floor with the same velocity as the air,in
order to correctly simulate the boundary conditions at the
ground. The wheel should then rotate in contact with the
floor,with, for no slipping 4 the circumferential velocity of
the wheel equal to the velocity of the floor. A subsidiary
problem is the elimination of the tunnel boundary layer ahead
of the moving floor in order to prevent the slower fluid in
the layer from going above the flccz' and creating the wrong
conditions there. This required simulation is shown in Fig,
(2-LI).

Previous investigators of the flow about the wheel

have not attempted this complete simulation, but have mount
the wheel with a clearance between itself and the floor to
enable balance measurements of the forces to be obtained. If
the wheel is run in contact with the floor the varying reacticn
forces betwoen the two can not be adequatoly allowed for in
 18

taking balance measurements, In order to appreciate the im-

plications of this incorrect simulation it is necessary to
look at the differences in the flow about bodies close to
the ground and those actually on it.

Potential theory indicates that the venturi effect

of air rushing through the gap beneath a body placed near the
ground generates large suctions on the undersurface and so
causes a lift force towards the ground. (see section (4 2 )
for cylinder and Mime Thompson (1968) p.562 for sphere), This
force increases with proximity to the ground and is also in-
creased by rotation (Fig. (L_5)), A rotating wheel is a lift-
ing body and can be expected to generate trailing vortices,
as was shown by Maccoil (1928) for a rotating [Link]
and Burns (1969) used the concept of an image horseshoe vortex
to simulate ground effect on a three dimensional wing (see
H. Glauert (19 L 8) for details). They showed that the main
effect of the image system was to increase the lift at any
given incidence, a and to reduce the induced drag. Thus dVdx
increases as the wing approaches the ground. This effect for
lifting bodies is additional to the venturi effect discussed
earlier and, for example, can mean that a wing at large posi-
tive incidence can have it upward lift increased by the pro-
ximity of the ground.

However, whether either of these potential theories

will have much significance for a bluff body, such as the
wheel, is extremely doubtful • For, in addition to these two

7
 19

inviscid effects, we find in practice that viscosity can

play an important part in determining the overall flow, either
through a change in the separation position of the boundary
layers on the body,or, with a small clearance, by directly
modifying the flow through the gap. Balance measurements on
bodies close to the ground show that the above effects can
have a decisive influence on lift and drag forces, but this
type of data can only give a limited understanding of the
flow and no detailed work has been done on the interaction
of viscous and inviscid effects in this situation.

The interrelation between these three effects, the

two inviscid and one viscous, for a more slender body than
the wheel, can be examined in the results of Stollery and
Burns (1969), who measured the forces on an idealized car-
like body, mounted above a stationary ground plane. As shown
in Fig. (i..i Lt), dCL /da increases when the body approaches the
ground but the lift curve as a whole is moved downwards due
to the venturi effect, As the incidence is decreased the
curves show a levellirmgoff and reversal of slope, the value of
a at which this occurs boing dopondont on the height above
the ground. The corresponding increase in the drag coeffi-
cients,at the same incidence, suggests that the boundary layer
on the undersurface has separated upstream of the trailing
edge. This would inhibit the flow beneath and thereby reduce
the venturi effects, decreasing the negative lift.

That separation effects can have a crucial influ-

once on the flow past a body near the ground is emphasised
 20

by the results of Wadcock (1970), who studied the two dimen-

sional flow past two circular cylinders in close proximity.
It might be thought that the potential solution for this flow,
given in section ( 1i.2.1),woulcl at least indicate the direction

and relative magnitude of the lift forces on the cylinders as

the two approach, yet, Wadcock results show that the ineasu-
red mean forces are in the opposite direction to that indica-
ted by the theory, being repulsive not attractive.(Fig.(4-)).
Mean pressure measurements around the cylinders indicate that
this is due to a difference in the 8eparation positions on
the two sides of the cylindex',which acts to oppose and over-
come the venturi effect.

Interactions between the vortices shed in the two

wakes of the cylinders have a significant effect on this two
dimensional flow, so that we need not necessarily have the
same situation in three dimensions, although the results of
Hyde (1972),for a circular cylinder of aspect ratio 1.7
mounted above a stationary ground plane,do show the same
repulsive forces as results. This resemblance is
superficial, however, since in hyde's case the separation
of the ground plane boundary layer ahead of the circular
cylinder was probably the most significant effect in produ-
cing the upward lift. This again illustrates the danger of
incorrect simulation of the ground boundary conditions for
curved bluff bodies like the wheel.
 21

(1.7) Effect of Ground Clearance on Wheel Flow Simulation

The above discussion on the effect of the ground

can be related to the wheel through the work of the Motor
Industry Research Association (MIRA) (1970) who carried out
tests on model wheels, both rotating and stationary,mounted
at various ground clearances.

A complete idealized Formula One car model was

used, the forces due to the wheels being obtained by deduct-
ing those due to the body alone. The model was mounted above
a fixed ground (i.e. the working section floor) with a boun-
dary layer thickness, 5 99 of about 2 ins. The wheels, made
of polystyrene were 6 ins in diameter and 2,3 and 4 ins, in
width all with square edge profiles. Obviously therefore the
boundary layer on the floor would have a large effect on the
flow under the wheels at the sniall ground clearances which
were used. Some tests were carried out with a moving floor
but with a gap under the wheels. This meant that the movementof the
floor
was of no advantage since it merely increased the flow through
the gap and produced greater error. At the test speed of
6i ft/sec the Reynolds number based on wheel diameter was
2x105 which is in the critical range of circular cylinder and
sphere flows.

The results MIRA designate as zero ground clearance

wore obtained by sealing the gap between the wheel and the
 22

stationary floor with a strip of paper, not by actually

having contact between the two (Fig (1-18)). Although this
will prevent flow under the wheel,it is still an incomplete
simulation of the true flow and will be subject to errors.
For one, the boundary layer on the stationary floor is going
to produce effects unrepresentative of the real flow. Also,
the incorrect boundary conditions at the junction of the
wheel and floor will mean that a unique and significant
feature of the flow will not be modelled correctly. In the
correct simulation, shown in Fig. (2_ 14), the wheel rotates
in contact with the moving floor and this means that both
the surface of the wheel and that of the floor are moving
together at the line of contact. This movement draws air in
the layers next to the surfaces into the corner between them.
Because the contact between wheel and floor prevents flow
under the wheel,this air must return back upstream as indica-
ted in Fig. (1-19). The above effects occur only in a
very small region under the wheel,but because of the large
pressures and appreciable velocities produced,they can have
a wider influence on the flow field. None of these effects
would be correctly simulated in the NIRA experiments, so
that their results at zero ground clearance must not be
thought to accurately represent the true picture when the
wheel is on the ground.

Despite these deficiences it is worth considering

the NIRA results,as they do give some insight into the real
behaviour of three dimonsional bluff body flows witli small
gaps, although it must bobcae in mind that the detailed
 23

results will only apply to the particular conditions of the

experiment.
The lift and drag coefficients for the 3ins
wide wheel found at various heights above the ground are
given in Fig. (1-15), the coefficients being based on the
projected frontal area of the wheel. The most important
feature of Fig. (1-15) is the behaviour of the lift coeffi-
cient of' both the rotating and stationary wheels. This beco-
mes more negative as the gap is reduced,but at some stage this
trend reverses and the coefficient ends up positive when the
gap is sealed. In the rotating case, the change from negative
to positive occurs at a very small gap,although the exact
position is doubtful,because of the wide spacing between
experimental points. This late changeover is despite the
fact that the floor boundary layer must have helped to in-
hibit the flow beneath the wheel,

MIRA also measured the static pressure distribution

around the centre-line of the wheels. However, this was done
with a static tube held as shown in Fig. (i-i6), which was
probably subject to some error in the thin highly curved
shear layer and mounted in the way shown must have considera-
bly disturbed the boundary layer on the wheel. At the Reynolds
number of the experiment (2x10 5 ) this might have led to un-
known transition effects,so altering separation and thus the
overall pressure [Link] point is that measure-
ments wore only made at every 300 around the wheel,so that
the position of separation points and other features cannot
 24

be fixed with any great certainty. Thus the results of these

tests are unlikely to have any great accuracy.

If the results are considered in a genera]. way, how-

ever, ignoring the detailed values they can be of use,if only
in demonstrating the striking differences in the pressure
distribution when the gap is sealed to when it is not. Figs.
(i-i6) and (1-17) have been drawn with the above consideration
in mind. The details of the NIRA results have not been in-
cluded but the general picture of the distributions is shown
and the different distribution with the gap sealed is empha-
sised. The large suction generated on the underside of the
stationary and, especially, the rotating wheel with open
gaps, agreeing to some extent with potential theory, con-
trast markedly with the more positive pressure over the same
region which occurs with the gap sealed. (The solid line in
Figs (1-16) and (1-17) shows the interpolation at e = 90 givai
by MIRA but if the gap were truly sealed it would seem more
reasonable to have the pressure tending to stagnation value
in front of the seal as indicated by the dotted line, although
it may be that the thick floor boundary layer has some influnce.
The differences in the pressures on the undersurface at the
various ground clearances is mostly responsible for the
changes in the lift coefficients of Fig. (1-15).

The other main point to note from the NIRA distri-

butions is tho difference in separation positions on the sta-
tionary and rotating wheels, at all grGund clearancos. If we
 25

examine the separation position on the upper and lower (except

9"d = o) surfaces,thon we find that the effect of rotation is
to suppress separation on the undersurface,which is moving
in the downstream direction, from e & 120° to ø200°, but to
promote early separation on the upper surface, from e 2100

to 2700. This is in agreement with Thom's results (Fig.

(i-io)) and also Prandtl's photographs for rotating cylinders,

The above work of MIRA has important implications

for the results presented by Morelli (1969) for the aero-
dynamic actions on an automobile wheel. A full sized racing
car wheel was used for these tests,at speeds giving a Reynolds
number range from 5x10 5 to 2x10 6 . Because the forces on the
wheel were to be measured by a balance, the wheel did not ro-
tate in contact with the floor, but was mounted so that the
bottom projected 15mm. into a slot cut in a fixed ground
board (see Fig. 1-18)). Measurements with a yawed wheel and
various fairings were obtained, but the basic result for an
unyawed wheel in free air was that the CL was about -0.1 and
the CD about 0.5. The negative lift force is especially signi-.
ficant as it is completely at odds with the results presented
in this thesis. The great weakness of Morel].i's system lies
intiegap of 10mm allowed under the wheel. This gives a gap/
diameter ratio of.016 and we see from MIRA's results (Fig.
(i-is)) that, for a rtating wheel, this can mean a negative
lift instead of the positive value obtained without the gap.
Norelli does give a correction for the suction experienced
by the wheel duo to the gap as,
 26

= pU 2 _og
________
( R ) __________(i)

where

p is the suction on that part of the wheel in the

slot

R is the radius of the wheel 31 ,5 cm

R is the gap i.e. 1 cm
Ti , the velocity of the wheel which is assumed
to be the velocity of the air in the slot between the wheel
and the ground.

The above expression will be small since R/R is

smUidNare1lithexfore assuned that the effect of the gap on the
lift is small,

All that this is saying however, is that the static

pressure gradient due to centrifugal effects is small in a
thin layer under the wheel; the normal boundary layer assumptii.
It ignores completely the fact that there will be a pressure
gradient between the front and rear of the slot which tends
to force air through, this tendency being helped by the rota-
tion of the wheel. A better approximate correction would be
obtained by ignoring these small curvature effects and assunng
that the transverse velocity components v, w are zero in the
slot, The problem then reduces to that of unidirectional flow
between parallel plates whose solution is given by Batchelor
(1967). It is wortliwhilo examining this solution hore,as it
 27

illustrates the interaction of the flow through the gap with

the external stream. The continuity equation reduces to

au = 0
ax

so u is independent of x and we assume independent of y • Then

the full Navier-Stokes equations reduce to

ax =I4 az 2'ay -0 =O

As the right hand side of the 1st equation is inde-.

pendent of x, we must have

= const = -C, say, through the slot.

Thus,if the pressure is known near the two ends of

the slot a linear pressure gradient could be assumed through
the slot. This would have been a far better approximate correct-
ion for Morelli to have used than the one he proposes.

A solution of the above equation,with boundary

conditions u = U when z = g and u = 0 when z = 0,i8

u = z(g-z) + iJz/g

80
volume flux
udy=C12p 1Jo
Q = • = -- + -
unit span 0 2

Now the volume flux depends on C, the pressure

 28

gradient determined by the external flow, but the external

flow itself will be influenced by the volume of fluid allowed
through the gap. It is this influence which makes any results
obtained with a gap under the wheel very doubtful, even if
some correction could be applied to the lift, for the drag,
as well as the lift,may be altered.

(1 .8) Conclusions

We feel that the features discussed in the intro-

duction have shown the importance of having a correct simula-
tion of the wheel flow. The Reynolds number of the experiment
must be, at least, in the same range as the real flow, since
otherwise the separation position and hence the pressure
distribution may be wrong. For similar reasons,the wheel must
be rotating at the correct velocity, since we have shown that
rotation can have significant effects on boundary layer
separation. The results of MIRA and Morelli, in particular,
have shown that the effects of even a small gap under the
wheel are such as to invalidate any application of the results
to the real flow. Furthermore, because the boundary conditions
near the wheel-floor junction give rise to important viscous
effects,it is necessary that these same effects should be
produced in the simulation. We therefore require a wheel
rotating in contact with the ground,which,for a wind tunnel,
implies a moving floor. Because the varying reaction forces
which are then generated preclude balance ineasureinents,the
 29

forces on the wheel must be o,tained from integrating static

pressure measurements over the wheel surface. This is the
arrangement adopted in tie experiments described herein,

Obviously, there are a large number of variable

factors which it would be possible to model in the experiment,
e.g. the tyre tread patterns, the deformation of the wheel
and the effect of various fairings around the wheel, However,
initially it was decided to look at the simplest case of an
isolated, non-deformable, smooth surfaced wheel,in order to
provide a basic starting point from which the effects of thp
different factors could then be examined as required.
 30

(2) SUMMARY OF EXI'ERINENTAL APPARATUS AND PROCEDURE

A detailed discussion of experimental techniques

is given in the Appendix, here it is merely intended to give
an outline description in order that the resui.t8 should be
understandable.

The initial interest in the flow around a wheel

arose out of a study of the aerodynamics of Qand cars,
which, because of the racing rules, have totally exposed wheels
As a consequence the shapes tested are representative of the
tyres and rims currently used for 'Formula One' racing
cars. The diameter of all the model wheels is 16.375 ins and,
at the test speed of 61 ft/sec this gives a Reynolds number
of 5.3x10 5, which is considered sufficiently high for the
tests to be accurately representative of the real flow (i.e.
in the supercritical range of cy2indere and spheres). Six
wheel shapes were tested in all; three different tread widths
(designated A,B, and C with increasing width) and two edge
profiles (designated 1,2). Details of the profiles and the
distribution of the pressure holes are shown in Fig. (2-1) and
(2-2). After being tested with a smooth rim, wheel C2 had
three grooves cut into the surface as indicated in Fig. (2-2)
and was thon retested. The wheels, which were machined from
light alloy, rotate freely in bearings and are supported
rigidly in contact with a moving floor by a system of adjusthb]e
 31

rods. Photographs of a wheel mounted in the tunnel are given

in Fig. (2-3), Checks with a stroboscope indicate that the
peripheral speed of the wheel is the same as the floor,so no
slipping occurs between them.

The wind tunnel boundary layer is bled off ahead

of the moving floor by a fairing and suction is then applied
through a 'Vyon' porous plastic sheet to prevent a new layer
forming (Fig. (2- It)). Originally the speed in the thin detached
shear layer shed from the trailing edge of the fairing fell
to about 80% of the free stream value at the wheel centre
position,but in its final form the fairing prevents the speed
falling below 95% (see Fig. (2-5)). As noted inthe introduction,
the flow in the corner region between wheel and floor is likelytobe siflcant,
so itis desirable that the above shear layer should be reduced,
as much as possible,to decrease any effect on this flow.

As the wheel rotates,the pressure sensing holes on

its surface experience a cyclic variation in signal, corres-
ponding to the distribution of mean pressure around the peri-
meter. Superimposed on this are random fluctuations arising
from the unsteady flow in the wake, The measuring system
must be able to respond sufficiently rapidly to resolve the
transient details of these fluctuations,and the data [Link]
equipment has to be capable of separating the random signals
from the mean cyclic variation, The pressure sensing instru-
mentation is shown in Fig. (2-6). A condenser microphone
(B.' K, 1/Zt tI), with its associated oscillator (DIsA No.51E32),
 32

is placed on the axis of rotation, to eliminate centrifugal

effects on it, and connected to one pressure port at a time.
The relatively high level signal is then taken from the rota-
ting assembly,through slip rings (using low noise silver gra-
phite bru8hes),to a DISA Reactance Converter (no. 51E01),the
output of which is proportional to the pressure acting on the
microphone.

Mounting the microphone on the axis meant that the

connecting tube between it and the sensing port had to be at
least 12 ins, long and was chosen to have 3/32 ins internal
diameter to lessen the effects of attenuation, As a conseluence
resonance was experienced at about 200 Hz (and at 3,5,7 etc.
times this), but by placing wire wool in the connecting tube
to act as a damping agent, the resonance was minimised and a
frequency response such as that of Fig. (2-8) obtained, Note
for comparison that the frequency of rotation of the wheel is
111.25Hz. The same length tube was used for all the connections
to the sensing ports to ensure a unique calibration. A correct-
ion is applied to all the resulting pressure measurements to
allow for the centrifugal pressure difference between the
microphone on the axis, and the surface of the wheel.

A problem, that was not anticipated, was the

unacceptable,large long term drifting of the transducer output.
This difficulty was overcome by making a branch in the tube
connecting the microphone to the sensing hole and leading this
pressure through narrow bore tubing and a rotating seal to a
 33

manometer, This instrument therefore indicates the mean level

of the signal and the microphone the fluctuating component.
The output signal from the microphone is processed by a
Hewlett-Packard Correlator (Type 3721A) which operates in the
'signal recovery' mode. This instrument is able to digitally
sample and average several thousand pressure histories; thus
accurately eliminating the random components and retaining
only the true cyclic mean distribution. A trigger signal to
initiate each sampling cycle is derived every revolution of
the wheel by an inductive pick up. The mean pressure distri-
butions are stored within the correlator, from which they are
recorded on an X-Y plotter and on punched paper tape for
subsequent computing of the forces on the wheel.

These arrangements could not be used for measuring

the pressures acting on the stationary wheel. These [Link]
were obtained graphically on an X-Y plotter using signals
from the pressure transducer and a potentiometer connected to
the wheel to serve as an angular displacement transducer. By
slowly rotating the wheel by hand the distribution for each
hole was plotted automatically.
 3k

() RESULTS

Note on Errors

A discussion of possible errors and where they

arise i8 included in the appendix but it was thought neceary
to give some indication here of the likely error limits of
the results given in this section. The general conclusion reached
in the appendix is that the results obtained for the lift and
drag coefficients of the wheels should be within 5% of the
true value (i.e. the true value of the forces on the experfmenl
wheels; the forces on the real road wheel will be affectd by
such factor as tread pattern and type of road surface. This
is discussed later). However, because the B wheels were the
first tested and measuring techniques were progressively im-
proved they may be subject to greater errors although within
about 10%. In particular the procedure of examining the
rapidly changing pressures near the ground at smaller inter-
vals, of 1/20 between points instead of the normal 50 was not
adopted until later, which wiflhave some effect on the results
for the lift. The B2 wheel also shows some variation in the
base pressure measured by different holes across the wheel,
which is difficult to account for,and the results for this
wheel must therefore be viewed with some suspicion. It i
unfortunate that there was insufficient time to repeat the
measurements for this wheel.
 35

( 3 . 1) Lift and Drag Coefficients

Two sets of lift and drag coefficients were calcul-

ated, the first (denoted by subscript i) are based on the
projected frontal area of the wheel (i.e. the 'X-sect. areas)
and the second (subscript 2) based on an area given by multi-
plying the wheel diameter with the width of tread in contact
with the ground. This tread width is Zt .25", 6.75" and 9.25"
for the A, B and C wheels respectively. Thus, although the
cross sectional area of the shape 1 and 2 wheels at any given
width will differ because of the change in edge shape, the
area based on tread x diameter will be the same. This second
area was felt to be of more practical interest,since it is
the tread width in contact with the ground which is the basic
length parameter in determining the road holding qualities
of the wheel.

The lift and drag coefficients for the six wheel

shapes tested are given in Table 1 and graphically in Fig.
(3-1). These coefficients were obtained by integrating the
static pressure over the external surface of the wheel, They
do not therefore includo any skin friction contribution to
the drag . However, this contribution is usually a small
part of the drag of bluff bodies because the separated region
produces such a large pressure drag. The coefficients do not
include any contribution from inside the hub either. These
are given in a separate column of table 1.
 36

Considering the subscript 1 coefficients first,we

see that both the lift and drag coefficients increase wi-jh
width. The coefficients for the A and C wheels are almost
the same for the two different edge shapes, whilst those
for the B wheels differ by about 10%, although this may be
due to the greater error associated with these wheels. In
changing to the second set of coefficients,these values are
multiplied in varying degree, depending on the ratio of the
cross-sectional area to the 'tread x diameter' area, to
yield the results shown. Surprisingly,the drag coefficients
no longer increase with width and,in fact,are almost constant
for the edge shape 1 wheels.

The above coefficients can be taken to represent

the forces on flush sided wheels,since they do not include
any contributions from the indented hub. These were obtained
by integrating over the hub surface using the pressures from
the holes shown in Fig. (2-2). The contribution to the drag
is greatest for the narrowest A wheels ,whilst that to the
lift is negligible for all the wheels. For this reason a
column of values for CL1 is not included in the table. We
thus have the surprising result that,when the hub contribin
is taken into account,CD2 is actually greatest for the
narrowest wheels.

(3.2) Comparison with Previous 'Work

If we now compare our results with those of previous

 :37

investigators as given in table 2, we will find that the

MIRA values are on the whole compatible with the present

results, but that }Iorelli's results are entirely different.

The reasons for the latter were given in the introduction

and a difference in values was only to be expected. We

suggest that Norelli's results are not at all representative

of the real flow about a wheel on the ground due to the gap
under his wheel and,when interpreted as such,are considerably

in error, even to the extent of producing a lift force of the

wrong sign.

In comparing with the work of MIRA,it should be

remembered that these results were obtained using a complete

idealised 'Formula One' racing car model, the wheels of which

were mounted above a stationary floor with the gaps underneath

sealed by paper strips, and, as explained in the introduction,

this arrangement is subject to a number of errors. Note also

that the MIRA wheels are square edged and flat sided, This means

that the coefficients based on actual x-sect. area and those

based on tread x diameter will be the same (i.e. CD2, CL2,

and CD 1, CL1, are the same) and there is no additional con-

tribution from the huh,

In table 2 the coefficients with subscript 1 are

for the forces due to the front wheels only, but those with

subscript L1 are for the forces due to all four wheels,

Surprisingly, the MIRA results for the front wheels only are

quite close to the results for the B wheels in the present

experiment,despite the differences in simulation. It is un-
fortunate, from our point of view, that the effects of wheel
width were only studied for te four wheels together, 8ince
it makes comparison with the present experiment difficult. It
is not possible to separate the effect of the width on the
individual wheels from its effect on the interference between
the front and rear wheels. However, the variation with width
is fairly similar to that of the subscript 2 coeffiôients of
the sharper edged shape 2 wheels.

It appears as though the NIRA results do give the

right order of values for the coefficients,despite the ina-
dequate simulation in their experiments, although the coef Lciait
for the wheel shape they tested may not be entirely correct.
It is impossible to quantify any errors,unless the complete
simulation is done for their wheel shape,and this is really
the weakness in their results.

(3.3) Variation of CL and CD across the Wheels

One advantage of finding the force coefficients

from pressure measurements is that much additional information
is also obtained. Thus,one not only finds the total forces1
but also the contribution to these forces from the different
parts of the wheel. This ability has been utilised to find
the variation in tLe lift and drag coefficients across the
wheels given in Figs, (3-2) to 3-5). In all these figures,
the distance from the centreline of the wheel, w, was non-
dimonsionalized by dividing by half the samo tread width, t,
as was used in calculating tho subscript 2 coefficnts.
 39

This does mean that the maximum value of w/t will be different

for each wheel, which would not be the case if the total

width of the wheel had been used. However, dividing by the

tread width does have certain advantages when interpreting

the figures in relation to the CD 2, CL2 results of table 1.

This is demonstrated by Fig. (3-2) which gives the

variation in CD across the wheels for edge shape 1. It is

immediately obvious from the figure how we obtain the rather

surprising result that CD2 is almost constant for the three

widths, The coefficents for the wider wheels are greater near

the centre-line, as might be expected, but this is balanced

by the more rapid fall near the edge of the wheel. This latter

fact is reasonable,when one cnsiders that the tread width

forms a much smaller proportion of the total wheel width of

the narrow wheels than it doeS of the wide wheels. By using

the tread width as our basic length therefore we are, in

effect, causing the wider wheels to have sharper edges than

the narrow wheels,

Note that the results were only taken for one•

half of the widest C wheels. This was because the length of

tubing, between the measuring holes and microphone, which

would have been required to obtain measurements over the

whole wheel, would have resulted in a resonance in the micro-

phone response at too low a frequency to have given satis-

factory results. In calculating the total CD and CL for

these wheels, therefore, it was assumed that the distribution

was symmetric about the centreline.

Fig. (3-3) gives the corresponding variation in drag

coefficients for edge shape 2 wheels. The most striking feature
is the behaviour of the B2 wheel, in particular the large wavy
variation near the centre].ine. This, and an equivalent wavi-
ness of the lift coefficient (Fig. (3-5)), are consequences
of the variation of pressure in the base region of the wheel
(illustrated in Fig. (3-17)). None of the other wheels show
such large changes in base pressure across the wheel and,as
suggested before,tliere must therefore be some suspicion that
the measuring system was in error in this case.

The behaviour of the lift coefficents given in Figs.

(3...1) and (3-5) is very similar to that of the drag already
discussed, including the negative contributions from the
edges of the wheels, with the exception that the B wheels
have value of CL very close to or exceeding(for B2) those of
the C wheels,even near the centre-line of the wheels. This
may well be a consequence of not examining the pressures near
the ground at smaller intervals,since the result of doing
this for the other wheels was a reduction of about3% in the
lift coefficients near the centre-line.

(3.L1) Pressure IDistributions for A Wheels

It is not intended to present all the pressuro

 41

distributions obtained for all the wheels since there is an

essential similarity between wheels, rather we shall give

detailed results for the A-wheels only and then concentrate

on pointing out any significant differences for the other

wheels.

Figure (3-6) presents the pressures measured around

the centre-line of the A-wheels. The most remarkable feature

of both curves is the sharp rise to a positive pressure

coefficient greater than 2 near the floor (e = 900). Now,

the c , cannot rise above the stagnation value unless extra

energy is being injected into the flow and this is produced

by the viscous actions referred to in the introduction and

associated with the moving boundaries. The pressure does not

fall immediately to a low value after 0 = 90° but takes about

40 to reach a c of zero and then oscillates briefly before

settling down at the base pressure. This might be an effect

connected with the response of the measuring system, but the

frequency of oscillation, 'about 800 liz, does not correspond

to any of the theoretical resonance peaks of the system

(nearest are 600, 1000 Hz). Instead we fool that this may

well be caused by the moving floor lifting in the low presstre

region behind the wheel, to toudh the surface, and

then falling back. This process would occur in the sma1]

region under the wheel,beliind the line of contact with the

floor,and would create the fluctuating pressure picked up by

the measuring hole.

 42

It would also account for the lack of the negative

pressure peak predicted by the theoretical solution for this

region given in section (4-22).

The positive pressure peak,which was found,will

contribute a significant amount to the lift coefficient of

0 0
the wheel. In fact the range of 6 from 75 to 10 contributes

almost 30% of the lift coefficient for this hole (.64),

In connection with the lift, the other important

feature is the position of separation from the top of the

wheel. Experiments to be described later,using flow visüa].i-

sation and Pitot-static traverses of the wheel boundary layer,

all suggest that separation occurs about e = 280_2900, i.e.

0
10 -200 in front of the top of the wheel. The pressure dis-

tribution shows a slight adverse pressure gradient before

_ 3000 to 280 0) and then a reversal of slope
this position Ce—

at around 280°. We shall take this as indicating the separatLon

position,although the mean pressures over the rear of the

wheel behind this position are not particularly constant.

Tlowever, the important point is the absence of any

large negative pressure over the top of the wheel, which

would occur if the boundary layer remained attached. This

early separation therefore acts to lower the lift coefficient

of the wheel, opposing the action of the viscous effects at

the bottom of the wheel.

Both these effects are associated with the moving

 413

boundaries of the rotating wheel and the moving floor, (it is

shown later that the early separation does not occur on the
stationary wheel),and so would not be produced in any simulation
which did not involve these features. Such simulation would
therefore be liable to errors in estimating the lift [Link]
and, since these special features will presumably have some
influence on the recirculating flow behind the wheel and hence
the base pressure, the drag coefficient might also be in error.

Note the great similarity between the curves in Fig.

(3-6), suggesting that changing the edge shape appears to pro-
duce little change at the centre-line even on the narrowest
wheel.

Fig. (j-7) and (3-8) give the pressure distributions

at holes 8 and 7 respectively, both of which are on that part
of the wheel which comes into contact with the ground. The
form of the curves for these holes and hole 9 is very similar;
all show the sharp rise in pressure near the ground and have
the separation position around 275_2800. The base pressure is
also similar and shows the same variation. The most noticeable
change is that for hole 7 the pressure no longer rises to near
stagnation value at the front, but this might be expected since
this hole is displaced by 1.6 ins from the centre-line.

Hole 6 (Fig.(3-9))is the first position where the

surfaco does not actually contact the floor at the bottom,
although Ihe gap in this case is vory small (about .05 ins).
 44

Ilowever,at e = 900 there issill a sharp rise in pressure,

althoughnottothe same extent as before, and now immediately

followed by a similar sudden decrease in pressure. Other
features of the distribution correspond to those of previous
holes.

The distribution shown for hole 5 is of a different

form to those already described. The pressure does not become
positive at any point on the curve and indeed becomes highly
negative between e = 500 and 80 0, for both wheels. For wheel
A2 the pressure is more negative over most of the front than
it is over the rear. Note that, for this hole and following
ones up to hole 1, the positions of the holes on the two
wheels no longer correspond (see Fig. (2-1)), 50 that agree-
ment between the two cannot be expected. Because of the more
neg,tive pressure acting in the lower front region, this hole
contributes a negative drag and lift to the total forces, for
both wheels.

The distribution for hole 4 (Fig. (j-io)) is similar

to that for hole 5,but with less variation near the ground.

Holes 3,2,1 at the side of the wheel give similar

resilts, with the variation in pressure over the rear half
growing larger from hole 3 to hole 1.

Fig. (3-12) shows the pressures measured by two

holes placed inside the hub at positions indicated in Fig.
(2-2). The positive pressures at 0 = i10° to 2000 from hole
Hi, suggests that the air is entering the hub in this region.
The contribution to the lift from these holes is not very
significant, but there is quite a large contribution to the
drag. In fact integrating over the surface of the hub using
just these two holes gives an addition of .12 and .14 to the

CD2 of .97 and .90 for Al and A2 respectively. There is little

point in obtaining this value to greater accuracy by using
more holes, since it will depend on the shape of the hub which
is likely to vary from car to car. However,since our hub
shapes are similar for all the wheels tested,we can directly
compare between the results. In the present case,the addition
to the drag is the same for both wheels and significant enough
-to 5uggest that a cover over the hub well, flush with the
wheel edge, would greatly reduce the drag of the wheel,although
there may be other considerations, such as brake cooling,
which would make this undesirable.

Only mean pressures were given in the above figures,

as obtained from the correlator. However,the fluctuating
part of the signal can contain some useful [Link]. For
example,the separation position could be deduced as the point
at which large scale pressure fluctuations begin,due to the
unsteady nature of flow in the separation bubble. An example
of the non-averaged signals is given in Fig. (3-i3) where
the signals for wheel Al are presented. Note that for holes
9,6 and 5 the major variations in the signal are over the
rear, boginning at about the point previously referred to as
 46

the separation point, and also near the ground at the front
( cf. Figs. (3-6) and (3-9)). A noticeable feature of the
signals from the three edge holes, is the negative pulse-lik.e
fluctuations that occur, of about 1 to 1 1/2 times the dynamic
pressure in magnitude, although these are not as frequent as
might be suggested by the photographs, which were chosen to
include this phenomeri in order to illustrate the very large
scale of the turbulent pressure fluctuation around the edge.

(3.5) Pressure Distributions for other Wheels

We shall now compare the above results with those

for the other wheel widths, doing this first for the shape
1 wheels, Fig. (3-14) presents the mean pressures given by

the holes an the centre-lines of the three wheels.

At the ground, 0 = 900, the B curve does not rise

quite as high as the other and then falls more slowly. Again,
this is , most likely, a consequence of using s° intervals
between measurements for this wheel, as opposed to 1/2° for
the other, since the other curves would also be more spread
out if a higher resolution had not been used. It is this
spread which suggests that the contribution to the lift from
this region may be too high for the B wheel, although this
will, be mitigated,somewhat,by the lower maximum value at 9=900.
The oscillation in the pressure after 9 = 90 will not show up
with the 50 intervals and,in interpolating botwoen points in
 C, hi

this region,we have drawn a smooth curve,ae this is the shape

assumed in the integration of the pressure distributions.

Major diff' in the pressure occur over the

lower front face,where the narrow A-wheel has a much lower
pressure than the other two wheels,and over the base region
where the widest wheel has the lowest pressure, as might be
expected. Near the separation position, 9 = 2800, the B and C
wheels behave in a similar way and both appear to have aepa
ration slightly further forward than does the A wheel. Over
the upper front of the wheel all three wheels behave alike.

The pressure distributions at holes 6 and 5 are

presente4 in Fig. (3-i5). The most obvious features of the
figure is that the complete change in appearance of the
distributions between these two holes is common to all three
wheels. For hole 5, the pressure distribution of the B wheel
is much closer to that of the C wheel than to that of the A
wheel, although for hole 6 this is not so noticeable.

If we look at the pressure distributions given by

hole 1 and by the two holes 1nsi1e the hub (Fig.(3-16)), we
again find that the B and C wheels behave alike, while the A
wheel is noticeably different, There are no results for 111
from the B-wheel as no hole was provided in this position
for either of the B-wheels. The greater variation irL pressure
inside the hub of the A-wheel suggests that there i more

interaction with the external flow for this wheel and this
 48

is reflected in the respective lift and drag coefficients

from the hub given in table 1.

The pressure distributions near the centre-line of

the shape 2 wheels show similar features to those for shape
1 given above. The only exception is the behaviour of the
pressures in the base region of the B2-wheel,which has been
mentioned previously as being anomalous • The large variation
in base pressure across the wheel is illustrated in Fig.
(3-17) which gives the [Link] at two adjacent holes.
The pressure distributions near the edge of the wheels are
given in Fig. (3-18). There is less evidence of any difference
in the behaviour of the A-wheel compared to theothers,espec-Ir11y
at hole 1. Inside the hub the variation in pressure is much
as before,although with a slightly greater variation for all
the shape 2 wheels.

(3.6) Effects of Rotation

In order to isolate the specific effects of wheel

rotation on the flow field,a comparison was made between a
stationary and a rotating wheel, including measurements of
total head. The wheel chosen for this comparison wa B2. It
is unfortunate that the B2 rotating results are not very
consistent with measurements taken of the other wheels,at a
later date, but it is felt that these inoonsistences are of
such a nature as to in no way invalidate any general conclusions
drai.n from the comparison with the stationary wheel.

The two sets of results [Link] in table 3, which

suggests that the effects of rotation are beneficial in

reducing both the drag and the lift of the wheel. If we turn

to fig. (3-19), which gives the mean pressure distribution

around the centre-line, we can see the reasons for these

reductions. The lift produced by the sharp pulse of high

pressure at the bottom of the rotating wheel is more than

balanced by the large region of low pressure over the top of

the stationary wheel, whilst the lower pressure acting over

the rear of the stationary wheel contributes towards the

groater drag. The figure shows that the early separation from

the wheel noted previously is a product of the rotation of

the wheel and not an intrinsic quality of the shape. This

early separation probably also plays a large part in deter-

mining the different base pressure, although the action of

the rotation on tho separated region itself may be [Link],

since one would normally expect a later separation to yield a

higher base pressure.

If we compare the iiressures in Fig. (3-19) with

those in Fig. (1-3) and (1-6) for a stationary sphere and

cylinder respectively, we see that the pressures over the top

of the stationary wheel boar a close resemblance to those of

the sphere and cylinder at an equivalc'nt Reynolds number

(5x1o),but the pressures over the top of the rotating wheel

resemble those of the sphere and cylinder at a lower Reynolds
 50

nuinbor,witli laminar separation. Despite this resemblanceit

is shom later (section ('i-i-i)) that the separating boundary

layer above the rotating wheel is indeed turbulent before

separation, which is in agreement with the results of GrifTiths.

and Na (1969), discussed in the introduction, who showed that

the rotating surface moving against the free stream provokes

transition and does not suppress it.

It is illuminating to consider a direct comparison

botwoon the pressure distributions obtained by NIRA for the

wheel with a sealed gap and the present results, for both the

stationary and rotating cases. This is given in Figs. (3-20)

and (3-21). Dealing with the rotating case first, we see that

the base pressures are in fair agreement , but,near the ground

and near separation,thore are wide discrepancies between the

two results,

The loi pressures given by NIR at e = 900 are

quite unrealistic for zero ground clearance,heing either clue

to poor interpolation,or poor simulation,as mentioned pre-

viously. Near separation,wo feel that the measuring technique

used by NLIA would have been in error, if only by disturbing

the boundary layer flow sufficiently to affect the separation

position. Note that the Ifl{A results imply a very steep

adverse pressure gradient before sepzration.

The two stationary wheels show much the sane

 51

differoncos. The agrccmont in base Pressure is not so Good,

but both results give the stationary wheel a lower base

pressure than the rotating one. The same remarks as before can

be made regarding the differences at the ground and at separation,

Note that although the IIIRA results do not agree with the

present ones near separation, they do show a greater region

of low pressure over the top of the stationary wheel, than

over the rotating wheel.

The great differences between our results for the

rotating and stationary pressure distributions occur right

across the wheel as is shown by Fig. (3-22), which compares

the pressures at hole 6 and hole 5, and also by Fig. (3-23),

which gives the variation in CD and CL across the wheel

Note that the stationary wheel xhibits the same form of wavy
variation in and CD noar• the centre-line as the rotating
0L
wheel, although not to quite uch a large extent. It appears
from the pressure distributions for the stationary wheel
that at w/t 3,(hole 9) the boundary layer remains attached

further around the wheel (r'ig.(3-24)) giving the greater

lift and lower drag at this point.

As well as the above results for the pressures and'

forces acting on tho wheel, the external flow field was

examined by measuring total head with a Kid tube (for des-

cription see Appendix ). A total head measurement below the

free stream value indicates that the fluid concerned has

undergone viscous action. Thc position of shear layers and

 52

wake regions can therefore be identified by a survey of total

head. The total head was measured in four nianes at right

angles to the flow direction at positions,shown in 11'ig.(3-25),

the results obtained for these planes being plotted in the

next four figures. These results are presented in the form

of contour lines joining points of equal total head value.

IT -
Only the contours for values of = 0.9 are shown ,

partly for clarity and partly because it was felt that further
in the wako the total head measurements would be affected by

the high intensity of the turbulence. The 0.9 contour can

loosely be thoughi of as a map of the edge of the wake. Only

the half of the flow field opposite to the supporting rods is

shown,s±nce their wakes would only confuse the picture,and

for the same reason measurements were not taken very close
to the ground due to the wake of the fairing at the front of

the floor.

Fig. (3-26) shows the .9 total head contour for plane

1, 1.7 ins, downstream ofthe wheel axis,Contours for the

rotating and stationary cases are compared (designated fi and

S respectively). Note that the li-contour is well away from

th top surface of the whoel,su;esting that the boundary

layer has already separated upstream of plane 1 • On the other

hand ,the S-contour shows the boundary layer still attached

to the stationary wheel, thus tending to confirm the pressure

measurements described earlier.

At the side of the whoelthe fl-contour is well away

 from the wheel outside the upper part of the recessed hub, but

comes quite close further down. Neasurements, using wool tufts

to give the flow direction, suggested that the air inside the

hub is on the whole rotating in the same way as the wheel,Thus,

the air in tho upper part is moving against the external flow,

whilst that in the bottom part is moving with the flow. We

would therefore expect there to be more viscous interaction

between the air in the upper part of the hub and the free

stream. This effect does not occur with the stationary wheel,

of course; the contour suggesting that there is little inter-

action.

Lower down, the contours in both cases move right

away from the wheel. This is the turbulent region produced

by air coming out from undex the front of the wheel. Rather

surprisingly, at first sight , the S-contour lies outside the

fl-contour. One might have expected the jet produced under the

rotating wheel would have created a greater disturbance.

- However, the wider spread around the stationary wheel is

probably due to the boundary layer on the stationary floor

separating ahead of the wheel and rolling up to form a horse-

shoe shaped vortex around the wheel.

Figs. (3-27), (3- 28 ) and (3-29) show the development

of the wake as it moves downstream, The main points of note

are t] o differences between the two contours at the top of

the wheel and the spread of' the large disturbed region at
the bottom. The U-contours at the top on the rotating caso
 5,'

•1

remain at a high level, just above wheel height, but the S-

contours suggest that the air is swept down close behind the
wheel, only breaking into a disturbed wake at just about
half wheel height, One interesting feature of the stationary
case is that the contour at about 1 ins from the centre-line
appears to remain closer to the wheel than it does at
neighbouring points. This is the position corresponding to
hole 9 and, as noted previously, the pressure distribution
measired by this hole also suggests the above. However, the
total head measurements for the rotating wheel do not show
any such non-uniformity in this region,even through the lift
and drag coefficonts have the same wavy variation as the
stationary ones. The disturbed region at the bottom of the
wheel spreads both outward and upward,the fl-contour remaining
inside the S contour at each station, and this region extends
halfway up the wheel and out to the edge of the moving floor
by plane 3.

As a further examination of the separation region

and the flow around the bottom ofthe wheel,soine flow visuali-
zation experiments were conducted using wool tufts and
smoke. Some of the photographs obtained are given in Figs.
(3-30) and (3-3i) The top smoo photographs clearly show
the earlier separation and higher wake region of the rotating
wheel. The wool tufts confirm this and also show the reversed
flow back over the top of the wheel. The spread of the flow
around the bottom of the wheel is shown in the two lower
photographs of Fig. (J-3O, Note tho greater spread in the
 55

non-rotating casc' confirr1ing the total head measurements.

The above results have enabled us to distinguish

those features of the flow which are specifically due to the

rotation of the wheel. The early separation over the top of

the wheel is seen to be caused by the rotation and not to be

an inherent quality of the wheel at the Reynolds number used.

The rotation also produces viscous actions near the ground

and these, together with the early separation, are the basic

agents in determining the pressures and hence the forces

acting on the wheel. The wake region is also to a large extent

determined by these effects, together with .any extra rotation

induced into the wako by the moving surface of the wheel,

(3.7) The Effoct of Grooves in the Wheel Surface

Three, 1/Li" wide, 1/8" deep,. grooves were machined

into the surface of the C2 wheel, at the centre-line and 2"

either side of it. It was hopea that these grooves would

approximately simulate the effects of tyre tread pattern on

the wheel flow. In fact, the grooves were deliberately made

larger than most tread patterns to exagerate any such effects.

Table 1i presents the results for the two configura-

tions of C2. From what has been said previously regarding the

consequences of leaving even a small gap undor the wheel, it

might have boon oxpoc Led tha L tho grooves would have a more
dramatic effect on the wheel but the decrease in the lift

coefficient CL1 is onl about 10, whilst the drag coefficient

CD1, remains the same. In generating the second set of

coefficients,the tread width in contact with the ground of

the grooved wheel is taken to be less than that of the normal

wheel by 3/4" (i.e. 8.5" instead of 9,25"), 50 the area on

which the coefficients are based will be less. Thus CD2 be-
comes greater for the grooved wheel, while CL2 is almost the
same for the two wheels. When viewed in terms of the amount

of wheel actually contacting the ground, therefore, the

grooved wheel is less efficient than its normal counterpart.

To seek the reason for this low efectivenoss,we must

consider tho individual pressure distributions. Fig.(3-32)

gives a comparison at hole 11 on the centre-line of the two

wheels and,therefore,within the central groove of the grooved

wheel. The first point of note is that, although the pressures

differ markedly in the vicinity of the ground, over the rest

of the wheel there is virtually no change at all. Our dis-

cussion of the effect of gap under the wheel leads us to

expect a low pressure in the groove at the bottom, but note

that the c rises to a value close to 1 at about 800 before

p
950•
falling to its lowest value of -1.2 at Thus we see that

the effect of the groove is confined to a fairly small region

at the bottom of the wheel, although this is an important

region for the lift,bocause of the very high pressures normaily

found there. Thus the lift for this particular hole is reduced

considorably,but because of the localized nature of the effect

 57

the drag is not altered.

To see if the groove effect is less localized across

the wheel,we compare the pressures at hole 10, midway between
two grooves (Fig.(3-J3).). It is immediately obvious that the
grooves have no effect on this hole, in particular, the high
pressure at the bottom of the wheel is as strong as ever.
This is not too surprising,if we remember that it is generated
by viscous forces which are themselves- very localized in action.
If we examine the pressures measured by the other holes, we
find that only those holes actually in a groove are at all
affected. This is best seen by looking at the variation of CL
and CD across the wheel for the two cases, given in Fig.
(3-3k). Here we have used the same value of t = 9.25/2 for
both wheels,as we wish to make a direct comparison of the
same position across the wheel.

It is surprising that the air passing through the

groove at the bottom does not have more effect on base pressure,
since even quite low rates of base bleed can reduce the nega.-
tive base pressure and hence drag by a significant amount.
It is possible that the viscous actions,still generated by
the surface which contacts the ground,act to reduce the effect
of the air passing through the grooves,since the action in
the rear corner will tend to induce any air to flow back into
the corner,as discussed in section (k-23).

It has been argued (e.g. Iiorel1i (1969)) that,

because the grooved tread pattern of tyres allows air to pass
 58

under the wheel, this is equivalent to having a slight gap

between wheel and floor. The major criticism of this practice

is that any errors introduced will not be quantifiable unless

the full simulation is carried out, Our results show that

the effect of ftio grooves is to decrease the lift in their

immediate neighbourhood only and that the viscous actions

associated with the contacting surfaces of the wheel and ground

will still occur and be important. This being so,allowing any

gap under the wheel, which will eliminate these effects, will

not give a realistic representation.

(3. 8 ) Variation in Wind Speed

In order to accurately model the wheel flow in the

wind tunnel it is necessary to equate the wind speed with

the speed of the floor, However, since the wind speed is

likely to drift slightly with changes in temperature and

other atmospheric conditions,it was decided to investigate

alteration of the wind speed from tho optimum equal value in

order to discover whether the flow was very susceptible to

such changes. If this was so, it would be a source of error

difficult to account for.

Fortunately, our results for wind speed values either

sido of the desired equal value shows that there is little

deviation from the' results already doscribed. Fig. (-35)

compares the three rossuro distributions at the centre-line

of the C2 wheel for v/U= .87, 1,0, 1.07 (v = rotational

 59

speed). The pressures given are uncorrected for blockage effecta,

since the correction used (due to Naskell,see Appendix) c1eper1s
on a knowledge of the mean base pressure across the whole of
the base region of the wheel and this was not available for
the difforent win ppeeds, only the pressures on the centre-
line being obtained. The pressures wore measured at 5.10
intervals around the w1iel only, so the interpolation near
9 = 900 may be inaccurate. Bearing this in mind the agreement,
within 2%, for the CD and CL is quite acceptable. The most
significant difference in the pressure distribution occurs
near the separation position, 9- 280 0, where it would appear
from the lower pressure that the separation occurs further
downstream at higher wind speed (i.e. lower v/U ), If,as has
been suggested, the early separation from the wheel is pro-
duced by the rotation, it would seem reasonable that a lower
v/U would decrease this effect.

To investigate this further,some more pressure

distributions were obtained at much lower wind speeds. It was
not possible to obtain any at higher speeds because of excessive
belt lift. Also, the wind speed was chosen to bo varied,because
changing the floor speed involves the fitting of new gear
ratios in the drive, for every speed change. The results,un-
corrected for blockage as above, are presented in Fig. (J-36)
The pulse of high pressure at 9 = 90°appear s greater at the
lower speeds although in absolute magnitude this is not so,
In fact this phenomenon is more closely related to the speed
of the boundaries, which is kept constant, than to the wind
 60

speed and, as the pressures at each wind speed are normalized

by the appropriate dynamic pressure, the pulse magnitude, in

coefficient terms, will appear greater at the lower speeds.

The position of separation at the higher values of

v/U is difficult to interpret,since the base pressures are

far from constant. It is probably best merely to say that the

low pressure over the top of' the wheel, is progressively re-

duced by the rotation, although the indentations in the

pressure distributions, around 8 = :300°, do suggest that the

phenomenon we have called separation is moving upstream as

the rotational speed increases, except that it no longer

gives rise to even approximately constant pressure. Despite

this reduction in pressure at the top of the wheel the CL of

the lower v/U values is the greatest because of the increase

in importance of the high pressure at the bottom. The CD re-

mains much the same for all the yalues of v/U tested, This

applies only to the uncorrected values, of course,ifacorrection.

for blockage is applied the above conclusion may be different.

It is felt that the findings of section (3-6),where

the rotating and stationary wheels were compared, have been

confirmed in the present seotion. It appears that the rota-

tion is responsible for reducing the pressure over the top

of the wheel (i.e. provoking early separation) and for

reducing the negative base pressure, the amount of reduction

in both cases being proportional to the speed of rotation.

 61

(3,9) Two flmc'nsonnl Flow

It was felt that the two dimensional flow past the

wheel (or, in two dimensions, therefore, a cylinder) would

prove of interest in a number of ways; it would provide a

bridge between the theoretical work which applies to the two
dimensional case and the experimental results for the wheel;

it would give an extreme value for the aspect ratio results;

and should enable those special features associated with ro-

tation to be more easily observed, since in a sense, these
are diluted in the three dimensional flow.

Unfortunately, in the time available, it was not

possible to use a cylinder of very high aspect ratio as this

would have meant major changes to theapparatus, so a com-

promise had to bereached by using the widest wheel available

with square edge j5rofiles, giving a width of 13 1/ li. ins., and

fitting it between endplates, The arrangement in the tunnel

is given in Fig. (3-37). Flat plates were fitted normal to

the air stream outside the end plates and flaps fitted to the

rear of the endplates,in order to balance the flow through

the three sections, The ends of the wheel were covered in

felt cloth and the bottom of the endplates with draught ex-

cluding strip in order to prevent undesired flow through any

gaps.

Although care was taken to achieve as uniform a flow

through the central section as possible, clearly, with an

 62

aspect ratio less than one, true two dimensional flow will

not be attained and significant three dimensional effects

were found. Our results must therefore be considered as not

truly representative of the two dimensional case, especially

in such features as separation and base pressure. The realisa-

tion that we were unlikely to attain very satisfactory

answers decided us against pursuing this project in detail,

instead measurements were restricted to obtaining the mean

pressure distribution near tho centreline of the cylinder and

taking some smoke photographs of the flow.

Anothersource of. possible error is contained in

the calculation of the correction which must be applied to

the dynamic head to allow for blockage, The blockage corrections

of the flat plates in the outer sections of the tunnel were

calculated using Maskell's method (which can be applied to

flat plates near the ground according to Gould (1970)) and

the results averaged. It was then assumed that,since the flow

in all three sections had been balanccd,the blockage correction

of the wheel in the central sdction would b the same as the

average of the plates. The addition to the dynamic head,

given by this process, amounted to almost 30%. Such a large

degree of bloc1 age would almost certainly affect the flow in

other ways,hosides increasing the effective free stream

velocity; in particular, the separation position may well

be altered. IIowevc'r, despite all the foregoing, the results

obairiec1 ar considered sufficiontly interesting to be in-

c1udd here.
 63

Fig. (3-39) givc's the average pressure distribution

around the wheel, the averae being taken over 5 holes spaced
at 1 ins, intervals with the mickle hole on the centre-line,
Some idea of the non-uniformity of the flow is given by
comparing the results from each hole as in Fig. (3-38).The
rotating distribution shows much the same features as with
the ordinary wheel,but the base pressure is lower and the
separation slightly further forward. There is now no stagna-
tion region near the centre of the wheel at the front,but
the pressure at the ground still rises to a high value,
although no higher than in e three dimensional case, again
demonstrating the local nature ot this feature. Note that
average pressures were only taken at every 5.10, 80 any
oscillation in pressure after e 90° will not show up.

The stationary results are considerably different

from those presented for the B2 wheel in section (3.3). One
major reason is that the boundary layer on the stationary
floor in front of the two dimensional wheel separates ahead
of the wheel and thus effectively 'streamlines' the flow.
Thus the pressure on the lower face is that in this separated
region. Note that separation from the wheel now occurs much
closer to the top, about e = 260 0 , which is probably the
reason why the low pressure over the top is no more than in
the three dimensional flow (see Fig. (3- 1 9)). The base prsure
is more constant in the two dimensional case and is also re-
inarkably similar to the rotating value, although this may
well be an effect of the upstream separation from the floor,
 6Li

rather than an indication that the rotation no longer affects

the base pressure. The results of 'Wacicock (1970) for two
stationary cylinders in contact, indicate a base pressure
around -15, but in our results the stationary base pressure
is little lower than the three dimensional stationary case,
at about -.7. However, Wadcock's figure was obtained at a much
lower Reynolds number, which might explain the difference
(2.5x10 against 5.2x1O).

Smoke photographs for the stationary and rètating

cases are given in Fig. (3-zIo). The earlier separation and
higher wake reon of the rotating wheel is demonstrated in
the figure, with ti-ic' difference from the three dimensional
flow that the separation from the stationary wheel now occurs
near the top, agreeing with the pressure distributions, The
bottom photographs are of the flow in front of the wheel.
Unfortunately, because of the separation in front of the
stationary wheel, the two pictures do not indicate a very
clear difference between the flotQs • however, one can distin-
guish that the flow leaving the floor in the rotating case
does so at a sharper angle. This flow above the movng floor
is almost certainly caused by the upstream moving jet emerging
from the cornerbetween the wheel and floor and is not a boun-
dary layer separation under adverse pressure gradient as in
the stationary case.

For comparison with the above, some more photographs

wore obtaiiied in a small smoke tunnel (26" x 2 1/4" section)
 65

of the flow around a simple image-system model, consisting

of two 2 ins, diameter cylinders in contact, with the wakes
being separated by a splitter plate, 20" length. Fig. (3L1.i)
presents photographs of the flow, with the cylinders rotating
in the same way as a Wheel and in the opposite direction, and
with them stationary. The influence of the rotation on the
separation position can clearly be seen. Note that the se-
paration appears further forward in the stationary case than
it does in Fig. (3-40),which is probably connected with the
lower Reynolds number of the image system (2x10 4 ),but may
also be influenced by the lack of separation in front of the
cylinder. When the cylinder is rotating as a wheel,there is
evidence of a large disturbed region at the front, The solution
for this flow in the corn Suggests that the jet emerges
midway between the bounding surfaces ,which would indicate that,
with the image system,the jet would point directly upstream,
whereas,with the moving floor,it would be angled upwards,In
this way,the image system is not 1 a complete analogy to the
wheel flow.
From the above observations of the flow, together
with the low Reynolds number solutions for the flow in the
corners, plus what has been conjectured about separation
over the rotating body, the likely mean streamlines of the
various two dimensional configurations have been draw-n
(Fig. (3-142)). It must be emphasised that these figures have
been constructed from the above information and are not te
results of specific measurements in the flow. They do, how-
ever, account for thct features shown in the smoke photographs.
 66

The figure of the cylinder on the moving floor illustrates

the unique properties. of the flow, including the jetting
at the front, the free separation point with its associated
complex boundary layer and the way in which the flow near the
ground at the rear affects the recirculating bubble. Note
that,in the mean,the fluid in the layer next to the cylinder
recirculates about the cylinder and separation region. In
this,it is compaxiable to the layer shown in Fig. (1-12),
which circulates around the rotating cylinder in a free stream.
In the three dimensional flow,the fluid which comes over the
top of the wheel is lost around the side in the forward going
jet and this mechanism must therefore extract fluid from the
base region. On the other hand,the action in the rear corner
should tend to induce fluid back into the base region and
would also contribute to the general increa8e in circulation.
in the separated region bel4nd the rotating wheel,

([Link]) Conclusions

We have usually drawn conclusions from the results

as they were presented and here we merely intend to summarize
the more important ones, Two sets of coefficients wpre derived
in nalyeing the results, in order to study them from diffe-
rent viewpoints. By basing the coefficients on an area equal
to the tread width in contact with the ground xtbe diameter,
we were able to show that,for a given tread vidth,the sharper
edged wheels were better as regards drag and, except for the
 67

B wheels, as regards lift also. The same set Of coefficients

showed no obvious increasing trend with increase in width,but
coefficients based on the projected frontal area of the wheels,
as is more usual, did show an increase with width in both lift
and drag. In all cases the lift force was directed upwards
contrary to the results of Morelli (1969).

Rotation of the wheel was found to have a beneficial

effect,in the sense that both the drag and the lift were de-
crBased. The pressure distributions showed that this was due
to a reduction in base presure,combined with a forward move-
ment of the position of' separation from the top of the wheel.
This movement of separationi and also the high pressure found
under the wheel are connected with the movement of' tile boun-
daries of the wheel and floor and would only be produced in a
complete simulation of the real flow. Because of their effect
on the overall forces acting on the vheel,these features are

examine4 in greater detail in the next chapter.

A smooth surfaced, non deformable wheel was chosen

for these tests,in order to present a basic result upon which
the effects of any other factors could be added. •T1e smooth
surfaced wheel has direct application to one practcal case
where the aerodynamics are of great importance, that of Grand
Prix cars which run with 'ead]ass tyres, but it was felt that
some attempt should be made to study the effects of tread
pattern and this was done by simulating the pattern with slots
in the wheel surface, It was found that the slots decreased
 68

the lift on the wheel,but had little effect on the drag. The
results also indicate that it is not possible to accurately
simulate the effects of tyre pattern by leaving a small gap
under the wheel,since the flow near the ground,produced by
the raised part of the surface,is still important.

No tests were made on the effects of deformation of

the wheel under an axle load,but it seems likely that the same
flow features will still occur. However, there will probably
be a small reduction in the value of the lift and drag coeffi-
cients,because the effective height of the wheel above the
floor will be decreased and also because there will be a finite
air
area in contact with the grôund,on which noApressures act.

An attempt was made to obtain some two dimensional

reeults,but,because of the low aspect ratio and high blockage
of the cylinder used,these are probably not very accurate.
The two dimensional flow may well, repay further study using
a better experimental arrangement,since some of the more in-
form of
teresting features of the flow, such as the specialAsepara_
tion, should be easier to study in two dimensions.
 69

(k) DISCUSSION OF FLOW FEATURES ASSOCIATED WITH THE WHEEL

The results of the last chapter have established

that the major factors in determining the wheel flow are the
rotation of the wheel and the presence of the ground. In this
chapter, it is intended to give some consideration to the spe-
cial features produced by these agents. Inevitably, a great
many simplifying assumptions have to be made in applying any
theory to such a complex flow and we do not claim that any
of the theoretical models cam acèurately . predict the real flow
about the wheel, Rather, the theory is used solely as an aid
in understanding the experimental work and in this sense bad
agreement with experiment can be as illuminating as good
agreement.

We 8hall first consider those features associated

with the rotation only, especially the early separation from
the top of the wheel. Then, the effects of the ground are
discussed, including those features which are due to a cfm-
bination of the rotation and the ground, Finally, a free atreaxz-
line theory for a cylinder on the ground is presented, which

takes into account the effects of separation on the pressure

distribution over the body.
 70

(14.1) Features Associated with Rotation

(14,1.1) The Boundary layer- on the wheel near separation

Because the early separation from the rotating wheel

has such an important effect on the lift,it was decided to
investigate the boundary layer rear the 8eparation position.
Unfortunately, the present experimental set-up does not lend
itself to this type of investigation. The main difficulty is
that the wheel is subject to a certain amount of vibration,
which appears to be initiated by the joint in the endless
belt of the floor passing between the wheel and the roller.
This occurs every 31/2 revolutions of the wheel. The thickness
of the belt at the joint is greater by about .1 mm and the
amplitude of vibration of the wheel is somewhat more than this,
whilst, at the Reynolds number of the experiment, the boundary
layer thickness, ö•9 on the front face of the wheel,is about

3 mm. This vibration introduces an amount of uncertainty into

the measurement of distance from the wheel surface and also
means that the very lowest portion of the boundary layer cannot
be observed, Since our main interest was in seeing whether
the velocity profile of the layer, near separation on the
centreline, resembled that conjectured for the two dimensional

flow in the introduction, it was felt sufficient to use pitot

and static tubes to measure this,rather than hot wires,vhich
would be easily damaged by contact with the wheel. A flattenod
 71

pitottube was used, Fig. (2-Th, because of the thinness of

the boundary layer. A correction was applied,to allow for the
displacement effect on the measurements in the shear layer ,
of.2 times the external normal dimension of the tube (i.e.
.lmzn). The above form of the correction really applies to
round tubes only,but it was felt that,because of the uncer-
tainty in norma]. distance measurement due to the vibration,
there was little point in trying to obtain a more accurate
correction. Measurements in the free stream indicated that
—Static
the Pitob,1 tube read high at low speeds, although no more than
5%, and another correction was applied for this.

The results for the velocity profiles of the layer

on the centreline of the wheel A 1 are presented in Fig. (11).
The important point is the overall shape of the profiles, ra-
ther than the actual values of the velocity, which may be
in error for the reasons given above. It i Unfortunate that
the reversed part of the layer on the rotating wheel is so
thii that we were unable to make measurements of teyarsed
velQcity except at = 27O at which po8ition the layer has
definitely separated. Note the inflexion in the profile at
this position, whereas at t?e other positions the shape is
basically similar, although yith some alteration and a definite
thickning at 6 = 2800 • The rotating layer, as a whole, is
thicker than the stationary one and the thickness increases
with movement downstream. The stationary layer, on the other
hand, remains much the same throughout the region measured,
confirming that separation does not occur on the stationary
 72

wheel until later.

Because the pressure distribution around the centre-

-line of the rotating wheel resembled that about a sphere with
laminar separation, as has been noted earlier, the question
arose of whether the wheel layer was laminar or turbulent.
All that was required was some indication of the relative
turbulent intensities in different parts of the layer. Thus,
comparative measurements of pressure fluctuation,such as those
given by a stethoscope,would prove sufficient. To obtain per-
manent recordings of the results, the BE. K 1/4i1 microphone,
used previously in the wheel, was now connected via flexible
tubing to a probe, consisting simply of a straight metal tube
1/16 11 inside diameter, which could be placed in the boundary
layer. Photographs were then taken of the signal from the
microphone displayed on an oscilloscope screen.

The results presthited :in Fig. (4-2) therefore are

representative of pressure fluctuations in the boundary layer.
Note that the first photograph does not indicate a high tur-
bulence level in the free stream, merely that the measuring
apparatus is somewhat 'noisy'. Near the front stagnation posi-
tion,there is no great increase above the free stream fluctuat-
ion, suggesting a laminar region. This region extends only
about 100 above the stagnation point and about 20° below; down-.
stream of these positions,a small higher frequency pressure
fluctuation becomes noticeatle, indicating a turbulent layer.
The amplitude of these fluctuations grows gradually over the
 73

top of the wheel, until at about 6 = 2800 a larger lower

frequency oscillation is also picked up. This is probably
associated with the pressure fluctuations in the wake and
thus 8uggests that the layer has separated at about this
position. Investigations on the stationary wheel indicate a
laminar layer until about 100 behind the top, then a short
separation bubble ,with turbulent transition in the shear
layer,which reattaches about 20-25° behind the top. The tran-
sition to turbulence in this case is quite abrupt and at re-
attachement the pressure fluctuations are as large as those
of the rotating layer just before separation. On the rotating
wheel no such abrupt change takes place, the pressui'e fluctua-
tions just increase slowly with mpvement downstream.

Although the wheel apparatus is not well suited for

investigations of the boundary layer, we feel that the above
resi4ts do indicate that the boundary layer near the centre-
line of the rotating wheel does have the form at separation
that was conjectured for the two dimensional flow past a ro-
tating circular cylinder. It wouLd appear to agree in basic
form with the work of Thom (Fig.(1-11)) and suggest that near
the centreline two dimensional ideas might be applied.(Note
that by two dimensional here we mean the rotating cylinder
in free air,as the boundary layer for the cylinder pn the
ground is more complex and less representative of the flow
near the centreline of the wheel). Moreover,although the layer
separates far forward on the whepl, it would appear to be
 74

turbulent before separation. This agrees with the work of

Griffiths and Ma (1969) on rotating cylinders and that of
MaccoIl (1928) and Davies (1949) on rotating spheres, all of
which indicated that the upstream moving surfaces provoked
early transition to turbulence within the boundary layer.
This early transition implies that there is likely to be
little change at higher Reynolds numbers than that of the
present experiment, since the transition is already so far
forward d,hence,the experiment will be representative of
wheel flow over a wide range of car speeds.

As discussed in the introduction,it appears that

separation from a rotating body is of unique forni, actually
being separation from a thin layer of air which remains attathed
to the surface and moves in the same direction as he surface.
This form of separation does not imply that the sheer stress
at the wall becomes zero at separation, but does appear to
be more easily provoked than ordinary separation, w4th hrdly
any adverse pressure gradient before separation. In the next
section we shall discuss some theories connected with the
special features of this type of separation.

(4.1.2) The two-dimensional boundary layer on a rotating

cylinder

Although, as discussed in the introduction,separa-

tion in three dimensions is in general of a different form to

75

that in two, near the critical position (see Fig. (1-13)), the
flow does show some two dimensional behaviour and our results
indicate that this also applies in the rotating case. Also,
the movement of the separation points with rotation seems
to occur whether the boundary layer is laminar or turbulent.
It would therefore seem reasonable to use some of the two-
dimensional laminar boundary layer theory available,as a
starting point in examining the boundary layer flow on the
wheel, in particular near separation.

Glauert, M.B.(1957),has given a general boundary

layer theory which has applications to rotating cylinders.
The theory is based on the assertion that if (u 1 ,v 1 , w1 , p1,
T 1 ), which are known functions of (x ,y, z, t),are a solution
of the general boundary layer equations, with given boundary
conditions, and there exists a second set of expressions
(u2 , v2 , w2 , p 2 , T 2 ),such that

u2 (x, y, z, t) - u 1 (x, y, z , t)

v2 (x, y, z, t) = v 1 (x, y, z , t)

y, z, t) = p1 (x, y, z +, t) (i)

T2 (x, y, z, t) = T 1 (x, y, z +, t)

w2 (x, y, z, t) aç Uia vi8

- at - F
x,y,z+, t
 76

(where= (x, y, t)
and h 1 dx, h2 dy are elements of length in the coordinate di-
rections)

thai the quantities with subs cr1 pt 2 al 80 Sat a fy the boundary layer
equations and so give a second solution,which will Satisfy
different conditions at z = 0 to the first solution. Glauert
proves the theorem by direct substitution of quantities (2)
into the boundary layer equations, proving that they satisfy
the equations, provided does not change so rapidly that addi-
tionai. curvature terms must be introduced.

In two dimensional, steady, incompres8ible flow

equations (i) reduce to

u2 (x, z) u1(x, z
(z)
v2 (x, z) = w1(x,z. ) - u 1 (x, z + ) d

where = (x)
or, in terms of the stream function ()

2' z) ((x, z + )

using Taylor's expression on the above gives

a
r3i 1
( )2 (x, z) = 'p1 +s— + + ••••.•
-:•
(3)
+ ul + 2 8u1 + . . . . . .
+
 77

= au
2
U2 U1
a2 +
(3)
aU 2
2
3U1 r3u1
= •- +'• a 2 + •.••....••... S..••

At z O,8olution isatisfies Lj). u1 0,

3u 1 -Cl
= (4)

where I is the skin friction

and

a2u
I_1J
I
12 (5)
az2 1z=0 L dx dx

from the boundary layer equations, where U(x) is the velocity

outside the boundary layer.

Glauert now assumes that (/)2<<1,i.e. the distance

is small compared to the boundary layer thickness 5. Then
at z = 0

1
= 0, U2 = .L' 12 - T. _pu()

from (3), (4), (5).

It is because u 2 is not nece8sarily zero at the

wall that the theorem can be applied to a rotating cylinder.
Thus,if the cylinder has a small tangential velocity, at
 78

the surface and solution 1 gives the flow with zero velocity
at the surface, choosing

= (6)

gives the required solution 2,with

T 2 = 1 - 1.1w (7)
dx

Note that the external velocity U,used for solution

1,must be that for the rotating cylinder. Alao that (/5) small
impliesuw/U small, so the solution will only apply to slowly
rotating cylinders, but,since we are basicalr interested in
the qualitative effect of rotation on the layer,this is not a
great disadvantage,sirtce our results on wind speed variation
suggest that effects which are connected with the rotation,
such as the forward movement of separation, are amplified
by an increase in the rotational speed,without undergoing
any basic change in character.

When 1.1w is negative, i.e. when the surface i8 moving

in the opposite direction to the external stream, the second
8olution will not be defined forzO,but Glauert over-
comes this difficulty by postulating an analytic continuation
of the original solution across z = 0 for the small distance

We are most interested in the solution near the

Beparation points,where,for the non-rotating cylinder,the shear
 79

stress becomes small. Thusbecomes large and the 8olUtiOfl

ceases to be valid. However if the terms in are retained
in (3), the boundary condition on u2 is satisfied by

- 1/2.ç 2 pu() (8)

Glauert now ubstjtutes the value of satisfying

(8) into (7) and so derives

1/2
2 =(112 - 2UpU() ()

which is equivalent to (7) when is not small,but does

not become large near separation.

Using (9),Glauertcaisiders the values of 12 at the

two separation points. If TJ is positive, so that the surface
i8 moving in the same direction as the external stream, then
12 has a finite value when = O,since dU is negative here.

The validity of the solution ceases at this point, so no further

deductions can be made, although Glauert supposes that 12 would
fall steadily to zero as x increases. When Uw is negative,
12 = Owhen

= 2UpU() (io)

i.e. when t.> 0, and the solution cannot be extended further,

because there is no point in solution 1,or its continuation
 80

below z = O,which could be taken as the new position of the

surface, Glauert thus assumes that remains zero beyond
this point.

It would be natural to assume that what Glauert shows

above is that separation moves forward on the upstream moving
surface, especially as this would agree with the experimental
results given previously. However, this is not true in the
sense in which we have been using separation, since a value
of the shear stress 12 equal toOred not produce any apparent
effect on the pressure distribution, as is true for a rotating
body. This is demonstrated by the work of Thom (1926 and 1931),
already discussed in the introduction (see Fig. (i-u)), al-
though his results are for a faster rotation rate than the
theory. Glauert, himself, is careful not to equate the shear
stress being zero with separation and this independence of the
two can be shown by drawing a further conclusion, which
Glauert fails to make. This is that,with 1 1 positive,U posi-
tive, and (dU/dx) positive, t 2 can still be zero from (9).
Thus 12 can be zero in a favourable pressure gradient and,in
fact,would be zero in the same regions as that indicated by
the results of Thom mentioned above.

On the surface moving against the external stream,

our picture of separation would suggest that 12 does not be-
come zero at all,whereas the theory suggests that it might.
However,at the lower rotational speeds to which the theory
is applicable,it may be that thespeed of the recirculating
 81

fluid in the separated region will be as great as that of the

rotating surface 1 in which case will become zero. Also note
that both the results of Thom and the present ones indicate
that at the higher rotational speeds (ciU/clx) hardly becomes
negative at all,so that T2 is unlikely to become zero by (10).

To summarise, the theory of' Crlauert is applied toa

cylinder rotating such that U/UZ1,but does agree with the
independence of the pressure distribution from the wall shear
atress,deduced from Thoms results for a higher rotational
velocity. This independence does mean that one cannot make
the assumptions, which Glauert does make,that will fall to
zero on the TJpositive surface,or that,ift 2 once becomes
zero ontheTJ-negive surface,it will remain zero,

We can take this examination a stage further,by

considering the flow near separation in a manner suggested by
Mime- Thompson (1968). for a stationary surface.
If

=c2(X,0) = vorticity at the wall,z = 0,

then

C-.)
J '.w - w

Therefore near to the wall, we assume

U
 82

From the equation of continuity

au
—=—-=ee -' z
az ax

where the dash refers to differentiation with respect to x.

Thus

= 1/2cc z2

since w = 0 when z = 0

=-1/2 (?/Q)z (i)

which is small for z small,provided (-) = 0(1),

but irQ = 0 it becomes infinite and we shall take
this to be an indication of separation. Thus,separation occurs
when au/az 0.

Now, if we try and apply the same analysis to a

moving boundary, with wall velocity U, then we must assume

= +

hence

aw = - au
OX = QWz

w = Qz2 as before, since w = 0 when z= 0

 83

w 1/2Z
so - = (2) which remains small for z small
U
1j
even if 0

Thus,again demonstrating that au/azL can be zero

on a moving boundary,without indicating any equivalent sepa-
ration to that shown above.

However, if instead,we consider the point in the

boundary layer above the moving wall,where the tangential
velocity u is zero, at height z 0 say. Then if = vorticity
at this point = assuming 'small,since it is of 0(5)

in the boundary layer,

we can let

u=-Q 0 (z-z 0 ) nearz

0

= - say.

Then from the equation of continuity

z
v = dz = Q' (z-z0 )dz = c? 0 (z 2 /2 - zz0)
0 j - 0

1/2. T' (z2 -2zz+z) - 1/2 Qz

002 1/2

22
z-h)
w 1,Q (0
50 -= (3)
U 2 °c? 0
h

This will, of courses tend to infinity as h tends

 84

to zero,since u has been specified to be zero at this point

whilst w need not be, but it will also tend to infinity for
0 and h 0.

This suggests that a new condition for separation,

which could be applied above moving boundaries,is that a/a
should become zero at the value of z at which u is zero. In
fact, this could be taken as a more general definition of
separation, since the stationary case is included,and would
be compatible with the idea of a free separation point above
the solid surface.

(11.1.3) The movement of separation position

The above theories, although indicating why the so-

paration positions might be different on a rotating cylinder,
do not indicate what sort of changes would occur. The theore-
tical determination of the separation position is obviously
a complex issue, involving a continuous interaction between
the external stream, the wake and the boundary layer and,in
fact,the problem has not been entirely solved for even a sta-
tionary cylinder. All that we can do here is to offer some
suggestions for the movement of separation, concentrating on
the case of most interest at present, that when the surface
is moving opposite to the external stream.

On a rotating body,the flow near separation is corn-

plicated by the presence of a reversed sublayer of fluid next
 85

to the body'8 surface. In our time mean picture of the flow

past a rotating cylinder (Fig. (1-12)), this sub-layer is
completely separate from the recirculating fluid in the wake
region and also separate from the external stream. The sepa-
ration point marks the limit of the upstream movement of air
in the recirculating region behind the cylinder. In a separa-
tion from astationarybody, the adverse pressure gradient has
first to decelerate the fluid near the wall against the influence
of the faster moving fluid in the outer part of the boundary
layer, but in the present case,which already has a reversed
sublayer,this should. not be necessary and it would seem likely
that even a slight adverse pressure gradient would produce
an upstream flow of the fluid in the separated region, espe-
cially as any greater circulation induced into the wake region
by the rotation of the cylinder would aid in this. It is not
surprising, therefore, that the experimental pressure distri-
butions obtained in the pre'sent experiment,and by Thom,show
little or no adverse pressure before separation.

Another physical effect, which is ignored by boun-

dary layer theory but which may be of importance near separa-
tion,has been discussed by Kuchernann (1967). This is that the
vorticity in the shear layer,when it is turned through some
non-zero angle, such as at separation, has an induced velocity
field with a component parallel to the body's surface which
acts to retard the flow before separation. If this effect is
strong enough,therefore,it may provoke separation further up-
stream. Kuchemann uses a simple model of flow with constant
vorticity to examine the above effect in separation from the
 86

trailing edge of an aerofoil,but the same mode], can be thought

of as a general separation from a plane surface with the flow
in the separated region ignored, as shown in Fig. (Li_3(ii)).

Far away from the separation point,at some arbitrary

distance 1, the constant vorticity model produces a straight
velocity profile with a non zero velocity, U0 , at the wall
(Fig. (4-3(i))), which may be regarded as an approximation to
a turbulent boundary layer profile without the laminar sublayer.
In this model,the condition for separation is taken as being
when the velocity at the wall falls to zero. This will always
occur at r = 0, the original separation point, but can occur
upstream of this point,to form a separation bubble,as shown
in Fig. (4-3 (iii)) and this is interpreted as a forward move-
ment of separation. The necessary condition for the bubble is
found to be

U0 <)ltan6 0

where w is the vorticity and 6 is the angle

the flow is turned through at separation

U0 is simply the difference between the true wall

velocity and the constant vorticity model velocity at the
wall. If we assign a velocity to the solid boundary and hold
constant,then we shall have a greater value of vorticity,
w, if the boundary moves against the external stream and less
if it moves with it. An increase in the value of w acts to
increase the chance of the above condition being fulfilled.
Thus,if we think of the solution with no enclosed bubble as
 87

being the conditions at separation on the stationary surface,

movement of the surface agninst the flow will tend to move

the separation position forward, which agrees with the experi-
mental results, 't is not suggested here that the vorticity
is necessarily a major agent in the movement of separation,
the special form of the separation when acted upon by the
changed external pressure distribution is probably more impor-
tant, but it is one possible cause,

(4.1. 11) Flow near the stagnation point on a moving surface

Another singular point in the boundary layer, which

is likely to have a different form when the surface is moving,
is the stagnation point.

The solution for the two dimensional flow near

stagnation with a moving surface of any velocity has been
given by Glauert, M.B. (1956),as a special case of an oscillat-
ing surface, The solution is derived from that when the stir-
face is stationary, given e.g. in Schlichting (1968) and
shown below.

The potential flow solution for flow against an in-

finite flat plate at z = 0 is given by

U = CX Si = -CZ (i)
 88

For a bluff body of any shape, the above is true in the neigh-
bourhood of the stagnation point, x = z = 0, (i.e. x z small
cf. the radius of curvature of body).To solve the flow with
viscosity included,we assume that in this neighbourhood

u = xf'(z) and w = -f(z) (2)

which satisfies the equation of continuity. Using

non-dimensional variables (2) becomes

u cxf'(c), w -(Vc) 1/2 r(ç) c 1/2

whereç= z(ç)

(V= kinematic viscosity)

If these values are substituted in the boundary

layer equation in two dimensions,with boundary conditions

u w = 0 at z = 0 u-U - cx as z - :

(or into the fullNaviex-Stokes,tvo dimensional equations)

then

f''' + ff" - f' 2 + 1 = 0 (3)

with boundary conditions (for a stationary surface)

f(0) = f'(0) = 0, f'(cD) = 1

(which is the Hiemenz equation)

With a boundary moving transverse to the flow with a velocity

 89

Uwwe look for solutions of the form

u = cxr'(c) + w = - (c\T)1'2 f(s) (")

where f andare as before and the continuity equation is still

satisfied.

The new boundary conditions are

u w = 0 at = 0; u-)'U = cx as-'Pcx

Substituting in the boundary layer equation now gives

c2x (fIt + ff " - ' + + u( 0 ,, + fØ' - f 'Ø) 0 ()

the expression in the first bracket is equal to zero by

equation (3) so

0" + fØ' - f'Ø = 0 (6)

with boundary conditions

0(o) =1 Ø(oc =0 (7)

If we differentiate equation (3) we have

f(IV) + fr" - f'f"= 0 (8)

therefore 0 = f" is a solution of (6)

Numerical integration of (3) Yields r"(o) = 1.2326

and f"(cD) = 0. Hence
 90

0 = t112326 = .8113f" satisfies the boundary conditions (8)

Thus u = cxf'(c) + .81i3U f"(c)

(9)
w _(ct)1'2 r(ç)

is the solution derived by Glauert

If '4) is the stream function , such that

and -=w

(io) (ii)

then

(io) implies '4) = ( cV)h/2 () 1/2f I

x±' + .8113 Uwc + g(x)

and

(ii) implies '4) = ( cV) 1'2 xf + ii (z)

choosing 0 vhenc= 0 (i.e. on the boundary ) gives

q)=u (V)1/2(Cx f + .8113f' )

we

In fig. (4-1 ),we have plotted xf + .8113f' = cons-

tant lines,iri order to give an idea of the form of the flow
given by the above. The streamlines shown,suggeet that the

flow is divided into two parts by '4)= O,although the two

halves of'4)= 0 do not meet, Streamlines to the left of(l)= 0
 91

in Fig. (L_Zi) go to the left,while those to the right eventuallr

go to the right. No streamline actually connects with the sur-
face and there is no singular point in the stagnation region,
at which a streamline divides. These features of the flow have
been used in proposing the picture of the two dimensional flow
past a rotating circular cylinder, given in Fig. (1-12) and
discussed in the introduction.

(k.i.) Conclusion

In the present section we have discussed separation

from a rotating body, in particular from the upstream moving
surface and have pointed out some of the unique features
associated with flow about such a body. Simplified mathematical
models have been used to demonstrate some of the features
although no attempt has been made to obtain any quantitative
theoretical results because of the complexity of the flow.
These models have shown the independence of the wall shear
stress and separation. A new criterion has been suggested,
namely that separation should be deemed to have occuredat
the position where both u and au/az are zero. The causes of
separation movement were discussed and one possible agent indi-
cated, although other effects will probably be as important.

The experimental results for the wheel indicate

that the special form of separation plays an important part
in determining the overall forces, especially in reducing the
 92

lift component, so that further study of the conditions near

separation would prove useful. The present experimental set-up,
which was designed, basically, for measuring the surface
pressure .distribution on the wheel,is not well suited to a
detailed study of the boundary layer,but such results as were
obtained do not disagree with the separation picture proposed,
at least near the centreline of the wheel. A study of the two
dimensional separation from a cylinder rotating in free air
would be a suitable starting point for any further investi-
gation, as this would eliminate the difficulties experienced
in taking measurements in the boundary layer and provide suit-
able results for a theoretical treatment. The same forward
movement of separation seems to occur over a wide range of
Reynold8 number provided the rotation rate is high enough to
overcome transition effects, so that lowornumbars might be
used initially to increase the boundary layer thickness and
thus make measurements easier.

(4,2) Features Associated With the Ground and Rotation

(4.2.1) Potential flow solutions for cylinder

In this section we shall give some potential flow

solutions for a cylinder above and on the ground, and with
rotation added. The results will then be compared with availa-
ble experimental data. Of course, a potential solution for
such a bluff body as a cylinder is unlikely to give realistic
results for the magnitude of forces but we might expect it to
 93

indicate correct trends with variation in height above the

ground.

The potential solution for flow past a cylinder

radius, a, in free air at z ic consists of a uniform stream
plus a doublet, thus

2
w = 0 + i(.)= Uz + z-ic

To model the ground we must include an image cylinder

at z = -ic i.e. an image doublet strength = Ua2 dt z = -ic.
But this in turn must have an image in the first cylinder to
keep the cylinder shape as a streamline i.e. a doublet of
strength 112 = a2/4c
2 at z = i(c-ci 2 ) where d 2 a /2c.
2 We
then require an image of this doublet in the image cylinder
and so on. Thus we generate an infinite series;

w=Uz+ Z (i)
zi(c-d )
n=1

where =Ua 2
2
n-i a (2)
- (2c-d n-i )2

. d2
=i
. . n n-i n-
jL

and = 0
(3)
d = a2
2C-dn-i
 94

The solution is given in the above form by Carpenter (1958).

From (1)

'In
- dw
dz- U - lv U -
___________ (4)
1 (zi(c_d))2

We are most interested in the pressures on the cylinder, and

if we take the radius a = 1,so that + (y-c) 2 = 1

v = 2Ux
1 (1+2yc2yo;2ydn_2cdn+dn2) 2

where X =
n
= X d2 and
n-in = 1 (6)

Terms

X(y+(c-d))
(1_2cd+d+2y(2c_d)) 2 + (1_2cd 1+d2
n+1 +2yd
n+1)2

can be combined,using (3) anc (6).as

d
(y-c) _1 + ..L)+ 2 n+1
(7)
2dn+ 1 2
(i
xn+i y)

A = i/ n then A n-i = d2xi An and A1 -

Define xi

(9) (io)
 9,

thus (7) becomes

(A n+1i-d2
n+1 A n+1 )+2d n+1 An+1
'1.y-c)
1+2dnnn
A y

Theuxzinbined term can be included in the summation in (5) in

the same form as the combined terms so that

(y-c)( 1 +d)A+2dA
v=2TJx Z
1 (1+2dnnn
A y )2

or V
u = 2xS (ii)

where S represents the summation.

The other velocity component, u, can be treated in

a similar way to yield

u/U = -2(y-c)S (12)

We can prove that S is a convergent series in the following

way:

=0 andifd_ 1 <1, thandIi = 1 <1

2 c-d n-i 2c-1

since c)1. Thus d<1, all n

Similarly,we can show by induction that 0 and d> d_1


96

for all n.

A180 -i (y-c)1
2d I
(1 +d) A [(y... c) + tfdJJ 2.A.3
so •<
(1+2d A y) 2 U fl
(2dAY) 2 2y2 A_1

Now

f(i/ia) 1 + 4 + 4d + ...(4...ci2)
n=1

(d,)' which is a convergent geometric

r=O
series since

IdI<1

Therefore Z (1/An ) is convergent and hence S is absolutely

convergent by comparison.

To take rotation into account, we must add vortices

inside the cylinders,so as to retain the cylinders as stream-
lines. Thus, if the vortices are added at z ib, bc, the
complex potential due to the vortices strength KU, -KU is

- z+ib'
w2 -
2Jt log' .1
(ia)

and
KU PA where PA, PB are defined in
Fig. (L_5)

thus,j constant,are streamlines and


97

PA (c+a-b b-.'(c-a) at C and D respectively

= (c+aj+b b+(c-a)

b2 (c+a)(c-a) or (c+1)(c-1) if a = 1

Thendw2
-= 2bKU (14)
dz z+b

and therefore on cylinder

v2 =- UxKb u 2 - bKU(c-y
y (15)
y

Adding these values to the previous solution for a stationary

cylinder, gives

_bK,/2) -2(y-c)S' say

U /U -2(yC)(S
(16)

v/ - 2x ( s_bK,2y) = 2xS'

Let
2 22
U +V -Z 2

then

= i-us'2 (i7)

and the forces on the cylinder can be found by integration.

To obtain the potential for a stationary cylinder

actually on the ground,we return to equation (i) and take
c=a,which gives
 98

2 + a 2/4 + ____
= iJ( zia
a z±iaf2 + ...... )

Combining terms gives

2
2az )
w=U(z+ z 2 22
n=1 a +n z

= aitU coth () 7W coth (--),for a=1

which is the same as the solution given by Milne-Thornpson,

(1968) p. 180.

Near the ground,the rotational part of the potential,

given by (13), as c-a or b0 becomes - by expanding
the logarithm in a Taylor series and ignoring terms o((b/z)3 ).
This term will give rise to a velocity which tends to infinity
at the origin, z=0. However, near the ground,ve might expect
viscous forces to modify the flow significantly, Indeed, the
low Reynolds number solution for the flow in the corner between
cylinder and grourxl, given later, BuggeSt8 that when the cylinder
is actually touching the ground, the moving boundaries give
rise to an upstream moving jet and the experimental results
indicate a c greater than 1 at the ground,which obviously
cannot be obtained in a potential solution. The jet might
be modelled by reversing the sense of the vortex doublet in
the solution with rotation,but, although this will give a flow
picture bearing a closer resemblance to the real one, it will
also give an infinite negative pressure on the cylinder near
the ground, Since all potential models are likely to be Un-
 99

realistic on a bluff body, it would seem that taking the sta-

tionary case for the cylinder on the ground will be sufficient
for the present purposes,i.e. the effect of touching the ground
is taken to be destroy the circulation about the cylinder.

The variation in the lift with height above the

ground is given in Fig. (k-5). This suggests that the cylinder
will experience a lift force towards the ground, which reache8
very high values near to the ground, and that rotation of the
cylinder exagerates this trend. Note that the values of rota-
tion used in the theory were chosen to give a free stream lift
force close to the experimental, values for rotating two dimensional
cylinders with v/TJ= 1 and for the MIRA wheel. Once the cylin-
der actually touches the ground, there is a complete reversal
of lift force and a large upward lift is generated.

The pressure distributions, given in Figs. (4-6)

and (47) for the stationary and rotating cylinders respecti-
vely, at different heights,shov how the downward lift force
is generated by the large suctions on the under-
surface of ie cylinder,caused by fluid rushing through tbe gap
between the [Link] also show how the increase in ne-
gative pressure is proportionally greater in the rotating
case. ihen the cylinder is in contact with the ground,the
theoretical pressures are of a completely different form (Fig.
(4-8)). The negative pressures on the undersurface have been
eliminated by the prevention of flow through the gap and have
been replaced by an extensive region of high pressure, which
together with the large suction over the top of the cylinder,
 100

produces a high upward lift.

Wadcock's (1970) mean pressure results are included

in Fig. (4-6) and show how different the experimental distri-
butions are, as well as indicating how the movement of the
two separation points produces a positive lift force. It
might have been thought that the theoretical modelvould most-
resemble Wadcock's two dimensional flow, but in fact the re-
suits of MIRA for the wheel (Figs. (1-16) and (1-17)) bear a
closer resemblance to the potential solution, probably because
of the elimination of wake interference effects. The conclusion
to be drawn is that we must be very careful about applying
potential solutions to bluff bodies near the ground,for they
may give completely wrong answers, even for the direction of
the forces.

Potential theory can be extended to three dimensions

for lifting bodies by using the Horseshoe Vortex model descri-
bed by Glauert, H. (1948). Here the body is replaced by a bound
vortex and two trailing vortices. Because the theory ignores
the body's shape,it obviously cannot take any separation effects
into account and so is unlikely to give very good results for

bluff bodies near to the ground. Thus,although this theory

has been applied to 'streamlined' bodies near the ground
(Stollery&Burns (1969)), with fair results at larger clear-
ances, it was thought unlikely to give very satisfactory re-
sults for a rounded bluff body such as the wheel, no more so
than in two [Link] the numerical potential methods
of the typodoscribed by 11088 &Smith (1968), are unlikely to be
 101

of great use for such a body and hence were not pursued. The
only way in which potential theory can be applied to bodies
with separation,with reasonable succoss,is in some form of
free streamline theory as discussed in section ( l4_3) possibly
extending to three dimensions by combining the numerical methods
with free streamline [Link] then those features of
the flow determined by viscous effects will have to be obtained
empirically or by a separate analysis.

(11,2.2) Viscous flow very near the ground

The flow in the neighbourhood of the line of contact

between wheel and floor will be dominated by the viscous forces
arising from the movement of the boundaries, We might there-
fore obtain some understanding of the behaviour in this region
by looking at a theoretical solution for the two dimensional
viscous flow,induced entirely by the movement of straight
boundaries meeting at a finite angle. (see Fig, (11_9)).

The Navier-Stokes equation is

+7(1/2v2 ) - = - !Vp -V•V (i )

where =V, is the vorticity (2)

and the equation of continuity is

(3)
 102

We can now merely assume that the inertial effects

are negligible,in which case the left hand side of (i) dis-.
appears, or we can make the approximation more forma]ly,by
introducing the Reynolds number Ul/v, and using non dimensional
variables defined thus

A
R(p-p)
t £t 2 = (4)
= PU 2

Then (1) becomes

A
R (s + ( i/2±2) - ) = -

where V is the gradient operator with respect to

4 A

S = Strouhal number = f. (r = frequency)

If we now let R->0 and assume that all the terms

in the bracket remain finite, (5) becomes

A A
Vp = VA

or dropping the "symbol

(6)
Vp =V,

and this even applies to the unsteady case provided the fre-
quency of variation is not great.

The equation of continuity i8 unaffected by the apro-

xiznation.
 103

Since our boundary conditions are expressed in v

alone, we will eliminate p from (6) by taking the curl of
both sides thus

VA ( V A W) = 0 (7)

Since V.w =V . (VA ) = o,(7) is equivalent to

= 0 (8)

In two dimensional flow, with polar coordinates

(r, 6), has only one- component in the direction normal to
the plane (r, 9), the z direction,say

1 8
c=—(rv) -;----

and defining a stream function such that

V = and v6= - (9)

wz=_V2(I)
82 1 1 82
)

80 from (8)

v 2 (v 2tp) = o (io)

We shall look for a separated solution of (io) in

the form

(= rf(0) (ii)
(where n arbitrary real or
complex number)
 10k

satisfying boundary conditions

= -U, = 0 , on 0 0
(12)

ar = 0 ,. on 0 = a

The condition = constant along the walls

is such as to suggest that 11 = 1,i.e.

(= rf(0) (13)

Then V2&j)= 1/r.(f+f)

and v2(v2) = l/r.(fT) + 2f" + f) = 0 (14)

Try f=euG +Oe'u9

Then (14) becomes

(p2+1)2ee0 + (p 2 +1) = 0

which is true if IL 80

f = AsinO + Bcos9 + Cesine + DOcosO (15)

Now let4)= Urf

80 that boundary conditions become

f=0, f'=-1 , on 0=0

(i6)

f=0, f'=-1 , on e=a

 1O

Applying these boundary conditions to (is) gives

1 simultaneous equations for the constant8, which can be 801-
ved to give

1-cosa sina
A a-sina , B = 0, C = cz-sincL' a-sina

so that,
r'• (-asin9 + (1-cosa).eBine + sinczoos e)
=

or
Ur
cL-sinct (Øsii(a9)(aQ)sinQ ) (17)

and it is immediately apparent that 4)= 0 when 8 = O,a and a/2.

From (17)

vr = a-sincL
U (sin(a-e)-ecos(cte) + sino-(a-9)coae)
(18)

v U
e cz-sina (esin(a-e)-(a-Q)sin9) (19)

The velocities given by (18) and (19) are independent

of r, so will be of the same form at any radius. (e.g. section
AB Fig. (4-9)). As an exaznple,they are plotted for a=30°, U=1
in Fig.(4-10) and the form of the flow field indicated. Note
that the two boundaries moving into the corner actually pro-
duce a jet of air moving outwards at a maximum velocity about
U/2.
The shear stress on Q=O is given by
 106

a 9
1(r - (-) +
LU
= (+c")

i.e. X hr
and the pressure from (6) as

3 1( 3 2Vr __
Vr - 2 )
= a2 -

= . (f lu + ft)
r

i.e. p - p hz'

Thus as r-0,the solution implies infinite values.

In practico,the boundaries do not make contact at a non-zero
angle .and there is never perfect contact between two bodies,
80 that the values remain finite.
Now

v ( 2sin(/2) - cLcos(ft/2) )
x'I (a-sinx)

and for small a,ve can substitute for sines and cosines from
from the appropriate Taylor expansions,with terms 0(a 4 ) ne-
glec ted.

- Thus
U[2(a/2_ (a/2)3)a(1(cz/2)2)]
Vr
e=a/2
(a—a+cr/3Z)
 107

= U/2 (to o(-)) (20)

50 that the jet emerges with half the velocities of the boun-
daries.

The motion is reversible,since the governing equations

are linear and homogeneous, so that the flow in the rear corner
between wheel and floor can also be modelled and this time
there would be an inward flow in the central region between
the boundaries (Fig. (11_9)).

Now, the assumptions regarding the dominance of the

viscous terms in the Navier-Stokes equation will only be true
in a very small region near the corner. However, we feel that
this form of flow does occur under the wheel and gives rise
to the high pressures measured there; not the infinite pressu-
res of the theory, probably because the curvature of the wheel
enables the actual contact with the floor to be tangential.
Noreover,because the central jet has a significant velocity
(o(u/2)) in a region of little external flow, the effects of
these viscous actions are noticeable over a much greater region
than that for which the assumptions are valid. Thus, in the
photographs of the two dimensional flow (Figs.(3_11O) and (3-111))
the effects of the forward moving jet can clearly be seen, al-
though in the three dimensional flow most of the jetting is
lost around the side of the wheel. Conversely, the flow in the
corner is relatively little affected by the external flow,
because of the localised nature of the actions producing it.
 108

Fig. (Z_i1) compares the pressures near the ground, with a

free stream velocity equal to that of the boundaries and with
no external velocity.

The solution in the rear corner indicates a flow

,with velocity of 0(U/2) into the corner,of fluid in the se-
parated region,which would appear likely to have an important
effect on this region and thus in determining the base pressu-
re. Our results do not show a large negative pressure here,
which might be expected from the theory, and this cannot be
entirely due to the response of the [Link],since the low
pressures in the grooves of wheel C2G were followed quite
easily. It would seem likely that the moving floor would lift
up to touch the wheelif a large negative pressure were applied
above it and this probably accouni for the oscillation of the
pressure in this region, which was discussed in section (3.1).
There may, therefore, be an error in the zesults from the ex-
perimental modelling with a moving floor, although a real do-
formable tyro might undergo a similar reaction in this rear
corner region; the tyre moving to meet the floor instead of
vice versa, In that case,the basic feature of a large pressure
near the ground would still occur. One could never model this
region very close to floor entirely,anyway, since on this scale
the different road surfaces encountered by the wheel will have
a significant effect.

This viscous flow,produced in a very localized region

under the wheelcan be significant in determining the overall
 109

flow field around the wheel and will be important in studying

such features as spray dispersal from a wet road. It is there-
fore necessary to duplicate these actions in the wind tunnel1
as far as possible, by having the wheel in contact with a
moving floor, if realistic results are to be obtained.

(k.3) Free Streamline Theory for Cylinder on Ground

(4.3.1) Free Streamline Theory

The basic suxnption underlying free streamline

theory is that the region of separated flow behind a bluff
body can be separated from the main stream by steady stream-
lines. These streamlines can be thought of as an idealized
form of the separated shear layers as the Reynolds
number tends to infinity; the shear layers become thinner
with rise in Reynolds number and hence could be considered
as lines of discontinuity, the free streamlines, on extra-
polating to infinity. In real flows,these shears layers soon
break down into the turbulent wake region, with entrainment
of fluid from the mainstream, violating the assumptions of
the theory. However, it might be expected that the free stream-
line concept will have some validity close to the body and,
in particular, in predicting the pressures on the wetted
surface of the body. (The surface before separation).

The theory cannot be used to predict base pressure,

or tie position of the separation point on curved bodies,since
 110

the8e properties depend on the mechanics of the wake and of

the attached boundary layer, both of which are ignored by the
theory. Consequently, the values of these items must be fed
into the theory from empirical data, or from separate mathe-
matical treatments of the boundary layer and/or wake, which
must then be matched with the theory.(for example Roshko(1955)
and Bluston & Paulson (1972)).

Having made the above assumption, one is faced with

the problem of finding the potential flow of the main stream,
given the direction of the velocity on the vetted surface and
its magnitude along the free streamline, the position of the
free streamline being unknown. This last point, the indetermi-
nancy of the free streamline position in the physical plane
(z plane), means that it is more convenient to work with the
complex potential,w = 0 + i4J,as the independent variable,
since streamlines are then represented by = constant in
the v-plane and, in particular, the streamline which stagna-
tes on the body, forms the wetted surface and then separates
to become the unknown free streamlines, can be simply represen-
ted asll)= 0. Because of the form of the mixed boundary condi-
tions described above,the most useful function of w to solve
for is

a= log0 (1/) log(UJqe8)

= log(UJq) + ie =T+ iO, say (i)

since G will be given on the vetted surface and I will be given

 111

on the free streamlinea,and this separation of the boundary

conditions enables the problem to be solved. Once 0= 0(w)

has been found for a given flow,other properties can be deri-
ved inunediately, from

1
z = -
J'edw (2)

II 2
q2= IdzI U2 e-2t (3)

and then from Bernoulli's equation

2t)
p = PØ; + 1/[Link]( 1-a (4)

The initial form of the theory, derived by Helm-

holtz and Kirchoff, applies only to rectilinear boundaries
with the velocity on the free streamlines taken as constant
and equal to the free stream velocity,tJ. The boundaries will
then form a polygon in the o plane, since 0 is constant on the
wetted surface, except for discontinuities, and I is constant
on the free streamlines (Fig.(4-12)). The boundaries can then
be transformed to the real axis of an intermediate 1-plane
by a Schwarz-Christoffel transformation. The boundaries in
the w-plane are of a similar form and can be transformed to.
the real axis of another intermediate plane, 2• The relation
between the and 2 planes can then be uniquely determined
from three pairs of corresponding points on the real axis,
usually the stagnation point and the two separation points.
In theory, 1 and 2 can be eliminated ando=0(w) foundbut in
practice one of the is usually retained as a parameter.
 112

The further development of the theory has been cqn-

cerned with two points; the extension to curved boundaries
and allowing non constant velocities along the free stream-
lines, either of which means that the simple form of the boun-
daries in the c--plane will be lost and,in general,the
Schwarz- Christofel]. transform can no longer be used, (Fig.
(.i4)).

Levi-Civita (1907) considered the case of flow past

a general bluff body. The transform from the v-plane to the
intermediate 2-plane is

= Lia( )2
W - 8ifl(/2) (5)

where a = (Øi/2 Øi/2?/6

(Ø1/2Ø1/2)
sin(>'2) = (Øi/2Ø1/2)

2 are the values of 0 at the separation points

and and
(The stagnation point is taken as 0 = 0 see Fig. (Z1_13)).

The 1'2 relation is defined as

= -½'( + hr ) (6)

and so

v = a( + Zsin(X/2) )2 (7)
+

(6) maps the upper half of the 2-plane onto the

 113

upper half of the unit circle in the 1 -p1ane,such that the

wetted surface lies on the semi-circle and the free stream-
lines lie on the real axis. It is now only necessary to spe-
cify the mapping from a-plane to the 1 -plane to complete the
solution. Levi-Civita assumes a mapping( 1 ), then uses (6),
(5) and (2) to obtain the corresponding boundary shape. His
theory is thus an inverse method only,since the boundary shape
cannot be specified in advance.

However, Brodetsky (1923) was able to obtain a nu-

mericai. solution for flow past circular and elliptic cylinders
by making use of Levi-Civita's method and the method was ex-
tended by Villat (192o),to form the basis of a direct method
of solution by an iterative process. Villat's method is in
principle similar to the method of Woods (1955), which is des-
cribed later, but suffers from the disadvantage of having to
specify a constant velocity along the free streamlines, (as
do all the methods mentioned so far).

The severity of this disadvanta can be appreciated

from a consideration of the base pressute acting on the sur-
face of a bluff body,down stream of separation. The base
pressure [Link] defined as pb_p/h/2pUQ and,in
real flows,.is always negative and virtually constant across
the base. A velocity of Uat separation implies, through
Bernoulli, that Cpb O,which will therefore lead to a much
lower value for the drag than the experimental value. To
obtain a negative base pressure,the velocity on the free
 ilk

streamlines at separation must be given as KU , where K> 1

and must be determined from empirical data,or from a considera-
tion of the wake flow. (The velocity must be the same on the
two streamlines because of the constant base pressure). Since
the velocity on the free streamlines should tend to U,far
downstream of the body, this implies that the velocity, and
hence pressure, must vary along the streamlines,

Another major defect of assuming constant velocity

is that,far downstream,the width of the wake tends to infinity.
Since,in the mathematical model,the fluid inside the wake is
assumed to have zero velocity, the wake width is equivalent to
the displacement thickness and,from a consideration of momentum
balance,the far downstream wake displacement thickness should
be finite and proportional to the drag of the body. Any thoory
which purports to give a realistic drag value ought, therefore,
to lead to a finite value of wake width at infinity.

Attempts to obtain a realistic base pressiire have

been made by Riabouchinsky(1920) and Gilbarg and Rock (1945),
who retained the constant velocity along the free streamline,
but with a closed wake cavity,and by Roshko (195k),who
allowed the velocity to vary along the free streamline.
Unfortunately, although Roshko's method is fairly straight-
forward for straight edged objects, he adapts Brodeteky's
numerical method for the circular cylinder,resulting in a
complex procedure which is unlikely to be applicable to a
general curved obstacle.
 115

For a general curved body the method of Woods (1955),

which is based on the work of Levi-Civita and Vi]lat, is much
to be preferred. Here the upper half 2-plane of Fig. (-i)
is mapped, not into the upper half circle of the 1 -plaxLe, but
into an infinite 8trip by

2=tanh1 (8)
, 1
(see Fig. (l1..15))

The boundaries now transform into = 0 for the wetted surface

and = 7t/2 for the fx1ee streamlines. Thus the function
C +i9 is such that Q is given on = 0 and t is given on
= it/2. Woods uses the properties of such functions to obtain
as

cT(5,) j ( 9 (b) cosech (5*_5_j ) +

t5) sech(5*_i) )do*

(9)
where 9(o) = 0(5,0)

'2' (o,/2)

Because the infinite range of integration makes

numerical treatment of the above integral difficult, it is

more convenient to transform to %=+ iY by

2 = tanh 1 - isini(X/ 2) (10)

(see Fig. (e-15))
 116

Here,the wetted surface isq 0, -,tYt and the

free streamlines are T(= it, 1<0 and= — it, y<O. Substituting
in ( 9 ) gives
It

a(X) = cosh(X12) 1 e0(Y') d(

^
211 —
[j•
-It
sin(i'2)+isinh(XI2)

0 0
-c,(if) dy t2(-c') dT
______ -i
1' I
cosh(/2) + j sinh(X/2) ^
(11)
-co cosh(Tc/2)H sinh(X/2)J
-'to
where suffixes 1 and 2 refer to the upper and lower streamlines
respectively, and integrating the 1st term by parts gives
It
Ln sin(
__________ —)' de0CY)
(x) i e(-it) - ^
1-c ' cos(YiX.) ''
•V-i-c
4
0
cosh(X/2) 1 1
t1(-) di + t2(i') [Link]*
1
2 It J
-co
LcoSh(i'2 )+ sinh(X/2) cosh(17)— L sinh(Xf2)j

(12)

w(%) is given from (5) and (10) by

w= L a(isiii (X/ 2 )+ sin(X/2)) 2(

13)

On the body where % = iY

deo ds 1 (ik)
dO o (y) ds dydYRq dydY

where s is the distance along the surface, R is

the radius of curvature of the surface and q0 is the velocity
on the surface.
 117

From (13)

ds
dl = = 4a(sin() - sin()) cos(') (i)

so on the body (iz) becomes

U
= lO(q(•))

it___
((%) - s i n(X/) ) cos() '•I
4a r 1 In 4 11I d'f
-
= E- J q(y) Jcos(i±Y)I
-it

sir(T')
0TLfl 4
- rc cos(Y)
4
(which is the contribution to the 1st integral in (12)
from the discontinuity of size eT, in Q0 at the stagna-
ion point y X.)

+ the terms in (12) involving t1(), 12(Y) and indepen-

dent of q0
(16)•

Now if some distribution of velocity on the free

streamlines is specified, so that 11 and 12 can be assumed
known, the above represents a complicated integral equation
or q0 . The velocity distribution chosen should be such as to
give a finite wake width far downstream, for reasons discussed
earlier, By expanding cr(X) in (i) at infinity, Woods shows
that the above condition requires 1 1, I to satisfy

118

O0 (Y) sin () dV - t2)cosh(1/2) dq = 0 (17)

Also,since the base pressure is constant, the pressure at the

two separation points must be the same,or

t 1 (r=o) = t 2 (i-. 0) = •C, say (18)

soc =1-e S

If 11 and 12 are required to have the correct form near infinity,

the wake width there must be set proportional to drag, but
since the main use of the theory is for calculating pressure
distributions on the body, the exact form at infinity is not
necessary and some simple mathematical form can be chosen to
satisfy the first two conditions only. Woods adopts

-Cs
12 (19)
= [Link](/2)
which satisfies (18) and (17) if

= j ii() dY (20)

If this value for 1 and 12 is inserted into (16)

the equation reduces to

jSIfl()
4
= __
4a _________ tn
cos()(sin(2)_5.(X/2))
c1;v)
-Ic L RcCY) J +
 119

x-v
_tn sin(---) ts
It cos(X'Y) I + [Link](72)
'4
(21)

We can attempt to solve for q0 (Y) in the above by

numerical iteration. A distribution q 0 (f) is assumed and (ii)
used to obtain a = s( y ). Then R = R( y ) can be found since
R = R(s) is known for any given body. These values of R(Y) and
and q0 ( y ) are then introduced into (21) and a new q 0 ('y) is
found,completing one iteration. There is no proof of the con-
vergence of this process and,although Woods found that for
circular cylinders the convergence was quite rapid, the present
author has found cases where there was no convergence. This
is discussed more fully in section (4.3.3.). A numerical
procedure for solving the equation in the above manner is given
by Woods in his paper (1955) and this method was followed in
devising a computer program to solve for the flow about a
cylinder on the ground (i.e. two cylinders touching). The
process is quite suitable for computerizing, so provided
convergence can be achieved,the lengthy and complex numerical
procedures required are not necessarily a problem. Much more
of a problem was the lack of convergence found for certain
cases.

Woods theory provides the formal solution to the

general problem of flow over curved surfaces, with 'estrictions
placed on symmetry and smoothness only. More recently, Parkinson
and Jandali (1970) have used a much simpler approach to the
 120

problem. They make use of the fact that the required wake
characteristics resemble those of a 'wake' produced by a source
in a uniform stream (The distribution chosen by Woods is also
similar to a wake source flow) and simulate the wake effect
by sources.

The flow model is set up in a c-plane (Fig.(Le-17)

and consists of the uniform flow past a doublet at the origin,
with the wake effect supplied by two surface sources, of
strength 2Q, symmetrically located at angles on the circuin-
ference of the circle. The complex potential for such a flow
is

w() = v(+ )(ln(-Re5)

it + 1n(-R 5 ) - 1n ) (22)

and the velocity comes from

v() ==v(i-)
1 + 1 1 ) (23)
it -Re iS -Re

The circle of radius R is a streamline of the flow

and has symmetrical stagnation points located at S 1 and S2 due
to the upstream flow of the two sources.

The aim now is to find a conformal mappin,z=f()(24)

,from the circle to a given body shape in the z-plane. This
mapping must preserve the direction of velocity at infinity
and have S 1 and S2 as critical points, at which f'() has
simple zeros. Hence angles of intersection of curves are doubled
 121

at S 1 and S 2 , when they are mapped into the z-plane. The

stagnation streamlines in the c-plane then become streamlines
tangentially leaving the body surface at S 1 and in the z-
plane and the circle actually maps onto the slit,S 1 AS2BS 1 , -
flow inside the free streamlines being ignored. If f() can be
found then velocity in the z-plane follows from

v(z) = dw
dii d v()
= (25)
=

and pressures would come from Bernoulli's equation .

The 8eparation points are given in the z-plane and

Q and ô must then be chosen to ensure that.v() = 0 at the
corresponding points in the -plane, thus providing one relation
between them. A second relation is obtained by equating the
velocity at separation to its given value, thus

v(l = K = (i-c )1/2 (26)

(Note that v(z) is finite at separation since both v() and

ft() are zero there).

The two relations enable Q and ö to be specified for

a given base pressure and separation point,and the problem
solved,provided f() can be found for the given body.

Parkinson and Jandali give mappings r() for a flat

plate, a wedge and a circular cylinder and also deal with
 122

elliptic cylinders by mapping from the circular cylinder

solution. The transform for the circular cylinder

z f() = - cota - 1
-cota (27)

where a is the angle of the stagnation points in the

c-plane and is related to the separation angle in the
z-plane by

a = (28)

Now f'() = 0 when

V. +
- i + cota, z = -+ 2i

which fixes the position of the singular points so that

R = coseca and h = 4cosec (29)

setting v() 0 at 8 = gives

(cosb-cosa) (30)
Q = 2itU sina

and substituting in (23) for Q, (25) becomes

I v(z)I = sine (1-2cosacose+cos2a)

U U cD
(31)
O n the wetted surface.
 123

If v(z) in the above is set equal to its desired

value, K at separation, 9 = a, then

sin3a
cos& = coscz + K (32)

which completes the solution.

Thus,the theory leads to a simple analytic form for

q0 on the wetted surface,compared with the complicated integral
equation of' Woods. When the two theories are compared with
the experimental values of pressure distribution for a circular
cylinder,the agreement of the Parkinson and Jandali method is
at least as good as that of the Woods method,despite the fact
that it is only a model of the flow and not a formal solution
of the problem. Thus,if one can find i() for the given body,
or transform from one of the known solutions for a circular
cylinder, it would be better to se the simpler method.

A free streamline solution,obtained by any of the

above methods,may be subject to the physical unrealities
shown in Fig. (li17), i.e. the free streamlines cross each
other producing a doubly covered region,or they intersect with
the downstream solid surface (although the former is not
applicable to the Parkinson and Jandali model).Woods (1955)
has examined both types of physical unreality •He considers
the first type by expanding o(%) on the free streamlines, far
downstream,to show that the distnce between the free stream-
lines remains positive and tends to a finite limit at infinity,
 1211

provided the condition given in equation (17) holds. The other

type of unreality is considered by examining the radius of
curvature near the separation point of the free streamline.
The condition that the streamline should have either finite
curvature or infinite curvature aiay from the body is shown by
Woods to imply

tan(i) dOjY) + J[coth(4) d 1 () - tanil4) 2(n)]
A - J

(33)

By then considering the pressure gradient just before separa-

tion , Woods shows that a value of A<O implies an infinite
adverse gradient, which in turn implies that the boundary
layer would have separated earlier. It therefore appears as
though A 0 is a necessary condition for physical reality
at separation. However, real flows are found for which A1&O,

suggesting that viscous boundary layer effects,neglected by

the theory,may be of importance in determining conditions
near separation. The conditions,given in equations (17) and
(33), must not therefore be thought to necessarily place
physical restrictions on real flows,but simply to e?camine
the reality of the free streamline solutions. Even than the
condition A(O cannot guarantee that the streamlines would
not intersect with the body further downstream,if the body's
curvature reduces downstream. It is therefore a necessary
condition,but not sufficient and must be checked
 12

by calculating the [Link] of the free streamlines.

When the Parkinson and Jandali model is applied to

a circular cylinder,the conditions on the separation point
reduce to a simple form, as they show by considering the
pressure gradient on the wetted surface near separation and
comparing with Wood's results. The condition is

+ve co K <3/2. sine1

ac
is finite if K= 3/2. sine1 (:34)
sepn.

-ye co K>3/2. sine1

corresponding to A<O. A = 0, A>[Link],for a given base

pressuro,the separation angle,9 11 is restricted to a range
of value, for which sinø?2K,3 and,if sine = 2K/ 3 ,tlie stream-
lines will separate with finite curvature.

(4.3.2) T ransform for cylinder on ground

The easiest method of obtaining a free streamline

solution for the flow past a cy)4nder on the ground seemed
to be a transform from the single cylinder to two cylinders
in contact, the second cylinder being the image of the first
in the ground [Link] found,such a transform can be used
with any of the methods discussed above that will give realistic
 126

solutions for a single cylinder, but,because of its simplicity,

the Parkinson and Jandali method was preferred. In addition,
Woods theory appeared to offer a method of tackling the problem
directly and it was decided to apply it to the two cylinder
case,both to compare with the results obtained by transforming
the cylinder solution and to see if his numerical routine
would converge to a solution for the bluffer obstacle presented
by two cylinders.

Considerable thought was given to the problem of

obtaining a suitable transform and, eventually, the solution
was suggested by the Icoaxalt coordinates of Milne-Thompson

(1968, p.176). If A and B are two points on the x axis at

(c, 0) and (-c, o) respectively, the coordinates of any point
P may be expressed in terms of (r 1 , (r2, °2•
Where (r 1 ,) and (r2,Q2) are the polar coordinates
of P,treatiig A and B as poles respectively (see Fig. ('i.-18)).
If we define 'coaxal' coordinates by

- B2 and •I= loge(r21)

then %=+ir is given by

ilog (!.±2\
'z-cI (i)
z= iccot(X/2)

The important point,from our point of view,is that all curves,

=constant,are circles passing through A and B,whose centres
are at (o, ccot),with radii of ccosec ; in particular =it/2
 127

is a semi-circle,radius c,centre at the origin, (Fig. (11-18).

Thus, if we transform a circular cylinder at the origin in
the z 1 -plane to the %-plane defined above, the wetted surface
transforms into two straight lines,which are parts of the
lines = Tt/2 and = -t/2, Fig. (4-18). A simple linear transform

= aX (2)

moves the lines to. lie on = ait/2, -alt/2, but these determine
two equal circular segments in the z 2-plane with the x-axis
as the line of symmetry (see Fig.(k-19)). If a tends to zero,
the shape in the z2 -plane tends to the desired two cylinders,
whilst a 4 2 produces a straight line on the x2 axis.

Thus,the required transform from the circular cylinder

solution of Parkinson and Janda].i consists of

= z-b (3)
z i +c
X =iiog( z1-c ) ( 11.)

= id cot (a%/2) (5)

which may be combined to yield

Z2 = d [kz_b_cna + (z_(b+c))al (6)

Lz_b_c - (Z_(b+C))aJ

although it is simpler in practice to use the separate trans-

forms. (3) transforms the cylinder, radius c, with centre at
z = b,into one at the origin. Thus,(28) and (29) of the pre-
viou8 section, determine c and b as
 128

c = = 2 cosec 2a (7)

and b = ccos8 1 = -ccos2cx = -2cot2cz (8)

(The separation points lie on the y axis in the z-plane)

The constant a in (5) must be chosen to give the

required shape in the z 2 -plane, in our case a-O, The constant
d is obtained from the requirement that the velocity should
tend to U as z-

We shall first deal with the general solution for any

a and then take the limiting case as a)'O. • In both cases,we
shall be aiming mainly to find the pressures acting over the
wetted surface.

The velocity in the z plane

v(z) = v()/f'() is known from the Parkinson and

Jandali solution for a cylinder, so

- !v(z)I (9)
Iv(z2)I dz1 dX
12 dzL
dz dz1

From (3)

VziI
iaI = 1 •(io)

129

From (4)

dX = .. 1 1
z1+c - z1-c

On the wetted surface z1 = ce101, so

dX - -2ic - -2ic - -2i

- z2 2
1 -c
- 2
c (e2ie1_1) - c(cos2e 1 -1+isin2e 1 )

80

dX r____________ 1 1/2
= csin91 (11)
aI1 = L20(
2 1-cos291) J

From (5)
dZ2 -
- - ida cosec 2 aX -ida
= ([Link]%)

k 2 I
da
1I [(1_cosa.cosha11 )2+(sinaLsinha )2J1/2

which reduces to

dz2 da (12)
= (cosha—cos(ait/2))
on the wetted surface, where from (4)

= rc/2
(13)
= -log(tan(Ot/2))
 130

Hence, on the wetted surface

ai
Iv(z 2 )I IV(Z)I.C. inej(cosha,i COS /2 )
(1li)
ad

At the front stagnation point in the z 2-plane, which

corresponds to , e 1 +n,we require Iv(z 2 )I -o.

Now

cosh ai = 1( 1 + (tan el/2)a)

2 (taneh/2)a

(1+cose1)^(l_cos1)a
1
- (89)a
(15)
L 2

and
lv(z)lfvsinQ1 from (31) of the last section,

so from (in)
2-a
( sine 1) a s

' S 1v(z2)1O provided a<2

Another requirement is that v(z 2 ) should tend to

U co as z -)'co
2
and this is best examined using % as the variable
then

dz 1 -1C dz -ida
d% = I-cos% and = (1-cosa%) (16)
 131

Thus
v(z ) c (1 -cosa%
Iv(z z )I = ad(1—cos%

= Iv(z)1 f(%)j , say

Now z 2 - is equivalent to XO and,using L' hospitals rule,

the 2nd derivative of f(%) gives ca2co saX
adcosXI =
ac
d

• • We choose d = ac to satisfy our requirement,since

tv(z)f • 1 as z-a
I i-
I QI

Hence, the cylinders in the z 2 plane are of radius

d ac with centres at ac cot i-)

a'rc/ 2
sin = sin alt/ 2

and as a-)O,the radius- . 2c/rC,with centre-.t 2,which gives

the two cylinders just touching.

The relation between the base pressures in the

and z 1 planes comes straight from (1 Li) as

K2 = (sinG1 (coshai - cos/2) ) (17)
8 e pn

where
K2 (1_c)h/2


132

It merely remains to find the relation between e1

and 9 2 ,so that at separation can be found for a given
separation angle, 025• From (13)

1 1
1z2+dl
=log(
t ane l/ 2
)=
lzz_d
80

1z2-dl 1/a
tan 1J2

which reduces to

tan i/ 11^5iflç_/2 1 (is)

= L1+sin z +a ) J

A more convenient relation for changing from to

02 is from ()

i_08a1, cosha
tan02 = (19)

which can be written as

tan02 =
sin alt/2((1 +co )a (i cose )a)

to show that at =

2 a = -cos
at/ 2
= 1 (20)
tane STAG -sin 2 2 a
.alt/ tan 2
 133

which fixes the value of a for a given body shape in the

plane.

To obtain the solution when a-O,it is simpler to

make an assumption of small a in transform (5), rather than
try to take the limits of the previou8 results. This then
becomes

= 2id = 2d (21)
2 aX. a ,z-b+o

Then
dz 2 -2id
=

so using (io) and (16)

1 caX.2
dz 1 .d%.dz2 = 2d(1-cos%)
dz dz 1 d%

z 2 -coas-O,so applying L'hospitals rule,we find that the

above expression -*? as %O and we therefore choose d = ac
as before to give !v(z2)I-TJ.

Then

z2 = 21c (22)

dz2I 2c 2c
an< (23)
= (2/k+q2

1311

on the wetted surface1

From (16)

(2k)
ldzii = C/h

Thus, on the wetted surface,

2 2
Iv(z)I (--+)
v(z2)I = .= 1v(z)1 f () (25)
2 coshr

or in terms of

= Iv ( z ) i 1(7t2/4 + (log(tan ))2)

(26)

= Iv(z)I f(81),say

To examine the behaviour of f at the stagnation

point write = - then

2 2
1 -O as
cosh _]_) 4. 1I
=

..1v(z2)t o asi--co

The relationship between the base pressure comes from (26) as

K
K1 = (27)

and the relationship between 01 and 2
 135

tan -it
= e= exp ( .
cos82 (28)
i+sin92

or
tan(e2/2)= (it/2)-'l - (/2)^1og(tan6h/2) (29)
- (lt/2)_log(tan0l/2)

To find the coordinates of the streamlines in the

z2 -plane we start with the streamline equation in the Parkinson
and Jandali c-plane.

Ir +
Im W) = -Q (30)
If we write = ji+i1 and substitute into the equation for
w() ( (j2) of section (4-3-i)) the above becomes

'F(ji 2 + .2 R 2 )
(22)
+
atan (_'F(2+'sf2-R2) =
(31)
(2+)(-2Rcos5)+J.LR2 U

which can be written in the form

- jitu (2f2) atan (32)

(
- (}L2+'f2_R2) (22)(2Ro)R2)

and g 2sin2a
K1 K 2Rcosô = 2cotcz1
+ = 2cotcx+
2sin2a ______

So once a and K 1 are known the above equation can be solved

by iteration to find f for a givenp..It then remains to trans...
form the coordinates into the z 2-plane. For the case when
a-'0 suitable transforms are (writing cotcz as Ct and tana as t)


136

z 1 = + iy1

= ([Link])
[
- i
(ct)2
21)
t
I (3)
1
y 1 =1 ( 1 + )
(-ct)2+%'2

x =+i.n

2cy1
= atan _____
x 1 -C - atan
1
atan( )
X1+C
(x1-o)(x1+c)+y2

(3')

r ( .—)+1
=
___

where c t/sin2cx

z2 = X 2 + iy2

X2 _______
it
= ( 22 ) ' X' (2.2) (35)

where r2= it
= radius of cylinders in z2-plane

The far downstream width of the wake can be obtained

in the Parkinson and Jaridali mode), from continuity considera-
tions as
 137

2Q=UH (36)

Substitute for Q from (4.3.1), equation (30)

H = (1-cose1)

Hence the wake width is finite downstream.

Hii: H _____ (1-cose15)

ç= ç = -- xl

where
h2 = total width of obstruction in z2-plane

2 x diameter of cylinder

h 1 = diameter of cylinder in z 1 plane

The condition,sin9 15 2,.K 1 ,for the free streamlines

not to intersect the cylinder in the z 1 -plane, also determines
the physical reality at separation in the z 2-plane. In terms
of z 2 variables,this is
cose2.
sech2( 2 1 + sinO
K2 3it
8(1+sin92s)

(l.3.3,) Comparison of results

Computer programs were written for the two transform

casea,of O<a<2 and a-4O,and also for Woods theory,using the
 138

numerical procedure given in his report (1955). It was found

that this procedure would not always converge to a solution
for the bluffer body shape presented by the two cylinders in
contact. This lack of convergence seemed to be connected with
the length of wetted surface, STOT; the rate of convergence
decreasing with increasing STOT until finally ceasing altogether
i.e. one iteration would yield a set of values for q 0 , which
when iterated returned to the previous set and so on, This
behaviour was not investigated very extensively, since Woods
theory was mainly being used as a comparison with the transform
theory and sufficient cases with convergence could be obtained
for this requirement.

Woods theory was therefore applied to some of those

shapes, with a between 0 and 1 for which the iterations con-
verged and compared with the transform results for the same
shape. It was felt that agreement between the two theories,
for these shapeswould be a good check on the correctness,or
otherwise,of the mathematics and programming involved, Some
of the results are shown in Fig. (k-zo). The fairly close
agreement for a = .333 suggests that both programs are working
as expected,although as a gets smaller the agreement between
the two decreases. This is probably due to the bluffer body
shapes emphasizing the differences between the two theories.
Our results indicate that,surprisingly, the Parkinson and
Jandali transformed model gives closer agreement with available
experimental data than does Woods theory,
 139

The results for the transform theory are first corn-

pared with the pressure distribution obtained by Wadcock (1970)
for the two dimensional flow around two cylinders in. contact,
since this flow is the closest experimental analogue to our
free streamline model, (Fig.(4-21)). The experimental pressure
coefficient at separation is -1.35 and the theoretical results
are plotted for two separation angles. The experimental results

appear to have separation at 9 = 280°, but a theoretical value

of 9 = 2900 appears to give the best overall fit. The
theoretical curves have an infinite adverse pressure gradient
at separation, since A in equation (11..3,.2.(33)) is less than
zero, and so cannot be expected to give good agreement with
real flows in this region; viscous effects not consjLdered in
the theory tend to smooth out the experimental [Link],
away from the separation region, the curve for = 290 0 is
almost coincident with the experimental curve, in particular,
showing the very large stagnation region between 9 = 300 and
900.

In fig. ( Zi-22),the theory is compared with the pressure

distribution obtained for the flow over the stationary wheel
with endplates. It should be remembered that the floor boundary
layer separated before the stationary wheel and therefore
stagnation pressure was not attained near the floor. This also
had the effect of 'streamlining' the flow over the wheel and
of thus smoothing out the expected pressure variation over the
body surface. Since the theory does not take these effects into
account tho agreement with the oxperiinontal pressure is not as
 111.0

good as before. The same considerations apply to the rotating

wheel (Fig.(L1.-23)), the viscous effects leading to the high
pressures and upstream jetting at the bottom of the wheel
being ignored by the theory, as are any special effects at
separation.

The position of the free streamlines after 8eparation

is shown for the three cases in Fig.( L1._24). In the stationary
wheel case s the solution for e 8 =270° is chosen,as this gives a
closer agreement for free streamline position with the smoke
photograph, Fig.(3-40), The other solutions can also be com-
pared with the appropriate smoke photographs (Figs.(3-40) and
(3-41)) when it will be seen that the streamline positions at
least show the same trends as the photographs, the 'otating
solution having the highest streamline.

The main use of free streamline theory in connection

with the wheel is that, having established that the theory
gives good agreement to the flow around two stationary cylinders
in contact, we can point out those areas of the two-dimensional
flow around the wheel on the ground which differ significantly
from the theory and thus from the simple flow, We can then
postulate reasons for the differences and test our conclu-
sions by further experiment and/or by other theoretical conside-
rations. If desired it may then be possible to use the theory
as a basis for a more exact theoretical treatment of the flow,
bringing in the boundary layer and the viscous effects in the
corner. However, even then,the theory in its present form
 liii

could not be applied to the three dimensional flow about the

wheel, although it is possible that three dimensional bluff
bodies could be dealt with by modelling the wake with sources
and using some form of numerical potential method to obtain
the flow on the wetted surface. However, this was thought
to be too difficult to attempt in the time available, so,
instead,we have used the theory in its simple form to help
in understanding the two dimensional flow, which in turn should
lead to a further understanding of the three-dimensional case.

(1.L) Conclusions

Of necessity, the theoz'etical work in this section

has had to be confined to simplified approaches to some of
the interesting features of the wheel flow and has,therefore,
given rise to a number of isolated models of these features.
B ecause the assumptions used are not the same in each case,
it is not generally possible to mathematically join the various
solutions together to form a whole. However, one can do this
intuitively to some extent and use has been made of the various
solutions in postulating the flow pictures of Figs. (1-12),
(1-19) and (3_L2). For example,in Fig .(3_ Z#2 (iij )) , we began
with the low Reyrnlds number solution for flow in the front
and rear corners, added the picture of separation that has
been built up, and then, joining up the mean streamlines to
 11l2

to prevent any starting or finishing in the flow, the picture

had to be as shown. The same type of process was used to form
Fig.(1-12); here, starting with the pictures of separation and
stagnation which had been deduced.

The extension from these two dimensional models to

the three dimensional flows about the wheel is more difficult,
although they provide a suitable basis for this,when allowance
is made for flow around the sides of the wheel. For example,
we would expect to quickly lose the upstream jetting from the
front corner around the side of the wheel,but the high pressure
under the wheel can still be explained by assuming a flow
similar to the low Reynolds number solution will still take
place, sufficiently close to the line of contact. Again,although
separation in three dimensions is different from that in two,
in general, near the centre]..ine,the resemblance will be close,
so the movement of separation above the wheel can be partly
explained in term of the special type of separation, deduced
for two dimensional flow.

Where the theoretical work has not been directly

applicable to the wheel, as in the case of free streamline
theory, it has been of use in comparing with the two dimensional
experimental results,as an aid to understanding then,and the
theory, as given here, could form a possible starting point
for a more extensive treatment. Thus, although the ideas in
this section have applied to diverse simplified models,they
have all helped to contribute to an understanding of the flow,
as revealed by the experimental results.
 113

References

Abernathy, F.H. 'The formation of vortex streets'.J.F.M.

Kronauer, R.E. Vol. 13, pp 1-20, 1962

Batchelor, G.K. 'An Introduction to fluid dynamic8'

C.U.P. 1967

Bearman, P.W. 'Some recent measurements of flow around

bluff bodies in smooth and turbulent
streams', Imperial College, Aero dept.
report No 72-09, 1972

Berry,A.J. 'An experimental investigation of ground

effect using a moving surface simulation'
Univ. of London [Link]. Thesis, 1968

Betz, A. 'Zeitschr des Verines deutscher inge-

nieure' 69, 1925
(quoted in Goldstein (1938)1E)

Bluston, M.S. 'A theoretical solution for Laminar

Paulson, R.W. flow past a bluff body with a 8eparated
wake' Jul. de Mecanique, Vol II, No.1,
pp161-18O, 1972
 144

Brodetsky, S. 'Djscontjnuoe fluid motion pa8t circular

and elliptic cylinders' Proc. Roy. Soc.
A,Vol. 102, p542, 1923

Carpenter, L.H 'On the motion of two cylinders in an

ideal fluid'. Jnl. of Research of the
National Bureau of Standards, Vol.61,
pp 83-87, 1958
(quoted in Wadcock (1970))

Chubb, T.W. 'The response of a narrow bore pressure

measuring system to step and oscillatory
pressures. RAE. Tech, rep. No.68010, 1968.

'Proceedings of the first symposium on

City University
road vehicle aerodynamics'. City Univer-.
sity, London, Nov.1969.

Davies, J.M. 'The aerodynamics of golf balls'

J. Appl. Phys. Vol.20, No.9, 191&9

Page, A. Experiments on a sphere at critical

Reynolds numbers', A.R.C. R N No.1766,
1937

Page, A. 'Further experiments on the flow around

Falkner, V,M. a circular cylinder', A.R.C. R N. No.
1369, 1931 (quoted in Goldstein (1938)11)
 1b5

Fail, R.,Lawford,J.A. 'Low speed experiments on the wake

Eyre, [Link]. characteristics of flat plates normal
to an air stream', A.R.C. R. M.
No.3120, 1957

Fink, P.T, 'Some low speed experiments with 200

Taylor, J. Delta wings'. A.R.C. Report No.1785l4,
1955

Flaschbart 'Hanbuchder experimental physik'

Vol. Ll, part2, 1932
(quoted in Goldstein (1938)1)

Gerrard, J.H. 'The mechanics of the formation region

of vortices behind bluff bodjes'
J I.F.M., VOl. 25, pp k01_Zi13, 1966

Gilbarg, D. 'On two theories of plane potential

Rock, D.H. flows with finite cavities' Memo U.S.
Naval [Link] Lab. 8718, 19k5
(quoted in Thwaites (1960))

Glauert, H. and airscrew theory'

c.u.., 2ndEd, 1948

Glauert, M.B. 'The laminar boundary layer on oscillating

plates and cylinders'
J.F.M.,Vol.I, pp97-11O, 1956
 1 hlr6

Glauert, M.B. 'A boundary layer theorem, with appli-

cations to rotating cylinders' J.F.M.
Vol.2, p89, 1957.

Goldstein, S. 'Modern Developments in fluid dynamics'

O.U.P. Vols.1 2, 19:38

Gottingen 'Ergebresse der aerodynamischen Ver-

suchsanstadt zu Gottingen' Vol.2. 1923
(quoted in Goldstein (1938))

Gould, R.W.F. 'Wake blockage corrections in a closed

wind tunn3, for one or two wall mounted
models subject to separated flow'
A.R.C. R M No. 369, 1970

Griffiths, R.T. 'Differential boundary layer separation

Ma, C.Y. effects in the flow over a rotating
cylinder'
[Link]. Aero. Soc., Vol.73, pp52k-526
1969.

Hess, J.L. 'Calculation of potential flow about

Smith, A.M.0. arbitrary bodies' Prog. Aero. Sci.
Vol.8 1968

Hyde, J.R.N. 'High speod land vehicles',

[Link]. Project report, Imperial College,
1972.
 147

Kuchemaxin,D. 'Inviscid shear flow near the trailing

edge of an aerofoil' [Link], Vol.15,
p 292, 1967

Levi-Civita, T. 'Scie e leggi de resistenzia'

B.C. Circ. Nat. Palermo, Vol.23,1907
(quoted in Mime-Thompson (1968)

Linke, W. Physik Zeitschr, Vol.32, 1931

(quoted in Goldstein (1938), II)

Maccoil, J.W. 'Aerodynamics of a spinning sphere'

[Link]. Aero, Soc. Vol.32, p777, 1928

Maskell,E.C. 'Flow separation in three

A.R.C. Report No. 18063, 1955

Maskeil, E.C. 'A theory of the blockage effect on

bluff bodies and stalled wings in a
closed wind tunnel' A.R.C. R N No.3400
1963

Mime -Thompson,L.M. 'Theoretical aerodynamics' 5th Ed.

Macmillan Co, 1968

M.I.R.A. Aerodynamic characteristics of exposed

rotating wheels' by W.R. Staleford and
G.W. Carr, MIRA report No. 1970/2, 1970
 148

Morelli, A 'Aerodynamic actions on an automobile

wlieel 5th paper at the First Symposium
on Road Vehicle Aerodynamics, City
University, 1969.

Morkovin, M.V. 'Flow around circular cylinder - a

Kaleidoscope of challenging fluid
phenomena' presented at the Symposium
of fully separated flows, The Am.
Soc. of Mech, Eng., Aim Arbor 1964

Olson, HF. 'Acoustical engineering' 2nd ed., Van

Nostrand, 1957

Olson, H.F. 'Dynamical analogies' 21 ed. Van

Nostrand, 1958

Parkinson, G.V. wake source model for bluçf body

Jandali, T. potential flow' J.F.M. Vol.40 pp577-
594, 1970

Prandtl, L, 'Applied hydro-and aeromechanic8'

Tietj ens ,O. Dover, 193L1-

Ried, E.G. 'Tests of rotating cylinders'

NACA T.N. 209, 1924 or Flight, Jan8,
1925

1 149

Riabouchinsky, D. 10n steadyfhiid motionwithe suzlaces'

Proc. Lond. Math. Soc. Vol.19, p206,
1920 (quoted in Thwaites (1960))

Roshko, A. 'A new hodograph for free streamline

theory' NACA TN 3168, 1954

Roshko, A. 'On the wake and drag of bluff bodies'

J. Aero Sci., Vol. 22, p12k, 1955

Roshko, A. 'Experiments on the flow post a circular

cylinder at very high Reynolds number'
J.F.M., Vol.10, p3 145, 1961

Schliohting, H. 'Boundary layer theory' 5 ad, MCGraw_

Hill, 196a

Sto].lery, J.L. 'Forces on bodies in the presence of

Burns, W.K. the ground' 1st paper, First Symposium
of Road Vehicle Aerodynamics, City Univ.
1969.

Thom, A. 'Experiments on the air forces On

rotating cylinders', A.R.C. R M 1018,
1925

Thom, A 'The pressures round a cylinder rotating

in an air current' A.R.C. R M 1082,
1926
 150

Thom, A. 'Experiments on the flow past a rotating

cylinder' ARC R M 1410, 1931

Thwaites, B 'Incompressible aerodynamics' 0.U.P.
1960

Villat, H. 'Apercu.s theoriques sur la resistance

des f]uides' Collection Scientia Gauthiers
Villars, Paris 1920
(quoted in Thwaites (1960))

Wadcock, A.J. 'The interference between a pair of

circular cylinders normal to a stream '
MSc project report, Imperial. College,
1970

Weisselberger, C. Derluftviderstand von Kugeln'

ZFM, Vol.5, ppl4O-144, 1914
(quoted in Schlichting (1968))

Woods, L.C. 'Two dimensional flow of a compressible

fluid past given curved obstacles with
infinite wakes' Proc. Roy. Soc. Lond,
vol. 227, p367, 1955
 151

Foot Note

After this thesis had been written,a report by Telionis

and Werle (1972) came to the author's attention. The report
discusses a numerical computation of the boundary layer above
moving boundaries. Models of the flow near separation, with
the boundary moving upstream and downstream, are given,which
agree with the models discussed in this thesis. It appears
that these were first proposed by Moore, Rott and Sears in
connection with unsteady separation from a fixed boundary. The
numerical results of Telionis and Werle show features that
have been discussed in this thesis. For example, the numerical
integration proceeds smoothly through a point of vanishing
skin friction without any singularities. When a singularity
of the type normally associated with separation does appear,
it occurs some perpendicular distance away from the boundary
and nonsingular flow continues underneath it. These results
lend support to the models Qf the flow near separation given
in this thesis.

Reference

Telionis, D.P. 'Boundary layer separation from moving

Werle, N.J. 1972 boundaries' Virginia Polytecimic Insti-.
tute and State University Report No.
VPI-E-72/1 3
 152

Appendix

Details of experimental apparatus

(A-i) The Wind Thnnel

The tunnel used is the Aeronautics departments 5t 41

closed return tunnel, with a speed range of 40-150 ft/sec.

The working section is 7 ft. long, of rectangular shape,with
corn fillets,which reduce the average cross sectional area
to 17.5 [Link].. The fillets decrease in size along the working
section,to allow for boundary layer growth. Breather slots
at the downstream end of the working section ensure that the
static pressure within the section is close to atmospheric,
although slightly less due o leakage around the tuwiel circuit.
When the moving floor is used, it is necessary to remove the
bottom tunnel boundary layer ahead of the floor. This means
that for a given speed the static pressure is lower and has
more variation along the working section, as given in Fig.
(2-7). A honeycomb and gauze placed upstre4m of the contraction
ensure a free-stream turbulence level of about 0.2%.

The tunnel speed was measured from the difference

in the static pressure before the contraction and at the be-
ginning of the working sections this difference having first
been calibrated against the speed in the empty working section
 153

at the wheel, centre position,given by a Pitot-static tube,

In calculating the results a correction is applied to the
speed measured in this way, to allow for blockage effects,

(A-2) The Moving Floor

A detailed description of the design and development

of the moving floor has been given by [Link] (1968). The
floor consists,basically,of an endless belt running under
tension between two rollers one of which is driven by a Cons-
tant speed AC motor.(Fig.(2-4 )). To control any sideways
tracking of the belt,the end rollers are crowned and the
bearings of the non-driven roller are movable,to provide un-
equal tensioning. The belt is 33 ins, wide and has ft.
between roller centres, these dimensions being chos9n to
enable the floor to fit under the bottom corner fillets., re-
placing the fixed tunnel flooring.

For use with the wheel a support roller is mounted

under the floor on bearings (Fig.(2-4)). The wheel is arranged
to press down on this roller, sandwiching the floor between
them, When the floor moves therefore it causes the wheel and
roller to rotate. Checks of the floor speed and wheel rotation,
with a stroboscope show that there is no slipping between them.

When the belt is subjected to a negative overpressure,

such as might occur beneath and behind a model mounted above
 1511

the belt, it will lift away from its supporting surface into
the tunnel. This can have important consequences for tests on
models mounted close to the floor, altering the flow on their
underside and hence possibly the whole flow field. In the
present experiment the wheel pressed down on the middle of the
belt, helping to prevent lifting. However, at high tunnel
speeds,the large negative static overpressure caused the belt
to lift excessively all around the wheel and especially in the
wake region. For this reason,it was felt desirable to run the
tunnel at as low a speed as possible,within the limitations
imposed by Reynolds number requirements.

In order to accurately model the real flow with the

moving floor, it is necessary to eliminate the tunnel boundary
layer before the floor,as completely as possible. Initially,
thi8 was attempted by fitting a fairing in front of the floor
to give a 2 ins. duct,through which the boundary layer could
bleed away. The duct was fitted with a variable venturi, to
ensure a favourable pressure gradient through it. This arran-
gement eliminated the tunnel boundary layer, but allowed a
new layer to grow over the top of the fairin, re8ulting in
a boundary layer over the moving floor of the form shown in.
Fig. (2-5(i)). To overcome this,a portion of the upper sur-
face of the fairing was removed and replaced with Vyon porous
plastic material, through which suction was applied. This
improved the situation, but not as much as had been hoped

resulting in the boundary layer profile of Fig. (2-5(ii)).It

was felt that this was due to the area through which suction
 155

was applied being too small and possibly by the blockage to

the duct caused by the suction chamber beneath the fairing,
The whole fairing was therefore redesigned to increase the
suction area, reduce the blockage and at the same time increase
its strength. This final version is shown in Fig.(2-5 (III)),
together with the much improved boundary layer profile.

(A-3) The Wheel

As explained previously, the diameter of the wheel

had to be such as to give a Reynolds number in the super-
critical range, when used at the maximum tunnel speed at
which the moving floor would not lift excessively. The choice
was for a wheel of 16/8" diameter, running at a speed of 60.8
ft/sec, giving a Reynolds number of 5.3 x 1O and a maximum
blockage ratio of 8%. The static pressure in the working
section at this speed is of the order of 0.05 ins, of water,
without the wheel present, and the maximum negative over-
pressure (in the wake close to the wheel) was less than O,5ins.
of water, which did not cause the belt to lilt excessively.

It was decided to have three different rim widths,

ins,, 6/4 ins and 9 1 /L& ins, all of 16/8 ins, diameter,
in 6rder to study the effects of varying aspect ratio. These
rims were to be combined with 2 edge profiles, representative
of tyres used on Grand Prix cars, to give a total of 6 models
to be tested. (fig.(2-l)). The different edge shapes give
 156

overall widths which are 4 ins, and 31/4 ins, greater than the
width of the particular rim used. To accomodate the various
overall widths needed,one inner hub (4 ins. wide),two outer
hubs (31/2 ins, and 6 ins.) and two spacers(31/4 ins, and
21/2 ins.) are used in various combinations, All the above
are cast from aluminium alloy.

Pressure measuring holes (i/i6 ins. dia.) are dis-

tributed over the wheels at the positions indicated in Figs.
(2-1) and (2-2). Flexible plastic tubing (/32 ins. l/D)
connects the holes to brass tubes in the hub well and the
transducer on the wheel axis is connected to one tube at a
time (Figs.(2-3) and (2-9)). The tubing is arranged to have
the same overall length for all the measuring holes 1 for reasons
connected with the frequency response of the 8ystem. (section
(A-4)).

The wheel rotates freely on bearings, the bearing

assembly being supported by 3/8 ins, diameter mild steel rods
each of which is [Link] in length to a small extent. Thus
the rods can be adjusted to allow the wheel to press down onto
the floor,with sufficient force to prevent any slipping between
the two. Outside the tunnel,the rods are connected to a heavy
girder framework bolted to the laboratory floor.

Fig. (2-9) shows the mounting arrangement for the

pressure transducer,which is placed on the wheel axis to avoid
centrifugal effects.
 17

The output from the transducer is led through a shaft of mild

steel tube to a DISA oscillator No 51E32,rotating with the
wheel. The oscillator output is taken from the rotating
assembly through two silver graphite brushes,running on silver
slip rings. (Fig.(2-1O)).

Because of difficulty experioncedwith drifting of

the D.C. component of the transducer signal,a small diameter
tube was teed to the tube leading to the transducer and the
mean pressure is led along the rotating shaft through this
tube and out via the pressure seals, as shown in Fig.(2-1O).
Another tube is led through the shaft to connect the air
behind the transducer diaphragm to the external atmospheres
to allow for long term variations in atmospheric pressure
The size of this tubing is sufficiently small to prevent short
term variations in pressure being transmitted with sufficient
power to affect the transducer.

(A-4) The Pressure Measuring System

The transudcer, mounted at the centre of the wheel,

is a B.& K. .1/ li. ins condenser microphone, type 11135. This
consists of a thin diaphragm (2k) mounted close to a rigid
back plate, the two forming the electrodes of a [Link]
chamber behind the diaphragm communicates with the external
air as described above. When the diaphragm is subjected to a
varying pressure on. its front face, it moves and so alters
 1

the distance between itself and the back plate.

the capacity thus
varies in step with the pressure. This capacitance forms part
of a resonant circuit,with a basic resonant frequency of around
5 MHz. The variation of the microphone capacitance changes
this frequency by a few KHz., producing a frequency modulated
signal.

The resonant circuit and the driving signal of 5MHz.

are provided by the DISA 51E32 oscillator,rotating with the
wheel. The oscillator is powered by and the resulting F.M.
signal returned to a DISA reactance converter, Type 51E01.
(Fig.(2-6)), The reactance converter amplifies and rectifies
the F.M. signal, producing a varying D.C. voltage at the out-
Put,depencient on the pressure acting on the transducer
diaphragm. Over the range of pressure used,this variation was
found to be linear (Fig.(2-il)).

The frequency response of the transducer in free

air is virtually flat between 6 Hz. and KHz., However, to
measure the pressure at the rim of the wheel,the transducer
has to be connected to the measuring holes by tubing around
13 ins, long. This tubing then determines the frequency cha-
racteristics of the system, since it both attenuates higher
frequencies and suffers from resonances.

The attenuation of pressure waves in narrow tubes

has been given by Olson (1957). If A is the amplitude at a
 159

distance x along the tube from where the amplitude is A 0 we

can write

A = A0 e -ax

Then for a tube length 1, diameter d, the attenuation is

determined by

ax = B.

where f = frequency
and B is a constant

In theory,therefore,attenuation can be overcome,

for any particular frequency required,by choosing a large
enough internal diameter for the tubing (in our case 3/32 ins.)
Howeverif a large diameter is used,there are problems with
resonance of the air in the tube. The simple theory for reso-

nance in a tube ,with one end closed,is given in many texts

([Link] -Thompson, 1968, p.462) and shows that resonances
occur at frequencies of

c (fundamental
r frequency)' 4 ' 4j etc.

where c = speed of sound.

In particular, the frequency of the first resonance

varies as 1/1. Thus,to have the fundamental frequency as high
as possible and to reduce phase change of the pressure waves,
in passing down the tube,we require 1 to be small. However,
 160

this would necessitate reducing the diameter of the wheel, d,

and the frequency of rotation of the wheel, Lw, varies as
From the above,the first resonance,f ,varies as l/d. So
fwj1 fr 's.i U co Re dNo.

Thus,to obtain the best definition in pressure

variation around the wheel would require choosing the wheel
diameter large,although this would increase the attenuation
in the tubing. Obviously some compromise ha8 to be reached
between these conflicting requirements and the other require-
ments, such as tunnel blockage.

For the chosen wheel size and speed, the tube length
of around 13 ins, gives a fir8t resonance peak about 200 Hz,
compared to the frequency of rotation of the wheel of lk.25Hz.
This resonance caused a pronounced ringing in the signal from
the wheel,aa the pressure hole passes the floor. Initially,
we attempted to eliminate this ringing by electrical filtering,
but found it very difficult to counter completely in thi8 way,
so eventually a combination of electrical filtering and
mechanical damping was used, The mechanical component consists
of a tuft of wire wool inserted into the tube leading from
the transducer to the connection in the hub. By varying the
compactness and amount of wire wool, it was possible to
obtain a much flatter frequency response, smoothing out the
resonance peaks. (Fig. (2-8)).

However, this does not eliminate all the unfortunate

 161

features of using a long tube. As described previously, when

the measuring hole passes the floor, it receives a sharp pulse-
like input, due to the rise in pressure in front of the line
of contact between wheel and floor, followed by the sharp
decrease just after. Thj pulse contains components at many
frequencies, including high frequencies, and, as these travel
along the tube, they suffer varying amounts of attenuation and
phase change. Thus,on recomposition at the transducer head,
the signal will no longer have its initial form. Thus,the sudden
rise and fall of pressure at the floor is not followed exactly
by the transducer and one cannot know the magnitude of the
error involved,since the true form of the signal is unknown.
The pulse response of the system was studied using the set up
shown in Fig.(2-11(1)). 1 t was found that the results, in the
sense of fastest rise and fall times, could be obtained by
making the frequency response as flat as possible up to as
high as possible.

As mentioned in section (4.2.2), in this region,there

will also be some error due to the lifting of the belt in the
low pressure region behind the line of contact. In the deter-
mination of the lift force on the wheel, however, the movement
of the separation position has most effect, since the pressures
near the ground act only on a very small area of the wheel's
surface (although because of the large values involved they are
significant). It is hoped, therefore, that the error in the
lift force for the model wheels is not greater than 5%.However,
the lift forces on actual wheels,running on different types of
road surface,will have some variation from our figures,because
of the effect on this region near the
 162

ground. The drag is little affected by the pressures in this

region and so should not have such a great variation.

An initial difficulty in operating the system was

the high rate of drift of the D.C. level, the rate being
variable and influenced by such factors as wind speed and
wheel rotation. To overcome this difficulty,it was decided
to measure the mean pressure directly with a manometer, as
described before, and to use the microphone to measure the
fluctuating pressures only. The fitting of the side branch,to
do this,will have an effect on the response of the transducer
to the pressure fluctuations passing down themain tube, as
can be seen by considering the electrical analogy of the
acoustic system.(see Olson(1958)). Because,at the frequencies
of interest,the wavelengths of the pressure fluctuations
are long compared with the tube length, we can use a lumped
impedance analogy, where electrical current is analogous to
volume current and e.m.f. to pressure. Then the acoustic
system can be representedasinig. (2-13) and we can apply
iCirchoff's laws ,in the normal way, to get four simultaneous
equations,which can be solved to give

e 0 z2z1&
= (z2 (z 1 +z3 +z 4 ) + z1z4+z1z3)

in particular, the dependance on z2 is in the form

e Az2
e = 13z2+C
 163

We should like the system to be as independent of

z2 as possible. The form of the above shows two ways of doing
this; (i) if is made small so that z 1 z -O then
3 e Z3+Zk

independent of z2 . Physically, this means taking the side branch

as close to the measuring hole as possible; (2) z 2 is made
large or, physically, the side tube diameter is made small

Because, for simplicity, the side branch was taken

from the tube joining the transducer to the connecting tubes
on the hub, it was not possible to achieve the first alter-
native above. A 5mall diameter (,i mm) side tube was therefore
used,but the time taken for the manometer to attain the mean
pressure was then excessive (being inversely dependent on the
fourth power of diameter (Chubb(1968))) and the tube was
susceptible to blockage. The simple answer was to use the
system,without the side branch,to measure the fluctuating
pressures, then to fit a wider side branch (1mm dia.) to
measure the mean pressure. This 1mm tubing does affect the
fluctuating signal received by the transducer,but this should
not affect the value of the mean pressure.

The mean pressure,obtained in this way,has to be

corrected for the effects of the centrifugal pressure gradient,
since we are taking the pressure out of the rotating system
at some small radius, rm, (fig.(2-lo)) and we wish to know
the pressure at the radius of the measuring hole, r 5 . The
correction is obtained by considering the pressure difference
across a small element of thickness,dr, i.e. pw 2 rdr, and
 164

integrating to give

U 2 co 22
r-r
s m
2 I. 2
0
man.
where =Aradius of wheel.

The mean pressure must also be adjusted to allow for

the static pressure at the wheel position. Thi5 is obtained by
measuring the static pressure of the undisturbed flow at the
beginning of the working section and then assuming the same
variation with distance along the working section as occurs
when the wheel is not present.

The calibration of the transducer was performed in

the range 15-800Hz using a B& K Pistonphone No.4220, which
produces a constant amplitude signal throughout this range
of frequencies. Measurements at higher frequencies were per-
formed by comparing the response of the measuring assembly with
a transducer placed at the source of the pressure waves (Fig.
(2-11)), the same set-up being used to compare the pulse
response of the system.

(A-5) Data Reduction

The fluctuating signal,emerging from the pressure

measuring system,is fed into a Hewlett-Packard Correlator,No.
3721A (Fig.(2-6)). By taking an average from several thousand
samples (usually 8K),the correlator is able to eliminate any
 i6

noise present on the signal; the 'noise' in this case being

mostly produced by the turbulence in the separated region of
the flow, very little electrical noise being present.

The correlator averages in tie following way; At each

revolution of the wheel,a trigger pulse is fed to the correlator
to initiate a sampling cycle, in which the incoming signal is
sampled consecutively at 100 points, the time interval between
points being adjustable between 1 [Link] and 1 sec. Each sample
of 100 points is divided by the total number of samples and
added to previous samples, In this way we eventually obtain
100 equally spaced points around the wheel,each of which re-
presents the average of the signal at that point. Any distur-
bance on the signalwhich is not correlated with the rotation
of the wheel, e.g. noise or hum, is therefore eliminated,
provided enough samples are taken. The corre].ator is used in
an AC mode, i.e. with all signal fluctuations below 1Hz.
filtered off. This is to eliminate the varying D.C. level,
refered to earlier. The correlator output can then be re-
combined with the reading of mean pressure from the manometer
to produce the complete averaged pressure signal.

The electronics for applying the trigger pulse to

the correlator,once every revolution of the wheel,are given
in schematic form in Fig.(2-12). The inductive transducer
head is mounted on the fixed wheel supports,close to the hub
(Fig.(2-9)). A small steel insert in the hub passes close
to the transducer head,as the wheel rotates, causing a change
 i66

in the inductance of the transducer head and hence altering

the signal from the bridge network. This signal is processed
by the rest of the system,to provide a suitable pulse to the
correlator every revolution of the wheel. An adjustable time
delay could be added to the system by inserting a second
monos table.

This time delay allows any particular portion of

the signal tobe picked out in greater detail. For example,
the basic signal from the wheel is repeated every 70 msec
and the correlator is usually set up with 1 msec between
points so that every two revolutions of the wheel yield 100
msec of the signal and the distance between points is 5.1°
around the wheel.(The second trigger pulse generated during
this sampling is ignored). If it is required to look at any
portion of this signal in greater detail, the time delay can
be set to initiate the sample just before the portion of
interest and the time interval between points reduced to give
the desired definition. Thus,setting the interval to 0.1 msec
one could examine 10 msec of the signal at every 0.50 around
the wheel. This ability was particularly useful in providing
small intervals between points near the floor. The error
from numerical integration in this region of rapid change
can be reduced to fall within the error due to the transient
response of the system.

The correlator was connected to a Facit L 070 punch,

in order to obtain the output on paper type for computing
 167

purposes. It was also connected to a H-P 703kA X-Y plotter,

to give a permanent visual record of the output display, for
use in comparing with the processed results, as a check for
any computing errors,

A program was developed to provide the pressure,

lift and drag coefficients for the wheel. As input,the pro-
gram requires the averaged pressure signal data for each hole,
from the correlator, with a specification of the angular
interval between the points of this data,and of the position
and direction of eaph hole. Also required,are an indication
of the position,within each set of 100 points,which corresponds
to the point of contact with the floor, the dynamic head
corrected for blockage, the mean pressure corrected for cen-
trifugal effects and tunnel static pressure, and a zero level
signal. The last, obtained from the rotating wheel with the
wind tunnel off and the transducer not connected to any of
the pressure measuring holes, is necessary to eliminate any
effects, such as varying resistance around the slip rings,
producing a fluctuating [Link] is correlated with the
wheel's rotation. The position corresponding to the point
of contact with the floor was determined by scrutinizing the
detailed X-Y. plots of the pressure near the ground and choosing
the position of highest pressure as the required point. As
a check, the effect on CL and 0D of displacing the floor
position 20 away from this maximum position was computed
for a typical signal. This produced a change of less than k%.
 168

For each ho].e,the program subtracts the zero level

data from the pressure data and then adds the appropriate
mean pressure and divides by the dynamic head,to give the
pressure coefficients around the wheel for that hole. Because
the signal from the wheel repeats every 70 msec only 70 of
the 100 points are needed, the otherB being discarded.

The integrals for the total lift and drag coefficients

are

CL o sin9 cosø rde

J

2t
CD c co86 cosø rd9 dw
f
0

where the symbols are explained in Fig.(2-14) and the area

chosen was either

W
2 rdw
(projected frontal Area)

or tread width x diameter.

In practice,the lift coefficient for each hole is

defined as

CL 1/2 cosø sinOdO

 169

and this is calculated using Simpsons rule. The normal inter-

Va]. between points is 5.10 , but near the ground a smaller
interval of 0.51° is used and the contribution to the integral
from this region is evaluated separately.

The total CL of the wheel is then

Jrcdw

and these integrals are evaluated by the trapezium rule,to

deal more simply with the unequal interval size,occasioned
by the unequal spacing of the pressure measuring holes across
the wheel.

The drag coefficient is evaluated in a similar

way.

The dynamic head,used in the above calculations,

must first be corrected for tunnel blockage, This is done
using a correction due to Maskell(1963), who uses a model
of the separated flow similar to that in the free streamline
theory discussed in section ( Li-3),but applied in three dimen-
sions. That is,he assumes that the separated region is divided
from the main stream by a surface of discontinuity,on which
the velocity is KU. By considering the momentum balance past
the body in the tunnel,Maakell obtains the rolation
 170

K2 CD %+ 2
- = 1+ - o (() )
K C 2 K2-1
C

or

K2 =+B+
J B2-4K2
C
2

t ,S, 2
neglecting terms 0 i

where

B = (K2+l---)

S a Area of the model u5ed in calculating CD

C a X-sect. area of wind tunnel working section

and subscript c denotes corrected values.

The correction for the dynamic head q= 1/[Link] comes from

2 1-c a
p
- =q

Gould (1970) has shown that the above can be

applied with succe8s to wall mounted flat plates, which follows
from the fact that the theory makes no use of images,but con-
siders only the momentum balance in the tunnel. The theory
does assume that the effect of blockage is a simple increase
in dynamic head,which is equivalent to saying that the distri-
bution of the static pressure on the model is unaltered by
blockage. Maskell shows that this is largely true for flat
 171

plates normal to the stream,but for more three dimensional

bodies,with rounded surfaces,the effect on the pressure dis-
tribution is likely to be greater. In such circumstances, it
is best to keep the blockage ratio, S/ ,small so that any
errors are minimized, The corrections calculated for the
wheels are given in table 5. The base pressure used is the
mean taken over the base region of the wheel.

(A-6) Additional pparatus, used in examining the external

flow

For the measurements of total head,given in section

(3-6),a Kiel tube was used because of its insensitivity to the
yawing of the flow. The dimensions of the tube are given by
Fink and Taylor (1955),together with a yaw calibration show-
ing the flat response to yaw angles between ± 30°. Despite
this,because of the large deviations in the flow directions
around the wheel, the tube had tp be mounted in the tunnel
with the ability to point in any direction and this was
achieved using the supporting sytem shown in Fig. (2-15),
The approximate flow direction was found by using a wool tuft
and the yaw insensitivity of the tube made use of to eliminate
any errors in the direction so obtained.

For measurements of the boundary layer profiles on

the wheel,a flattened Pitot tube and a static bewt5the
sizes are given in Fig. (2-16). The special problem connected
 172

with this work,was due to the vibration of the wheel and the
supports of the 1itot or static tube had to be fairly stiff
and well damped,in case the wheel should contact the tube
and set it vibrating. The arrangement shown in Fig. (2-16)
was found to be satisfactory. At each station where a profile
was required, this arrangement was set up with the Pitot or
static tube parallel to the wheel's surface and then the
tube and supporting rods traversed normal to the surface.


(%sJ
I-
-J 0 0 0 0 C 0
o
4 c.j . LU
C ' 0 C ' 9 C'
.0 ______ _______
'- 0 - (.0 cc (Y)
c) 99 999
o

U, C') - 0) IC)
- 'ii - IC) (C) '.0 (0 IC)

- -o cc CD IC)
r0) C) C) C? C)
OODi-

> cc 0 cc I-
U) - - j -

_co
• 0
L.._ _________ _________ .0
C4 C') 0 CO 0
rj
'- ,, LU (0 trn- IC) LU F-
(9 V 0 L.i . .

oaJ(n ________ ________

0 . cc - LU CO
-1 0 N. 0
,;: o- • .
I- U)

CO IC) C') C4
0)
C .
CC?)
L. Q ,.
><<u,

- o - ((4 0 _ (0 (0 (0
IC)
rrj IC)

- . 0 (C) (0 (0
( LU (C) Co (0

(••.4 e.J
0. 0
 Table 2

Total Tread Four Wheels Front Wheels

Width Width
_____ Dia. Dia. C D CL Co1 CL,

M.I.R.A. 33 •33 1 . 11 •50 -

•5O •50 1•00 . 60 . 64 •40

'67 •67 1•06 •44 - -

Morelli 39 .28 - '45 -1

Table

Wheel No. CD, CL, Co1 CL2

B2 Rotating • 58 •44 85 •65

B2 Stationary -77 •76 1q13 1•12

Table 4

X-Sect Tread
Wheel Area x Dia.
Co1 CL, CO2 CL2
No. sq. ft. [Link].

C2 142 105 '70 '43 94 '57

C2G 1'42 .97 '70 38 102 -56

 LC)
IC)
U OD

c'J

C,)
eJ N

IC)

LC)
(0
I— o

co

o 0
I-

C
0

__ S
 0-.-'
.,

0.

a
-o
o
0
.- (n
—

0
DC
o
< c C

E
C
(I,
(t
E - C
ni C w
C-
0)
L. 0

r - C C
. rtl 0
>
?
).__
a' E
0 o rtl
_f: s)
a)

4-

IL
 & U'I
Lfl 0
• 1.. >1
•0- — >1
c - a -'.
U. C .0 > 0.
Q rrtn
1 4- ci
rC cu ($ ,
-b.0 c L
x o—— .-a: 1.
(0
IC) '-5- °.c
— , .0 _eJ,—U
c) - C)
cs..
V 4- - .5-4 u
' L. C
0) V • 2 2G >
ct oc_ a) 0
In DLC C r - 0 -x
cj
V 0.
0
0 OC 0 c
o

I
o U
[Link]
L.
x
LC
C,.)
I-
O
0.— 0i
1nU)
c c
rØro
-.0
°.2

O
C E
0
(0 I.. '- L. i...
4-
•0 Q .- • O o

U,

' 'I E
0)
' l.

4-
( c' 0
- 0 •1c1
0'-. U-
O) — U
C I_uI .4-
,vJ
-
V 4-
V. '-5- '•- -D
-tJ
2''
V
CD
a-,
151 :

.0 n,- E 0
>,
.0I
0 .0
,V)
0 o
e
I (hf

Li

0
0
S.
P

0
(f) a) .0,
V

I
V a)
a)
V . (l
CD
V U
0) cb
E
cv,
21
CV)
U


C')..

C )
C.,
C
0
4-.
'•0 C. -
9^ C .0 o0
-.
- ,- '3.
U,-.
- I 'J -

•u_ u't
El

I 'c c " 'c 'c C

I tP1
)C ) )< X
I I
o I
I-I
C'4 - c3 O cc CL
U I 1 , ,t it

Q1 oOUJCL
C-) -Q I I

E -S
s__I_S - -
z _I
I
C If, 5-..-.-

-o zr u_L- \ I
I .5
I.-.
0 S. .5....
0 0) _cJ•) S S.
5%
C
C
>
0 I % J)
.1—
'11
:3 0
1
; cs)
oJ
L. S.
\ 0
'1::
0,
0 9- Li_I
I
/
/

U
C)
.0

-
-
,
.-.-- -- --.-
6
1
I

- -.-
-

•.;3
Co
C

t_
--

- U)
CD
U-
c
I-
.a

II VI

S-nv)

I I I I I
L() 0 Lf) CD 9
(sJ
9
C4
I I I I
 z.
-0

ED

0
-o
E
:3
z

I')
Co

C -
>. _'•-,l.
0
c1 L..C)
I 0'-'
C —
•,-.
c1
•nE2
C C, 0. 0-
=;—.. .
C)
=

.4-
0
u_i
o ,

I-
4-
'.4-

E 0
o--
.4 C
>
o _c
(-4 - C .c
, U
C = O
C> ,
C, •
- (1)
e
e o '
a_i c °

C,
£3 I-

I I

Ln • U, .4 c) C.4


1.Q
)c ,c )c
4-.
t- cj

It
, )
> cr cro:
-
U,

C)
C
1.C) •; •
C) (I)

L.. 0 0 -
>1
uJ
0
C .—. -o
cij ', C
C4 c..j
'—. .—^ > 0
0 U
c ':
> 9 'S
r 0) Qi
—J
(j
\\\ —. —
•'
LL0Q
—
4-' 4-'
E
t
0 C
0
0 0
L I-
4-.
(ti
C U)
=
>.. n
, , (_,
.1-D —
U 1' r5 cj ,_

I'
0 :, c '0
CM
'ti 'I —
— C) =
rj
Li.. I' LL. ()
I'
0
0
I
4-, I, 'C
U I'
c
'4-
LU
'I
-c
1- 1'

('1
I!)
1
I'
(.
ti I-
-c
I' 0
1 U)

L'
d
± r-. '.0 C:,
 C
0
..
0
(I)

0 .0

z-

' tI

o '- U) C4
o 1 c _
.u, I
II ti Ii it
S.
'S
od r0000

..
S. S.
.

E .3 3-.

ZR
C,,
U)0) I.

-
I-

,
0
C,
S
.353•35
0
3.-, I
0) I
)
, /
/

, ,
— — - ,
r- '4-
.. ,_- ,,
0 -_ -- .-
.0
'I-
U)
C

cj

(0
C',

(
LL

I I I'
L) 9 U,


CL

1•4

Arrows Show Average Trend

with Increasing Reynolds
Number.

1•0

AU Re. No. Between

4xOand 4 6x10'

&p_ ynotds
Rto Number
(,57. 6 3x103- 6x10
Re:3dO Thom (925) ç29 . 6 41014x10
C Bi 9x10- 6x10
Reid (1924) 13'3 i.x1O-1.2x10
Btz (1925) 47 5'2x10E.P
Griffiths and 20 74x10t46x105
4a (1969) E.P
E.P End Rates Fitted
I I
FO 12 14 £..
Re.:46x 0' U0Q

FIG.(1-7) ffects of Rotation&. Speed

-.4 .and Reynotds Number on CL
pf Circular Cylinder

-6
3610r
Re.:
 V
C U,
- L

E- a
V
Zc
z(o
0) Li1 -
r.n• >

U
o L.
N
> 'I
U

0 cD

f5-
9- N
'I
, t-

9-

H
0

'4- S
0

IC

'4- U
'4- -
c:4

I.-

5D q

L


C,
,- L()
S.-
(N
(N
cc (N
(N >
O•)
C) 0
z
-
0 (N
-
C-) C_) oW
>w
o —•t: cc a.
,-. U
EE 0
—J
U) - - -4 Lt)
D cD D D cD (0 I-
z - - - ,- ,_ 0 1

) xxxxxx uJ

UI-

'—Ui
\ 0 -

(N
-

DC-n
0
,_
Zj

-J a.
I—U)
e cu
—J

d
I U
U-

U
0
0

• ( Lf) C (N '- 0
0
U
 • (0
D cj
C )<

EI
o
S)
c.
--C
- 0
-J

- .4-

c.o
I I I I I I


001

-a •1)I V)

U c ;z

Cr)
1-

< v-
U COO (0
%.. t_

000c'4

0
C,)

C- '5:;
(Q) Cu) E
J) -oo
U U
110< I•

I-.
/)
3t-
L.
o-
01 G <
0 0.. LLI>
-T
C..J

I I-'

(.

U 0.,

'I

E S.
C
IS)

- -"S
-z 0
(53
- S - - - - - -

- - - -

' 0
- '.0

>\\
F

(0
 11

.2
>'

c
iii

-t
C

>
0,
C
0

> ctj
0
0
3:
c-fl
IL '4-
0
0
c-fl

rc-i

Q
E
rti

3:

rt3

U)
0

C4
I-
 .- Surface of
Separation Bubbte
Viscous Region—
in External Stream

(1)

p. i
•,
I'I(v
_7\ ! ,
4L.
I-
I ,'T /
'p
Surface of
SoLid Body
Ahi ,i,

Litniti ng
Streamli nes
n Surface
of Body

ne

Limiting Streamlines

FIG.(1-13) Forms of Separation in Three DimensionaL FLoW



CL,CD

.4

CD ©

L

-8 6 -.4 --2 a

gi FIG.(1-14)
iirfr-
c ...cAgainst Incidence for
/8 Various Ground Clearances
(Stotlery and Burns (1969))

:o
® •110
-1 • 2 © •200

-14
Cc
L' D

FIG.(1-15) Variation in Lift and Drag

10 Coefficients of Wheel with
jght Above the Ground.
(After MIRA(1970))
•8

CD4

.4\ Dstar

•2

.3
-
C
V
-

x Rotating V,CJD.:l
0 Stationary

CL
Rot.


..: 2 •$• C')

0

0
0
tO

'I
0
z

a
-a
C"
-U
c
0

-.T.-- 'I
Eu,

0
cE tfl a • tO
0(t5 CD c.'J
;:>
CC) -
— (ti'-

- a
C'-

I- 0e- cr5
'.-
,

•Q. __
— 0 -
- –

(r)rLl
- cv,

.. F
.• I

(0 0 ('41')
T I-
I

L Q 0 L 0 L
I .— ç:J cJ
I I


°D

0
Cr)

o
0::
0 34(j) 0
0
rt tn - C,) J d
x

It

-i 0
z

Cjo
C—
0
--4- f
•-4--• '-J'--

, r
(n'•I -
s C')
cn QQi-C)
a
• co

tflr
U)>—' 0,-C1C') -

0
- C,)

(-.4

I-

iç o Ic) Ic) Q Ic) 9

9 çZ.j
c'J
I I
 M.I.R.A.

u1*, U',',

Seating Stri

Stationary Rotating

Morelli

FIG.(1-18) perimentaL Configurations



a)
C.)
C Iii
I-
.5
C-
______ —
/ IL
•1
o
C 1) \\' i ,
CD
0 —
z
CU

\'t\ cJJ C
fftrmm
/
/
I-
F-
Li
(ti /
I- /
w CU
0
a) 0
1)
z
I-z
// 50
//
/ %._

0
> z
'0

C) -
0 F-
/ -
a)

0
x U
(n
U
=
/ / I-
V
/
IL
(2 C 0
0
a). -I

/ I
-
'a

>.
/I I
- I
-I
-I
-I
/I
-
IL

F
 () (2
Centre-line Positions For
Three Rim Widths

C),4 ( Prof ite Shape 1

(full scale)

(Pressure Holes Marked L Are Common To All Rim Widths.

Hole Positions Near Centre-line Vary From Rim To Rim)

(2

©/
( Profile. Shape 2

FIG. (2-1) ge Profiles And Pressure,

Hole Fbsit ions
 LL (full scale)

PP I

(V () © Q9) QQ:
L • ©
FIG,(2-2) Pressure Hole Positions, Near Centre-tine
of Rim and Inside Hub

®and©

(full scale)
(i) View from Upstream

(ii) Side View

Fig.(23) Wheel in Wind Tunnel



>
L.

> I
- I 00.
• II
II .0

• O _
_
0
I,-

I-

ci)
0
0
4-'
I-
0
a
a

(I) 0J

a)
ci)
-c

a)
-c
3:

L. C4
0
0
IL.
0,
c
0


.
F_
8' .
4+

_____

iii
4 +
\-

-
0 -
4'
I_fl.>',
,-. U *
. 0

if 1:

AJ&/J\
Ic,
C!sJ.

C
IL
I

9 -r c4
_ ,_ e- -


-ci
(Ti
-
0 ___ ci ..,'
o
t'-._ ::::•:: (TiL.:'4
..
•,._, (Y) ci
- 00
1
C ..-
Q...
E
L.
a)
(I)
>
t,0

(U C
L.
0 0)
('j
('j
a) E
m U) 0
C
a)
(U
C
0)
0 -
N I
0 1-n
C
0)
-4
U..

U, 0 0
(0.
U
a cN (Ii
Qçv)
If, ____= E
a)
a
U
Ui
0.-
.c
I.. I- P
C'J
C- a)
00 0
- - &
00) U,
C
0D
CC c- I—
s.-sCi 0) •:•

c--C
-.05

-04

03

-02

-I 01

FIG.(27) Static Pressure in Working Section

dB

Hz.
0 1 20 o 100 1000

FIG.(2-8) Iy.pical_Frequency Response



0 •- C..0
t.-o

OC

Z C)
—

C
0
n$tn

t)> .
I C
ãY C0
7/'

(A
0)
C
I..

(A

LI)

FU HF
H III
I-

C
n E
Ct)
nit*
L1h
I' I I II
hgi
I)
I0
4-,
E
4..
t) I-in
I_
c

Nh • th

C
I- c t) C)
.U)

u U 0c
CL.
A .>l
th I..
t)Qt I o
OCU
00
—'-- I I

\.__)
 U

- V
'I)
C
tn ctl
(1) Lfl C
cL cE
cn
- o -XI--
C
(j V)LL U
ro

LU

I-
+9
-. I,
LLiIar
I rtL) U
U,
I'
00
rQ
[Link] II
DZ

It
 --

Microphone

rtrL
Pis
_____ Sieaker
LJ_ Reference Sinusoid
Microphone Cam
'-- B&KV4"

1) Pu'se Response And Frequency (if) Frcquency Response

Response For Frequencies > 800Hz. 14- 800 Hz.
Using&K Pi stonphone

Pressure
mm.H0

40

iii Amhtude Response And

lypicat Resutt
Pres
30

25
Dynamic Head
20 Microphone

15

10
Betz—
Manometer
5

250 3 put
mV.

FIG. (2-li) Methods Of Cati brat ing jransd ucer

 0
•1

I-
( t,
>.
(ti- C -
c >
:•- 0, t
0 L)—JV)
c'j

0
C)

U)
C)
C
0

I
CsJ

CD
FIG.(2-13) Electric AnaLogy of Acoustic System_

.uo.

FIG. (2-14) 5ymbols for Inter atio n over Wheel's Surface



tn
:i)

0
fr-_c
D
C h-*--------- -
t
, III
III U) >.
0' rj
*-
U,
U)E >
0 (0
- :i-. ' L )

41_L
•441
'-.. I I — •v1
c
mU) c(

_H
If)
-4
I cir3I
iZ<,rcIo
DL

• C
0
U,
.
U, If)

E s-
0
(_) C
C
4-
.4-
0
2IIfl rIi •
[Link]
.-
.4-
0
0.
0.
—a-
U)

.4-

•0
t. U)
0

It
IC)
0 IT
c.i
cr
If) — •0 • -

LL<I

LU
o ' tflt
_____ -
fl

it 6cr < 2-. o

itOC
CD - - - - - - -% - - - - - ---K
,0
/
/
#"
--0-- -
Variation of CL and CD
FIG.(3-1) with width of wheel.

.7

.5
A C

- Coefficients based on X-sect. area.

Coefficients based on tread x dia.
x Edge shape 1.
o Edge shape 2

CL

.5

.4

.3

A B C

a cv-, •0
— U LO (0
. cJ
- r4
. .
Or-. co C-
o;;p

.. - -,-
I.-.'

I.
/

ii

I.-. •

(_I) (•
\
P
-
5z.
7) S

a
I_i_i •
2

0 '7
I.'
.2 t
. S

— • A.
ç(D
•z:
'C
cr5 C
c..1
,>>

_
9
c •) •

Sc.

c•.4
2 cS4 o LL

C4 -1 C'J
4 a' C)

__
- • - . - .L...
. ._
•0•

-- _/ •

I U • I
. I

c c.p •.st
0 c4

• .• • . -

o _• -.- I.. -S - • -. -
U O ._._.__f!
.
Lii . - .
l.
• .
 a

-c
U)

U,
U•)

2U

—J
U

C
0

cr5
>

C,,
 4
co '.t r'

0
cJ
N
C Lt)

C'4

-Jo
C) - 0

C4

rj
cp
ULID

-c
o1J

.
0*
— tn

(n-c

9
(S%J
C,,
0
 IC) 0 IC) 0 0 IC) 0
0.
U
 L) L)
0.
U
LC 0

c:J
LC) 0 0 9
I
 LC) 0
0 c.J
 U)

0
=
rj
U)

0
c ou,
OLL) c
.0 ccj
'I
I
0 U)
(0(0
O9
I.
U)
(1)
C4 0
<4. (0
I-.
I
3-
/
/

0)
(0
0 0
=

Ip
o.
0 9


(Ii

—1'
ocr, ocD..
U)

0 - 0
-(0
Cr., I
S

rj4
U)
C
I
(i
I

U) (I)
0 0
-N. .—
a —.,
c.,1

U) -
U) /
/
• L.
F I
I
•

1
I
0 0
Go -Go

Ooo
,-
F (1)
Oco
O [Link]) '-

I
ii• . 0-
p LL

U'-99-
-j

Oçç
J0L()
I
I

t 'I

•0 •0
C) C'4'
/
•cZ<
p

I -- I' - S I
I
C)
I ' p
I
1
1

51 I
V

___ I ___ I

0 LC) CD
4.)
 -o
£

OcD
f-f)

o- 0•
-(0
o w
CV)
/ =
I

/
S•
• I
/
. •o •
I 0£ I,

• I •.;:; ni
- :3w- •

L.

o V)—
•- U)

C1 0 f-s)
)
-
I :3 — •
•
/ ;J) •

/ U)
F

I a. • (

-S
.5
0 0
- 'I
C) —

/ 0
S.-.

- S
1 •
1

0C')

090 S a00
- '• •

__J (54
-
0 I
-J
0 0 0
(_) C)
Ooo I

(54'-

C4 1i, - II

ji
I
t
)

- II I

—I
o I

I =(

I I
(31. 0 19
9 (JQ 9
 =

C
rj

U,

c4, 0
cj V)

U, I
-o
C
0

-o
(1, (n
0

-C o
N. \
ci)
U,

1
l,
L.
C?, y,
'1
0'
GD'
-1•
LL

o
()O0 0)

I I

C•s1
=

-ö

L) 0 L o tp
c-f-
U,
a) a)
0
I 0
I

.>

(.0
cy)
c1
0 a)
I 0
I
> .1
nj
1

Ia)
U)
UIC
(0

0)
G) (7)

0 0
(V)
I
>


C,,

.–J

C (
0-C
C4
U,
, e,
• c ci
-L-w

•9-
0 C
..-,
.7(1
( (I
I
0 I

•.;; I'
—

cD
9-

It -4•
-
U OC-.

'- ._ ,_
J C_,

C" I I•
.'i I I I
I
,

IL
(V.._ I
I
I
I
I

I I
o LL)
0 LC) 0 .Lf) 0 L

U c .J ,. I-


ocr,'f

N' Cl,

I
j
IC)

0
/
(0

0 - N •
'I •6 ci
C-, Lc?tP:t
iii
'I'I
I' •
4-'

• I'
'I
.

I_n , I, - a,
_J (OLC) 0 •1'
U I

/ /
• 0-cr
o •) I I I
I.- 1
_ ii
I, ft - .- -
I • II
'I
U)
°
•1
°I H .
U),
1 '* Lou' • I
I I.'
.I iH • 'I I I I
=1
.1 •'I I i •iH1
0
-•

,?LC)
.-
/ ___

,, ,,

/
0
'-4
/---
LL

(1"
' 1 ' I
IC) 0
9 C)
0. 0 If) C)


ocD

o
w
• 0
(0
(Y) C') C')
1.
= I-
I. • I

ii -
I. .1
:1 Ct,
I )
• I I
/ I
II I I
I. I I
if
o v) 0 0
'N-

0
I.

'1 M- I
/
-CI /
/
•
I! 0 /
— ,,I 4.
/ )

,w
4.
I
I
o 0 0
Ot

4 'I
( (;
I.
\

• e3)
1
I)
\ t

•0 .0 •o
0) 0) 1

'—I •
CD
I-
=1 • I
=1

I I c91
C) —I

-I 01
=1
'I

IL
'I
1;
I,J <I -(
I)
I
0 L) 9 (_)LC) C L)
f.
T
 a
0)

(ti
Q

Cr)

-- c
— —

1)
0
-OQJ N.
'.
—
-,

L
I

I S

. I
(I.

)
o C'
1

I-, i'
I I I'
II J o 0
I_I =
#1
'.
IC)
4-.
r rØ
4 —
o 0

c4 CsJ

I,
I,
I,

"
St

LO 0 L 0 1) 0
I-


(.4

I-
0
OcD

0
1'
c C%
?5 1(
I' I I
li
I,
/ // 1%
I, • IN
*1) _ (/
( '
it
I'
I
I If
i-. f- - II
I . :
11 S
c_fl

L.
N- 5 I
c•.4
'S I'

\ 'I

1,
1,
J,
— //
/ ,I/•
/
i 0

ii
I •1
'1/
K
I

8
\1
I

0 0
I?
Il 0
=
I /
I.
< I Il___ -( 'p
I
0 LC) O0 c_C)
I o
 >
rj
c
0
•
1-
o rtl
'-
(I)
—

.'-
t; - - -
C -

o rtl
0 - - - /
cr ,
S

--

.4
.4
t .4

1 •4 W
S () Nm-

C, F
/
/
cY,

0 (4

U c-
'.-
ft. ,-

4.
>4
.4-

0 (Ii
C) . C
4- 0
rj

o —
0 (I)

C4 ('J

I i

I I

LC) L o L 10
(_) (\


(
C

(fr Q 9
cz J I-
L)
p
0 LU
-c

0. 0 U) U) 0
o . I
 >

C
0

4-,
U)

(ti
U,
C
0

4-,

4-. 0
U)
C

U,
(I,
L
0

cJ
cv,

U-

0 9
Lfl 0 9
 cI.
[0

(0

- ---.- I-
- - - -

cp

I I -

(N

o •:tt:

0 L-

(_) .,' 0
0
9- C o r
o
(N •.;:; I-
C4 C4

(I-)

, -0
cy) fi C C [ci
> '-I-,

IL


(n.Q
C.,-,
orTj
-I.-, j) >

- 0
I C CJ 0
- . 'I
(I,
ci-
I'
Ii
1... i
OC)
U)
ci, ,
I- _6 0
= =
-p --- - - - - -

(1 C)
U)
ci)

0
0
U

-J
Cd
ci)

cJ

LL
 20
19
18
C
17
a16
C-,
C
15
Cl)
14
13
12
11
10
9
8
7
6
5
4
3
2
1

0 12345676 9101112131415
Distance in

Fig.(3-26) Total Head Contours at Plane 1.

17 in from wheel centres
 20
19
18
17
a)
o 16
C.
LI)

14

13
12
11

10
9
8
7
6
5
4
3
2
1

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Distance in

Fig.(3-27) Total Head Contours at Plane 2.

57 in from wheel centre.
• 20
19
18
C
•— 17
a.' 16

eel
14

12
ii
10
.9
6
7
6
5
4
3
2
1

0 1234567 89101112131415
Distance in

Fig.(3-28) Total Head Contours at Plane 3.

85 in from wheel centre.
 2Q

.E 18
17
C
(U
16
15
a
14
13
12
11
ip
9

7
6
5
4
3
2
1

0 1234567 89101112131415
Distance in

Fig.(3-29) Total Head Contours at Plane 4.

14.2 in from wheel centre

(i) Side View, Rotating (ii) Side View, Stationary

(iii) lop View,Rotating (iv) Top View, Stationary

FIG. (330) Smoke Visual isation of Flow past Wheel

 EE

(I) Top of Wheel ,Rotating (ii)Top of Whee(,Stationary

(ii i) Behind lop , Rotating (iv) Behind , Stationary

FIG.(3-31) Wool Tufts Showing Flow over Wheel

 U)
,-. >
'-. 0
0
L.

Co
.2
4-. o L° tO
099
L 0
4-'

_j .-

C4
0.0
I
0

IC) 0 0 0 L() 0
p. 0 IC) 0 U) 0 Lf)
9
cz4 I


C4
U

-c

0
°cDI
eI,
to
C IC')
-j

L.

U
0
-c
U) .10
L
, _3 0 -
.rJ
c to cP

C C (_) i.-.
,- ,- I-.

C
0

-Q >— 00 0
I- I- •

U)
I I
C :$

1.

C')
U)

I-
0

LC)

LL

• 11

I I.
I
In

L()
9 1 0 In 9
I-
4 1-

oo_.
 -J

p.
C —
.4-
00

,) c_fl
_C )
-'-'
':' rj
C>
o-
C

N
•
-
.t
to c'C))
, —j u, ci v

_p__I

rn L.J O(DC)

(I) C..J cU —QO

c0

- c,)
L.

—'-C
f_fl

(0 -C')'
L. h... > lo C C')

C4

(0
(V) I.
• i,J

'11

C',
E —
0 >
x /
a-
U)
Co I
LI,
C
,c •-

— 11)
'Ti _Cfl.
-x -CI
U J)J:2
o b
I I
(1)1
I (1)1
I Of
L.I
UI
rTjI
CII
I <I I
Oi
•.;::It
ctil 1 Q)
—I, (Ti
(f)I
C')
>.

—j

If
ô
>

0
\c-)
e tC) o\
- - CII
- - I
ci
'C 0 .21
CtI
--I
C I_I
.
v
I
c
— ' LI)
C —
oD
•;;; Ci •— a
0 C"
"
C LI <ncr I cv,I
N.
E<i CI
)f(c) (5
—
did cr1 'IL
C
(C
— QIL)
IL j— I—

0 0)


1/,

rcj

C
C

I.
I

o o.9 9 9
Lf) Lii 0 Lii Lii

Cs.J I

(i) Side View,Rotating (ii) Side View ,Stationary

(iii) View of Front,Rotating (iv) View of Front, Stationary

F!G.(3-40) SmoIVisua1isation of 'Two-Dimensional

Flow past Wheel
(i) Cylinder Rotating as Wheel, V/11

V
(ii) Cylinder Stationary, "u? 0

(iii) Cylinder Rotating Opposite to Wheel , v/U_1

FIG.(3-41) Smoke Visualisation of Two Dimensional

Flow past Cylinder, Re = 2 x 10'
FIG.(342) Two Dimension&, Mean [Link]

(I) Stationary (Mirror Image)

(ii) Rotating (Mirror Image)

-
( jj ) Rotating (Moving FLoor)
 C'l CD LC) .T C') c
C') - 0 0)
 a)
(Y) a)
U,
x
0 XL
a
a 00
(Tjrj
cDLr

>

0
O

a)
cLL-) cJ

0
E I-
a
a
C-
C', 4-.
a,C
a)
>

C
U
r'j

0
°Lt)
0
(Y)
E
C',
a)
1
0—i
a
a
C',
a)
a)

>
Li
 FIG.(4-3) Kuchemann's Modet of Separation

(I) TypicaL Turbutent Boundary
Layer Profile

Constant Vorticity
Model Profile
z. §

•1 x
I 1/

(ii')
 lo

-c

- 0

•-• 0
w-; •4>
(11 '

01 U)
r >-
- t- ,
(fl(I:5 >
-0 U)
%- L C
• 0
- e 0 L
LLZDJF

>__
+

L
t.
.cJ

0'

>
C FIG.(4-5) Variation in Lift Coefficient With
L
+1 ight Above Ground

(Potential Solution at g, = 0

0 'S CL=+45)
'h o. -

o 4
K0

K=-025

-1
068

-2
d=2ct

-3
U00
-

Potential Solution
--°---- Wadcock (1970)

-4


o 10 U,
O LcD
C
-a U)
>
a a 0 0I
-_.
- ----- >
a 0
u)
I
-
a. a £
I a 3: 0
I a
o S .2
0 LI.. (t$
/ U)
o (I,
'4-
I -
—
C D 0
I -
o --
o U,
- - 0
-

--- I
>
U,
--
- I 0 C
0
0,0Lç) -o
L
4-
1... '4-

U, 0
a..- 0
r
- T (U

u-I
U)

I.-
I
0

#
-
:
±;: 0

(0

a.—
- -

() LC) (0
0
I I I I I I


U,

11)

'%._•% I
0
's.;
L 0.,) LL.

I
r1
I—
C
L
I-
0
I- a-

o
C- .("4
- u,-cZ4
I

(I_i
U)
_
I I
c_fl

—
a
a >1 [Link]
'- a - .1-, U

1
I

o
I I
I I

r: :i :
(4 (0
I I I I
c

C
C

£
L
.4-
(I
C

U
(I

F
.4
C

.4
C
C

0•
cx

Cl,
0
I I
I
 ._t I

U,

0
U -D
0

Ct,

0
IL

'1

.0
E

U,

0
c
>

3:
0
-I

C,

-o
0
IL
U-'
0
 FIG.(4-1O) Results for Flow in Corner
With a=30°

.5

-.5

-1'O

u,,u

•05

.-sQ5
 U,

I-

V
C

I O.. 0)
£L .
>
II 0 a

-c

a, C
LL

•1-'
L

II
0
I—I
U-

L) L)
0 0 L 0
U C'4 I
d

W-Ptane

________D
A B 5'

crPLane
B

D
AD
B
S/

rLP(ane
A

5' B 5 D

FIG.(4-12) Kirchoff Theory Transform Planes

 Z- Ptan
D

D'
W-P[ane

0 S -D

A B SI

çP[ane

-i 1-
.35 D

-PEar

S DD' S'

FIG.(413) Levi-Civita Transform Planes

 t
S

FIG.(4-14) 0 -Plane For Curved Obstacles

1Fjane

S S
1I2TC

SI S

'I

•11

.-III.

FIG.(415) Transform Planes For Wood's

Tftey
 'I
)

Z-Ptane

FIG.(416) Parkinson & Jandp[j

Transform Ptares

- 0

- -

FIG.(4-17) Possible Unrealities Of Free
Streamline Solutions
 I

G) Coordinates in Z- PLane

0•

asing

ing

(ii) Lines constant =

FIG.(4-18) Coaxat Coordinates

 Lone D

x.

B B B

22PE

i)

FIG,(419) Transforms To Z2-Ptane



t_. -
C
ed,

00

EO
0
Cr)
H

a
a) cij
( C
4'
-)

El y v
01 C

C)

I,
IL

9 L) 0
I- I I •

00 • E
'I) >
H 0
a
a) p—F-
•V) .0
I i
.. I H
0
a

cv,
cv?
II
a

9 1) 0
111111111±^r
IC) 9 IC) C) 10
() I 7


•'0

x.
LJJ
LC) [Link].Q

cOO o_ -
("C
ci
o OI

'—a'
"-0
0_
Lfl a) (1, rn
V)V)
0
'I) (no
o ci)
I I (ii
•U, >>
L. L
00
OIL
°c c
o
Do
I
()
I I
I I

IL

•s\S. %,
•.s\

4'
I

'I,
_0•
(
—.

,. 0 L) 0 0 L) 0
4—.
,- I I I
I-
 >

I',

a..'
.
0•. o

nj a
a.'
— LLLI
COc,o (flJ

Lc. '.0
c±c

'4- >

0
C)
• I-_ k..
. 0 0 (ti 'J)I
_) a, Cl
EI
cI
I I
0
0
I
I
I
I •
C\J
C•s4
(r)

I i I 0'•
'-'4
• IL.

o
C.',
Cr)
%%
%% %%

%%.'• ••

•
"4%

0
o
C',.
I
/
_6•
.,
r -- - -- -
.-
0 I

• I

9 L) 0 L) 0 L p
I '- T C"sJ
 >

ba)0
o)
0)
0)0)

U-

• C
.0
o•

Ct,
a
'.- -

U,
ci)
LC)
a (0

-c

> C)
II

a
ci,
:

C,
C >

0.
(_)
9
I-
1
9
T
C
0
tn
0

C)

0
4-,
0
E
>
If)

CI,,
Iii

0-u,
a,