(Swanson 1961) Swanson Magnus Effect Review Paper
(Swanson 1961) Swanson Magnus Effect Review Paper
(Swanson 1961) Swanson Magnus Effect Review Paper
SWANSON
Associote Professor o f M e c h a n i c a l Engineering,
W a s h i n g t o n U n i v e r s i t y , St. Louis, M o .
Assoc. M e m . A S M E
Introduction
cover so many different geometr'es that they are only briefly con
sidered here.
M a n u s c r i p t received at A S M E
Historical Notes
The first record of the drift deviation of a spinning body was
described b y G. T . Walker in 1671. The body was a "sliced"
tennis ball. G. Magnus [l], 1 in faying to account for the drift of
spinning projectiles, performed some crude experiments with
musket balls and with a cylinder in an air jet. He correctly
ascribed the drift to an aerodynamic force produced by the
interaction between the rotation and flight velocity which gave
rise to an unsymmetrical pressure distribution produced by the
Bernoulli effect. His projectile drift experiments were performed
b y laying musket balls with their center of mass not coincident
with the geometric center in a musket with their center of mass
in a known orientation. If the center of mass were to the
right, an impulsive clockwise rotation about a vertical axis would
be produced b y the center of pressure acting through the geometric center and the inertial reaction acting through the center
of mass. When fired, the ball should drift to the rightit did.
Lord Rayleigh was the first to set up the ideal flow representation. In 1877 he published a paper "On the Irregular Flight of a
Tennis Ball" [2] in which he presented the mathematical model
of the flow as the classical potential flow around a cylinder with
circulation. A t the time he also stated that it was not possible
to give a complete mathematical formulation of the actual physi1
.Nomenclature,
a
L = lift
p =
- pressure coefficient =
drag
i = imaginary quantity
1
K = a factor relating velocity ratio and circulation, Ka
1
a
[/
jU =
pressure
aUw
SEPTEMBER
19 6 1 /
461
cal process since no mathematical methods were available to express the manner in which friction between the fluid and the
rotating cylinder would produce circulation.
The first quantitative data on record were determined by Lafay
in Paris between 1910 and 1912 [3, 4]. He also obtained negative
lift forces at low rotational speeds.
Many attempts have been made to use the high lift forces o b tainable on a spinning cylinder in the wind, or in an air stream,
but none has ever led to a financially successful venture. The
most notable attempt was made by Anton Flettner in Germany
during the period after the First World War. Flettner consulted
with Prandtl and the Gottingen group on the idea of replacing a
ship's sails with rotors. In a cross wind, the Magnus effect would
produce a thrust many times that for equivalent sail area. It
was, of course, necessary to drive the rotors, but the power was
only a small fraction of the power required for screw propulsion.
Ackeret and Busemann [5, 6, 7] conducted a series of tests on
cylinders with and without end plates which indicated that the
method was feasible. T w o Flettner Rotor ships, the Buckau and
the Barbara [8, 9, 10, 11) were built and one sailed across the
Atlantic. Although this method of propulsion was quite inexpensive, the speed and reliability of screw propulsion was more
than competitive.
The Madaras Rotor Power Plant, patented in 1926, was proposed as a rather complicated method of generating power on a
large scale using large vertical spinning rotors propelling themselves around a round or oval track by means of the Magnus
thrust. Generators attached to the wheels generated the power
output.
Attempts have been made to employ spinning cylinders as
portions of high lift wings, but the difficulties of complicated
drives and excessive weight have eliminated this application.
The first Magnus investigation in this country was made by
E. G. Reid in 1924, at the Langley Field N A C A Laboratory [12],
using a single cylinder projecting through both sides of the fiveoot diameter tunnel. The most interesting of these results were
never published.
The most complete experimental work was done by A. Thom
at the University of Glasgow and reported in his doctor's dissertation and in five Reports and Memoranda of the British Aircraft Research Council covering a period of nine years from 1925
to 1935 [13-18]. The effects of Reynolds number, surface condition, aspect ratio, end conditions, etc., were investigated.
Pressure, velocity, and circulation data were also obtained.
A comparison of the Magnus lift force coefficient as a function
of peripheral to free-stream velocity ratio as obtained by several
investigators under a number of conditions is shown in Fig. 1.
One general trend indicated by these data is that due to aspect
ratio: A finite and smaller aspect ratio produces a smaller lift
at a given velocity ratio and also gives a peak lift which has not
been obtained with an infinite aspect-ratio cylinder.
7T
/'d
Reynolds
no.
Aspect
ratio
Curve investigator a n d
ref.
CL v e r s u s
velocity-ratio
Remarks
( I d e a l fluid)
oo
CO
Thom [ 1 8 ]
12.5
and 2 6
5.3 103
8.8
Reid [ 1 2 ]
13.3
3.9 104
1 1.6
(Gottingen) [21]
4.7
5.2 X
104
Thom [ 1 4 ]
104
Swanson
OO
1.6 X
Thom [ 1 7 ]
5.7
3 -
9 X
104
Thom [ 1 7 ]
5.7
3 -
9 X
104
Smooth s u r f a c e
(Gottengen) [37]
4.7
5.2 X
Schwartzenberg
4.5
5.4,18.6 X
Swanson
5 X
X cyl-dia
disks
end
104
104
104
Unpublished (Case
Inst.)
Continuous
sections
end
-NomenclatureX
X
y
Y
velocity ratio, a =
r =
cely
circulation
462 /
Subscripts
v0/U
p = density
<p = stream function
co = angular velocity
SEPTEMBER
196 1
(i.e., at
Superscripts
Transactions of the AS ME
One of the first attempts at obtaining an analytical representation was made by W. G. Bickley [19] in 1928. This is probably
one of the most useful analyses with respect to providing an explanation for some of the observed flow phenomena. Bickley
considered the potential flow resulting from a vortex in the neighborhood of a cylinder with circulation in a free stream.
The most complicated mathematical model was formulated
and worked out by Torsten Gustafson [20] and published in 1933.
In a certain sense it may be considered as an extension of Bickley's method. Instead of a single vortex used to represent the
vorticity shed in the wake, " . . . a flow where vortices of constant
intensity are distributed on the logarithmically deflected streamlines in the wake" is considered. This method is based on
Oseen's methods of approximate analyses of viscous flows as extended by Zeilon. Gustafson's method is quite tedious and requires numerical solution as well as determination of several
parameters from experimental data.
A recent work by E. Krahn [21] considers the very low
Reynolds number region in which no wake is formed. A laminar
boundary-layer analysis produces the circulation as a function of
rotational speed. The method is carried through only for the
case where the t wo stagnation points coincide at the bottom of
the cylinder.
The foregoing investigations have been carried out primarily on
the case of a cylinder normal to the flow direction. The Magnus
force and tumbling moment produced on spinning projectiles
and missiles at small yaw angles are also of interest and have been
the subject of much investigation [22, 23, 24, 25, 26, 27, 28, 29].
The Magnus effect is one of the primary factors contributing to
the drift and dispersion of spinning missiles.
More recently Glauert [30, 31, 32] and Moore [33] have made
some analyses for the small perturbation case of low peripheral to
free-stream velocity ratio.
As previously mentioned, Lafay was the first to make quantitative measurements of Magnus forces. He also made measurements of the pressure distribution around the rotating cylinder.
The extensive data gathered by Thom for several end conditions
including various end shapes and combinations of end disks give
a variety of data for many basic shapes for the case of 90-deg
yaw. These results along with those of other investigators are of
primary interest in indicating the effect of finite aspcct ratio.
The smaller the aspect ratio, the smaller the maximum lift obtained and the smaller is the velocity ratio at which this maximum
is reached. Leakage flow and consequent pressure equalization
around the ends of the cylinder is responsible for this aspect-ratio
effect. The best approximation to two-dimensional flow (i.e.,
independent of end effects) is obtained by extending the cylinder
through the tunnel walls with a very small clearance. The only
investigation using such an apparatus seems to have been made by
Reid [12] in 1924. Many attempts have been made to approach
two-dimensional or infinite cylinder end conditions by adding end
disks. End disks, however, give entirely different flow conditions
from those of an infinite aspect-ratio cylinder so that the flow
pattern will be one due to a finite cylinder plus that due to the end
disks resulting in secondary axial flows and end flows around
the disks.
Under such circumstances, no combination of disks
on a finite cylinder woidd be expected to produce conditions
similar to those for ail infinite cylinder.
The variation of the lift and drag at low Reynolds numbers and
velocity ratios is also interesting. The variation in behavior with
Reynolds number is attributed to the nonsymmetric boundarylayer separation from the top and bottom surfaces of the cylinder
as explained by Krahn [21] and Swanson [34]. As one example,
consider curve g of Fig. 3 for which the Reynolds number is subcritical (with respect to laminar-turbulent boundary-layer transition) at 1.52 X 105.
At low cylinder to free-stream velocity
ratios, a, between 0 and 0.1, both the upper and lower boundary
layers remain fully laminar up to their separation points. The
upper separation point moves rearward (in the direction of the free
stream) and the lower separation point moves forward. The additional length of boundary attachment on the top surface gives a
greater region of negative pressure coefficient here than over the
bottom where the length of attached boundary layer has decreased. As a consequence of the pressure distribution, a positive
lift is produced.
SEPTEMBER
19 6 1 /
463
/
/
F
a
\p
' A
<
"N.
4
.5
Fig. 2
Group 1
Group 2
Group
SEPTEMBER
1 9 6 1
.8
.6
.9
1.0
Lift c o e f f i c i e n t Cl v e r s u s v e l o c i t y r a t i o a
Curve designation
464
"""
Reynolds
3.58
4.9
6.07
7.91
9.9
12.8
15.2
18.15
22.5
26
29.5
32.5
36.5
no.
104
42
45
50.1
Transactions of the AS ME
a
F i g . 3 D r a g c o e f f i c i e n t Co
R e y n o l d s n u m b e r s as Fig. 2 .
v e r s u s v e l o c i t y r a t i o a.
Letter c u r v e d e s i g n a t i o n s
imply
same
Experimental
V,
(Eouation ( 7 )
\ x a = 2.89, c/a = 3.51, T = 242.1
(Eouation ( 7 )
( K a = 2 . 8 9 , c/a = 4 , y 224
>
/
/
0
7
/
\
j
i
\
Fit;. 4
\
9
10
11
Lift c o e f f i c i e n t C i v e r s u s v e l o c i t y r a t i o
12
15
14
15
16
17
a.
I
1/
//
V
N
f ,
ii / I
/
i
i t
/J
w
/
Experimental
(Equation ( 7 )
.
l K a = 2.89, c / a - 3 . 5 1 , Y = 242.1
(Eouation ( 7 )
U 0 - 2 . 8 9 , c/a = 4, r = 224
./
0
3
Fig. 5
10
11
12
13
14
SEPTEMBER
I 96I
/ 465
5
B x
10"
F i g . 6 D r a g c o e f f i c i e n t C o v e r s u s R e y n o l d s n u m b e r R p l o t t e d o n a l i n e a r s c a l e . T h e letters
a, c, e, e t c . , c o r r e s p o n d t o t h e f r e e - s t r e a m R e y n o l d s n u m b e r s f o r t h e c u r v e s s h o w n i n F i g . 2 .
velocity over the top of the cylinder is decreased, but it is increased over the lower half. It is useful to introduce here the concept of a "relative Reynolds number." Where the surface is
traveling with the free-stream velocity, the transition can be expected to be delayed, and conversely. The relative velocity R r o i
= R ( l a) ( + sign for bottom and sign for top). The
boundary-layer behavior can then be correlated with the relative
Reynolds number and a qualitative estimation of the relative portions of laminar and turbulent boundary layer and total length
of boundary layer can be obtained. Referring back to curve g
of Fig. 3, for a increasing from 0.1, the upper relative Reynolds
number decreases so that conditions in the upper boundary layer
can be considered to be similar to those for Reynolds numbers
less than 1.5 X 10E on Fig. 6. The lower boundary-layer conditions correspond to Reynolds numbers greater than 1.5 X 105.
For example, for a = 0.2, the upper relative Reynolds number
would be 1.5 X 10s (0.8) = 1.2 X 106 and the lower would be 1.8 X
106. Fig. 6 then indicates a longer attached boundary layer on the
bottom than on the top. This effect counteracts that previously
mentioned and the lift begins to decrease. A t higher rotational
speeds, a greater portion of the lower boundary layer becomes
turbulent and therefore more reattached. The greater reattachment on the lower surface produces the negative lift effect shown
and. also results in a lower drag as shown in Fig. 3. The lower
boundary layer will finally reach a fully developed turbulent state
near the point of maximum negative C L .
At this point it is helpful to introduce the concept of a "boundary-layer origin.'' The positions of the boundary-layer instability
and separation points are a function of a boundary-layer length
Reynolds number. For a stationary surface such as a flat plate,
airfoil, or nonrotating cylinder, the boundary-layer length is
measured from the front stagnation point. The shear (or rotation) in the boundary layer has opposite direction (or is of opposite sign) on opposite sides of the body from this point (Fig. 7).
It seems logical to define the beginning of the boundary layer, or
a boundary-layer origin, with respect to such a condition if we
accept the definition and idea of a boundary layer as a shear layer.
For a stationary surface, this boundary-layer origin is coincident
with the stagnation point; however, for a moving surface this
boundary-layer origin will, in general, not coincide with the stagnation point. On the rotating cylinder, the forward stagnation
point is translated in a direction opposite to the direction of rotation. (In addition, it is interesting to note that the stagnation
point as a point of zero velocity can no longer lie on the surface
466 /
SEPTEMBER
of the body.) If the boundary or shear layers are now investigated, we find that the upper and lower layers conforming to the
required conditions for a boundary layer begin at a point where
the cylinder surface velocity is equal to the fluid velocity (i.e.,
where the relative velocity is zero). The boundary layers are
then as shown in Fig. 7. It is seen that the boundary-layer
origin is translated in the same direction as the direction of motion of the surface (direction of rotation) but opposite to that of
the stagnation point. Boundary-layer profiles obtained for the
cylinder rotating at two different velocity ratios are presented in
Figs. 8 and 9. The outlines of the boundary layers are shown.
It is interesting to note that, for a = 1, both boundary layers are
about the same length. In order to calculate characteristics of the
boundary layer, it is necessary to have a set of initial conditions.
The previously described boundary-layer origin is the most logical
position at which to attempt to start a boundary-layer solution
for a geometry where the wall is in a nonuniform motion with
respect to the external flow.
As the rotational speed is further increased (beyond a value
corresponding to a ~ 0.5) the boundary-layer origin moves
further back on the upper surface with a consequent increased
length of flow attachment producing more Magnus lift. It is
believed that both boundary layers are in the fully developed
turbulent state at a velocity ratio of 1.0. An idea of the flow
pattern may be obtained by referring to Fig. 8. Although the
relative velocity between the surface and free stream is much
196 1
Transactions of the AS ME
Fig. 8 B o u n d a r y - l a y e r p r o f i l e s o b t a i n e d a t p o s i t i o n s e v e r y 3 0 d e g a r o u n d
the cylinder for a = 1. T h e dotted line represents t h e a p p r o x i m a t e limit
f o r t h e t h r e e b o u n d a r y l a y e r s . T h e v e r t i c a l s c a l e o n t h e p r o f i l e s is t e n
t i m e s t h a t f o r t h e c y l i n d e r . S c a l e s a r e v/toro v e r s u s y / r o .
Fig. 9
S a m e as Fig. 8 ; a =
For velocity ratios greater than one, the force behavior is indicated to be essentially independent of Reynolds number for 2 X
10' < R < 5 X 106. In this region, both boundary layers are
apparently fully developed. Since the surface and fluid velocities
over the top are in the same direction producing a low relative
velocity, it is possible that the upper boundary layer in its fully
developed state at a > 1 is laminar. However, boundary-layer
instability and transition, and the transition from laminar to
turbulent flow, in general, are functions of Reynolds number based
on absolute velocity (considering the Reynolds number and its
consequences to be representative of the ratio of inertial to viscous
forces). Over the upper surface the absolute velocity is quite
high so that, for free-stream Reynolds numbers corresponding to
laminar separation for the static cylinder, it is possible that the
upper boundary layer is turbulent for the case of the rotating
cylinder. One of Brown's photographs for 1 < a < 2 (Fig. 14,
ref. [36]) favors the second argument: The smoke filament on
the upper surface indicates a turbulent boundary layer. Similar
arguments can be presented for the behavior of the lower boundary layer.
The knee of the lift curve near a 3 is produced when the
boundary-layer origin reaches the top of the cylinder. At any a
greater than this value, the cylinder is everywhere traveling at a
velocity greater than that attainable by the external flow; consequently the boundary-layer origin as defined previously no longer
exists. Prandtl considered that the flow pattern produced by the
stagnation point coincidence at the bottom of the cylinder would
produce a flow pattern and force system that would not change
by any further increase in rotational speed. He reasoned that no
further vorticity could be shed; therefore the CL would level off
at a value of 4ir. The foregoing condition where the boundarylayer origin reaches the top may be considered to pose a similar
limitation, except that the CL is seen to continuously increase, but
at a slower rate than before. As the rotational speed is further increased beyond a value corresponding to a = 3, the vorticity shed
in the upper boundary layer goes from positive (clockwise) to
negative (counterclockwise) while that shed in the lower boundary layer is always negative. In a real fluid, the lift and shed
vorticity will always have opposite signs (for the co-ordinate
system orientation shown) and there will be a correspondence between the magnitude of the shed vorticity and the lift. As a
is further increased, the vorticity shed in both boundary layers
is negative and increases in magnitude. The field induced by this
shed vorticity has the effect of rotating the flow pattern around
the cylinder in a counterclockwise direction. (An approximate
calculation of this effect will be given later.) The pressure distribution produced by this pattern is seen to increase the lift. If
this is the primary effect producing the lift rise beyond the knee,
it appears as if some limit should exist as the wake formed by the
separating boundary layers rotates around to a position near the
front of the cylinder. N o such limiting condition was reached
within the limits of data taken.
The drag behavior is also indicated as a consequence of the
foregoing flow behavior. As a increases beyond 1, the drag, surprisingly, increases to a value much greater than the drag on the
nonrotating cylinder, even though the wake area is decreasing.
This large drag increase with increasing a is produced by flow reattachment over the rear of the cylinder accompanied by a movement of the rear stagnation point and the wake in a counterclockwise direction into the region near the bottom of the cylinder.
The drag peaks in the region where the lift knee occurs. The
boundary-layer origin is at the top of the cylinder and the separation points and wake are near the bottom of the cylinder. An increase of a as described produces a further rotation of the
wake toward the front of the cylinder. The resultant flow pattern
and pressure distribution produce a decrease in CD.
SEPTEMBER
19 6 1 /
467
dF
W > - U > - i V > - - - - U
_ t(r +
K)
t
(l
jk /
a2 \
-)
i_\
2ir \ 2 zA
2TZ
z zB )
d<f> -
~dT dz
The first contour integral on the right-hand side of (3) gives the
convective or W2 contribution to the force and the second term
gives the unsteady or time-dependent contribution. The contour
of integration is the cylinder surface, z = a.
After performing the necessary calculations, the lift and drag
coefficients can be determined as:
S T . (' "
- -
idir.
0 ^ 0
( ' - ? )
" ^ f " H
Ctt7- -
i r ) " "
<4
>
( >
vortex"
Tr
a2 \
i(T + k)
in
2 - ZA
(1)
a2
468 /
SEPTEMBER
196 1
(f)
a y
<6'
Transactions of the AS ME
(7)
cos y / a \ ,
- - ^ T (7)
Even considering that this model presents a good approximation to the actual flow condition, the parameters c/a, 7 , and Ka
would be expected to be functions of or; consequently, if these
relationships could be determined, a better " f i t " to the experimental results would result. Unfortunately, there is no way by
which c/a(a), 7 ( a ) , and Ka(a) can be determined.
It is interesting to note that the field induced by the trailing
vorticity has the effect of rotating the entire flow pattern in a
counterclockwise direction about the center of the cylinder (in
the same direction as the lift circulation T). The forward stagnation point is then located closer to the front (6 = 0) of the cylinder and the rear stagnation point is closer to the bottom (6 = 90
deg).
For example, for the values of c/a = 4, 7 = 224 deg, and
Kaa = 2.89 cited in the foregoing example, the stagnation points
for a = 1 are located at 9 deg and at 180 + 21 deg; for a = 2,
at 19 deg and 180 + 50 deg. This predicted effect is in agreement with visual observations by Brown [36] and Gustafson
[20] and measurements by Swanson [34] (Figs. 8, 9). In reality,
the rear stagnation point does not exist because of the wake region
formed. It is of significance that for the range of Reynolds numbers investigated (3 X 104 to 5 X 106) the wake is always present
at all values of a. Therefore, for a real fluid flow at high Reynolds
numbers, the condition where the two stagnation points can
coincide will not exist and a separated wake will always be present.
Indeed, some smoke studies by Gustafson [20] at very large a (5
\
\
/>
/
A
//
Experimental
(Y = 210
/ c/a - '<0
jy = 22'*
1 c/a = 3 . M
\ y = 2'l2.)
References
Fig. 11
L i f t - d r a g p o l a r p l o t w i t h c o m p a r i s o n s f r o m Eq. ( 6 )
1 G. Magnus, " O n the Deflection of a Projectile," (a) "Abhandlung der Akademie der Wissenschaftern," Berlin, Germany, 1852,
English translation in Taylor's Scientific Memoirs, 1853. (b) " P o g gendorffs Annalen der Physik und Chemie," vol. 88, no. 1, 1853.
SEPTEMBER
1 96 1 /
469
470 I
SEPTEMBER
1961
Transactions 01 the AS ME