, Ann. Rev. Fluid Mech. 1986.

18: 367-03
Copyright 1986 by Annual Reviews Inc. All rights reserved
MARINE PROPELLERS
Justin . Kerwin
Department of Ocean Engneering, Massachusetts Institute of Technology,
Cambridge, Massachusetts 021 39
INTRODUCTION
Propellers produce thrust through the production of lift by their rotating
blades. Propeller hydrodynamics is therefore part of the broader feld of
lifing-surface theory, which includes such varied applications as aircraf,
hydrofoil boats, ship rudders, and sailboat keels. Air and water propellers
have much in common from a theoretical point of view, particularly if one's
attention is restricted to air propellers operating at low Mach numbers
(where compressibility efects are negligble) and to water propellers
operating without cavitation. The cross sections of most lifting surfaces are
also similar in appearance, being designed to produce a force at right angles
to their motion through the fuid (lift) with a minimum force parallel to their
direction of motion (drag).
In spite of these fundamental similarities, air and water propellers
generally look very diferent. The reason is that propellers for ships are
limited, for practical reasons, in diameter, and they are also limited by
cavitation in the amount oflif per unit blade area that they can produce. As
a result, marine propellers have blades that are much wider in relation to
their diameter than would be found in aircraft propellers. In addition,
propellers are generally located in close proximity to the stern of a ship.
This choice is based both on consideration of propulsive efciency and on
such practical matters as machinery arrangement and vulnerability to
damage. Since the fow near the stern is nonuniform, an inevitable
consequence is the development of vibratory forces on the propeller blades
and on the hull. Decisions concerning the number of blades and the shape of
the blade outline are infuenced to a great extent by the need to minimize
this excitation.
As an example, Figure 1 shows a photograph of a recently designed
propeller for a seismic exploration vessel. The computational model used in
J0J
00661 89/86/01 15-367$02.00
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Quick links to online content
Further
ANNUAL
REVIEWS
368 KRW
Figure 1 A highly skewe controllable-pitch propeller installed on a seismic exploration
vessel. (Photogaph courtesy of Bird-Johnson Company.)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
PROP8IL8R8 169
Figure 2 Vortex lattic representation of the propeller shown in Figure 1.
its design, which we discuss later, is illustrated in Figure 2. The complex
blade shape is required because this propeller must have very low levels of
vibratory excitation and be completely fiee of cavitation under certain
operating conditions.
The complete feld of marine propeller hydrodynamics is far too broad to
cover adequately in a single paper. In this review we restrict our attention to
single-unit propulsors, as illustrated in Figure 1. Multicomponent pro
pulsors consisting of pairs of counterrotating propellers, combinations of
rotors and stators, or propellers combined with fxed or rotating shrouds
are all of current interest but are not covered here. Propeller cavitation is an
extensive feld of its own, which we also do not cover except as a motivation
for determining accurate pressure distributions on the blades. However, the
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
37 KRW
reader should be aware that computational techniques for noncavitating
fows, which we do describe, have been extended to the case of cavitating
fows. Recent work in this particular area is reviewed in Van Houten et al.
(1983).
Another important aspect of propeller hydrodynamics that we do not
cover here is the interaction of the pressure feld of the propeller with the
hull. The published literature in this feld is extensive, and the interested
reader might possibly start with publications by Breslin et al. (I982),Vorus
(1976), and Vorus et al. (1978).
In this review we frst discuss the onset fow to the propeller, which must
be known before one can proceed with the solution of the propeller
problem. We then formulate briefy the problem of the fow around a
propeller in general terms, at which point we look specifcally at the
problems of designing a propeller for a gven distribution oflif, analyzing a
given propeller both in circumferentially uniform fow and in the unsteady
fow resulting from a nonuniform onset feld.
THE PROPELLER ONSET FLOW
Except under artifcially contrived laboratory conditions, propellers
operate in a fow feld infuenced by the presence of the ship, where the
boundaries of the fuid domain may include nearby portions ofthe hull, the
free surface, and appendages such as the rudder. The coupling between
the propeller and hull fow is generally considered sufciently weak to prmit
separation of the two problems. Thus, the propeller is assumed to be
operating in an unbounded fuid, but in the spatially varying fow feld
generated by the ship. This fow feld can be represented as a combination of
the potential fow of the hull moving in the free surface and of the wake fow
containing the residue of the hull boundary layer. For most ships, the
deviation from free stream of the fow entering the propeller is largely due to
the viscous wake. This wake fow can in some cases be extremely
complicated. Figure 3 shows equivelocity contours of the longtudinal wake
feld for a supertanker. In this example, a point near the tip of a blade will be
subjected to an onset fow varying between 5 to 85% of the ship speed
during each revolution.
In addition to the hull infuence on the propeller, the propeller induces a
pressure feld on the hull whose mean component is termed "thrust
deduction." The oscillatory component of the propeller-induced hull
pressure, although small compared with the mean thrust, is nevertheless
extremely critical from the point of view of hull vibration. Again if weak
coupling is assumed between the hull and propeller fows, propeller
induced forces acting on the hull can be found by solving the problem of the
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MAI PROPELLERS 371
hull alone, in the presence of the free-space fow feld generated by the
propeller. The assumption that the hull and propeller fows can be
separated in this way breaks down if the presence of the propeller
signifcantly alters the fow around the hull. For example, if the regon of
boundary-layer separation on the hull were changed by the propeller's
induced fow, a large modifcation of the infow to the propeller could result.
Fortunately, ships are generally designed to avoid fow separation as much
as possible, so that the infuence of the propeller on the viscous fow around
the hull can generally be ignored.
However, the wake fow in which the propeller is operating contains
vorticity generated in the hull boundary layer, and the fow feld of the
propeller interacts with this rotational fow. The result is that the fow into
the propeller is not the same as it would be Hthe propeller were not there
and is altered by the presence of the propeller. This altered fow feld is
Figure 3 Equivelocity contours of the longitudinal component of the wake feld of a
supertanker. The circle indicates the axis of rotation of the propeller, while the line in the
. upper-right comer shows a portion of the hull surface. The numbers indicate the velocity
defcit as a fraction of ship speed. From Holden et al. (1974).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
372 KRWN
termed the "efective wake," as contrasted to the "nominal wake" which
exists in the absence of the propeller. This complex interaction results in
changes to the propeller onset fow that are functions of both space and
time. The temporal variation is due to the fact that the propeller's induced
velocity feld is a function of time in a fxed coordinate system containing
the wake as a result of both the rotation of the propeller and its unsteady
loading.
Brockett (1985) suggests that the efective wake be defned as the total
velocity at any point in the fuid with a propeller operating minus the
potential component of the propeller-induced velocity. With this defnition,
the propeller problem is reduced to one of fnding (a)the velocity potential
in an unbounded fuid for a fow that satisfes the kinematic boundary
condition on the propeller surface and (b) a kinematic and dynamic
boundary condition at the trailing edge and on the trailing vortex sheets
behind the blades. Since the kinematic boundary condition involves the
efective onset fow, the propeller problem has not been separated at all,
except in the sense that one can hope to iterate to a converged solution or
possibly settle for a simple approximation to the efective wake. We now
consider these two possibilities.
Several decades passed afer the experimental discovery that the efective
wake and the nominal wake are diferent before any attempt was made to
develop a theoretical explanation. It is customary in developing a major
ship design to test a model together with a propeller in a towing tank. One
of the quantities determined is the thrust identity wake fraction, which is
defned as
(1)
where V is the speed of the ship and VA is the speed in uniform fow at which
the propeller would produce the same thrust as that measured behind the
model. The value of VA determined in this way is generally greater than the
average value of the velocity measured behind the model in the position of
the propeller. If one wishes to use the measured velocity distribution in the
design of the propeller, it would frst have to be scaled in order that its mean
value be equal to VA' If this were not done, the propeller design would prove
to be incorrect. This means that a propulsion experiment must be
performed in order to obtain the information needed for the propeller
design. However, it has been found that the propeller used in this test need
not correspond to the fnal design. A stock propeller of roughly the same
characteristics will yield essentially the same value of the thrust identity
wake.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MU PROP8LL8R8 171
Several theoretical treatments of the efective-wake phenomenon have
appeared in recent years
.
These include contributions by Goodman (1979),
Huang &Groves (1980), Dyne (1980), and R. J. Van Houten (see Breslin et
al. 1982). In addition, the introduction of the laser-Doppler velocimeter has
made it possible to measure fow felds just ahead of an operating propeller,
from which the detailed structure of the efective wake can be derived.
The basic idea is presented in Figure 4, taken from Huang & Groves
(1980), who treat the case of a propeller operating behind an axisymmetric
T
e 8
1.6
1.4
1.2
1.0
Rp
0
.
8
0.6
0.4
0.2
0.0
0.3
/
/
i
/

" 0.65._

4.
1.07
| '\ U
"X
-.mom|oa|
V

g
q EffatiMa
/ ,
V
.
.
/
. /
( 1eta|

" 0.
3
70
.
.
,
.
"
.

4.
"
T
F
^ `
_
.
U
.
.. . .. . 1ota| _
v. V > _
ttaa ~

_--~
.
0.5 0.6 0.7 0.8
0.9
1.0 1.1
Figure 4 Typical total and efective aa velocity profles computed from the measured
nominal axial velocity. Results are given for two diferent values of nondimensional thrust
coefcient CT, corresponding to two values of the advance coefcient J
Y
o These coe
f
cients are
defned as follows:
where T is the propeller thrust, V the ship speed, p the fuid mass density, Rp the propeller
radius, and H the number of propeller revolutions per unit time. From Huang & Groves (1980).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
374 KRWIN
body. The axial component of the infow velocity in this typical example
ranges from about 30% of free stream at the hub to 80% at the tip. If we
assume that axial velocity gradients are small compared with radial
gradients, the fow feld can be represented by a volume distribution of
circumferential vorticity whose strength, in the absence of the propeller, is
independent of the axial coordinate.
The propeller induces an axial velocity feld that can be considered as a
frst approximation to be the diference between the total velocity and the
infow. This propeller-induced fow accelerates the fuid as it approaches
the propeller, and as a result, the circumferential vorticity that it contains is
stretched. This reduces the magnitude of the radial velocity gradient in the
infow. Huang &Groves (1980) fnd this by solving the vorticity equation
with the boundary condition that the fow at large radii is unchanged. The
result is the efective wake, which is also shown in Figure 4. It is evident from
this result that the mean onset fow is increased, and also that its
distribution over the radius is now diferent. A simple scaling of the nominal
wake to produce the correct mean is clearly insufcient. Dyne (1980) cites
experimental evidence that propellers designed on the basis of a scaled
nominal wake have too small a value of blade angle at the inner radii, which
is consistent with the trend shown in Figure 4.
Propeller onset fows are generally not axisymmetric, and thus one must
deal with a much more complex problem in which circumferential gradients
are present. In addition, the alteration of the infow by the propeller varies
over the axial extent of the propeller, so that an efective wake is three
dimensional in nature, even if the nominal wake can be regarded as
independent of longitudinal position. This is an active feld of research at
present, as is evident from the Report of the Propeller Committee (1984) of
the Seventeenth International Towing Tank Conference. An example of
analytical progress in this area is the current work of Brockett (1985).
FORMULA TION OF THE PROPELLER
POTENTIAL-FLOW PROBLEM
As shown in Figure 5, we consider a propeller consisting of K identical,
symmetrically arranged blades attached to a hub that is rotating at
constant angular velocity about the x-axis. The hub is either idealized as
an axisymmetric body as shown or ignored completely. The geometry of the
blades and hub is prescribed in a Cartesian coordinate system rotating with
the propeller. The y-axis is chosen to pass through the midchord of the root
section of one blade, which we designate the key blade. The z-axis completes
the right-handed system. An equivalent cylindrical coordinate system in
which is the radial coordinate and 0 " 0 on the y-axis is also used here.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
H8 PROP8LL8R8 375
The blade is formed starting with a midchord line defned parametrically
by the radial distribution of skew angle 0_(r)and rake x_(r).By advancing a
distance _|c(r)along a helix of pitch angle _(r), one obtains the blade
leading edge and trailing edge, respectively, and the surface formed by the
helical lines at each radius form the reference upon which the actual blade
sections can be built. These sections can be defned in standard airfoil terms
by a chordwise distribution of camber ](s)and thickness t(s),where sis a
curvilinear coordinate along the helix.
The propeller is operating in an unbounded, incompressible fuid, in a
prescribed efective onset fow, as described in the preceding section
.
This
fow is defned in a fxed coordinate system in which the and Xo axes are
identical, and the y and Yo axes are coincident at time t .If we ignore the
variation of the efective onset fow, both with respect to time and
longitudinal position XQ, and make use of the cyclic nature of the fow, we
can write down the velocity components in the following generally accepted
Figure 5 Propeller blade geometry.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
376 KRWN
form:

'
L Ax.r.t)(r) cos nOo
+
L Bx.r.t)(r) sin nOo.
hFM hF
(2)
Transformation from the fxed to the rotating coordinate system simply
involves replacing 00 with 0-wt in the argument of the trigonometric
functions, thus introducing a periodic time dependency in the fow.
The governing equations for the velocity potential representing this fow
are well known, since they are the same for any incompressible fow around
a three-dimensional lifting body. The velocity potential at any point on the
surface of the body can be expressed in terms of a surface integral over the
body and wake using Green's formula:
..
0 1 o.(,) 1
"
.,
on . , . ,
q
o

where
bb_
+
. .

.,
n . ,
q
o

,
w
feld point where the velocity potential is to be calculated,
, source point where the source or normal dipole is located,
. distance between points (x,y,t)and , q,()
(RJ(X_)2+(Y_1)2+(Z_()2),
normal velocity at the body boundary
V. a 1_'a,,
a unit normal vector outward from the body,
U
'
" velocity vector of the undisturbed onset fow,
, , integral over the body surface o,excluding an infnitesimal
b^b_
region ocontaining the singular point p,
, , " integral over the wake surface,
w

. potential jump at the wake surface.

(3)
Equation 3 can be interpreted as a distribution of sources and normal
dipoles over the body, and as a distribution of normal dipoles over the
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MARIE PROPELLERS 377
wake. Alternatively, the normal dipoles can be replaced by an equivalent
distribution of vorticity whose strength is equal to the derivative of the
strength of the dipoles. The wake then consists only of sheets of vorticity, an
interpretation that many fnd more physically intuitive.
To complete the formulation of the problem, the Kutta condition must
be imposed, which requires that the velocity be fnite at the trailing edges,
and that the dynamic boundary condition of zero pressure jump across the
trailing vortex wake must be applied. The latter condition requires that the
vorticity in the wake be everywhere convected by the local fow, thus
establishing, in principle, the position in space of the vortex sheets.
The problem as formulated so far does not consider the action of viscous
forces. Following the usual boundary-layer approximation appropriate to
high Reynolds numbers, we may regard the boundary of the potential-fow
regon as consisting of the physical boundaries of the blades and hub,
augented by the displacement thickness of the boundary layers. As a frst
approximation, the displacement thickness can be ignored, which thus
returns us to the original problem, but with a rational basis for adding
viscous tangential stresses in the fnal determination of forces.
Formulating the problem in such a general way is obviously much easier
than solving it! The combination of the nonuniform onset fow, the complex
geometry of the blades, and the need to establish the geometry of the free
vortex sheets makes the solution of the propeller problem extremely
difcult. We next review the progress made in the solution of this problem,
considering in turn the design of propellers for a gven load distribution, the
related problem of analyzing a gven propeller in steady fow, and fnally the
analysis of a propeller in unsteady fow.
PROPELLER DESIGN
Stated simply, the hydrodynamic design of a propeller is accomplished in
two steps. One frst establishes a radial and chordwise distribution of
circulation over the blades that will produce the desired total thrust, subject
to considerations of efciency and cavitation. In the second step, one fnds
the shape of the blade that will produce this prescribed distribution of
circulation.
Betz (1919) frst developed the basis for determining the radial distri
bution of circulation that would result in optimum efciency for a propeller
operating in uniform infow. He found that the optimum propeller in this
case developed a trailing vortex system that formed a rigd helicoidal
surface receding with a constant axial velocity. However, it was not until
Goldstein (1929) that the potential problem posed by Betz was actually
solved. Goldstein's work, however, opened up the way for the development
of a propeller design method following Prandtl's concept of the lifting line.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
378 KRWIN
If the aspect ratio of the blades, i.e. the ratio of their span to mean chord,
is high, Prandtl ( 1921) deduced that the three-dimensional problem could
be solved by concentrating the circulation around the blades on individual
lifting lines, and that the fow at each radial section could be regarded as two
dimensional in an infow feld altered by the velocity induced by the free
vortex system shed from the lifting lines. Goldstein's solution for the
optimum propeller in uniform fow provided the means to calculate the
velocity induced by the free vortex sheets. By combining this information
with theoretical or experimental two-dimensional section data, one could
design an optimum propeller.
This approach was extremely successful for aircraft propellers, which
generally had very high-aspect-ratio blades and operated in front of the
aircraft in relatively uniform fow. However, marine propellers are gener
ally forced to have low-aspect-ratio blades, since their lift coefcient must
be severely limited to prevent excessive cavitation. As a result, marine
propeller designs based on lifting-line theory could not be expected to be
satisfactory. In addition, the onset fow to a propeller, as indicated in the
previous section, is generally quite nonuniform.
It was recognized at an early stage that lifing-line theory could be made
applicable to lower-aspect-ratio surfaces by introducing a correction to the
camber of the two-dimensional sections to account for the induced
curvature of the fow. An intuitive explanation of the presence of induced
curvature is that the velocity induced by the trailing vortex sheets is greater
at the trailing edge than at the leading edge. Approximate calculations of
camber correction factors for a few limited cases were made by Ludwieg &
Ginzel ( 1944), but it was not until 1 7 years later that precise results obtained
by computer were published by Cox (1961).
It is therefore reasonable to conclude that prior to the 1950s, analytical
methods for propeller design were not yet ready for practical application.
As a result, marine propellers were inevitably designed on the basis of
systematic series of model experiments. A textbook of that time period by
Baker ( 1951) states: "In all marine work, propeller design is based on
experimental data . . .. There is no theory extant which will enable the
efciency and the capacity to absorb power of a given screw to be calculated
for actual ship conditions, either from purely theoretical data, or from the
usual experimental data for aerofoil blades."
The situation soon changed. The extension of Goldstein's lifing-line
theory to the case of propellers with arbitrary radial distributions of
circulation in both uniform and radially varying infow was presented in a
landmark paper by Lerbs ( 1952). While initial acceptance was slow because
of the intricacy of the theory and the lengthy calculations required, the pro
cedure was computerized in the late 1960s. The Lerbs method is still the
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MAR PROPELLERS 379
universally accepted procedure for establishing at the early design stage the
radial distribution of circulation and the resulting thrust, power, and
efciency of a propeller.
During this time period design methods were developed, based on a
combination of the original Goldstein or the more general Lerbs lifing-line
theory, with lifing-surface corrections to camber and angle of attack.
Notable contributions at this stage were made by van Manen (1957) and by
Eckhart &Morgan (1955). However, this was also the time period for early
developments of a true propeller lifting-surface theory, and it was becoming
clear that the blade outline, the skew, and the form of the radial distribution
of circulation all had a major infuence on the correction factors, which had
initially been portrayed by a single graph. More extensive correction
factors were computed and published by Morgan et al. (1968). These would,
in principle, enable the designer to do a more accurate job. By this time,
however, it was also becoming evident that the problem was too
complicated to be reduced to a simple tabular/graphical hand calculation,
and that the growing availability of computers would soon render this type
of procedure obsolete.
In the meantime, numerical lifting-surface methods were evolving as a
direct consequence of the growing availability of digtal computers.
Actually, Strscheletzky (1950) and Guilloton (1957) published numerical
methods together with hand calculations, but their methods were probably
considered to be too laborious for wiespread adoption. Sparenberg (1959)
formulated the basis for a propeller lifting-surface theory, which would later
be programmed. Then, a sudden burst of publications of computer-based
propeller lifting-surface design methods occurred in 1961-1962. This
included contributions by Pien (1961), Kerwin (1961), van Manen &
Bakker (1962), and English (1962). However, these initial eforts all involved
simplifying assumptions of various sorts, which have since been found to be
unnecessary as a result of rapid advances in computer hardware and in the
development of efcient computational methods.
We therefore jump to the present time and describe two current lifting
surface computational methods that are essentially equivalent in their basic
formulation and provide almost identical results in a comparative calcu
lation, even though they use very diferent numerical methods. The two
methods are PROPLS, developed by Brockett ( 1981), which evaluates the
resulting singular integrals by direct numerical integration, and PBD-lO,
developed by Kerwin, which uses a vortex-lattice procedure. The design
method developed by Kerwin and an analysis procedure developed by
Greeley were published jointly by these two authors in 1982 (Greeley &
Kerwin 1982).
The presence of the hub as a solid boundary is ignored in both of these
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
380
KERWN
theories and has generally been ignored in the past. This may seem
surprising until one realizes that the inner radii contribute little to overall
propeller forces as a result of the low rotational velocity in this region.
However, an extension of the vortex-lattice method in which the hub
boundary is accounted for is discussed later in this review.
It is further assumed that the blades are thin, so that the singularities
distributed on both sides of the blades in accordance with Equation (3)
merge, in the limit, into a single sheet of sources and either normal dipoles
or vortices. The source strengths are directly proportional to the stream
wise derivative of the thickness function, whereas the vortex strengths are
prescribed. Diferentiation of (3) with respect to the three coordinate
directions results in singular integrals for the components of the induced
velocity on the mean surface representing the blade. The resulting
expressions are obviously lengthy owing to the complicated nature of the
geometry involved and are not reproduced here.
In the design problem, the geometry of the blade surface is only partially
known. Specifcally, the radial distribution of chord length, rake, and skew,
and the chordwise and radial distribution of thickness are prescribed in
advance. The radial distribution of pitch and the chordwise and radial
distribution of camber are to be determined. However, the source and
vortex distributions representing the blades and wake must frst be placed
on suitable reference surfaces in order that their induced velocity feld can
be calculated. In linear theory, the perturbation velocities due to the
propeller are assumed to be small compared with the onset velocities, so
that the blade and wake can simply be projected onto stream surfaces
formed by the undisturbed fow. However, in most practical cases the
resulting blade surfaces deviate substantially from this, and thus linear
theory is generally not sufciently accurate.
The procedure employed in PBD-lO is to start with some initial
prescription of pitch and camber, compute the total fuid velocicy at a
number of points on the surface, and then adjust the surface in such a way as
to annul its normal component. The process is repeated using the adjusted
surface as the new reference surface until convergence is obtained. The
trailing vortex wake is similarly aligned with the resultant fow, as
illustrated in Figure 6. The details of the vortex-wake alignment procedure,
which is also employed in the equivalent steady-fow analysis procedure,
are gven in Greeley &Kerwin (1982).
One therefore obtains an exact inviscid solution for a set of zero
thickness surfaces representing the blades and vortex wakes, upon which a
linearized thickness solution is superimposed. Perhaps the term "exact" is
an overstatement, since a discretized representation of the propeller
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MI PROPELLERS 381
, used and the alignment of the trailing vortex wake involves some
approximations.
In PROPLS, the blade reference surface is helicoidal, with an arbitrarily
specifed radial distribution of pitch. It is therefore equivalent to PBD-lO
with a specifcation of zero camber. Since the maximum camber of propeller
sections is generally of the order of 2-3% of the chord, this diference is
minor. The trailing vortex wake consists of constant-radius helical lines
whose pitch may be chosen to correspond either to that of the undisturbed
onset fow or to the pitch of the blade reference surface.
Brockett's (1981) procedure for evaluating the induced velocities on the
blade is one of direct numerical integration. Since the integrals over the
other blades and the trailing vortex wakes are nonsingular, the integrands
are ftted by trigonometric polynomials over a prescribed set of chordwise
and radial intervals. The integration for the induced velocity and the second
integration required to obtain the mean line shape are then performed
analytically using precomputed weighting functions.
The integral for the induced velocity at a point on the key blade contains
a Cauchy principal-value singularity. The integration is therefore frst
TRANSITION
WAK E
(0) WAKE FOL LOWING UNDISTURBED INFLOW
ULTIMATE
WAKE
-
(b) WAKE ALIGNED WITH FLOW
Figure 6 Illustration of vortex lattic representation of the trailing vortex wake before and
after alignment with the local fow. Note the substantial increase in pitch afer aligment.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
182 KRWlN
performed in the radial direction, and the singularity is then factored out of
the remaining chordwise integrand. The regular part of the integrand is
next ftted with a cosine series, which then yields a series of integrals whose
Cauchy principal value was derived by Glauert (1948).
In the vortex-lattice scheme employed in PBD-I0, the continuous
distributions of vortices and sources are replaced by a set of concentrated
straight-line elements, whose end points lie in the mean blade surface.
Velocities are then computed at suitably placed control points between the
elements. This avoids the difculty associated with the evaluation of
singular integrals and reduces the problem to a geometric one of fnding
points on the mean surface and of calculating the velocity feld of a simple
line vortex and source. The latter step involves only the evaluation of
elementary integrals. An example of the vortex/source lattice arrangement
used in PBD-I0 is given in Figure 2.
One must be careful in setting up the geometrical arrangement of lattice
elements and control points or the method may not converge properly.
Vortex-lattice methods are generally very robust, in that a wide variety of
spacing algorithms do converge to the right answer, and if they do not, the
error is generally local. Understandably, those individuals with more
rigorous inclinations have frequently been suspicious of the accuracy of
vortex-lattice methods and would prefer a direct approach as exemplifed
by Brockett's (1981) method. However, James (1972) and Lan (1974) both
provided rigorous proofs of the convergence of vortex-lattice methods in
two-dimensional fow. James treated the case of constant spacing of
vortices over the chord and proved that the commonly used 1/4-chord 3/4-
chord arrangement of vortices and control points within each subinterval
was correct. He also showed that the local pressure obtained from the
solution of the vortex element closest to the leading edge approached a
value that was 1 1.4% too low as the number of elements became large.
However, it is important to recognize that this is not a case of false
convergence, since the value of the local pressure at a gven position near
the leading edge would converge to the right answer as the number of
elements increased, while the place where the answer was inaccurate would
move closer to the leading edge.
Lan (1974) showed that the arrangement of vortex locations Xv and
control point locations Xc represented by
1 (a,
x_a)=_ I-cos
N
'
a 1, 2, . . . ,N (4)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
n PROPELLERS 383
gave exact results for the total lif of a fat plate or parabolic camber line and
was more accurate than the constant-spacing arrangement in determining
the local pressure near the leading edge. This choice, commonly referred
to as cosine spacing, can also be seen as related to the conformal trans
formation of a circle into a fat or parabolically cambered plate by the
Joukowski transformation.
Similar spacing arrangements can be used in a vortex-lattice represen
tation of the lifing-line problem, whose exact solution is well known. Table
1 shows the convergence of the calculated induced velocity at the tip panel
of a lifing line with elliptical circulation using cosine spacing. Also shown is
the total induced drag obtained by summing the elementary drag forces
over all the panels. The-convergence is approximately quadratic in this case,
and it is evident that 10 to 20 elements yield results that are accurate enough
for any practical purpose. In the case of elliptical loading, the error in the
computed induced velocity is constant over the span, so that the values of
induced velocity given in Table 1 for the tip panel apply to all of the panels.
Table 1 also shows what happens in a vortex-lattice scheme if the control
points are not located in the correct position. The values labeled
"midpoint" are the results obtained by keeping everything the same as in
the previous calculation except the position of the control points, which are
now moved to the midpoints of the intervals between vortices. The induced
velocities at the tip panel are seen to be completely wrong and diverge as the
number of panels is increased. However, the results over the rest of the span,
which are not tabulated, are not as bad; this is indicated by the fact that the
total induced drag appears to be convergng to the right answer.
There is no exact solution to compare with in the case of the propeller,
Table 1 Vertical velocity inducd at tip panel and total induced
drag for an elliptically loaded lifting line using a vortex lattice with
cosine-spaced vortices
Velocity at tip panel Total induced drag
Panels Cosine Midpoint Cosine Midpoint
5 -0.9836 0.5441 1.5198 1.2443
10 -0.9959 2.3357 1.5579 1.3948
20 -0.9990 5.3357 1.5676 1.4787
4 -0.9997 12.6831 1.570 1.5236
80 -0.9999 26.4017 1.5706 1.5469
160 -1.000 53.8123 1.5707 1.5588
Exact (-1.000) (1.5708)
The control points are either cosine spaced or at the midpoints of the
panels.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
384 KRWN
but one can be reasonably confdent that if a vortex-lattice arrangement is
used that converges to the correct answer for both two-dimensional fow
and a lifting line, then one can believe the converged solution. The
convergence of a propeller vortex-lattice method calculation with increas
ing numbers of chordwise and radial elements is illustrated by Greeley &
Kerwin (1982).
We close our discussion of the propeller lifting-surface design problem by
comparing the results obtained by the two methods discussed. The test case
is a fve-bladed, highly skewed propeller whose geometry is similar to that
of the propeller illustrated in Figure 10. The detailed geometry of the test
case is given by Brockett (1981). To make a consistent comparison, trailing
vortex-wake alignment was suppressed in PBD-10, and its pitch distri
bution was set to conform to PROPLS. However, the blade reference
surface in the PBD-10 calculation was automatically adjusted to its
converged value, and this represents a diference between the two methods.
Figure 7 shows the radial distributions of pitch and camber obtained by
these two methods. The results are very similar, although small dis-
1.6
1.4
P/O
PBO -10
0
d
1.2
PROPLS
+
0
I
< I.
0:
0:
w

I
L
w %

< .6 .0
30
0
I
^ <
:
.4 .02 0:
u
PBO -10
I
0:
a.
w
(
.2 .01

<
0
0
U
.2 .4 .6 1.0
NON -DIMEN SIONAL r/R
Figure 7 Comparison of radial distributions of pitch/diameter and camber ratios obtained
by current lifting-surface methods by Brockett (PROPLS) and Kerwin (PBD-IO).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
PROPELLERS 385
crepancies exist. The largest diferences are in the camber at the inner radii,
and it is possible that these are due to the iteration of the blade reference
surface by YBO-i. In any case, diferences as small as these would be
almost impossible to detect by means of a model test.
The neglect of the hub in these design methods, while justifable from the
point of view of overall propeller performance, results in substantial local
errors in section shape at the inner radii. These sections are therefore not as
good as they could be, from the point of view of cavitation and viscous drag
.
In addition, neglecting the hub actually makes the determination of the
local shape at the inner radii more difcult. This is because the inner
boundary of the blade becomes, in efect, a free tip with a large squared-of
chord. The exact shape required to achieve a prescribed circulation
distribution for this artifcial tip may be very complex, requiring fnely
spaced elements to obtain an accurate solution to the wrong problem! If the
hub boundary condition is to be ignored, it would seem better to recognize
that the solution in the immediate vicinity of the hub will be incorrect, and
therefore that one should not attempt to calculate section shape in this
region but instead should smoothly extrapolate the results obtained over
the rest of the blade.
Of course, a better approach is to include the hub in the problem. This
results in a mixed design/analysis problem. The hub is a body of revolution
of known shape on which the normal component of the total fuid velocity
must vanish. On the other hand, as before, the circulation on the blades is
specifed and their shape is to be determined.
This problem has been recently treated by Wang (1985), who combined
Kerwin's vortex-lattice method with a surface-panel representation of the
hub. The surface-panel elements were also chosen to be concentrated
vortices, which are aligned with the corresponding elements on the blades
at the hub juncture. An iterative solution is used in which the velocity feld
generated by the initially hubless blades is treated as a given onset fow for
the hub solution. The velocity feld thus generated by the hub is then
similarly added to the onset fow in the next iteration of the blade solution.
Since the blade shape changes during each iteration, the hub is continu
ously repaneled to match the blade at the hub juncture.
This procedure generally converges within three or four iterations. The
computing times are substantially greater than for the hubless case but
would not be considered excessive for a fnal propeller-design calculation.
As might be expected, the efect of the hub is negligible over the outer
portion of the blade but results in a reduction in pitch and camber in the
immediate vicinity of the hub. In some cases, a large negative camber is
required to generate the desired circulation distribution in this regon, and
the chordwise distribution of camber may have a pronounced "s" shape.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
386 KRW
Since the blade geometry in the hub region is very sensitive to the
prescribed distribution of circulation, the abrupt change in shape in this
region may be impossible to build. On the other hand, since the method for
arriving at the prescribed circulation distribution near the hub may be
arbitrary, it is questionable whether this type of design approach is always
appropriate. An alternative would be to design the shape of the blade for a
prescribed circulation distribution, either without accounting for the
infuence of the hub or with a simple hub image approximation. One would
then modify this shape, unecessary, to insure that the shape was smooth
and buildable. The smoothed design would be subject to an analysis, and
the resulting pressure distribution near the hub would be examined to
determine if it is acceptable from the point of view of cavitation and/or
boundary-layer characteristics. If the pressure distribution were not
acceptable, the shape would then be systematically altered in a smooth way
and the analysis repeated.
It would also make sense if a surface-panel method were used for this
kind of analysis, rather than a lifting-surface method as used in the design.
The reasons are that the blade sections near the hub tend to be thick for
structural reasons, and that the spacing between blades at the hub juncture
is of the same order of magnitude as the blade thickness. In addition, fllets
are generally present, so that the actual geometry is quite diferent from that
assumed in present lifing-surface procedures. This is discussed further in
the next section.
ANALYSIS IN STEADY FLOW
Background
In the analysis problem we are gven the geometry of the propeller and wish
to determine the fow feld that it generates. The govering equations are
the same as in the design problem, but the unknowns are reversed. The
circulation distribution over the blades, which was prescribed in the design
problem, is now the unknown, whereas the shape of the blade is now given.
The singular integral that yields the velocity induced by a known
distribution of circulation in the design problem becomes an integral
equation in the analysis problem. While the latter is, in principle, more
difcult from a mathematical point of view, this diference becomes
relatively unimportant once a numerical solution is employed. In that case,
the singular integral equation is inevitably replaced by a system of linear
algebraic equations whose solution presents no problems if the number of
unknowns is not excessive ..
One of the earliest analysis procedures consisted simply of inverting the
lifting-line design methods of van Manen (1957) or Eckhart & Morgan
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
n PROPELLERS 387
(1955). One such method was developed by Kerwin ( 1959), who employed
an iterative solution to match the two-dimensional section characteristics
at each radius with the approximate induced infow obtained by interpo
lation of tabulated values of the Goldstein function. The method worked
fairly well for simple, unskewed blade shapes, but it would obviously o
unable to handle current complex propeller forms.
A lifting-surface analysis method was published by Yamazaki (1962),
although the results given were obtained by hand and involved some
simplifcations. Tsakonas et al. (1968, 1973, 1983) developed a procedure for
propeller analysis in both steady and unsteady fow based on the
acceleration potential. His approach was to represent the unknown loading
by a summation of chordwise and radial mode functions whose amplitudes
could be determined. The earlier versions of Tsakonas' procedure were
based on a strictly linearized theory, which was found to introduce much
larger errors in the steady solution than in the unsteady solution. However,
later refnements improved the accuracy of the steady solution, as indicated
by Tsakonas et al. (1983).
The mode approach was combined with a vortex-lattice representation
of the blades by Cummings (1973) to solve the steady-How analysis
problem. The procedure was simplifed by restricting the chordwise modes
to two; one consisted of the loading form of a two-dimensional Hat plate,
and the other was chosen to be the form of the two-dimensional loading of
the propeller's camber line. While this method worked fairly well, the error
introduced by the limited representation of the chordwise load distribution
could not be readily evaluated.
Current Partially Linearized Analysis Method
We consider in this category methods that are linearized to the extent that
the How feld is constructed from singularities located on the mean blade
surface, but where induced velocities are not necessarily considered small
compared with the velocity of onset How, and where the positions of the
blade and trailing vortex wake are allowed to deviate from a stream surface
of the undisturbed fow. This is in contrast to boundary-element methods,
in which the fow feld is constructed from singularities located on both
sides of the actual blade surface.
Represented in this category are methods by Tsakonas et al. (1983),
Kerwin & Lee (1978), van Gent (1977), and Greeley (1982). However, we
limit our review to the essentials of the procedure developed by Greeley,
which he has designated as PSF-2.
The PSF -2 program uses a vortex-lattice representation of the blades
that is identical to the design procedure described earlier. Rather than using
spanwise and chord wise mode functions to describe the unknown circu-
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
388 KRWIN
lation distribution, each spanwise vortex element is treated as an unknown
that is to be found by collocation using an equal number of control points
on the blade. This avoids convergence difculties, which must receive
careful attention in a mode collocation scheme, but at the expense of a
larger number of unknowns. While this would be a disadvantage if the
number of vortex elements became very large, it has been found that
converged results can be obtained with roughly 10 to 200 elements. The
time required to solve a linear system of equations of this size is much less
than the time needed to compute the required infuence functions, so that
economization in the number of unknowns is not signifcant.
PSF-2 uses a trailing-vortex-wake alignment scheme that is identical to
that used in the PBD-lO design program. However, in this case, two levels
of iteration are required. The distribution of circulation on the blades, and
hence in the wake, is frst found based on an assumed geometry of the wake.
Keeping this circulation fxed, the wake is aligned with the fow in an
iterative way. When this has converged to a specifed tolerance, the
circulation distribution is recomputed and the entire process repeated until
no further changes occur.
Additional considerations enter into the analysis problem if the propeller
is operating of-design, particularly as the angles of attack of the thin
outboard sections are increased beyond their design value. As illustrated in
Figure 8, a vortex sheet tends to form not from the tip, but from the leading
edge starting at some radius farther inboard. The mechanism for the
formation of this vortex is believed to be similar to that for a highly swept
wing, and is governed by the viscous behavior of the fow near the leading
edge.
The presence of a leading-edge vortex has two important consequences.
The overall lift of the tip regon of the blade is increased because of the
reduced induction of the vortex as it moves of the blade surface; in
addition, the local pressure reduction at the leading edge is attenuated,
which thus delays the inception of cavitation relative to that which would
be predicted on the basis of inviscid attached fow.
The frst efect can be inferred from the fact that propeller thrust and
torque measured under conditions of high angle of attack are generally
greater than the values calculated assuming an attached vortex sheet. It is
therefore necessary to incorporate some form of detached-vortex-sheet
model in a propeller analysis procedure.
The feld of vortex-sheet separation is currently an active one, with
numerous marine and aerodynamic applications. However, anything close
to a rigorous solution, involving both the proper alignment of the free
vortex sheet and the treatment of the viscous efects that initiate it, is not yet
at hand; even if it were, the computing efort would be so great as to make
such an analysis scheme impractical for routine design studies.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
UE PROPELLERS 389
A simplifed representation of a separated leading-edge vortex sheet was
devised by Kerwin & Lee (1978) and incorporated by Greeley (1982) in
PSF-2. As shown in Figure 9, the actual blade tip, which is generally
rounded, is replaced by a vortex lattic with a fnite tip chord. The spanwise
Figure 8 Illustration ofleading-edge vortex formation made visible by cavitation. (Top) The
proplIer is operating at its deig point, and the vortex leaves from the tip. (Botom) The
propelIer is operating at a low advanc coefcient and a leading edge vortex is evident.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
390 KRWN
Figure 9 Il1ustration of a simplifed leading-edge vortex separation model. For clarity. the
magnitude of 1 has been exaggerated.
vortex lines in the tip panel are continued by free vortex lines that depart
from the surface of the blade and join in a "collection point," which then
becomes the orign of the outermost element of the discretized vortex sheet.
The position of the collection point is establishe
d
by setting the pitch angle
of the leading-edge free vortex equal to the mean of the undisturbed-infow
angle and the pitch angle of the tip vortex as it leaves the collection point.
This relatively crude representation of the leading-edge vortex sheet
generally results in substantially improved correlation with experimental
data. However, occasional discrepancies exist, which may be due to
defciencies in the theory, inaccuracies in model manufacture, or Reynolds
number scale-efect problems in model tests. Discrepancies between tests of
the same propeller model in diferent facilities and lack of repeatab;lity of
tests of the same model (due possibly to deterioration of the blade leading
edges) make it difcult to draw any defnite conclusions at present. This
situation is illustrated in Figure 10, which shows experimental results for
the same model conducted at two diferent facilities, together with the
results of the theory. Good agreement exists near the advance coefcient for
which the propeller was designed, but large discrepancies occur at low
advance coefcients. While most comparisons are not this bad, this one is
included to show that problems still exist in of-design analysis.
As a next step in the refnement of separated leading-edge vortex fow,
Greeley (1982) developed a semiempirical method for predicting the point
of leading-edge separation. He found that existing data for swept wings
could be collapsed reasonably well by expressing a critical nondimensional
leading-edge suction force as determined from inviscid theory,
(5)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
nE PROPELLERS 391
as a function of a local leading-edge Reynolds number
K

U
"
r
n

,
V
where
Fs = suction force per length of leading edge,
U
n = component of the infow normal to the leading edge,
r n radius of curvature of the leading edge in a plane normal
to the edge,
U" = free-stream velocity,
v = kinematic viscosity of the fuid.
A plot of this empirical relationship is reproduced in Figure 1 1 .
(6
)
Figure 10 Comparison of calculated and measured propeller characteristics for the propeller
illustrated operating in a uniform onset fow. The nondimensiona thrust and torque
coefcients are

K
_
12nR
where

is the propeller torque and all other symbols are as defned in Figure 4 legend.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
392 KRW
| 0

U
..

:I OO

m
L
80
z
w
u
i 60

C 40
2
Z
o 20
U

COMPUTED Cs g FLOW BREAKDOWN V5

RL
E
( BEST FI T TO AVAI LABLE DATA)
LONG (BURST ) BUBBLE
OR SERRATED FLOW
LI MI TING SUCTION FORCE
( 20 AND DATA )
LOSS OF L. E, SUCTI ON
DUE TO BUBBLE
BREAKDOWN
'
_SHORT BUBBLE OR
ATTACHED TURBULENT
FLOW
RESIDUAL SUCTION
AFTER BUBBLE BURST
(20 DTA ONLY) m
-
-
-
-
-
-
W
-
OLEADINDGE5RBLEN-- --
O && _ &&& & &&&& &&
| 0

3 x | 0

| 0
4
3 x | 0
^
| 0
LEA0| W0 E00E REYW0L0S WJVBER, R
g
Figure 11 Empirical relationship betwen the value of the leading-edge suction force
coefcient at the point of fow breakdown W a function of leading-dge Reynolds numbr. '
From Greeley (1982).
Applying the same criterion to two distinctly diferent propellers, Greeley
(1982) found reasonable correlation between the predicted and observed
radial positions ofthe initiation of the leading-edge vortex. The observation
of the vortex sheet in the experiments was made by reducing the tunnel
pressure to the point where the sheet was just starting to cavitate.
Once the starting point of the sheet is established, one still needs to trace
its path over the blades. As a frst step, Greeley (1982) developed a "frst
order" model in which the free vortex sheet was placed at a height equal to
the blade boundary-layer thickness, and the resulting change in the
predicted chordwise pressure distribution as compared with that of the
attached fow was then found. However, this was only a frst step, and one
must consider that this is still a feld for active research.
A Partially Linearized Method Including the Hub
Incorporation of the hub in the lifing-surface analysis problem was
recently accomplished by Wang (1985), who used the same vortex-lattice
representation as in the PSF-2 and PBD-lO programs. In this case, the
analysis problem is in some respects simpler than the design problem, since
the position of the blade surface is fxed and the blade and hub paneling can
be established at the outset. The number of unknowns is increased when the
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
PROP8LL8R8 393
hub is introduced, and one has the choice of solving a larger matrix or using
an iterative technique. Wang chose the latter option in order to minimize
changes in the existing PSF -2 code. He found that the process converged
rapidly as a result ofthe weak interaction between the portions of the blades
and hub that were not in the immediate vicinity of their intersection.
Figure 12 shows the radial distribution of circulation for a particular
propeller obtained both with and without the inclusion of the hub. Shown
in addition is the radial distribution of circulation obtained experimentally
by measurement of the circumferential mean tangential velocity just
downstream of the blades using a laser-Doppler velocimeter. It is evident
that the inclusion of the hub in the theory increases the predicted
circulation at the inner radii, which is in bette agreement with the
measurements. The sharp spike in the measured results near the hub is
60
Vortex l at tice cal cul at i ons
r
0 Symbol Hub Cordwi se Spnwi se

E l ements E l emen t s
m

Ignored 8 8
0
c
A Ignored 18 9
I:
C L I ncl uded 8 8
"
j.
(
I ncl uded 18 9
-
<
40

.
Q "
:
u
<

30
u
l.
0
Z
0 A
i
20 (
L
L

Measured by LOA j ust

X

downs tream of bl ades

I
e
0
t o
l
.
<
0

a
0.20 0.

0 0.60 0. 8 0 | . 00
DI STANC E FROM S HAFT CENTER r/R
Figure 12 Calculated radial distribution of circulation both with and without the inclusion of
the hub compared with experimental resuts obtained with a laser-Doppler velocimeter. From
Wang (1985).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
394 KRWIN
circulation induced by the rotating hub by the action of viscous stresses.
This, of course, is not included in the theory.
Boundary-Element Method
The two principal shortcomings of the lifing-surface approximation to a
propeller blade are the local errors near the leading edge and the more
widespread errors near the hub, where the blade thickness is large and
where the blades are in close proximity.
The frst problem can be overcome by the application of a local
correction due to Lighthill (1951), in which the fow around the leading edge
of a two-dimensional, parabolic half-body is matched to the three
dimensional fow near the leading edge of the blade obtained from lifting
surface theory. The accuracy of the Lighthill correction is greatest for thin
sections, which makes it particularly suitable for the outer part of the
propeller blade. It is fortunate that it is in this region that accurate pressure
distributions are needed for the prediction of cavitation inception.
However, it is not known whether the Lighthill correction remains
accurate as the slope of the leading edge increases toward the tip. In
addition, the error introduced by the lifting-surface approximation to the
thick hub sections will also not be reduced by a leading-edge correction
except in a very local sense.
As a result, there is current interest in the application of discretized
boundary-element methods, generally referred to as panel methods, to the
propeller analysis problem. Panel methods are currently being applied to a
variety of problems, including fows around complete aircraft confgura
tions and ship hulls. The number of diferent panel methods is rapidly
growing. A good single source for a derivation of a variety of panel-method
algorithms is a recent text by Moran ( 1984).
A panel method for propellers has recently been developed by Hess &
Valarezo (1985). Their method is an adaptation of the one developed by
Hess & Smith (1967) for nonlifting bodies and extended by Hess (1975) to
include general lifting bodies.
As illustrated in Figure 13, the blades and hub are represented by a large
number of quadrilateral panels. The appearance is superfcially similar to
the vortex-lattice representation of a propeller shown in Figure 2. However,
in this case, quadrilateral panels are located on both sides of the blades as
well as on the hub.
A distribution of sources with constant density is placed within each
panel. In addition, the panels on the blades contain distributions of normal
dipoles that are constant over an entire chordwise strip. The dipole
distributions are extended into the wake by an equivalent distribution of
discrete trailing vortices, the latter being essentially the same as the vortex-
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MRTE PROPELLERS 395
lattice representation discussed earlier. The hub is considered a nonlifting
element, and thus its panels are exclusively sources.
The strengths of the individual source panels and the dipole strengths
associated with each chordwise strip are found by requiring that the normal
component of the total fuid velocity vanish at a centrally located control
point within each panel. While the representation of the trailing vortex
wake is fundamentally the same as for a vortex-lattice method, Hess &
Valarezo (1985) have initially simplifed the geometry of their transition
wake to that of a pure helix. Their ultimate wake is modeled as a ser
infnite cylindrical wake of an infnitely bladed propeller, whose velocity
feld can bfound in closed form. This is quite diferent from the ultimate
wake representation used by Greeley &Kerwin (1982), in which the vortex
sheet is rolled up into one helical vortex line from each blade. The latter may
be closer to physical reality, but the former is computationally more
efcient and may well be equally accurate.
A typical chordwise pressure distribution for a section of a ship propeller
obtained by Hess &Valarezo (1985) is shown in Figure 14, together with
results computed by Kim & Kobayashi ( 1984) and measurements by
Versmissen &van Gent ( 1983). Kim & Kobayashi's results were obtained
from their extension of the PSF-2 vortex-lattice program to include the
Figure 13 Illust
r
ation of propelJer blade and hub panelling. From Hess & Valarezo (1985).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
396 KRWN
computation of local surface pressures. However, their procedure does not
include a Lighthill correction at the leading edge and can therefore not be
expected to properly capture the local pressure minimum. The panel
method result shows the presumably correct pressure minimum, which
Hess &Valarezo point out as being an advantage of the panel method.
However, some caution is called for at this point. Thin sections develop a
sharp pressure peak located close to the leading edge, so that extremely fne
paneling may bneeded near the leading edge in order to capture the peak
value. On the other hand, Lighthill's rule will gve the correct pak value in
two-dimensional fow in the limit of small section thickness. Consequently,
a panel method may not be obviously superior to a vortex-lattice method
augmented by Lighthill's rule in this particular case. This is clearly an
important area for future research.
-0. 3
-0. 2
-0.1
0. 0
Cp(lOCAl)
0. 1
0. 2
0.3
0.4
0.0
PRESENT METHOD
----- PSP METHOD
-" 0- EXPERIMENT 1
-' 6- EXPERIMENT 2
0.6
FRACTION OF CHORD. lie
0. 8 1 . 0
Figure 14 Comparison of clculated and measured chordwise pressure distributions. The
solid line (identifed as PRESENT METHOD) was obtained by Hess & Valarezo (1985) using
their surface-panel code. The dashed line (identifed W PSP METHOD) was obtained by K
& Kobayashi (1984) using a vortex-lattice method based on PSF-2. The two experimental
curves were obtained by Versmissen & van Gent (1983) using pressure transducers embedded
in a OA8-m diameter propeller model. The diference between the duplicate experimental
results is indicative of the difculty in carrying out t
h
is type of experiment.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
UE PROPELLERS 397
UNSTEADY PROPELLER FLOWS
Background
We now come to the important problem of unsteady propeller forces, which
unfortunately we are able to treat only briefy. The unsteady problem is
complicated by the presence of shed vorticity in the wake that depends on
the past history of the circulation around the blades. The fundamental
problem of an airfoil or hydrofoil in unsteady fow has an extensive
literature, including recent contributions by McCroskey (1982) and
Crighton (1985
) in this series.
With the onset fow represented in terms of its circumferential harmonic
components, and with the assumption that the propeller responds linearly
to changes in the onset fow, the problem can be reduced to one of fnding
the response of the propeller to each harmonic. The nondimensional
parameter that characterizes the degree of unsteadiness of the fow is the
reduced frequency k, which is defned as the product of the frequency of
encounter and the local semichord, divided by the relative infow speed. For
a typical marine-propeller chord length, the reduced frequency correspond
ing to the frst harmonic of the onset fow is of the order of one half, while the
value for the harmonic corresponding to the number of blades will be of the
order of two or three. From the classical two-dimensional olution for an
airfoil traversing sinusoidal gusts, it is known that the unsteadiness of the
fow becomes signifcant for values of the reduced frequency of roughly one
tenth or higher. Thus the response of a propeller to all circumferential
harmonics of the onset fow is unsteady, in the sense that the lift is
considerably smaller than the equivalent quasi-steady value and is shifted
in phase relative to the infow.
Early attempts to calculate propeller unsteady forces used a variety of
approximations, ranging from a purely quasi-steady approach to ones that
employed two-dimensional unsteady airfoil results. A number of such
semiempirical methods were applied to a specifc case in an interational
cooperative study conducted by Schwanecke (1975) ; the study showed that
a large spread existed in the results obtained by the diferent methods.
Current Liting-Surface Method for Unsteady Flow
One of the frst investigators to publish a complete theory for the unsteady
problem was Hanaoka (1962), although numerical evaluation of his theory
was not published until 1969 (Hanaoka 1969). The theory developed by
Tsakonas et al. ( 1968, 1973), which was discussed earlier in connection with
the steady-fow analysis problem, became widely used during this time
period for the prediction of unsteady propeller forces. Figure 1 5, taken from
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
398 KRWN
Boswell et al. ( 1983), shows good correlation between Tsakonas' theory and
measurements for a series of propellers with varying blade area. Also shown
in Figure 1 5 are results obtained by various approximate theories, which
can be seen to give results either far above or far below the measurements.
Results obtained with the unsteady lifting-surface theories of Hanaoka,
Tsakonas, and others (also compiled by Schwanecke) were found to be in
much closer agreement than the semiempirical methods.
Kerwin & Lee (1978) used a diferent approach, with the blades
represented by a vortex lattice in a manner similar to the steady-fow
problem described earlier, and with the fow solution obtained in the time
domain rather than in the frequency domain. The problem was solved as an
initial-value problem starting from the steady solution, with the propeller
rotated in discrete angular increments through three or four complete
revolutions until a steady-state oscillatory solution was obtained.
The motivation for using a time-domain solution was largely in
|
W
::
a
J I
I r
> ::
U I
z :
w I
::
z
O C
w _
a r
u w
w O
0
c
m
0
1 .0

0.6
0.4
0.2
UNSTEADY 20
"- _ UNSTEADY LL
,
"
-

.
.

, ULS (PLEXVAN)
.y J , - "":;5 (QUASI/
0.3

o .. -
ULS (PPEXACT)
* ULS ( PUF2)
'

' - ' ' - : '

- 1- '
"
EXPERIMENT
L (T:AA'HI l
LOW ASPECT RATIO
06 0.9 1 .2
EXPANDED AREA RATIO. A
E
/A
O
Figure 15 Blade-frequency thrust correlation over a range of expanded area ratios a,Ja.
The various programs are identifed as follows : QUASI (McCarthy 1961), PPEXACT
(Tsakonas et a. 1973), PLEXV AN (Tsakonas et a. 1983), PUF2 (Kerwin & Lee 1978),
TANIBAYASHI (Tanibayashi 1980), LOW ASPECT RTO (Brown 1981), UNSTEADY
2D (Boswell & Miller 1968), UNSTEADY LL (Brown 196). From Boswell et a. (1983).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
n PROPELLERS 399
anticipation of extending the approach to the solution of propeller fows
with intermittent cavitation. A frequency-domain solution cannot be used
in this case because of the highly nonlinear relationship between fow
conditions and the length of the cavity. As indicated earlier, this interesting
problem is beyond the scope of the present review.
Results obtained by this theory were added to those compiled by
Schwanecke (1975) and were found to lie in the middle of the group of
results obtained by the other unsteady lifting-surface theories. The
combined results may be found in Kerwin & Lee ( 1978), with additional
favorable comparisons in a discussion to that paper prepared by Tsakonas.
While Figure 1 5 shows good agreement between theory and experiment,
there are many contradictory results in the published literature in this area.
It would therefore be desirable to have a defnitive set of experiments that
could be used as a basis for deciding whether or not current theories for
unsteady propeller fows are accurate. Unfortunately, there are two
considerations that make this task more complicated than one might think.
First, unsteady propeller force measurements are extremely diffcult to
make, since the propeller shaft and measuring system must be carefully
designed and dynamically calibrated, and the resulting output signals must
be processed to remove noise. Second, if calculations are to be made
corresponding to a given experiment, the efective onset fow must be
determined. It is not clear at present how much the harmonic content of the
wake feld difers from that of the nominal wake. It is possible that a pure
single harmonic wake generated by a screen in a water tunnel will not be
altered appreciably by the induced velocity feld of the propeller. On the
other hand, the harmonic content of the complex wake feld shown in
Figure 3 could be expected to be considerably diferent. A weak link in the
process of predicting unsteady propeller forces may well be in the
prediction of three-dimensional efective wakes ; the current interest in this
topic is certainly justifed.
The latter problem is not a concern if the nonuniform infow is generated
as a result of the inclination of the rotation axis relative to the fow
direction. This is the case in Figure 16a, taken from Boswell et al. (1981),
which shows substantial disagreement between Tsakonas et al. and Kerwin
&Lee in the prediction of the once-per-revolution alternating thrust force
on a single blade. Both theories underpredict the force compared with
experimental results, although to varying degrees depending on the
advance coefcient of the propeller.
Improved agreement at low advance coefcients is obtained by a
refnement introduced by Kerwin ( 1979), in which the axis of the propeller
slipstream is allowed to depart from the axis of propeller rotation. This
introduces a nonlinear coupling between the mean loading and the once-
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
.

35
,

l
w
F l

w
w

m
`
Ep
25 1
'` 0
w
w
w
w
MUanhy _
2
1
(OUP
15
10 1 _
/
/
-
-
-.a. al
".
M7ACl
m m m
-
-
a
.4 0.0 0. T.0 T.Z 1 .4 1 .6 T. Z.0
J V (I-WMl/nD
r - ,r_
4

_ xptmt
0
0
0
0
0
0
PHOPELEH%T
" EO
WTHMHEEN
o
~
_-
0
0
0 K_n & LlMF2l
.^
4
w
Mcy IQUAn
......
"
"
" "-

, ,' -- .

, "
.
. -.-. -..-

,
w..
+
"
"
T

/ / TWKm.. a lMAMC
T0

Ig4
,
I
I
0
1 I I
I
I I
0 0.8 a.B T.0 1 T.
J ^ VJ -w_)nO
T.0
U
I.B
-
ia
Figure 1 6 Correlation between theory and experiment for frst harmonic thrust fuctuation. The results in (a) are for a 10 shaft inclination in uniform
fow. The results in (b) are for a scren-generated axial wake. Progam identifcation is the same as in Figure 15, with the addition ofPUF2IS (Kerwin
1979). From Boswell et a. (1981).
s

A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Mr PROP8LI8R8 401
per-revolution time-dependent loading, which in this case seems to be
additive.
The quasi-steady method of McCarthy (1961) is shown in Figure 16a to
give quite good results in this case ; however, it does overpredict the once
per-revolution unsteady thrust by roughly a factor of four in Schwanecke's
(1975) survey. There seems to be no obvious explanation for this
contradictory behavior.
Figure 16b, also from Boswell et al. (1981), shows a similar correlation
between theory and experiment for a circumferentially varying infow
generated by a screen. Here, all three theories considered agree well at the
design advance coefcient of the propeller, but none follow the experi
mental trend at low advance coefcients. This may result from either the
infuence of the efective wake (since the onset fow contains vorticity in this
case) or possibly the asymmetry of the slipstream.
Another phenomenon that may afect unsteady blade forces is the
formation of a leading-edge vortex during part of one revolution of each
blade. It is known that this vortex formation increases lif at high angles of
attack in steady fow, as discussed earlier ; it is possible that this could also
af ect unsteady fows. H so, this would introduce a strongly nonlinear
Reynolds-number-dependent complication to an already complicated
problem!
In conclusion, present unsteady lifing-surface theories are accurate
enough to be of signifcant help to a designer wishing either to predict
unsteady vibratory forces generated by a propeller or to estimate fuctuat
ing loads on an individual blade as input to a structural analysis. The efect
of changes in blade shape on unsteady forces can be readily studied, and this
type of analysis is frequently used in the optimization of a design. However,
we see that the reliability of unsteady force predictions is by no means
perfect. The most likely sources of error are in the prediction of the three
dimensional efective wake, in the modeling of the unsteady trailing-vortex
wake geometry, in the neglect of unsteady leading-edge vortex separation,
and in the method of application of the Kutta condition at the trailing edge.
ACKNOWLEDGMENT
The writing of this review was supported by the Ofce of Naval Research
Special Focus Program in Ship Hydrodynamics.
Literature Cited
Baker, G.8.1951. Ship Design, Resistance and
Screw Propulsion. London: Birchall. 206
pp. 2Od ed.
Btz, A. 1919. Schraubenpropller mit
geringstem Energieverlust. K. Ges. Wiss.
Gittingen Nachr. Math.-Phys. Klasse
1919: 193-217
Boswell, R.!.,Miller, M. L. 1968. Unsteady
propller loading-measurement, correla
tion wt theory, and parametric study.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
402 KERWIN
NSRDC Rep. 2625, Nav. Ship. Res. Dev.
Cent., Washington, DC. ix+79 pp.
Boswell, R. J., Jessup, S. D., Kim, K.-H. 1981.
Periodic single-blade loads on propellers
in tangential and longitudinal wakes. Proc.
SNAME Propellers '81 Symp., Virginia
Beach, Va., pp. 181-202
Boswell, R. J., Kim, K.-H., Jessup, S. D., Lin,
G.-F. 1983. Practical methods for predict
ing periodic propeller loads. Paper pre
sented at Int. Symp. Pract. Des. Shipbuild.,
2nd, Tokyo
Breslin, J. P., Van Houten, R. J., Kerwin, J. E.,
Johnson, C.-A. 1982. Theoretical and ex
perimental propeller-induced hull pres
sures arising from interittent blade cavi
tation, loading, and thickness. Soc. N avo
Archit. Mar. Eng. Trans. 90: 1 1 1-51
Brockett, T. E. 1981. Lifting surace hydro
dynamics for desig of rotating blades.
Proc. SNAME Propellers '81 Symp., Vir
ginia Beach, Va., pp. 357-78
Brockett, T. E. 1985. Shear fow efects on
propeller operation : Efective velocity,
loads and vortex sheet geometry. Rep. 294,
Dep. Nav. Archit. Mar. Eng., Univ. Mich.,
Ann Arbor
Brown, N. A. 1964. Periodic propeller forces
in non-uniform fow. Rep. 64-7, Dep.
Ocean Eng., Mass. Inst. Techno!.,
Cambridge
Brown, N. A. 1981. Minimization of un
steady propeller forces that excite vibra
tion of the propulsion system. SNAME
Pub!. S-7, pp. 203-31 . New York : Soc.
Nav. Archit. Mar. Eng.
Cox, G. G. 1961. Corrections to the camber
of constant pitch propellers.

. Trans. R.
lnst. Nav. Archit. 103 : 227-43
Crighton, D. G. 1985. The Kutta condition in
unsteady fow. Ann. Rev. Fluid Mech.
1 7: 41 1-5
Cummings, D. E. 1973. Numerical preiction
of propeller characteristics. J. Ship Res.
1 7: 12-18
Dyne, G. 1980. A note on the design of wake
adapted propellers. J. Ship Res. 24 : 227-31
Eckhart, M. K., Morgan, W. B. 1955. A
propeller design method. Soc. N avo Archit.
Mar. Eng. Trans. 63 : 325-74
English, J. W. 1962. The application of a
simplifed lifting surface technique to the
deSign of marine propellers. Rep. SH
R.30/62, Ship Div., Nat!. Phys. Lab.,
Feltham, Eng!.
Glauert, H. 1948. The Elements ofAerofoil
and Airscrew Theory. Cambridge : Cam
bridge Univ. Press. Reissued by Dover. ii
+228 pp.
Goldstein, S. 1929. On the vortex theory of
screw propellers. Proc. R. Soc. Lndon Ser.
A I23 : 5
Goodman, T. R. 1979. Momentum theory of
a propeller in a shear fow. J. Ship Res.
23 : 242-52
Greeley, D. S. 1982. Marine propeller blade
tip fows. Rp. 82-3, Dep. Ocean Eng.,
Mass. Inst. Techno!., Cambridge
Greeley, D. S., Kerwin, !.E. 1982. Numerical
methods for propeller design and analysis
in steady fow. Soc. Nav. Archit. Mar. Eng.
Trans. 90 : 415-53
Guilloton, R. 1957. Calcul des vitesses in
duites en vue du trace des helices. Schifs
technik 4: 101-9
Hanaoka, T 1962. Hydrodynamics of an
oscillating screw propeller. Proc. Symp.
Nav. Hydrodyn., 4th, pp. 79-124. Washing
ton, DC: Nat!. Acad. Press
Hanaoka, T. 1969. Numerical lifting surface
theory of d screw propeller in non-uniform
fow (Part 1 : Fundamental theory). Rep.
Ship Res. Inst., Tokyo 6(5) . 1-14 (In
Japanese)
Hess, J. L. 1975. Review of integral-equation
techniques for solving potential-fow prob
lems with emphasis on the surface-source
method. Comput. Methods Appl. Mech.
Eng. 5 : 145-96
Hess, J. L., Smith, A. M. O. 1967. Calculation
of potential fow about arbitrary bodies.
Prog. Aeronaut. Sci. 8 : 1-137
Hess, J. L., Valarezo, W. 0. 1985. Calculation
of steady fow about propellers by means
of a surface panel method. AI AA Pap. No.
85-0283
Holden, K., Sontvedt, T., Ofsti, O. 1974. On
stability and volume of marine propeller
cavitation and corresponding spectral dis
tribution in hull pressure felds. Proc.
Symp. High Powered Propulsion Large
Ships, Wageningen, N eth.
Huang, T., Groves, N. 1980. Efective wake :
theory and experiment. Proc. Symp. Nav.
Hydrodyn., 13th, Tokyo, pp. 651-73
James, R. M. 1972. On the remarkable accu
racy of the vortex lattic method. Comput.
Methods Appl. M echo Eng. I .59-79
Kerwin, J. E. 1959. Machine computation of
marine propeller characteristics. Int. Ship
build. Prog. 6: 343-54
Kerwin, J. E. 1961. The solution of propeller
lifting surface problems by vortex lattic
methods. Rep., Dep. Ocean Eng., Mass.
Inst. Techno!., Cambridge
Kerwin, J. E. 1979. The effect of trailing
vortex asymmetry on unsteady propller
blade forces. Rep., Dep. Ocean Eng., Mass.
Inst. Techno!., Cambridge
Kerwin, J. E., Lee, C.-S. 1978. Prediction of
steady and unsteady marine propeller per
formanc by numerical lifting-surfac
theory. Soc. Nav. Archit. Mar. Eng. Trans.
86 : 21 8-53
Kim, K. H., Kobayashi, S. 1984. Pressure
distribution on propeller blade surface using
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
numerical lifting-sur!ace theory. Paper pre
sented at SNAME Propellers '84 Symp.,
Virgnia Beach, Va.
Lan, C. E. 1974. A quasi-vortex-lattice
method in thin wing theory. J. Aircr.
1 1 : 518-27
Lerbs, H. W. 1952. Moderately loaded pro
pellers with B fnite number of blades and
an arbitrary distribution of circulation.
Soc. N avo Archit. Mar. Eng. Trans. 60 : 73-
1 17
Lighthill, M. J. 1951. A new approach to thin
aerofoil theory. Aeronaut. Q.3 : 193-210
Ludwieg, H., Ginzel, J. 1944. Zur Theorie der
Breitblattschraube. Rep. 44/A/08, Aero
dyn. Versuchsanst., Gottingen, Ger.
McCarthy, J. H. 1961. On the calculation of
thrust and torque fuctuations of propel
lers in nonuniform wake fow. DTNSRDC
Rep. 1533, David Taylor Nav. Ship Res.
Dev. Cent., Bethesda, Md. iii H33 pp.
McCroskey, W. J. 1982. Unsteady airfoils.
Ann. Rev. Fluid Mech. 14: 28531 1
Moran, J. 19&4. An Introduction t o Theoreti
cal and Computational Aerodynamics. New
York: Wiley. 4 pp.
Morgan, W. B., Silovic, V., Denny, S. B. 1968.
Propeller lifting-surface corrections. Soc.
N avo Archit. Mar. Eng. Trans. 76: 309-47
Pien, P. C. 1961. The calculation of marine
propellers based on lifting surface theory.
J. Ship Res. 5(2) : 114
Prandtl, L. 1921. Application of modem
hydrodynamics to aeronautics. Nat!.
Advis. Comm. Aeronaut. Ann. Rep., 7th, pp.
157-215
Report of the Propeller Committee. 1984.
Proc. Int. Towing Tank Conf, 1 7th, Gite
borg, 1 : 140-270
Schwanecke, H. 1975. Comparative calcu
lations of unsteady propeller blade forcs.
Report of the Propeller Committee. Proc.
Int. Towing Tank Con{., 14th, 3 : 441-47
Sparenberg, J. A. 1959.
A
pplication of lifting
surfac theory to ship screws. Proc. K. Ned.
Akad. Wet.-Amsterdam, Ser. B 62(5) :
286- 98
Strscheletzky, M. 1950. Hydrodynamische
Grund/agen zur Berechnung der Schif s
schrauben. Karlsruhe : G. Braun. 257 pp.
Tanibayashi, H. 1980. Practical approach to
unsteady problems of marine propellers by
quasi-steady method of calculation. Mit-
U PROPELLERS 403
subishi Tech. Bull. No. 143, Tokyo
Tsakonas, S., Jacobs, W. R., Rank, P. H. Jr.
1968. Unsteady propeller lifting-surface
theory with fnite number of chordwise
modes. J. Ship Res. 12 : 14-45
Tsakonas, S., Jacobs, W. R., Ali, M. R. 1973.
An "exact" linear lifting-surface theory for
a marine propeller in a non-uniform fow
feld. J. Ship Res. 17 : 196- 207
Tsakonas, S., Breslin, J. P., Jacobs, W. R.
1983. Blade pressure distribution for a
moderately loaded propeller. J. Ship Res.
27: 39-55
van Gent, W. 1977. On the use of lifting
surface theory for moderately and heavily
loaded ship propellers. NSMB Publ. No.
536, Neth. Ship Model Basin, Wageningen
Van Houten, R. J., Kerwin, J. E., Uhlman, J.
S. 1983. Numerical solutions of lifting
surfac sheet cavitation. Proc. Gen. Meet.
Am. Towing Tank Cor, 20th, Davidson
Lab., Stevens Inst. Technol., Hoboken, N.J.,
1 : 393-30
van Manen, J. D. 1957. Fundaentals of ship
resistance and propulsion. Part B : Propul
sion. Publ. No. 132a, Neth. Ship Model
Basin, Wageningen. 14 pp.
van Manen, J. D., Bakker, A. R. 1962.
Numerical results of Sparenberg's lifting
surface theory for ship screws. Proc. Symp.
Nav. Hydrodyn., 4th, Washington, DC, pp.
63-77
Versmissen, O. O. P., van Oent, W. 1983.
Hydrodynamic pressure measurements on
a ship model propeller. Proc. Symp. Nav.
Hydrodyn., 14th, pp. 787-822. Washing
ton, DC: Nat!. Acad. Press
Vorus, W. S. 1976. Caculation of propeller
induced hull forces, force distributions,
and pressures ; Free surface efects. J. Ship
Res. 20: 107-17
Vorus, W. S., Breslin, J. P., Tein, Y. S. 1978.
Calculation and comparison of propeller
unsteady pressure forces on ships. Proc.
Ship Vib. Symp. '78, Washington, DC, p. 1-1
Wang, M.-H. 1985. Hub eff ects in propeller
design and analysis. PhD thesis. Dep.
Ocean Eng., Mass. Inst. Techno!., Cam
bridge
Yamazai, R. 1962. On the theory of screw
propellers. Proc. Symp. Nav. Hydrodyn.,
4th, Washington, DC, pp. 1-28
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
8
6
.
1
8
:
3
6
7
-
4
0
3
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
N
I
V
E
R
S
I
T
Y

O
F

M
I
A
M
I

o
n

0
1
/
1
1
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.