JOURNAL OF PROPULSION AND POWER

Vol. 10, No. 5, Sept.-Oct. 1994

Design of Optimum Propellers

Charles N. Adkins*
Falls Church, Virginia 22042
and
Robert H. Liebeckt
Douglas Aircraft Company, Long Beach, California 90846
Improvements have been made in the equations and computational procedures for design of propellers and
wind turbines of maximum efficiency. These eliminate the small angle approximation and some of the light
loading approximations prevalent in the classical design theory. An iterative scheme is introduced for accurate
calculation of the vortex displacement velocity and the flow angle distribution. Momentum losses due to radial
flow can be estimated by either the Prandtl or Goldstein momentum loss function. The methods presented here
bring into exact agreement the procedure for design and analysis. Furthermore, the exactness of this agreement
makes possible an empirical verification of the Betz condition that a constant-displacement velocity across the
wake provides a design of maximum propeller efficiency. A comparison with experimental results is also
presented.

a
a'
B
b
Cd
C,
Cp
CT
Cx
Cy
c
D
D'
F
G
/
K
L'
n
P
Pc
Q
R
r
T
Tc
V
v'
W
wn
wt
x
a
j8
F
s

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

17
A
g
\e
0
p
a
<f>
<f)t
fl

Nomenclature
axial interference factor
rotational interference factor
number of blades
axial slipstream factor
blade section drag coefficient
blade section lift coefficient
power coefficient, P/pn3D5
thrust coefficient, T/pn2D4
torque force coefficient
thrust force coefficient
blade section chord
propeller diameter, 2R
drag force per unit radius
Prandtl momentum loss factor
circulation function
advance ratio, VlnD
Goldstein momentum loss factor
lift force per unit radius
propeller rps
power into propeller
power coefficient, 2P/pV37rR2
torque
propeller tip radius
radial coordinate
thrust
thrust coefficient, 2T/pV27rR2
freestream velocity
vortex displacement velocity
local total velocity
velocity normal to the vortex sheet
tangential (swirl) velocity
nondimensional distance, lrlV
angle of attack
blade twist angle
circulation
drag-to-lift ratio

= displacement velocity ratio, v'lV

= propeller efficiency
= speed ratio, V/flR
= nondimensional radius, r/R = AJC
= nondimensional Prandtl radius
= nondimensional hub radius
= fluid density
= local solidity, Bc/2/nr
flow angle
= flow angle at the tip
= propeller angular velocity

Superscript
' = derivative with respect to r or , unless otherwise
noted
Introduction
N 1936, a classic treatise on propeller theory was authored
by H. Glauert.1 In this work, a combination of momentum
theory and blade element theory, when corrected for momentum loss due to radial flow, provides a good method for
analysis of arbitrary designs even though contraction of the
propeller wake is neglected. Although the theory is developed
for low disc loading (small thrust or power per unit disc area),
it works quite well for moderate loading, and in light of its
simplicity, is adequate for estimating performance even for
high disc loadings. The conditions under which a design would
have minimum energy loss were stated by A. Betz2 as early
as 1919; however, no organized procedure for producing such
a design is evident in Glauert's work. Those equations which
are given by Betz make extensive use of small-angle approximations and relations applicable only to light loading conditions. Theodorsen3 showed that the Betz condition for minimum energy loss can be applied to heavy loading as well.
In 1979, E. Larrabee4 resurrected the design equations and
presented a straightforward procedure for optimum design.
However, there are still some problems: first, small angle
approximations are used; second, the solution for the displacement velocity is accurate only for vanishingly small values (light loading), although an approximate correction is
suggested for moderate loading; and third, there are viscous
terms missing in the expressions for the induced velocities.
These viscous terms must be included in the design equations
if they are to be consistent with the classical propeller analysis.
This approach is given later.

Presented as Paper 83-0190 at the AIAA 21st Aerospace Sciences

Meeting, Reno, NV, Jan. 10-13, 1983; received July 23, 1992; revision received July 6, 1993; accepted for publication Dec. 15, 1993.
Copyright 1994 by the American Institute of Aeronautics and
Astronautics, Inc. All rights reserved.
*Consulting Engineer. Member AIAA.
tMcDonnell Douglas Fellow. Fellow AIAA.
676

ADKINS AND LIEBECK:

The purpose of this article is to correct these difficulties

and bring the design method into exact agreement with the
analysis. It is then possible to verify empirically the optimality
of the design. This work was initiated at McDonnell Douglas
in 1980 in response to a requirement for simple estimates of
propeller performance. In-house methods, if they existed, had
been irretrievably archived. An early version was presented
as AIAA Paper 83-0190. Continuous requests for copies of
the paper plus some refinements to the method have motivated its publication in the Journal.

VORTEX FILAMENT
AFTER TIME
INCREMENT, At

AXIS

VORTEX FILAMENT (t = 0)

Momentum Equations
Detailed axial and general momentum theory is described
by Glauert,1 and only a brief summary is given here to emphasize several important features. Consider a fluid element
of mass dm, far upstream moving toward the propeller disc
in a thin, annular stream tube with velocity V. It arrives at
the disc with increased velocity, V(l + 0), where a is the
axial interference factor. At the disc, dm exists in the annulus
27rr dr, and the mass rate per unit radius passing through the
disc is 2irrpV(l + a), neglecting radial flow. The element dm
moves downstream into the far wake, increasing speed to the
value V(l 4- &), where b is the axial slipstream factor. Axial
momentum theory determines b to be exactly 20, whereas the
general theory (which includes rotation of the flow) determines b to be approximately 2a. Using the axial approximation, which is generally accepted, the overall change in
momentum of the element is 2VaF dm where F, the momentum loss factor, accounts for radial flow of the fluid. The
thrust per unit radius T', acting on the annulus can now be
expressed as

Fig. 2 Definition of displacement velocity v' in the propeller wake.

develops the Betz condition for heavy loading by including

the contraction of the wake. He shows that sufficiently far
downstream in the contracted wake, the vortex sheet must
be the same regular screw surface for a propeller of minimum
induced energy loss. This optimum vortex sheet acts as an
Archimedean screw, pumping fluid aft between rigid spiral
surfaces.
At the blade station, r, the total lift per unit radius is given
by

L> = f = BpWT

(2)

and in the wake, the circulation in the corresponding annulus

is
BY = 27rrFwf

i rri

T = = 27rrpV(l + a)(2VaF)

(3)

(la)

By similar arguments, the torque per unit radius Q' is given

by

Q'lr = 2irrpV(l + a)(2Slra'F)

677

DESIGN OF OPTIMUM PROPELLERS

(Ib)

Flow geometry about a blade element at the disc is shown in

Fig. 1, where W acts on the blade element with a, and acts
on the disc at <j>. F goes from about 1 at the hub (where the
radial flow is typically negligible), to 0 at the tip, and is not
unlike the spanwise loading of a wing. The functional form
of this factor was first estimated by Prandtl1-2 and a more
accurate, though more complex, form was determined by
Goldstein5 and Lock.6"8

Setting the circulation F in Eq. (2) equal to that in Eq. (3)

will ultimately determine that circulation distribution F(r) that
minimizes the induced power of the propeller.
In order to obtain F(r), it is necessary to relate vvr to a more
measurable quantity. Figure 2 shows the wake vortex filament
at station r and the definition of the various velocity components there. The motion of the fluid must be normal to the
local vortex sheet, and this normal velocity is wn. Therefore,
the tangential velocity is given by

Wt = wn sin </>
However, for a coordinate system fixed to the propeller disc,
the axial velocity of the vortex filament would be

Circulation Equations

v' = H^/COS (/>

At each radial position along the blade, infinitesimal vortices are shed and move aft as a helicoidal vortex sheet. Since
these vortices follow the direction of local flow, the helix angle
of the spiral surface is ((>, shown in Fig. 1. The Betz condition
for minimum energy loss, neglecting contraction of the wake,
requires the vortex sheet to be a regular screw surface; i.e.,
rtan <t> must be a constant independent of radius. Theodorsen3

where the increase in magnitude of v' over wn is due to rotation of the filament. This is analogous to a barber pole where
it appears that the stripes are translating in spite of the fact
that only a rotational velocity exists. It will become clear that
it is convenient to use v', and the corresponding displacement
velocity ratio, = v'lV. The tangential velocity is then

Wt = V sin cf> cos 4>

and the circulation of Eq. (3) can be expressed as
F - 27rV2G/(Bty

G = F x cos (/> sin </>

and x is the speed ratio given by
DISC PLANE

Fig. 1 Flow geometry for blade element at radial station r.

(4)

(5)

678

ADKINS AND LIEBECK:

DESIGN OF OPTIMUM PROPELLERS

Constraint Equations
For design, it is necessary to specify either 7, delivered by
the propeller or the power P, delivered to the propeller. The
nondimensional thrust and power coefficients used for design
are

Tc = 2T/(pV2irR2)
3

(9a)
3

Pc = 2P/(pV 7rR ) = 2Q(l/(pV >7TR )

(9b)

and using these definitions, Eq. (6) can be written as

T'c = I( - Itf

(lOa)

P'C=J[{ + J&

(lOb)

where the primes denote derivatives with respect to f, and

DISC PLANE-

Fig. 3 Force diagram for a blade element.

/; = 4G(1 - e tan 0)

(lla)

The circulation equations for thrust T , and torque Q', per

unit radius can be written by inspection of Fig. 3 as

1'2 = A(/;/2)(l + e/tan </>)sin <j> cos </>

(lib)

/; - 4fG(l + e/tan 0)

(lie)

T = L' cos 0 - D' sin 4> = L' cos 0(1 - e tan <)

(6a)

J'2 = (/;/2)(l - s tan </>)cos 0

(lid)

Since is constant for an optimum design, a specified thrust

gives the constraint equations

Q'lr = L' sin < + D' cos 0 = L' sin </>(! + e/tan

(6b)
where e is the drag-to-lift-ratio of the blade element. Next,
using Eq. (2), L' can be replaced by F(r) which, in turn, is
related to conditions in the wake by Eq. (3). Based on the
flow in the wake, F(r) is given by Eqs. (4) and (5), and T
and Q'lr are reduced to being functions of <f> and the displacement velocity, = v'lV. The local flow angle <f> will
clearly be a function of the radius; however, at this stage of
the analysis, the optimum distribution (r) is not yet determined. Several diagrams and an excellent photograph of the
vortex sheet can be found in a 1980 work by Larrabee.9

Condition for Minimum Energy Loss

At this point, a departure from Larrabee's4 design procedure is made, and the momentum equations, Eqs. (1), and
the circulating equations, Eqs. (6), are required to be equivalent. This condition results in the interference factors being
related to by the equations
2

a = (/2)cos ((l - tan </>)

(7a)

a' '= (/2*)cos (/> sin <(! + e/tan </>)

(7b)

where Eqs. (4) and (5) have been used to express L' in terms
of , and the terms in epsilon correctly describe the viscous
contribution. Equations (7), together with the geometry of
Fig. 1, lead to the important simple relation
tan < =

/2)/x - (1 + /2)X/(

(8)

Here, A is a constant, and varies from 0 at the hub to unity

at the edge of the disc. The relation between the two nondimensional distances and the constant speed ratio is
x

= (r/R)/\ = f/A

Recalling the Betz2 condition, r tan </> = const, Eq. (8)

proves that for the vortex sheet to be a regular screw surface,
must be a constant independent of radius. This is the condition for minimum energy loss. It should be noted that Eq.
(8) results from Eq. (7) whether viscosity is included or not.

= (A/2/2) - [(A/2/,)2- TC/I2]2

(12)

PC = J^ + / 2 f 2

(13)

Similarly, if power is specified, the constraint relations are

= ~(JJ2J2) + [(V2/2)2 + PJJ^'2

(14)

Tc = U - I22

(15)

where the integration has been carried out over the region
f = & t o f = 1.

Blade Geometry
For the element dr of a single blade at radial station r, let
c be the chord and Cl the local lift coefficient. Then, the lift
per unit radius of one blade is

= Pwr
where F is given by Eq. (4). It follows directly that
We = 47rXGVR/(ClB)

(16)

Assume for the moment that is known; then the local value
of cj) is known from Eq. (8), and the above relation is a
function only of the local lift coefficient. Since the local Reynolds number is We divided by the kinematic viscosity, Eq.
(16) plus a choice for C/ will determine the Reynolds number
and , from the airfoil section data. The total velocity is then
determined by Fig. 1 as
W = V(l + fl)/sin

(17)

where a is given by Eq. (7), and the chord is then known from
Eq. (16). If the choice for C/ causes to be a minimum, then
viscous as well as momentum losses will in most cases be
minimized, and overall propeller efficiency will be the highest
possible value. For preliminary considerations, it is usually
sufficient to choose one Ch the design Ch for determining
blade geometry. (Any Cl specification is permissible as long
as the optimum blade loading distribution, cC/(r), is maintained.) Since a is known from C, and Reynolds number, the
blade twist with respect to the disc is j8 = a. + <f). G is zero

ADKINS AND LIEBECK:

679

DESIGN OF OPTIMUM PROPELLERS

at the edge of the disc, and the tip chord is therefore always
zero for a finite lift coefficient.

Design Procedure
Either F or K, relation for the momentum loss function can
be selected. For the sake of simplicity, only the Prandtl relation is described as

F = (2/77)arc

(18)

f=

(19)

where

and </>, is the flow angle at the tip. From Eq. (8)
DISC PLANE-

tan <t>t = A(l + /2)

(20)

so that a choice for determines the function F as well as <f>

by

tan $ = (tan <,

Analysis of Arbitrary Designs

The analysis method is outlined here in order to discuss
problems of convergence for off design and for square-tipped
propellers in general, and to point out two minor errors in
Glauert's work. Figure 4, which is simply an alternate version
of Fig. 3, shows the relation between the propeller force coefficients, Cy and Cx, and the airfoil coefficients, C, and Cd. The
equations are

Cx = C, sin </> + Cd cos </> = C7(sin </> + e cos </>)

and the relations for the thrust 7" and torque Q' per unit
radius are then

T = ($)pW2BcCy

(21)

which is simply the condition that the vortex sheet in the wake
is a rigid screw surface (r tan < = const). For an initial value,
= 0 will suffice.
The design is initiated with the specified conditions of power
(or thrust), hub and tip radius, rotational rate, freestream
velocity, number of blades, and a finite number of stations
at which blade geometry is to be determined. Also, the design
lift coefficientone for each station if it is not constant
must be specified. The design then proceeds in the following
steps:
1) Select an initial estimate for ( = 0 will work).
2) Determine the values for F and <j) at each blade station
by Eqs. (18-21).
3) Determine the product We, and Reynolds number from
Eq. (16).
4) Determine e and a from airfoil section data.
5) If e is to be minimized, change C, and repeat Steps 3
and 4 until this is accomplished at each station.
6) Determine a and a' from Eq. (7), and Wfrom Eq. (17).
7) Compute the chord from step 3, and the blade twist
)3 = a + (f>.
8) Determine the four derivatives in / and / from Eq. (11)
and numerically integrate these from = 0 to = 1.
9) Determine and Pc from Eqs. (12) and (13), or and
Tc from Eqs. (14) and (15).
10) If this new value for is not sufficiently close to the
old one (e.g., within 0.1%) start over at step 2 using the
new .
11) Determine propeller efficiency as TJPC, and other features such as solidity.
The above steps converge rapidly, seldom taking more than
three or four cycles. An accurate description of viscous losses
can be obtained by creating another design with e equal to
zero and noting the difference in propeller efficiency.

Cy = C, cos <f> - Cd sin < = C,(cos </> - e sin </>)

Fig. 4 Force coefficients for propeller blade element analysis.

Q'lr = ($)PW BcCx

(22a)
(22b)

Again, it is required that the loading Eqs. (22) be exactly

equal to the momentum result Eqs. (1). With the use of the
flow geometry in Fig. 1, this requires the interference factors
to be
a = o-KI(F - <rK)

(23a)

a' = <rK'/(F + aK')

(23b)

K = Cy/(4 sin2 (/>)

(24a)

K' = Cx/(4 cos 0 sin </>)

(24b)

where

and cr is given by

er = Bc/(2m)
Equations (23) correct the placement of the factor F used by
Glauert in his equations (5.5) of Chapter VII as identified by
Larrabee.4
The relation for the flow angle is obtained from Fig. 1 and
Eqs. (23) as
tan <t> = [V(l + a)]/[nr(l - a')]

(25)

For determining the function, F, in Eq. (18), Glauert suggests

the relation sin $, = f sin 4> be used in Eq. (19). It is recommended that Eq. (21) be used instead, i.e.,
tan </>, = tan </>

which is exact for the analysis of an optimally designed propeller at the design point.
The analysis procedure requires an iterative solution for
the flow angle </> at each radial position, . An initial estimate
for </> can be obtained from Eq. (8) by setting equal to zero.
Since j3 is known, the value for a in Fig. 3 is /3 <, and the
airfoil coefficients are known from the section data. The
Reynolds number is determined from the known chord and
W, which is obtained from Fig. 1 and Eq. (23a), and the new
estimate for < is then found from Eq. (25). A direct substitution of the new <f> for the old value will cause adequate
convergence for an optimum design which is being analyzed
at the design point. However, for analysis off-design and for

680

ADKINS AND LIEBECK:

DESIGN OF OPTIMUM PROPELLERS

nonoptimum designs, some recursive combination of the old

and new values for <f> is required to cause adequate convergence. Under some conditions (usually near the tip), convergence may not be possible at all due to large values for
the interference factors, a and a', in Eq. (23). Since Fis zero
at the tip and a is not for a square tip propeller, the value
for a is - 1 and a' is +1. Such values are physically impossible
since the slipstream factors are approximately twice the values
at the rotor plane. Wilson and Lissaman10 suggest empirical
relations for resolving this problem, whereas Viterna and
Janetzke11 give empirical arguments for clipping the magnitude of a and a' at the value 0.7 (a/F at the tip is finite at the
design point for an optimum propeller).
For analysis, the conventional thrust and power coefficients
are

CT = TI(pn2D4)
Cp = P/(pn3D5)
Using Eqs. (22) and (24), the differential forms with respect
to f are given by

c; =
When these have been integrated from the hub to the tip, the
propeller efficiency is

77 = CTJ/CP

Airfoil Section Data

The proper accounting for blade section (airfoil) characteristics has proven essential for accurate and reliable esti2.0-.

-20

70

80

NACA4415

RN = 0.5 x10 6
-1.0 J

0.2-,
"C/4

-30

0.1-

10

-20

20

30
40
a (DEG)

50

(26)

where the effect of profile drag on thrust has been ignored.

In order to study this problem empirically, consider a general
function H(x) and the two first-order terms of its Laurent
series, IIx and x, and describe the displacement velocity distribution as

= 0

where/ = V/(nD) is the advance ratio. Propeller performance

is typically described by plots of Cr, Cp, and 17 vs /.

-30

Empirical Optimality
In Chapter VII of Glauert's work, his equation (2.20) shows
that when blade friction is neglected, the most favorable distribution of circulation is where the displacement velocity is
constant across the wake. Here, the term x2/(l + ;c2) is the
small-angle approximation of G, given by Eq. (5) in this article. The effect of profile drag is shown by Glauert in his
equation (3.5), which states that the optimum distribution for
the displacement velocity ratio is
f = Io -

<rK')cos <t>}2

C'T =

mation of propeller performance, and the extent of airfoil

data required for propeller analysis exceeds that which is typically used in analysis of wings. Between takeoff, climb, and
cruise, propellers typically operate over a relatively wide range
of blade section Reynolds numbersa variation by factor of
5 is not uncommon. Also, during takeoff, climb, and windmilling, some portion of a propeller blade is likely to be stalled,
either positively or negatively. An example set of characteristics for the NACA 4415 airfoil is given in Fig. 5. These
would be supplemented by additional drag data in the unstalled region for a range of Reynolds numbers.

70

80

-.03-.04 J

Fig. 5 Example of airfoil section characteristics including normal and

chord force coefficients.

32/x

(27)

which includes the case of Eq. (26). It is desired to find values

for 8l and 82 which maximize propeller efficiency subject to
the constraints of Eqs. (10). To solve this problem, in Eqs.
(10) is replaced by Eq. (27). Then, a choice for 5, and 52 will
enable a determination of 0 and a calculation of overall propeller efficiency. A systematic study of various propeller conditions was undertaken using the design and analysis procedures of this article. Nonzero values for 5t and 82 that caused
an increase in propeller efficiency could not be found under
any conditions. Therefore, it was concluded that a constant
displacement velocity is at least locally optimum whether profile drag is considered or not. Momentum and viscous losses
are then uncoupled; former is minimized by constant displacement velocity, the latter by choosing a C, distribution so
that the drag-to-lift ratio is a minimum everywhere.
Some may criticize the authors for including viscous terms
in the development of the optimized circulation distribution
for the propeller design problem. Certainly, this is contrary
to classical lifting-line wing theory where the analogous elliptic
lift distribution is obtained inviscidly. However, it has been
shown that the combined momentum-circulation equations,
Eqs. (7), produce the screw surface equations, Eqs. (8), whether
viscosity is included or not. It is evident in Eq. (11) that the
momentum and viscous terms are directly separable for use
in Eq. (10). The viscous terms, when integrated in Eq. (10),
account for the difference in power and thrust between an
inviscid and viscous propeller design. If one accepts the classical blade element analysis equations as a measure of performance, the momentum and viscous losses are indeed uncoupled, and the viscous design equations produce a propeller
with minimal losses.
Alternatively, if one is still concerned about including viscosity in the design procedure, a propeller could be designed
inviscidly and the blade section drag simply added to the
inviscid design. This would require that the analysis procedure
be applied with drag set equal to zero when solving for the
induction velocities. The performance of the propeller would
then be obtained by adding drag to this inviscid solution without altering the induction velocities. However, this would

ADKINS AND LIEBECK:

681

DESIGN OF OPTIMUM PROPELLERS

Table 1 Propeller design solution

0.3424
0.4605
0.4269
0.3569
0.2796
0.1913
0.0000

0.5000
0.8958
1.2917
1.6875
2.0833
2.4792
2.8750

ft
58.3125
41.8645
32.2669
22.2978
18.7971
15.9619
13.8552

*
54.8118
38.3637
28.7661
22.7927
18.7971
15.9619
13.3552

RN

a'

0.4449
0.8104
0.9834
1.0295
0.9740
0.7830
0.0000

0.0348
0.0644
0.0804
0.0890
0.0938
0.0968
0.0000

0.0633
0.0365
0.0219
0.0142
0.0098
0.0072
0.0000

Input: brake horsepower = 70, 2 blades; hub diam = 1 ft, tip diam = 5.75 ft; blade section: NACA
4415, C, = 0.7, velocity = 110 mph, rpm = 24001.
Output: thrust = 207.61 Ib, 77 = 0.86996.
Note: a and a' have been set equal to zero at the tip.

Table 2 Propeller analysis solution

c,

*
54.8116
38.3638
28.7661
22.7927
18.7971
15.9619
12.5862

0.5000
0.8958
1.2917
1.6875
2.0833
2.4792
2.8750

0.7000
0.7000
0.7000
0.7000
0.7000
0.7000
0.7000

RN

a'

0.4449
0.8104
0.9834
1.0295
0.9740
0.7830
0.0000

0.0348
0.0644
0.0804
0.0890
0.0938
0.0968
0.0000

0.0633
0.0365
0.0219
0.0142
0.0098
0.0072
0.0000

Input: propeller geometry from Table 1; r, C, and /3 at the same radial locations; velocity =110 mph,
rpm = 2400.
Output: brake horsepower = 70, thrust = 207.61 Ib, 17 = 0.86996.
1.6
1.4

1.2
1.0

0.8

"7 DESIGN

0.6
0.4
0.2

DISC PLANE

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

rR

/ T.P

Fig. 6 Force coefficients for windmill blade element.

Fig. 8 C, distributions for example propeller.

0.08

0.8

0.06

0.6

0.20

1.0

CP.CT
0.8
0.04

0.15

0.02

0.05
0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

REF: NACA TN 1834, PROP MODEL 5

Fig. 7 Example of propeller performance.

require an additional layer of iteration to achieve a specified

design thrust or power. In light of the favorable agreement
between the present theory and the experimental results given
later in this article, it is argued that such an increase in complexity is not justified.

Windmills
All of the analyses described in this article are directly
applicable to the windmill problem after a minor adjustment
in the angle definitions of Fig. 1. The corresponding flow

Fig. 9 Comparison of theory and experiment.

geometry for a windmill is shown in Fig. 6, where the primary

distinction is that the blade section is inverted (as compared
with a propeller), and the local angle of attack is measured
from below the local velocity vector. Corresponding relations
for the angles are
windmill

a = </> - /3

propeller a = ft $

682

ADKINS AND LIEBECK:

0.3

DESIGN OF OPTIMUM PROPELLERS

Conclusions and Recommendations

J = 0.914
THEORY WITH PRANDTL F
- THEORY WITH GOLDSTEIN K
0 NACATEST

0.1

REF: NACA TN 1834, PROP MODEL 5

0.2

0.4

0.6

0.8

1.0

r/R-np

Fig. 10 Comparison of propeller analyses thrust coefficient.

as shown in Figs. 6 and 1, respectively. In these figures, C,

for the windmill is negative with respect to that for the propeller, and this sign change together with the angle definition
will convert the propeller methods to the windmill application.
For the design case, the input Pc value should be negative,
and the resulting values of v' (and the interference factors a
and a') and Tc will also be negative. (Thrust is of less interest
for a windmill since it typically represents the tower load and
is not a main performance parameter.) Similarly, the analysis
results for a windmill rotor will yield negative values for both
Pc and Tc.

Examples
As a sample calculation, the design of a propeller for a light
airplane is considered. The design conditions and the resulting
design are described in Table 1 which gives for each radial
station: blade chord, blade pitch angle, local flow angle, local
Reynolds number, and the interference coefficients a and a'.
This propeller geometry has, in turn, been analyzed at the
design condition and the result is given in Table 2. Agreement
is virtually exact. Analysis over a range of values of the advance ratio, J = V/(nD), provides the typical propeller performance plots which are shown in Fig. 7, and Fig. 8 gives
the blade lift coefficient distribution over a range of /s where
the design condition is the C, 0.7 and is a constant line at
J = 0.7.
A calibration of the method is given by comparing its theoretical prediction with experimental results. Reid12 has evaluated several conventional propellers extensively by experiment, and one of these has been chosen for comparison.
Figure 9 gives Cp, CT, and 77 vs / for both Reid's experiments
and the corresponding theoretical prediction. The agreement
here is quite good, with most of the disparity occurring after
the blade is stalled. This propeller uses NACA 16-series airfoils, and no poststall data were available.
Figure 10 gives the comparison of the blade thrust coefficient distribution as measured by Reid and calculated by the
method. Two theoretical results are shown: one using F, and
the other using the more complex (from a calculation point
of view) K. In principle, the accuracy of the method should
be better with the Goldstein factor for a propeller with few
bladesthis example had three bladesand the two factors
should give similar results as the number of blades is increased. The results of Fig. 10 confirm this trend, and the
overall comparison for both factors is regarded as quite good.

The propeller theory of Glauert has been extended to improve the design of optimal propellers and refine the calculation of the performance of arbitrary propellers. Extensions
of the theory include 1) elimination of the small angle assumptions in the optimal design theory; 2) accurate calculation of the vortex displacement velocity which properly accounts for the blade section drag; and 3) elimination of the
small angle assumptions in the Prandtl momentum loss function for both design and analysis. These extensions bring the
design and analysis procedures to exact numerical agreement
within the precision of computer analysis.
The primary approximation remaining in both procedures
is the use of the axial momentum equations which require the
increase in wake velocities to be twice those at the disc. Under
certain conditions this approximation is not good and gives
rise to the unnatural conditions and convergence problems
described in the analysis section. Improvements might be made
by replacing the axial momentum equations with relations
more closely aligned with the general theory, particularly in
those differential stream tubes in which "heavy loading" exists. Such conditions appear to be more prevalent in the analyses at off-design conditions than in the design itself, and,
when combined with poststall misknowledge, can lead to large
errors in analysis. However, for design and analysis within
the conventional operating regime, both procedures are simple, accurate, and reliable. This method has been extended
by Page and Liebeck13 to the design and analysis of dualrotation propellers. A favorable comparison between theory
and experiment was also observed.

References
l
Glauert, H., "Airplane Propellers, "Aerodynamic Theory, edited
by W. Durand, Div. L, Vol. 5, Peter Smith, Gloucester, MA, 1976,
pp. 169-269.
2
Betz, A., with appendix by Prandtl, L., "Screw Propellers with
Minimum Energy Loss," Gottingen Reports, 1919, pp. 193-213.
3
Theodorsen, T., Theory of Propellers, McGraw-Hill, New York,
1948.
4
Larrabee, E., "Practical Design of Minimum Induced Loss Propellers," Society of Automotive Engineers, Business Aircraft Meeting and Exposition, Wichita, KS, April 1979.
5
Goldstein, S., "On the Vortex Theory of Screw Propellers," Proceedings of the Royal Society of London, Series A, Vol. 123, 1929,
pp. 440-465.
6
Lock, C., "The Application of Goldstein's Airscrew Theory to
Design," British Aeronautical Research Committee, RM 1377, Nov.
1930.
7
Lock, C., "An Application of Prandtl's Theory to an Airscrew,"
British Aeronautical Research Committee, RM 1521, Aug. 1932.
8
Lock, C., "Tables for Use in an Empirical Method of Airscrew
Strip Theory Calculations," British Aeronautical Research Committee, RM 1674, Oct. 1934.
9
Larrabee, E., "The Screw Propeller," Scientific American, Vol.
243, No. 1, 1980, pp. 134-148.
10
Wilson, R., and Lissaman, P., "Applied Aerodynamics of Wind
Power Machines," Oregon State Univ., NSF/RA/N-74-113, PB2318595/3, Corvallis, OR, July 1974.
H
Viterna, A., and Janetzke, D., "Theoretical and Experimental
Power from Large Horizontal-Axis Wind Turbines," Proceedings from
the Large Horizontal-Axis Wind Turbine Conference, DOE/NASALeRC, July 1981.
12
Reid, E. G., "The Influence of Blade-Width Distribution on
Propeller Characteristics," NACA TN 1834, March 1949.
13
Page, G. S., and Liebeck, R. H., "Analysis of Dual-Rotation
Propellers," AIAA Paper 89-2216, Aug. 1989.