Design of a Quiet Propeller for an Electric Mini

Unmanned Air Vehicle

Ohad Gur

and Aviv Rosen

TechnionIsrael Institute of Technology, Haifa 32000, Israel

DOI: 10.2514/1.38814
Designing a quiet and efcient propeller is a demanding task because these two goals often lead to contradictory
design trends. The task becomes even more complicated when additional constraints are introduced. This paper
presents a multidisciplinary design optimization process for designing a quiet propeller under various constraints.
The acoustic model of the propeller is presented in detail, including its validation. The newprocess is used to design a
quiet propeller for an electric mini unmanned air vehicle. This design is subjected to power, structural, and side
constraints. The design variables include blade geometry, blade cone angle, propeller radius, number of blades, and
rotational speed. It is shown that the motor characteristics have an important inuence on the optimization, and it is
therefore important to take theminto account during the design process. The inuences of the various parameters on
the design and propellers characteristics are presented and discussed in detail.
Nomenclature
o = speed of sound
C
L
, C
D
= vehicles lift and drag coefcients
C
l|
= cross-sectional design lift coefcient
c = chord
JS = surface element of a noise source
E
|
= component of the aerodynamic force that
acts on the noise source
] = function
f = noise frequency
|, ], k = indices
I
in
= drivers input current
K
V
= motors speed constant equal to the inverse
of the back-electromotive-force constant
M = Mach number
M = Mach number vector
M
r
= projection of M onto r
rel
N
l
= number of blades
P
B
= power extracted from the battery
P
P
= propellers required power (equal to
motors output power)

AFFT|
= |th harmonic of an A-weighted pressure
wave

FFT|
= |th harmonic of a pressure wave

ref
= reference pressure

2
= mean square average of sound pressure

2
A
= mean square average of sound pressure (A-
weighted)
R = propellers radius
R
o
= motor resistance
r = radial/spanwise coordinate
r
rel
= position vector of an observer relative to
the noise source
r
rel
= magnitude of r
rel
^ r
rel
= unit vector in the direction of r
rel
SPL = sound pressure level
SPL
harmonic |
= |th harmonic of a sound pressure level
SPLA = A-weighted sound pressure level
SPLA
harmonic |
= |th harmonic of an A-weighted sound
pressure level
T = thrust
T = Lighthill stress tensor
i = time
th = thickness
i
C
= time duration of a noise cycle
i
l
= loitering time
V
in
= drivers input voltage
V
E
= ight velocity
v = velocity vector of a noise source
x = position vector of a noise source
[ = pitch angle
; = cone angle

dBAdB
= difference between SPL
harmonic |
and
SPLA
harmonic |
= perturbation of static pressure
p
|]
= generalized stress tensor

thick
,
loading
= thickness and loading noise pressure
perturbations
o = Kroneckers delta function
j
S
= motor systems efciency
j
D
= drivers efciency
,
o
= air density
o = von Mises stress
t = retarded time

0
= volume of a noise source
= propellers rotational speed
I. Introduction
T
ODAY, renewed attention is being focused on the rst aero-
nautical propulsion device: the propeller. This is due to the
increased use of unmanned air vehicles (UAVs) [1,2], the growing
market of general aviation [2], the increasing interest in ultralight
categories or light sport air vehicles, and the growing importance of
environmental issues that have led to the development of all-electric
emissionless aircraft [2,3].
One of the most disturbing problems of propeller-driven aircraft is
their noise, which may limit the aircrafts operation. Many airports
around the world impose strict limitations on noise level permitted
during day or night. The acoustic signature of military aircraft has a
signicant effect on their detection. The importance of the noise
signature of propeller-driven air vehicles was already noticed during
Presented as Paper 3073 at the 14th AIAA/CEAS Aeroacoustics Con-
ference, Vancouver, Canada, 57 May 2008; received 29 May 2008; revision
received 4 June 2008; accepted for publication 27 January 2009. Copyright
2009 by the American Institute of Aeronautics and Astronautics, Inc. All
rights reserved. Copies of this paper may be made for personal or internal use,
on condition that the copier pay the $10.00 per-copy fee to the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include
the code 0748-4658/09 $10.00 in correspondence with the CCC.

Researcher, Faculty of Aerospace Engineering; ohadg@aerodyne.

technion.ac.il. Member AIAA.

Professor, Faculty of Aerospace Engineering; [email protected].

ac.il. Fellow AIAA.
JOURNAL OF PROPULSION AND POWER
Vol. 25, No. 3, MayJune 2009
717
the 1960s [4] and led to the development of the YO-3A Quiet Star
that later became an acoustic research aircraft [5]. Today, with the
increased use of propeller-driven UAVs, there is a renewed interest
in reducing the noise of propellers [6].
In propeller-based propulsion systems, the main sources of noise
are the engine and the propeller. Reducing the engine noise can be
achieved by developing a suitable exhaust pipe or by using electric
motors. Many of todays mini unmanned aerial vehicles use electric
motors [710].
Reducing the propellers noise requires special attention during its
design. Noise reduction can be achieved by a systematic design of the
propellers geometry and aerodynamic characteristics. Yet, these
same characteristics determine the performance and efciency of the
entire propulsion system. In many cases, efciency and noise
reduction result in contradictory design trends. This makes the
design of an efcient and quiet propeller a task that requires a
multidisciplinary design optimization (MDO) approach [11].
In many cases, the propeller is the simplest and most inexpensive
element of a propeller-based propulsion system, yet its design has
an enormous inuence on the efciency and noise level of the
entire vehicle. Furthermore, in many cases, the conguration of the
propulsion system is determined at an early stage of the design,
except for the propeller, which can be improved quite easily at later
stages of the project or even during operation. Thus, the propeller
design is an important element in the process of developing a new
quiet air vehicle or reducing the noise of an existing one.
Most of the traditional methods for propeller design are based on
the well-known work of Betz [12] from 1919. An example of
applying this method is presented in [13]. This approach is based on
optimizing the propellers geometry at a certain specic operating
condition (a certain combination of airspeed, altitude, and propeller
rotational speed), such that the power that is required to obtain a
certain propulsive force at these operating conditions is minimized
(or the thrust produced by a certain power is maximized). An
example of sucha design is the propeller of Rutans Voyager (the rst
nonstop, nonrefueled ight around the world) [14].
Betzs method considers only the aerodynamic efciency of the
propeller. It does not consider the propeller noise signature. Thus, for
designing a quiet propeller when using Betzs method, a serial design
process is used. First, the optimal aerodynamic propeller, with a
maximum efciency, is dened. Then this propeller is modied
to reduce its acoustic signature. Such a process has been used
extensively for designing quiet propellers [15,16]. The main problem
of such an iterative process is that it does not ensure that the nal
design will be optimal. It is also difcult to introduce additional
constraints into this serial design process, such as side or structural
constraints.
As already mentioned, the best approach for solving such a design
problem is to use MDO. MDO ensures that all the different
disciplines are addressed simultaneously. Thus, the best compromise
between contradictory goals of an efcient and a quite propeller
will be reached. This kind of approach has been already applied
to propeller designs. Ormsbee and Woan [17] presented an opti-
mization procedure of the circulation distribution along a propellers
blades to minimize noise signature under performance constraints
using simplied models. Succi [15] and Miller and Sullivan [18]
showed several ways to decrease a propellers noise while main-
taining its efciency. Burger et al. [19] used a vortex lattice approach
and genetic algorithmto design an efcient and quiet propeller. Most
of the past investigations did not consider the characteristics of the
power plant itself (the motor). This may have a signicant effect on
the efciency of the entire propulsion systemand thus on the optimal
propeller design. References [2022] present examples of propeller
designs for specic vehicles, which represent a compromise between
aerodynamic and acoustic constraints. Most of the previous
investigations used a very limited number of design variables rather
than the full range of the parameters that are under the authority of the
designer. The design methods already presented considered the
inuence of the noise level on the propeller efciency with minor
(and sometimes no) consideration of the structural integrity of the
blades.
The paper will present a process for designing a quiet propeller
while taking into account performance and structural constraints. It
will be shown that the motor characteristics play an important role
in the design. First, the analysis tools (including aerodynamic,
structural, and acoustic models) will be described. Because noise is
the subject matter of the present paper, the acoustic model will be
described in more detail and its validation will be discussed. Then the
optimization procedure and design process will be presented. The
application of the method for designing a quiet propeller for an
electric UAVwill be presented in detail. The inuence on the optimal
design of performance constraints, structural constraints, and
number of blades will be shown and discussed.
II. Analysis Tools
Analysis tools for propeller design include an aerodynamic model,
a structural model, and an acoustic model. The selection of the
appropriate model is based on the following two considerations:
1) It should be sufciently accurate.
2) It should be efcient, because the optimization procedure
includes a very high number of iterations.
The aerodynamic model calculates the distribution of the
aerodynamic loads along the propeller blades. This data is used to
calculate the propellers thrust and required power. In addition,
the distribution of the aerodynamic loads is used as an input for
the structural and acoustic analyses. Consequently, an inaccurate
aerodynamic model may result in inaccuracies in the other analyses
of the propeller. In the present case, a momentum/blade-element
model is used.
For regular design conditions of propellers (axial uniformowand
high advance ratios), the blades cross sections do not experience
stall and the momentum/blade-element model gives results of
high accuracy [23]. The blade-element/momentum analysis can be
extended to include the inuence of rotation on the aerodynamic
behavior of cross sections after reaching stall [24]. Momentum/
blade-element models are usually much more efcient than other
models, and thus they are suitable for the present MDO analysis.
Astructural analysis is essential to ensure that the propeller blades
will be able to withstand the aerodynamic and inertial loads that act
along them. Common tools for the structural analysis of blades are
nite element models [25]. Yet, nite element models require
relatively large computer resources and computing time. Thus, a
more efcient rod model using a transfer-matrix formulation will be
applied [26].
The rod structural model describes the propeller blades as a series
of straight segments located along the blades elastic axis, which
is not necessarily a straight line. The structural cross-sectional
properties are uniformalong each segment and equal to the structural
properties of a representative cross section of that segment. The
transfer-matrix problemis solved using the boundary conditions of a
cantilevered rod (clamped root and free tip). The solution procedure
is very efcient and the results are very accurate [26].
III. Acoustic Model
A. Noise Equations
Most of the models used to calculate a propellers noise signature
are based on a solution of the constrained wave equation, known as
the Ffowcs-Williams/Hawkings equation [27]:
1
o
2

d
2
()
di
2

d
2
()
dx
2
|
=
d
2
T
|]
dx
|
dx
]

d
di
_
,
o
v
|
o(])
d]
dx
|
_
V
_

|]
o(])
d
]
dx
]
_
(1)
where o is the speed of sound, is the perturbation in the static
pressure (relative to the undisturbed pressure), i is the observer time,
and x is the position vector of the noise source relative to a stationary
system of coordinates (with components x
|
). T is Lighthill stress
tensor (with components T
|]
), p is the generalized stress tensor
718 GUR AND ROSEN
(with components
|]
), v is the source velocity vector (with
components :
|
), o is Kroneckers delta function, and ] is a function
that denes the surface of the body producing the pressure wave (in
the present case, it denes the surfaces of the blades):
](x) =
_
<0 xinside the body
0 xon the body surface
>0 xoutside the body
(2)
There are three forcing terms on the right-hand side of Eq. (1):
vortex, thickness, and loading terms (from left to right) [28]. It was
shown that for thin blades with cross sections operating in subsonic
or transonic conditions, the vortex term can be neglected [29]. Thus,
the solution of the Ffowcs-Williams/Hawkings equation includes
two terms: the thickness noise
thick
and the loading noise

loading
:
(x. i) =
thick
(x. i)
loading
(x. i) (3)
The solution of the Ffowcs-Williams/Hawkings equation is
obtained after applying Greens function [27]. This solution is then
discretized [30] and the following approximations are adopted:
1) The velocity of the noise source is subsonic.
2) The noise source is much faster than the observer.
3) The propellers rotational speed and the velocity of the air
vehicle do not vary with time.
The nal expressions for the loading and thickness noise become
[15]

loading
(x. i) =
1
4

k
_
_
F ^ r
rel
F ^ r
rel
[(
_
M ^ r
rel
),(1 M
r
)|
r
rel
o (1 M
r
)
2
................................,,................................
far field

F ^ r
rel
[(1 M M),(1 M
r
)| F M
r
2
rel
(1 M
r
)
2
...................................,,...................................
near field
_
k
(4)

thick
( ~ x. i) =
,
4.

k
_

0
r
rel
(1 M
r
)
3

_

M
r
1 M
r
3
_
_
M
r
1 M
r
_
2

_
M
r
o (1 2 M
r
)
r
rel
(1 M
r
)
2
_
M
r
o
r
rel
_
2
__
k
(5)
where indicates integration over the entire body that generates
noise, r
rel
is the location vector of the observer relative to the noise
source, r
rel
is the magnitude of r
rel
, ^ r
rel
is a unit vector in the direction
of r
rel
, M is the Mach vector, dened as
M =
v
o
(6)
M
r
is the projection of M onto r
rel
, F is the aerodynamic loading
force acting on the element, and
0
is the volume of the blade
element.
Adot accent indicates a time derivative at the noise source system
(differentiation relative to the retarded time). If i is the time as
measured in the observers frame of reference, the retarded time t
indicates the time when the pressure wave left the noise source:
t =i
r
rel
(t)
o
(7)
The loading noise [Eq. (4)] is composed of two components: the
far-eld and near-eld loading noise. These two terms differ by the
power of r
rel
in the denominator. The far-eld termis proportional to
1,r
rel
, the near-eld term is proportional to 1,r
2
rel
, and thus the last
term becomes relatively small at large distances from the noise
source.
Calculations of the noise produced by a propeller are based on
dividing the propellers blades into small elements (similar to the
aerodynamic blade elements). The pressure at a certain point in space
at a certain time is obtained by summing up the contributions of all
the elements according to Eqs. (4) and (5). Then the same process is
repeated for the next time step, in which the new positions of the air
vehicle and propeller blades at that time are considered. The pressure
wave is dened after repeating this process for a complete cycle (e.g.,
for a two-bladed propeller, one noise cycle is equivalent to half of a
revolution).
B. Spectral Analysis
If the pressure wave is known, its spectral analysis denes the
sound pressure level (SPL) and its different harmonics. The mean
square average of the sound pressure,
2
, and the resultant SPL
(given in decibels) are dened as follows:

2
=
1
i
C
_
i
C
0

2
(x. i) (8)
SPL dB =10 log
10

2

2
ref
(9)
where i
C
is the time duration of a noise cycle. The reference pressure
for air is usually chosen as
ref
=20 10
6
Pa.
Using fast Fourier transform (FFT), the different harmonics of the
pressure wave
FFT|
and the sound pressure harmonics SPL
harmonic |
are calculated:

2
=

|
1
2

2
FFT|
(10)
SPL
harmonic |
dB =10 log
10
_

FFT|

ref
_
2
(11)
The noise sensitivity of the human ear depends on the noise
frequency f. It is common to dene sound harmonics representing
the human ear sensitivity through the A-weighted sound pressure
level SPLA
harmonic |
. The differences between SPL
harmonic |
and
SPLA
harmonic |
are given by the empirical function
dBAdB
(f) [31]:
SPLA
harmonic |
=SPL
harmonic |

dBAdB
(f) (12)
Similar to Eqs. (10) and (11), the following terms are dened:
SPLA dBA =10 log
10
_

2
A

2
ref
_
=10 log
10
_

|
1
2

2
AFFT|
_
10 log
10
_
1

2
ref
_
(13)
SPLA
harmonic |
dB =10 log
10
_

AFFT|

ref
_
2
(14)
Substitution of Eq. (14) into Eq. (13) results in
SPLA dBA =10 log
10
_

|
1
2
10
[SPLA
harmonic |
10log
10
(
1

2
ref
)|,
10
_
10 log
10
_
1

2
ref
_
(15)
C. Model Validation
The model validation includes comparisons of the model results
with wind-tunnel [32] and ight-test [33] results. The wind-tunnel
test included a two-bladed propeller of radius R =1.015 m. The
pressure wave and the sound pressure level were measured at
different locations relative to the propeller.
GUR AND ROSEN 719
Table 1 presents a comparison of SPLfor three runs (designated as
BC-61, BC-2, and AC-1) and three microphone locations (1, 4, and
9). The exact denitions of the test conditions and microphone
locations are given in [32]. Comparisons between calculated and
measured results exhibit good agreement. The maximum difference
does not exceed 12 dB. Figures 1 and 2 present two examples of the
pressure wave and harmonics.
The ight tests reported in [33] included a Twin Otter aircraft
equipped with two microphones (fore and aft). The aircraft has
two three-bladed propellers. Four representative ight conditions
(designated as cases 14) are considered, which are dened in [34].
Table 2 presents a comparison between SPL calculations and
ight-test measurements. The differences at the fore microphone are
12 dB, and at the aft microphone, the differences reach 5 dB. The
higher differences in the aft microphone are probably due to wake
inuences.
Figures 3 and 4 present the SPLharmonics for two cases (1 and 4).
Similar to the results presented in Table 2, the comparison between
the fore microphone results and model calculations exhibits good
agreement, whereas in the case of the aft microphone, the agreement
deteriorates.
IV. Optimal Design of a Quiet Propeller
Any optimal design problemcan be described mathematically as a
search process for a design conguration that minimizes (or
maximizes) a specic cost function that represents the design goal.
This search process is usually carried out under certain design
constraints.
Any design process requires an a priori denition of design
variables, design constraints, and cost function. The design variables
are parameters that are determined by the designer. In the general
case of a propeller design, these variables are divided into three
categories: general design variables, blade design variables, and
cross-sectional design variables. The general design variables affect
the global conguration of the propeller system and may include the
following parameters: number of propellers, number of blades N
l
,
propeller radius R, and rotational speed . The blade design
variables are parameters dening the geometry and structure of each
blade [namely, the distribution along the blade of the following
parameters: pitch angle [(r), chord c(r), sweep angle, dihedral angle
(cone angle in the simplest case), mass and inertia, and structural
properties].
The blades cross-sectional design variables dene the cross-
sectional airfoil geometry as a function of a chordwise coordinate. In
the following examples, the cross sections will belong to the NACA-
16 airfoil family [35]; thus, the geometry of each cross section is
dened by two parameters that vary along the blade: thickness ratio
th,c(r) and design lift coefcient C
l|
(r).
The design constraints may include any kind of constraint: for
example, minimum or maximum chord size, allowable maximum
stress, etc.
The cost function is a quantitative measure of achieving the design
goal: namely, the cost function presents a measure of the quality of
the design.
A penalty method is used to solve the constrained optimization
problem [36]. A new constrained cost function is dened and
minimized to nd the optimal design, subject to the constraints.
A. Optimization Scheme
The optimization scheme used here includes a mixed-strategy
approach combining three different optimization methods:
1) A heuristic search uses a simple genetic algorithm (SGA) [37].
2) An enumerative scheme uses Nelder and Meads [38] simplex
scheme.
3) A derivative-based scheme uses the steepest-descent method
[36].
The SGA [37] simulates a natural selection process, similar to an
evolution process. It includes three major elements: reproduction,
crossover, and mutation. Starting from an initial random population
of designs and using a genetic scheme leads to an improved
population.
The next stage of the optimization scheme includes an enu-
merative simplex scheme [38]. This scheme is capable of dealing
efciently with a large number of design variables, including
smoothing constraints of the distributions of variables along the
blade.
Table 1 SPL (in decibels) as obtained by calculations and measured in the wind tunnel [32]
Run BC-61 Run BC-2 Run AC-1
Calculations Test Calculations Test Calculations Test Microphone location
116.6 119.5 103.3 102.5 102.3 101.1 1
131.2 129.6 111.8 109.1 110.3 108.1 4
129.7 128.6 110.4 108.7 108.9 105.9 9
S
P
L
H
a
r
m
o
n
i
c
-
i
[
d
B
]

Harmonic Number t[sec]

p
[
P
a
]

60
70
80
90
100
110
120
0 5 10 15
Calc.
Test
-20
-15
-10
-5
0
5
10
15
0 0.005 0.01 0.015
Calc.
Test
Fig. 1 Comparison between calculations and wind-tunnel results [32]
for run AC-1 and microphone 4.
S
P
L
H
a
r
m
o
n
i
c
-
i
[
d
B
]

Harmonic Number
t[sec]

p
[
P
a
]

-20
-15
-10
-5
0
5
10
15
20
0 0.005 0.01 0.01
Calc.
Test
50
60
70
80
90
100
110
120
0 5 10 15
Calc.
Test
Fig. 2 Comparison between calculations and wind-tunnel results [32]
for run BC-2 and microphone 9.
720 GUR AND ROSEN
A derivative-based method is used only during the nal stage of
the entire search, to pinpoint the nal design. Derivative-based
methods are extensively used [36] and are very efcient in unimodal
cases (in which the cost function has only one minimum) or in the
neighborhood of the global minimum.
The use of a mixed-scheme strategy for optimization (starting with
SGA, then using the simplex method, and, nally, applying the
steepest-descent method) exploits the various advantages of each
scheme [39], as indicated by Table 3. It ensures a thorough search for
the global minimum. A detailed description of this strategy and an
example of its use are presented in [40].
B. Design Process
The rst stage of the design process includes denitions of the cost
function (design goal), design variables, and constraints. In cases of
complex problems, it is often recommended to start with a simpler
problem that includes a simpler cost function and a reduced number
of design variables or constraints. This simpler problem helps to
better understand the actual problem and to investigate the optimal
design trends. It can also be used as an initial design point for the
more complex problem.
The design usually starts with a SGA. This scheme does not
depend on a specic initial design and is capable of searching
through a very wide range of designs to nd the global minimum.
One of the weaknesses of this scheme is the serrated nature of
the distributions of the various design variables along the blade.
These distributions are often impractical and should be smoothed.
Smoothing can be imposed as a constraint on the absolute magnitude
of the second derivative of the variable with respect to the spanwise
coordinate. Yet, the SGA scheme becomes inefcient as the number
of constraints is increased. Thus, smoothing constraints are usually
added only during the second stage of the optimization, which
includes the simplex scheme. The smoothing constraints should be
imposed in a gradual manner; otherwise, difculties in convergence
are encountered.
At the end of the simplex schemes application, the design is
usually very close toits nal form. In most cases, the steepest-descent
scheme results in only small changes of the cost function and its
effect is mainly to imply nal smoothing.
Note that the design process should not be treated as a black box
and is by no means fully automatic. The designer plays a major role in
the entire process. He or she decides when to switch fromone scheme
to the other. He or she should followthe results during the application
of each scheme and should determine the next steps based on these
results. In certain cases, the designer may decide to return to a
previous scheme (e.g., when it seems that the solution has converged
to a local extremum) or to ease certain constraints to increase the
exibility of the search process and thus examine a wider range of
options.
The next section will present the process of designing an optimal
propeller for an electric UAV.
V. Design of a Quiet Propeller for
an Electric Mini UAV
An electric mini UAV is considered. The total mass of the vehicle
is 5.5 kg. The wing area is 0.72 m
2
, and the drag polar of the vehicle
is
C
D
=0.03 0.033 C
2
L
(16)
where C
L
and C
D
are the vehicles lift and drag coefcients,
respectively.
A direct drive is used (the system does not include a gear
box); namely, the rotational speed of the electric motor is equal to
that of the propeller, . The parameters that dene the charac-
teristics of the electric motor are [8] 1) motor speed constant
K
V
=0.0045 V s,rad,

2) motor resistance R
o
=0.1 [|, and
3) drivers efciency j
D
=0.9.
The electric systems efciency j
S
is dened as the ratio between
the propellers required power P
P
and the power that is extracted
from the battery, P
B
:
j
s
=
P
P
P
B
(17)
where j
S
is a function of the input voltage V
in
, input current I
in
, and
electric driver efciency j
D
:
j
s
=j
D

P
P
V
in
I
in
(18)
For a certain rotational speed and power P
P
, V
in
and I
in
are
calculated by the following equations [8]:
V
in
= K
V

P
P
R
o
k
V
(19)
I
in
=
P
P
K
V
(20)
The main task of the UAVis to loiter at lowaltitude above a certain
area; thus, the goal is to design an optimal propeller for loitering at a
speed V
E
=10 m,s at sea level. The required thrust under this ight
condition, based on the preceding data, is
T =3.5 N (21)
Because of geometric constraints, the radius of the propeller is
limited:
Table 2 SPL (in decibels) as obtained by calculations and measured during ight tests [33]
Case 4 Case 3 Case 2 Case 1
Fore Aft Fore Aft Fore Aft Fore Aft
Test 113.6 115.9 106.5 112.4 106.7 108.3 115.4 119.3
Calculations 113.4 111.0 107.6 107.2 104.9 105.0 115.2 114.1
Fore
Aft
S
P
L
H
a
r
m
o
n
i
c
-
i
[
d
B
]

Harmonic Number
S
P
L
H
a
r
m
o
n
i
c
-
i
[
d
B
]

Harmonic Number
80
90
100
110
120
0 5 10 15 20
Test
Calc.
70
80
90
100
110
120
130
0 5 10 15
Fig. 3 Comparison between calculations and ight-test results [33] for
case 1.

It is also common to dene instead the torque constant (also known as the
back-electromotive-force constant), which is the inverse of K
V
.
GUR AND ROSEN 721
R 0.15 m (22)
NACA-16 cross sections are used. The database for this airfoil
family is limited to the following range of thickness ratio th,c and
design lift coefcient C
l|
[35]:
0.04 <
~
i
c
< 0.21 (23)
0.0 <C
l|
<0.7 (24)
For practical reasons, the chord is also limited:
c <0.052 m (25)
The design variables of the current design problemare the propellers
radius R; distribution of chord along the blade, c(r); distribution of
pitch angle along the blade, [(r); distribution of thickness ratio along
the blade, th,c(r); and distribution of design lift coefcient along the
blade, C
l|
(r).
The distribution of the various parameters is dened by their
values at 17 stations along the blade; thus, there are 69 design
variables. At a later stage of this study, another design variable will
be added: the cone angle of the blades. The energy source of the UAV
is a lithium polymer battery with a mass of 2 kg and a capacity of
250 W h. From this capacity, 150 W h are allocated for loitering.
The goal of the propeller design is twofold: 1) maximize the
loitering time and 2) minimize the noise produced by the UAV. As
will be shown in what follows, these are contradictory goals.
To understand the design trends and the physical phenomena
behind them, the design will be carried out in a gradual manner.
Initially, two-bladed propellers will be considered and the number of
design constraints will be increased gradually to understand their
inuence on the optimal design. Then the inuence of the number
of blades on the optimal designs will be investigated. Although
complex congurations of the blades may offer certain advantages
[41], the present study is conned to straight blades.
A. Two-Bladed Propeller
As a rst step in studying the problem, two different propellers will
be designed for the following two extreme goals:
1) For maximum loitering time, maximize the loitering time i
l
subject to the constraints of Eqs. (2125), without considering any
other goals or constraints. The cost function to be minimized is the
power extracted from the vehicles battery, P
B
.
2) To minimum noise at loitering, minimize the noise of the
vehicle as heard by an observer located 50 m beneath the UAV,
subject to the constraints of Eqs. (2125), without considering any
other goals or constraints. The cost function is the SPLAproduced by
the propeller at the observer point.
The preceding described optimization process has been applied
for the design of these two propellers and the results are presented in
Table 4 and Fig. 5.
Figure 5 presents the geometry of the optimized propellers:
namely, spanwise distributions of the pitch angle [, chord c, and
cross-sectional thickness th. The design lift coefcient C
l|
is equal (in
both cases) to its upper limit value C
l|
=0.7.
The propeller for maximum loitering time (case 1) operates at a
relatively high rotational speed (=13. 590 rpm) and extracts
from the battery an electrical power of P
B
=54 W, which means
a loitering time of i =160 min. Its noise intensity is SPLA=
36.0 dBA.
The propeller for minimumnoise (case 2) operates at a much lower
rotational speed (=2233 rpm) and results in a noise intensity of
only SPLA =18.4 dBA, which means a signicant noise reduction
of 17.6 dBA. This reduction in noise comes at a cost of a signicant
reduction in the efciency of the entire propulsion system. This
results in an increase in the power extracted from the battery
during loitering (P
B
=357 W), leading to a loitering time of only
i
l
=25 min. The signicant reduction in the loitering time is due to
the increase in the required power by the propeller (from46 to 54 W)
and a signicant decrease in the efciency of the electric motor (from
0.81 in case 1 to 0.15 in case 2).
In both designs, the propeller gives a thrust of T =3.5 N. Because
of the very large decrease in the rotational speed, in the case of the
propeller for minimum noise, there is a signicant increase in the
pitch angle and chord length (see Fig. 5). In fact, case 2 is impractical
and does not represent a feasible design, because the chord length
reaches its limit [Eq. (25)]. Even if design 2 were practical, the
loitering time is too short. Thus, to offer a reasonable loitering time, a
constraint of the magnitude P
B
will be imposed in the next
subsection.
B. Two-Bladed Propeller Under Power Constraints
In this case, the cost function remains the SPLA. In addition to the
previous constraints, the power that is extracted fromthe battery, P
B
,
is limited. Three different cases are considered: P
B
65. 70. 75 W.
In Table 4, a summary of the design results is presented. As the
allowable P
B
increases, the noise decreases, but the loitering time
decreases as well. The decrease in the noise is mainly a result of the
decrease in the rotational speed, . Unfortunately, this also results in
a decrease in the electric motor efciency and a reduction of the
loitering time.
The geometry of the propellers is presented in Fig. 6, together with
the geometry of the minimum SPLA propeller of Fig. 5. Because all
of the propellers have to produce identical thrust, the decrease in the
rotational speed is accompanied by an increase in the pitch angle and
chord length along the blade. The increase in chord length also leads
to an increase in the thickness of the cross section.
C. Two-Bladed Propeller Under Power and Stress Constraints
From Fig. 6, it is clear that the blades are very thin, and thus they
may develop high stresses that will endanger their integrity.
Therefore, structural constraints are added in what follows. It is
assumed that the blade is made of aluminum 7075-T6. The yield
stress of this material is 5.03 10
8
Pa. The maximum von Mises
stress o will be limited to
o < 1.5 10
8
Pa (26)
In the current example, the stresses are calculated during loitering.
If necessary (see [40]), it is possible to calculate the stresses under
other critical ight conditions such as takeoff or maximum speed.
The inuences of the stress constraints on the characteristics of
the propeller are presented in Table 4 (denoted as with stress
constraints). The SPLA increases by 2.9, 1.0, and 0.6 dBA for the
cases P
B
65. 70. 75 W, respectively, as a result of introducing the
Fore
S
P
L
H
a
r
m
o
n
i
c
-
i
[
d
B
]

Harmonic Number
S
P
L
H
a
r
m
o
n
i
c
-
i
[
d
B
]

Harmonic Number
Aft
70
80
90
100
110
120
0 10 20 30
Test
Calc.
70
80
90
100
110
120
0 5 10
Fig. 4 Comparison between calculations and ight-test results [33] for
case 4.
722 GUR AND ROSEN
stress constraints. The increase in the noise is mainly due to the
fact that to cope with stress constraints, the rotational speed of
the propeller is increased (by 23, 6, and 3% for the cases
P
B
65. 70. 75 W, respectively), whereas the radius is decreased
(by 8.7, 2.0, and 1.3%, respectively). The increase in the rotational
speed increases the tensile force along the blade, which results in
(due to nonlinear effects) a decrease in the cross-sectional bending
moments along the blade (that is caused by the aerodynamic loads).
Areduction of the blades length (reduction of the propellers radius)
also results in a reduction of the cross-sectional bending moment
along each blade.
Another method of reducing the bending moment along the blade
and thus reducing the stress level is to introduce cone angles at the
blades attachments to the hub. The cone angle is the angle between
the blades axis andthe plane of rotation(this plane is perpendicular to
the axis of rotation). When there is a positive cone angle, the inertial
bending moment (due to the centrifugal forces along the blade)
oppose the aerodynamic bending moment (due to the aerodynamic
forces). By choosing an appropriate cone angle, the resultant bending
moment can be minimized along the blade. In the present case, the
cone angle is added as another design variable; thus, the appropriate
angle is dened during the optimization process.
The results for the introduction of cone angles are shown at the end
of Table 4 for the same three cases: P
B
65. 70. 75 W. It is shown
that introducing cone angles leads to a small reduction in the noise,
due to a small reduction in the rotational speed, compared with the
case of stress constraints with zero cone angle. The optimal cone
angles are small and decrease with an increase in P
B
.
The geometry of the optimal blades for P
B
70 W is shown in
Fig. 7 for three cases: 1) optimal blade without considering stress
constraints (identical results also appear in Fig. 6); 2) optimal blade
when stress constraints are included, with zero cone angle; and
3) optimal blade under stress constraints, when the cone angle
becomes one of the design variables.
It is clearly shown that adding stress constraints results in a
signicant increase of thickness along most of the blade. There is also
a reduction of the pitch angle at the outer 40% of the blade, which
contributes to the reduction of the bending moment at the root region.
D. Inuence of the Number of Blades
One of the common methods of reducing propeller noise is to
increase the number of blades [5,15,21]. Thus, in addition to the
various optimal designs of the previous subsections that dealt with a
Table 3 Advantages and disadvantages of the various numerical optimization schemes
Heuristic scheme Enumerative scheme Gradient-based scheme
Capability of nding the global optimum Good Limited ability Poor
Pinpointing the local optimum Poor Limited ability Good
Reaching the feasible region subject to a high number of constraints Poor Good Good
Number of iterations needed for convergence High High Low
Required computer resources High Low Relatively low
Table 4 Optimal two-bladed propellers under various design goals and constraints
Design description SPLA, min i
l
, min P
B
, W P
P
, W j
S
, rpm R, m ;, deg
Maximum loitering time 36.0 160 56 46 0.81 13,600 0.150 0
Minimum noise 18.4 25 357 54 0.15 2230 0.150 0
Minimum noise, P
B
65 W 30.7 138 65 43 0.66 7300 0.150 0
Minimum noise, P
B
70 W 29.6 129 70 43 0.62 6470 0.150 0
Minimum noise, P
B
75 W 28.7 120 75 43 0.58 5900 0.150 0
Minimum noise, P
B
65 W, with stress constraints 33.6 138 65 46 0.71 8973 0.137 0
Minimum noise, P
B
70 W, with stress constraints 30.6 129 70 44 0.63 6872 0.147 0
Minimum noise, P
B
75 W, with stress constraints 29.3 120 75 44 0.59 6086 0.148 0
Minimum noise, P
B
65 W, with stress constraints and cone angle 33.3 138 65 46 0.71 8799 0.146 3.3
Minimum noise, P
B
70 W, with stress constraints and cone angle 30.5 129 70 44 0.63 6852 0.149 2.2
Minimum noise, P
B
75 W, with stress constraints and cone angle 29.0 120 75 44 0.58 6002 0.150 1.4
Table 5 Comparison between the inuence of performance constraints, structural constraints, and number of blades for
propellers for minimum noise
Design description N
l
SPLA, min i
l
, min P
B
, W P
P
, W j
S
, rpm R, m ;, deg
Without performance constraints 2 18.4 25 54 357 0.15 2230 0.150 0.0
P
B
65 W 2 30.7 138 43 65 0.66 7300 0.150 0.0
P
B
70 W 2 29.6 129 43 70 0.62 6470 0.150 0.0
P
B
75 W 2 28.7 120 43 75 0.58 5900 0.150 0.0
P
B
65 W, with stress constraints and cone angle 2 33.3 138 46 65 0.71 8799 0.146 3.3
P
B
70 W, with stress constraints and cone angle 2 30.5 129 44 70 0.63 6852 0.149 2.2
P
B
75 W, with stress constraints and cone angle 2 29.0 120 44 75 0.58 6002 0.150 1.4
Without performance constraints 3 1.7 17 56 535 0.10 1810 0.150 0.0
P
B
65 W 3 21.7 138 43 65 0.66 7350 0.150 0.0
P
B
70 W 3 19.7 129 43 70 0.62 6500 0.150 0.0
P
B
75 W 3 18.1 120 43 75 0.58 5940 0.150 0.0
P
B
65 W, with stress constraints and cone angle 3 27.8 138 47 65 0.73 9510 0.140 6.0
P
B
70 W, with stress constraints and cone angle 3 21.9 129 45 70 0.64 7070 0.149 2.8
P
B
75 W, with stress constraints and cone angle 3 19.5 120 44 75 0.59 6130 0.150 1.2
Without performance constraints 4 22.1 15 54 608 0.09 1650 0.150 0.0
P
B
65 W 4 13.1 138 43 65 0.66 7190 0.150 0.0
P
B
70 W 4 9.5 129 43 70 0.62 6480 0.150 0.0
P
B
75 W 4 7.3 120 43 75 0.58 5940 0.150 0.0
P
B
65 W, with stress constraints and cone angle 4 24.0 138 47 65 0.73 9540 0.111 16.7
P
B
70 W, with stress constraints and cone angle 4 13.5 129 45 70 0.65 7210 0.146 3.6
P
B
75 W, with stress constraints and cone angle 4 9.7 120 45 75 0.60 6290 0.148 1.4
GUR AND ROSEN 723
two-bladed propeller, optimal propellers with three and four blades
were also designed. The cost function remains the propellers SPLA,
and the design cases include 1) a quiet propeller without power or
structural constraints; 2) a quiet propeller with three kinds of power
constraints: P
B
65. 70. 75 W, without considering structural
constraints; and 3) the same as case 2, but with structural constraints
and cone angle as an additional design variable.
Table 5 gives the various optimal designs, and data for a two-
bladed propeller from Table 4 is also included for ease of
comparisons.
The signicant effect of the number of blades on the noise level is
clear. The case without any performance or structural constraints
0
0.0002
0.0004
0.0006
0.0008
0.001
0.0012
0.0014
0 0.2 0.4 0.6 0.8 1
5
10
15
20
25
30
35
40
0 0.2 0.4 0.6 0.8 1
w/o Stress Const.
With Stress Const. and Zero Cone Angle
With Stress Const. and Cone Angle
0
0.03
0.06
0 0.2 0.4 0.6 0.8 1
Chord
Pitch
Thickness

[
d
e
g
]

r/R
r/R r/R
t
h
[
m
]
c
[
m
]

Fig. 7 Inuence of structural constraints and cone angle on the
minimum noise design with P
B
70 W for a two-bladed propeller.
S
P
L
A
[
d
B
A
]

P
B
[W]
5
10
15
20
25
30
35
65 70 75
4 Blades
3 Blades
2 Blades
Fig. 8 SPLA of optimal propellers as functions of performance
constraints and number of blades for a quiet propeller design with stress
constraints.

0
10
20
30
40
50
60
70
80
0 0.2 0.4 0.6 0.8 1
Min. SPLA
Min. P
0
0.03
0.06
0 0.2 0.4 0.6 0.8 1
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0 0.2 0.4 0.6 0.8 1
Chord
Thickness

[
d
e
g
]

R / r R / r
r/R
t
h
[
m
]
Pitch
c
[
m
]

Min. SPLA
Min P
B
Fig. 5 Geometry of an optimal two-bladed propeller with minimum
noise and minimum power propellers.
0
0.03
0.06
0 0.2 0.4 0.6 0.8 1
10
20
30
40
50
60
70
80
0 0.2 0.4 0.6 0.8 1
P=65[W]
P=70[W]
P=75[W]
Minimum Pb
P
B
65[W]
P
B
70[W]
P
B
75[W]
Minimum P
B
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0 0.2 0.4 0.6 0.8 1
Chord
Pitch
Thickness

[
d
e
g
]

r/R
r/R r/R
t
h
[
m
]
c
[
m
]

Fig. 6 Geometry of an optimal two-bladed propeller with minimum
noise under various power constraints.
724 GUR AND ROSEN
(denoted as without performance constraints in Table 5) exhibits a
decrease of more than 20 dBA as a result of adding a blade.
Unfortunately, loitering time is also reduced signicantly, from
25 min (for a two-bladed propeller) to 15 min in the case of four
blades. This decrease in loitering time is a result of an increase in the
power required by the propeller and decrease in the electric motor
efciency.
When performance constraints are added, the SPLA decrease due
to an additional blade becomes 9, 10, and 11 dBA for P
B

65. 70. 75 W, respectively. Note that the rotational speed of the
optimal propeller depends on the magnitude of the power constraint,
but the number of blades has little inuence on that speed. In all
cases, the optimization leads to the maximum allowable radius of
R =0.15 m.
As shown in Table 5, the introduction of stress constraints has a
very signicant effect on the optimal design for different numbers of
blades, in comparison with an optimal design that does not include
stress constraints. This effect depends on the magnitude of power
constraints.
Figure 8 presents the SPLA of the optimal propellers as functions
of the magnitude of the power constraint P
B
for various numbers of
blades N
l
. It is shown that the reduction in SPLAas a result of adding
a blade increases with the allowable P
B
.
The geometry of a blade of the four-bladed optimal propeller with
and without stress constraints for P
B
70 W is presented in Fig. 9.
This gure is similar to Fig. 7, which presents the same comparison
for a two-bladed propeller. The inuences of the stress constraints on
the optimal design of a quiet propeller were discussed in the previous
section. In both cases (two- and four-bladed cases), the inuence of
the stress constraint on the blades chord is small, because both
blades have to produce the same thrust. The major difference appears
in the thickness of the blade, which increases substantially due to the
introduction of stress constraints. Whereas the two-bladed case
exhibits an increase of about 100%, the four-bladed case increases by
more than 200%.
The thicknesses of the optimal two-bladed and four-bladed
propellers are similar, but the chord of the four-bladed propeller is
smaller than the chord of the two-bladed propeller, and thus the
thickness ratio of the four-bladed propeller is higher. This makes the
four-bladed propeller less efcient; therefore, an increase in its
rotational speed is required. In addition to the increased thickness,
to avoid high stresses as the number of blades is increased, the
following trends reduce the magnitude of the bending moment along
the blade (see Fig. 10):
1) The rotational speed and cone angle are increased.
2) The radius is decreased.
The preceding trends are especially strong in the case in which
P
B
65 W. Here, the radius is reduced to 0.11 m, and the cone angle
becomes 16.7 deg. Such a high cone angle is required to counteract
the bending moment due to the aerodynamic forces along the blade.
To understand the very strong inuence of the number of blades on
the optimal design when P
B
65 W, another optimum search was
performed. In this case, the optimal propeller for a minimumP
B
was
designed without imposing any acoustic constraints and with a zero
cone angle. The minimal P
B
value as a function of the number of
blades is presented in Fig. 11. It is shown that for four and ve blades,
the minimal P
B
is larger than 65 W. By introducing cone angles, P
B
0
0.0002
0.0004
0.0006
0.0008
0.001
0.0012
0.0014
0.0016
0.0018
0.002
0 0.2 0.4 0.6 0.8 1
0
5
10
15
20
25
30
35
40
45
0 0.2 0.4 0.6 0.8 1
w/o Stress Const.
With Stress Const. and Cone angle
0
0.03
0.06
0 0.2 0.4 0.6 0.8 1
Chord
Pitch
Thickness

[
d
e
g
]

r/R
r/R r/R
c
[
m
]

t
h
[
m
]
Fig. 9 Inuence of structural constraints and cone angle on the
minimum noise design with P
B
70 W for a four-bladed propeller.

B
P [W]
B
P [W]
B
P [W]
0.1
0.11
0.12
0.13
0.14
0.15
0.16
65 70 75
R
[
m
]

[
R
P
M
]

6000
6500
7000
7500
8000
8500
9000
9500
10000
65 70 75
4 Blades
3 Blades
2 Blades
0
2
4
6
8
10
12
14
16
18
65 70 75

[
d
e
g
]

Fig. 10 Radius, rotational speed, andcone angle of the optimal propellers as functions of performance constraint P
B
andnumber of blades N
b
for a quiet
propeller design under stress constraints.
GUR AND ROSEN 725
can be decreased to 65 W, but this comes at the cost of increasing
noise (as shown in Fig. 8).
The optimized blades are very narrow (small chord) at the tip
region (see Fig. 7); in some cases, the chord gets below 1 mm.
Although these blades withstand structural constraints, such a small
chord may not be practical because of manufacturing, maintenance,
and reliability aspects. To study this problem and investigate the
inuence of additional constraints on the optimal design, a side
constraint was added that does not allow chord lengths smaller than
6 mm, and few optimizations were run.
In spite of an increase of the chord length in the tip region by
almost an order of magnitude, in the cases that were investigated, the
relative changes in the propeller parameters, such as required power
or rotational speed, were of the order of 10% or less. This may imply
that the trends and results that have been presented are also applicable
for designs that include additional other constraints (see similar
results in [42]).
E. Flyover Noise
The noise in all the previous examples (SPLA) was calculated
when the UAVies 50 mdirectly above the observer. Yet, in reality,
the UAV ies over the observer. Thus, the question arises if
optimization for the case in which the UAV is at the zenith point is
representative for the entire yover case. To answer this question, the
yover noise is studied.
Figure 12 presents the SPLA at the observers point during a
yover of a UAV with a two-bladed propeller. Three optimal
propellers are considered, designed under the following three power
constraints: P
B
65. 70. 75 W. Results for optimizations without
and with stress constraints are shown. In each case, the UAV ies at
an altitude of 50 m and the SPLA is presented as a function of the
horizontal distance (the horizontal projection of the distance between
the observer and the vehicle). As shown in the gure, when this
distance is positive, the UAV is ying toward the observer, whereas
negative distances indicate that it is ying away from the observer.
In both cases, the maximum noise is not obtained when the
UAV is at the zenith point, but later, at a horizontal distance of
1520 m. This distance increases as the constraint becomes stricter
(namely, the allowed P
B
decreases). For optimal propellers without
stress constraints, the maximum SPLA represents an increase of 2.0,
1.3, and 0.9 dBA for P
B
65. 70. 75 W, respectively, above the
noise at the zenith point. For optimal propellers in the presence of
structural constraints, the increase is 2.3, 1.3, and 1.0 dBA for
P
B
65. 70. 75 W, respectively. In all cases, the trends at the
maximum SPLA point are identical to the trends at the zenith point.
Note that whereas the noise is higher for lower values of allowed
P
B
for the case of a UAV ying away from the observer, the trends
may be reversed for an approaching UAV. To further understand the
behavior, the various contributions to the resultant noise for a yover
of two optimal propellers (P
B
65. 70 W) for an optimal design
without stress constraints are presented in Fig. 13. It is shown that the
far-eld loading noise has the largest contribution and practically
determines the resultant SPLA. Although the thickness noise is fairly
symmetrical for approaching and receding vehicles, the far-eld
loading noise decreases with the horizontal distance much faster for
an approaching UAV than for a receding one. This unsymmetrical
behavior is typical for propeller noise [15,4345]. The decreasing
rate with increasing horizontal distance of the far-eld loading noise
of the receding UAVchanges only slightly as a result of changing the
allowable P
B
. On the other hand, the rate of decrease with horizontal
distance of the far-eld loading noise of the approaching vehicle is
very sensitive to the allowable P
B
, and it increases as the allowable
P
B
decreases. This is the reason that the trends for an approaching
vehicle are reversed compared with a receding UAV. Similar trends
have been observed for yover noise of UAVs with optimal three and
four blades.
-60
-50
-40
-30
-20
-10
0
10
20
30
40
-100 0 100
-60
-50
-40
-30
-20
-10
0
10
20
30
40
-100 0 100
Thick
Far
Near
Total
Flight
direction
Horizontal distance [m] Horizontal distance [m]
P
B
70[W] P
B
65[W]
S
P
L
A
[
d
B
A
]

S
P
L
A
[
d
B
A
]

Fig. 13 Noise components of an optimal two-bladed propeller ying at
an altitude of 50 m for two different performance constraints (without
stress constraints).

59
60
61
62
63
64
65
66
67
2 3 4 5
b
N
P
B
[
W
]

Fig. 11 Minimum P
B
as a function of the number of blades under the
inuence of stress constraints (zero cone angle).
-15
-10
-5
0
5
10
15
20
25
30
35
40
-100 0 100
-15
-10
-5
0
5
10
15
20
25
30
35
40
-100 0 100
P<65[W]
P<70[W]
P<75[W]
Flight
direction
w/o Stress
Constraint
With Stress
Constraint
Horizontal distance [m] Horizontal distance [m]
S
P
L
A
[
d
B
A
]

S
P
L
A
[
d
B
A
]

Fig. 12 Inuence of different performance constraints on the yover
SPLA of an optimal two-bladed propeller ying at an altitude of 50 m.
726 GUR AND ROSEN
VI. Conclusions
The paper presents a comprehensive method for the design of quiet
propellers that is based on a multidisciplinary design optimization
approach. The design of a quiet propeller usually includes a high
number of design variables (6970 design variables in the preceding
described examples) and constraints. Thus, to cope with the
complexity of this design problem, three different optimization
schemes are used that enable the design process to reach a global
optimum. The designer is responsible for the denition of the cost
function and constraints. In addition, the designer decides on
transferring from one optimization scheme to the other.
Often, the design trends of a quiet propeller contradict the design
trends of an efcient propeller. Thus, a tradeoff is required to arrive at
a practical design. The efciency of the propulsion system does not
depend solely on the propeller, but it is affected signicantly by the
motor efciency and the vehicle characteristics. Thus, it is important
to include the motor characteristics in the design process.
Based on the design of a quiet propeller for an electric mini UAV,
the following conclusions can be drawn:
1) For a given number of blades, the main mechanism of reducing
the propeller noise involves a reduction of the propellers rotational
speed.
2) Dening the propeller SPLA as the cost function without
applying additional constraints may result in an impractical design of
blades with a very large chord and high required power. In the present
example, the loitering time for such a design was reduced from
160 min in the case of a maximum loitering time design (without
considering any other constraints) to only 25 min for a very quiet
propeller deign (without applying any performance constraints).
3) Imposing power constraints (namely, imposing a limit of the
extracted power from the battery) increases the propeller noise
signicantly (compared with a design without such constraints). In
the case of a two-bladed propeller, limiting the power to 75 Wresults
in an increase of 10.3 dBA, and a stricter power limit of 65 Wresults
in an increase of 12.3 dBA. This increase in noise is mainly due to the
increase in the rotational speed that leads to an improvement in the
efciency of the propulsion system.
4) Imposing stress constraints leads to an increase in the cross-
sectional thickness and the rotational speed. Because of the increase
in the rotational speed, the SPLA is increased as well (by 13 dBA
for a two-bladed propeller). It has been shown that by introducing a
cone angle, the aerodynamic bending moment is reduced and the
optimal propellers SPLA is decreased by 0.10.3 dBA.
5) As the number of blades increases, the noise signature is
decreased. The reduction in noise is a function of the power
constraint and can reach 10 dBA.
6) The inuence of stress constraints on the optimal design
depends strongly on the number of blades.
7) During yover, the noise of a UAV varies as a function of the
location of the vehicle relative to the observer. The maximum SPLA
is obtained for a receding vehicle shortly after it passes the zenith
point. Nevertheless, optimization for a minimum noise at the zenith
point reduces the noise signature for most of the yover case,
especially along the noisier parts of the path.
8) For most cases, as the performance constraint becomes more
strict (smaller power is allowed), the noise signature increases. Only
for an approaching vehicle at certain distances fromthe observer can
this trend be reversed.
9) The far-eld loading noise presents the main contribution to the
propeller noise. The contributions of the thickness noise and near-
eld loading noise are usually much smaller.
The results of the present paper emphasize the importance of a
multidisciplinary approach during the design of a quiet propeller. It is
important to address all the disciplines simultaneously along the
design process. Ignoring one of these disciplines may lead to an
impractical design.
References
[1] Holmes, J. B., Durham, H. M., and Tarry, E. S., Small Aircraft
Transportation System Concept and Technologies, Journal of
Aircraft, Vol. 41, No. 1, Jan.Feb. 2004, pp. 2635.
doi:10.2514/1.3257
[2] Carey, B., Outlook 2007, Avionics Magazine [online journal],
Jan. 2007, http://www.aviationtoday.com/av/categories/commercial/
7615.html [retrieved 25 Mar. 2007].
[3] Alexander, D., Lee, Y.-M., Guynn, M., and Bushnell, D., Emissionless
Aircraft Study, 38th AIAA/ASME/SAE/ASEE Joint Propulsion
Conference and Exhibit, AIAA Paper 2002-4056, July 2002.
[4] Hoehne, V. O., and Luce, R. G., The Quieted Aircraft as a Military
Tool, AIAA Paper 1969-792, 1969.
[5] Cross, J. L. and Watts, M. E., In Flight Acoustic Testing Techniques
Using the YO-3A Acoustic Research Aircraft, 2nd AIAA/AHS/IES/
SETP/SFTE/DGLR Flight Testing Conference, Las Vegas, NV, AIAA
Paper 83-2754, Nov. 1983.
[6] Fulghum, A. D., Stealth Now, Aviation Week and Space Technology,
Vol. 162, No. 13, May 2005, p. 38.
[7] Keennon, M. T., and Grasmeyer, J. M., Development of the Black
Widow and Microbat MAVs and a Vision of the Future of MAV
Design, AIAA/ICAS International Air and Space Symposium and
Exposition: The Next 100 Years, Dayton, OH, AIAAPaper 2003-2901,
July 2003.
[8] Lawrence, D. A., and Mohseni, K., Efciency Analysis for Long-
Duration Electric MAVs, AIAA Infotech@Aerospace, Arlington,
VA, AIAA Paper 2005-7090, Sept. 2005.
[9] Nagel, A., Levy, D. E., and Shepshelovich, M., Conceptual
Aerodynamic Evolution of Mini/Micro UAV, 44th AIAA Aerospace
Sciences Meeting and Exhibit, Reno, NV, AIAA Paper 2006-1261,
Jan. 2006.
[10] Logan, M. J., Chu, J., Motter, M. A., Carter, D. L., Ol, M., and Zeune,
C., Small UAV Research and Evolution in Long Endurance Electric
Powered Vehicles, AIAA Infotech@Aerospace, AIAA Paper 2007-
2730, May 2007.
[11] Sobieszczanski-Sobieski, J., and Haftka, R. T., Multidisciplinary
Aerospace Design Optimization: Survey of Recent Developments,
Structural Optimization, Vol. 14, 1997, pp. 123.
doi:10.1007/BF01197554
[12] Betz, A., Schraubenpropeller mit Geringstem Energieverlust, Vol. 2,
Knigliche Gesellschaft der Wissenschaften, Gttingen, Germany,
1919, pp. 193217; also Screw Propeller with Minimum Energy Loss,
Technical Translation 736, translated by D. A. Sinclair, National
Research Council Library, Ottawa, 1958.
[13] Adkins, C. N., and Liebeck, R. H., Design of Optimum Propellers,
Journal of Propulsion and Power, Vol. 10, No. 5, Sept.Oct., 1994,
pp. 676682.
doi:10.2514/3.23779
[14] Roncz, G. J., Propeller Development for the Rutan Voyager, General
Aviation Aircraft Meeting and Exposition, Wichita, KS, Society of
Automotive Engineers Paper SAE-891034, Apr. 1989.
[15] Succi, G. P., Design of Quiet Efcient Propellers, SAE Business
Aircraft Meeting, Society of Automotive Engineers Paper SAE79-
0584, Apr. 1979.
[16] Patrick, H., Finn, W. R., and Stich, K. C., Two and Three-Bladed
Propeller Design for the Reduction of Radiated Noise, 3rd AIAA/
CEAS Aeroacoustics Conference, Atlanta, GA, AIAA Paper 97-1710-
CP, May 1997.
[17] Ormsbee, A. I., and Woan, C. J., Optimum Acoustic Design of Free
Running Low Speeds Propellers, AIAA Aircraft Systems and
Technology Meeting, Seattle, WA, AIAA Paper 77-1248, Aug. 1977.
[18] Miller, J. C., and Sullivan, P. J., Noise Constraints Effecting Optimal
Propeller Designs, SAE General Aviation Aircraft Meeting and
Exposition, Wichita, KS, Society of Automotive Engineers Paper 85-
0871, Apr. 1985.
[19] Burger, C., Harteld, J. R., and Burkalter, J., Performance and Noise
Optimization of a Propeller Using the Vortex Lattice Method and a
Genetic Algorithm, 48th AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, Honolulu, HI, AIAA
Paper 2007-1883, Apr. 2007.
[20] Patrick, H., Finn, R. W., and Stich, C. K., Two- and Three-Bladed
Propeller Design for the Reduction of Radiated Noise, 3rd AIAA/
CEAS Aeroacoustics Conference, Atlanta, GA, AIAA Paper 97-1710,
1214 May, 1997.
[21] Drack, E. L., and Wood, A. L., DESIGN AND Analysis of Propellers
for General Aviation Aircraft Noise Reduction, 21st ICAS Congress,
Melbourne, Australia, International Council of Aeronautical Sciences
Paper 98-5.11.3, Sept. 1998.
[22] Janus, J. M., General Aviation Propeller-Airframe Integration
Simulation, Journal of Aircraft, Vol. 43, No. 2, Mar.Apr. 2006,
pp. 390394.
doi:10.2514/1.15354
GUR AND ROSEN 727
[23] Gur, O., and Rosen, A., Comparison Between Blade-Element Models
of Propellers, The Aeronautical Journal, Vol. 112, No. 1138,
Dec. 2008, pp. 689704.
[24] Gur, O., and Rosen, A., Propeller Performance at Low Advance
Ratio, Journal of Aircraft, Vol. 42, No. 2, Mar.Apr. 2005, pp. 435
441.
doi:10.2514/1.6564
[25] Yamamoto, O., and Aug., R., Structural and Aerodynamic Analysis of
a Large-Scale Advanced Propeller Blade, Journal of Propulsion and
Power, Vol. 8, No. 2, Mar.Apr., 1992, pp. 367373.
doi:10.2514/3.23487
[26] Rosen, A., and Gur, O., A Transfer Matrix Model of Large
Deformations of Curved Rods, Computers and Structures, E-rst
publication Feb. 2009.
doi:10.1016/j.compstruc.2008.12.014
[27] Ffowcs, J. E. W., and Hawkings, D. L., Sound Generation by
Turbulence and Surfaces in Arbitrary Motion, Philosophical
Transactions of the Royal Society of London, Series A: Mathematical
and Physical Sciences, Vol. 264, 1969, pp. 321342.
doi:10.1098/rsta.1969.0031
[28] Magliozzi, B., Hanson, D. B., and Amiet, K. R., Aeroacoustics of
Flight Vehicles Theory and Practice, edited by H. H. Hubbard, NASA
TR 90-3052, Aug. 1991.
[29] Hanson, D. B., and Fink, M. R., The Importance of Quadrupole
Sources in Prediction of Transonic Tip Speed Propeller Noise, Journal
of Sound and Vibration, Vol. 61, No. 1, 1978, pp. 1938.
[30] Farassat, F., and Succi, G. P., A Review of Propeller Discrete
Frequency Noise Prediction Technology with Emphasis on Two
Current Methods for Time Domain Calculations, Journal of Sound
and Vibration, Vol. 71, No. 3, 1980, pp. 399419.
doi:10.1016/0022-460X(80)90422-8
[31] Sanford, F., Pearsons, K. S., and Bennett, R. L., Predicting Aural
Detectability of Aircraft in Noise Background, U.S. Air Force Flight
Dynamics Lab. TR 72-17, WrightPatterson AFB, OH, July 1972.
[32] Dobrzynski, W. M., Heller, H. H., Powers, J. O., and Densmore, J. E.,
DFVLR/FAAPropeller Noise Test in the German-Dutch Wind Tunnel
DNW, Federal Aviation Administration Rept. AEE 86-3, 1986.
[33] Magliozzi, B., The Inuence of Forward Flight on Propeller Noise,
NASA CR 145105, 1977.
[34] Succi, G. P., Computed and Experimental Spectra for a Wing Mounted
Microphone on a Light STOLAircraft, NASACR165725, May 1978.
[35] Borst, V. H., Summary of Propeller Design Procedures and Data,
Vol. 1, U.S. Army Air Mobility Research and Development Lab.,
TR 73-34A-VOL-1, Nov. 1973.
[36] Boyd, S., and Vandenberghe, L., Convex Optimization, Cambridge
Univ. Press, New York, 2004, p. 716.
[37] Goldberg, E. D., Genetic Algorithms in Search, Optimization, and
Machine Learning, Addison Wesley, Reading, MA, 1989, p. 412.
[38] Nelder, J. A., and Mead, R., A Simplex Method for Function
Minimization, Computer Journal, Vol. 7, No. 4, 1965, pp. 308313.
[39] Van der Velden A., Tools for Applied Engineering Optimization,
AGARD Rept. R-803, Neuilly-sur-Seine, France, Apr. 1994
[40] Gur, O., and Rosen, A., Optimization of a Propeller Based Propulsion
System , Journal of Aircraft, Vol. 46, No. 1, Jan.Feb. 2000, pp. 95
106.
[41] Sullivan, J. P., Chang, L. K., and Miller, C. J., The Effect of Proplets
and Bi-Blades on the Performance and Noise of Propellers, SAE
Business Aircraft Meeting and Exposition, Wichita, KS, Society of
Automotive Engineers Paper 810600, Apr. 1981.
[42] Gur, O., and Rosen, A., Optimizing Electric Propulsion Systems for
UAVs, 12th AIAA/ISSMO Multidisciplinary Analysis and
Optimization Conference, Victoria, BC, Canada, AIAA Paper 2008-
5916, Sept. 2008.
[43] Hanson, D. B., The Inuence of Propeller Design Parameters on Far
Field Harmonics Noise in Forward Flight, AIAA 5th Aeroacoustics
Conference, Seattle, WA, AIAA Paper 79-0609, Mar. 1979.
[44] Herkes, W., An Experimental Study of the Noise Generated by a
Pusher Propeller Due to Wake Entering the Propeller Disc, von
Karman Inst. for Fluid Dynamics, Rept. 1979-17, Rhode-St-Gense,
Belgium, Nov., 1979.
[45] Heller, H., Propeller Aircraft Noise-Certication and Flight-Testing,
Deutsche Forschungs- und Versuchsanstalt fr Luft- und Raumfahrt,
Paper 82-16, Oct. 1982.
C. Tan
Associate Editor
728 GUR AND ROSEN