Composite Structures 89 (2009) 206–215

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Optimization and experiment of composite marine propellers

Ching-Chieh Lin a, Ya-Jung Lee b,*, Chu-Sung Hung c
a
Material Research Laboratories, Industrial Technology Research Institute, 5F., No. 75, Gangqian Road, Neihu District, Taipei City 114, Taiwan, ROC
b
Department of Engineering Science and Ocean Engineering, National Taiwan University, 73 Chou-Shan Road, Taipei City 106, Taiwan, ROC
c
Atech Composites Co., Ltd., No. 3, Yeong Kuang Street, Kaohsiung, Taiwan, ROC

a r t i c l e i n f o a b s t r a c t

Article history: This paper demonstrates the experimental result of changeable pitch propellers in composite material.
Available online 24 July 2008 The use of this composite material is to apply its characteristic of bend-twist coupling; this characteristic
is applied to better performance design requirements of the propeller. Two stacking sequences are con-
Keywords: sidered: the first one is a quasi isotropic sequence, while the second one is an optimum sequence
Fiber-reinforced plastic obtained using a genetic algorithm. Experiments are designed considering two original propellers man-
FRP ufactured by the first and the second stacking sequence, respectively, and a pre-deformed propeller with
Propeller
the second sequence. Experimental results correspond to the same trend as the calculations and confirm
Optimization
Genetic algorithm
the method of optimization.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction
Fiber-reinforced plastic (FRP) is extensively applied in various
structures, because of its light weight, high strength and corrosion
resistance. Bend-twist coupling effect is another unique character-
istics of composite material. Structures can be stiffened or de-
formed in a certain direction by arranging the orientation of the
fibers.
Most marine propellers are made of metal material such as
bronze or steel. The advantages of replacing metal with an FRP
composite are that the latter is lighter and corrosion-resistant. An-
other important advantage is that the deformation of the compos-
ite propeller can be controlled to improve its performance. Fig. 1
schematically depicts inflow angles at different ship speed. Propel-
lers always rotate at a constant velocity that maximizes the effi-
ciency of the engine. When the ship sails at the designed speed,
the inflow angle is close to its pitch angle. When the ship sails at
a lower speed, the inflow angle is smaller. Hence, the pressure on
the propeller increases as the ship speed decreases. The propulsion
efficiency is also low when the inflow angle is far from the pitch
angle. If the pitch angle can be reduced when the inflow angle is
low, then the efficiency of the propeller can be improved.
Bend-twist coupling is a unique characteristic of FRP composite:
structures twist when they are bended. This effect appears when
the laminating sequence is unbalanced, meaning that the number
of positive angle plies is not equal to the number of negative angle
plies [1]. This effect has been exploited in aerodynamics to solve the
* Corresponding author. Tel.: +886 2 233665773.
E-mail addresses: d87525005@ntu.edu.tw (C.-C. Lin), yjlee@ntu.edu.tw (Y.-J.
Lee), JackHung@horizonyacht.com (C.-S. Hung).
problems of fluttering and divergence ([2,3] for example). It has also
been applied in forward swept wings [4] and propellers ([5,6]) to
control deformation. The same effect can be employed in the mar-
ine propeller. The water pressure results in bending and twisting,
which, combined with the arranged ply angles, changes the pitch
of the propeller and thus changes its performance.
Our previous study employed the bend-twist effect and ‘pre-de-
formed’ design to produce a changeable-pitch-propeller [7]. The
optimum stacking sequence has been obtained using a genetic
algorithm [8,9]. The deformation and performance of the propeller
are calculated iteratively. The calculation is performed using the
finite element software ABAQUS and the propeller calculation pro-
gram PSF2 [10,11]. The thrust, torque, efficiency and pressure dis-
tribution are determined using the program PSF2 (developed by
the Massachusetts Institute of Technology), which applied numer-
ical lifting-surface theory for marine propellers as a practical tool
in the solution of both steady and unsteady flow problems. When
the pressure that acts on the blade is obtained, it is then input to
the finite element software ABAQUS to calculate the deformation
of the blade. Then, the deformed blade geometry is input to the
PSF2 again. After several iterations, a converged pressure and
deformation are obtained [7]. This paper introduces the optimiza-
tion method and performs experiments to verify the optimization
design of the composite propeller.
2. Optimum propeller
2.1. Definition of optimum propeller
The machinery that propels a ship must be selected to yield an
optimal solution for deadweight, storage and several other factors.

0263-8223/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2008.07.020
 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215 207

nance costs in keeping the engine at peak performance throughout

Fig. 1. Inflow angle at various ship speeds.
The machinery is selected such that the necessary propulsion
power is produced as efficiently as possible. Any change in one
of the part systems may seriously reduces the efficiency of the pro-
pulsion systems. Therefore, the characteristic curves of the opti-
mum propeller must match the operating area of the main engine.
The diesel engine has a general characteristic of the type pre-
sented in Fig. 2, on which a propeller demand curve is superim-
posed, and in this instance passes through the maximum
continuous rating (MCR) of the engine (the point of 100% power
and 100% revolution in Fig. 2). If the pitch of the propeller is se-
lected incorrectly, then the propeller will be either over-pitched
(curve A) or under-pitched (curve B). In either case, the maximum
power of the engine is not reached, because, in the case of over-
pitching, the maximum power attainable will be point X at a re-
duced RPM, which result is governed by the engine torque limit.
In the alternative under-pitching case, the maximum power that
can be reached is point Y at 100% RPM, since the engine speed limit
is the governing factor. In practice, the derated engine power is
used as the basis for propeller design, to prevent excessive mainte-
its life. Therefore, the propeller is generally based on a normal con-
tinuous rating (NCR) of between 85% and 90% of the MCR condi-
tion; Fig. 2 shows a typical propeller design point.
Propellers are designed to operate at a constant ratio of axial to
rotational velocity, at which point the torque, thrust and efficiency
are optimal for the propulsion system. Hence, the first limitation
on the design of a composite propeller is to maintain the originally
designed torque, thrust and efficiency at the designed axial
velocity.
When the propeller operates at a low axial velocity, it is in an
over-pitched condition, and the torque exceeds the tolerance of
the engine. Accordingly, the second rule that governs the design
of an optimum composite propeller is that the pitch must decrease
as much as possible as the inflow angle is reduced, to keep the tor-
que within the narrowest possible range.
The characteristics of the propeller are generally presented
using the following coefficients:
Va
advance coefficient J ¼
nD
Q
torque coefficient K Q ¼
qn2 D5
ð1Þ
T
thrust coefficient K T ¼
qn2 D4
TV a KT J
efficiency g ¼ ¼ 
2pnQ K Q 2p
where D = diameter of the propeller; Va = axial velocity; n = rota-
tional velocity (rps); q = fluid density; T = thrust and Q = torque.
Herein, the propeller DTNSRDC 4498 [12] is chosen as the de-
sign target; Table 1 presents its geometric parameters. The design
advance coefficient is J = 0.889 and the corresponding torque coef-
ficient is KQ = 0.05204. These values represent the design point at
which a composite propeller should optimally function. The other
point considered is J = 0.6, with a corresponding KQ of 0.07311. The
KQ of the optimally functioning propeller should be as close as pos-
sible to the designed value of 0.05204.
Therefore, the performance of the optimum propeller is defined
as follows:

(1) KQ at J = 0.889 must be equal to the designed KQ (0.05204).

(2) KQ at J = 0.6 must be as close as possible to the designed KQ
(0.05204).

The goal is to design a propeller with the same pitch as that

originally designed at J = 0.889, and with the smallest pitch at

Table 1
Geometric parameters of propeller DTNSRDC 4498 [11]

Pitch Rake Skew Chord Camber Thickness

r/R P/D Rake/D Degree C/D Camber/Ci Thickness/D
0.2 1.566 0 0. 0.174 0.0402 0.0434
0.25 1.539 0 4.647 0.202 0.0408 0.0396
0.3 1.512 0 9.293 0.229 0.0407 0.0358
0.4 1.459 0 18.816 0.275 0.0385 0.0294
0.5 1.386 0 27.991 0.312 0.0342 0.0240
0.6 1.296 0 36.770 0.337 0.0281 0.0191
0.7 1.198 0 45.453 0.347 0.0230 0.0146
0.8 1.096 0 54.245 0.334 0.0189 0.0105
0.9 0.996 0 63.102 0.280 0.0159 0.0067
0.95 0.945 0 67.531 0.210 0.0168 0.0048
1.0 0.895 0 72.000 0.000 0.0000 0.0029

Number of blades: 5; diameter: 0.305 m; hub-diameter ratio: 0.2; expanded area

ratio: 0.725; section meanline: NACA a = 0.8 and section thickness distribution:
Fig. 2. Engine characteristic curve. NACA 66.
208 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215

J = 0.6. The pitch of the optimum propeller changes automatically

with the axial velocity.

2.2. Optimization of the propeller

First, the stacking sequence that yields the least slope of the tor-
que coefficient curve at J = 0.6 must be found. The optimum stack-
ing sequence is found using the method proposed in [8,9].
Unfortunately, the optimized stacking sequence not only reduces
the slope of the curve at J = 0.6, but also significantly reduces the
KQ at J = 0.889, which is supposed to be fixed at the designed value
(see the original geometry curve in Fig. 3). Therefore, this propeller
does not fulfill the first requirement of optimization.
A pre-deformed propeller is then adopted to ensure that the tor-
que coefficient remains at the designed value. Pre-deformed de- Fig. 4. Pre-deformed propeller.
signs are commonly employed for high deflection structures,
such as wings. Traditional marine propellers are made of metal
and their deflection is small enough that pre-deformed design is
unnecessary. However, the deflection of composite propellers can
be larger with certain stacking sequences and a pre-deformed de-
sign may help to solve the optimization problem. A pre-deformed
design means that a blade is displaced in the direction opposite to
its deformation under pressure. When the propeller operates under
pressure at the designed speed, it returns to the originally designed
geometry, such that the torque, thrust and efficiency are at the de-
signed values. Fig. 4 presents an example of the pre-deformed pro-
peller. The upper left-hand section displays the original designed
geometry. Because of the thinner tail, the pitch decreases when
the propeller operates under pressure (shown in the upper right
section in Fig. 4). An opposite displacement increases the pitch,
as shown in the lower left part of Fig. 4. When the pre-deformed
propeller operates under pressure, it deforms to the originally de-
signed geometry (shown in the lower right section in Fig. 4). Thus,
the torque at J = 0.889 is the designed torque; when the propeller is
Fig. 5. Cavitation tank.
operated at J = 0.6, the higher pressure reduces the pitch and tor-
que as required.
In the experiment conducted herein, the diameter of the propel-
ler was 20 cm, rotational speed 7–20 revolutions per second (rps),
and the axial fluid speed 3.55 m/s. The optimum stacking sequence Table 2
was [452/902/452/452/452/452/02/02/02/02/02/452]s. Fig. 3 plots the Geometric parameters of pre-deformed propeller
calculated torque coefficient curves for the pre-deformed propeller.
Pitch Rake Skew Chord Camber Thickness
r/R P/D Rake/D Degree C/D Camber/Ci Thickness/D
0.2 1.571 0.000 0.035 0.174 0.0402 0.0434
0.25 1.538 0.000 4.650 0.202 0.0408 0.0396
0.3 1.509 0.000 9.305 0.229 0.0407 0.0358
0.8 no deflection 0.4 1.462 0.000 18.822 0.275 0.0385 0.0294
0.5 1.393 0.000 28.001 0.312 0.0342 0.024
original geometry
0.6 1.309 0.000 36.778 0.337 0.0281 0.0191
[452/902/452/452/452/452/02/02/02/02/02/452]s 0.7 1.215 0.000 45.465 0.347 0.023 0.0146
predeformed 0.8 1.124 0.000 54.259 0.334 0.0189 0.0105
0.7 0.9 1.033 0.000 63.129 0.28 0.0159 0.0067
[452/902/452/452/452/452/02/02/02/02/02/452]s 0.95 0.980 0.001 67.557 0.21 0.0168 0.0048
1 0.916 0.003 71.974 0 0 0.0029

Number of blades: 5; diameter: 0.305 m; hub-diameter ratio: 0.2; expanded area

10KQ

0.6 ratio: 0.725; section meanline: NACA a = 0.8 and section thickness distribution:
NACA 66.

0.5

Table 3
The material constants of the propeller

0.4 Young’s modulus (longitudinal) E1 = 1.12  1011 Pa

Young’s modulus (transverse) E2 = 1.05  1010 Pa
0.6 0.7 0.8 0.9 Shear modulus G12 = 8.4  109 Pa
J Poisson’s ratio 0.32
Ply thickness t = 1.55  104 m
Fig. 3. Calculated torque coefficient curves of propellers D = 20 cm.
 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215 209

Fig. 6. Mold of blade.

Fig. 7. Finished blade.

Fig. 9. Superposition of images of deflected and undeflected blades.

Fig. 10. Partially magnified image of the blade.

Table 4
Experiment result of [452/902/452/02/452/902/452/02/452/902/452/02]s propeller
without pre-deform design (n = 13)
Fig. 8. Composite propeller.
J T (kg) Q (kg cm) KT 10 * KQ g
The torque coefficient at J = 0.889 is close to designed value and the 0.5 12.289 44.377 0.4469 0.8070 0.4407
0.6 10.912 40.706 0.3969 0.7403 0.5120
slope is smaller than that of the original propeller.
0.7 9.669 37.371 0.3517 0.6796 0.5765
0.8 8.273 33.584 0.3009 0.6107 0.6273
3. Experimental design 0.889 7.251 30.740 0.2637 0.5590 0.6675
1 5.843 26.586 0.2125 0.4835 0.6996
1.1 4.541 22.747 0.1651 0.4137 0.6989
The purpose of the experiment is to verify the results of numer-
1.2 3.110 18.267 0.1131 0.3322 0.6503
ical analysis and optimization. Therefore, the performance and
210 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215

Table 5 Table 10
Experiment result of [452/902/452/02/452/902/452/02/452/902/452/02]s propeller Experiment result of [452/902/452/452/452/452/02/02/02/02/02/452]s propeller with
without pre-deform design (n = 7) pre-deform design (n = 20)

J T (kg) Q (kg cm) KT 10 * KQ g J T (kg) Q (kg cm) KT 10 * KQ g

0.5 3.496 13.975 0.4387 0.8765 0.3983 0.889 18.625 78.380 0.2862 0.6022 0.6724
0.6 3.095 12.786 0.3883 0.8019 0.4624 1 15.354 68.751 0.2359 0.5282 0.7108
0.7 2.781 11.916 0.3489 0.7473 0.5201 1.1 12.630 60.574 0.1941 0.4654 0.7300
0.8 2.418 10.771 0.3033 0.6755 0.5717
0.889 2.140 9.920 0.2685 0.6221 0.6106
1 1.721 8.657 0.2159 0.5430 0.6328
1.1 1.351 7.496 0.1694 0.4701 0.6310 1
1.2 0.923 6.122 0.1158 0.3840 0.5761 Original Geometry n = 13 rps
[-452/902/452/02/-452/902/452/02/-452/902/452/02]s

0.8
Table 6

KT, 10KQ, and Efficiency η

Experiment result of [452/902/452/452/452/452/02/02/02/02/02/452]s propeller without
pre-deform design (n = 13)

J T (kg) Q (kg cm) KT 10 * KQ g 0.6

0.5 12.385 44.142 0.4505 0.8027 0.4466
0.6 11.072 40.563 0.4027 0.7376 0.5213
0.7 9.693 36.851 0.3525 0.6701 0.5861
0.8 8.401 33.376 0.3056 0.6069 0.6410 0.4
0.889 7.330 30.360 0.2666 0.5521 0.6832
1 6.030 26.657 0.2193 0.4848 0.7201
10KQ
1.1 4.658 22.580 0.1694 0.4106 0.7223
1.2 3.361 18.453 0.1222 0.3356 0.6957 0.2 KT

Table 7 0
Experiment result of [452/902/452/452/452/452/02/02/02/02/02/452]s propeller without
pre-deform design(n = 7) 0.4 0.6 0.8 1 1.2
J
J T (kg) Q (kg cm) KT 10 * KQ g
Fig. 11. Experimental result for [452/902/452/02/452/902/452/02/452/902/452/
0.5 3.834 13.811 0.4809 0.8662 0.4419
02]s propeller without pre-deform design (n = 13 rps).
0.6 3.433 12.788 0.4307 0.8020 0.5128
0.7 3.128 11.854 0.3924 0.7435 0.5880
0.8 2.797 10.927 0.3509 0.6853 0.6520
0.889 2.501 10.020 0.3137 0.6284 0.7064
1 2.102 8.763 0.2637 0.5496 0.7637 1 Original Geometry n = 7rps
1.1 1.727 7.563 0.2167 0.4743 0.7996 [-452/902/452/02/-452/902/452/02/-452/902/452/02]s
1.2 1.282 6.178 0.1609 0.3874 0.7929

0.8
KT, 10KQ, and Efficiency η

Table 8
Experiment result of [452/902/452/452/452/452/02/02/02/02/02/452]s propeller with 0.6
pre-deform design (n = 13)

J T (kg) Q (kg cm) KT 10 * KQ g

0.5 13.889 50.666 0.5052 0.9214 0.4363 0.4
0.6 12.584 46.870 0.4577 0.8523 0.5128
0.7 11.239 42.918 0.4088 0.7805 0.5835 10KQ
0.8 9.912 38.996 0.3605 0.7092 0.6472
0.889 8.656 35.469 0.3148 0.6450 0.6906
0.2
KT
1 7.321 31.425 0.2663 0.5715 0.7416
1.1 6.025 27.360 0.2191 0.4975 0.7710 η
1.2 4.669 23.074 0.1698 0.4196 0.7729

0
0.4 0.6 0.8 1 1.2
Table 9
J
Experiment result of [452/902/452/452/452/452/02/02/02/02/02/452]s propeller with
Fig. 12. Experimental result for [452/902/452/02/452/902/452/02/452/902/452/
pre-deform design (n = 7)
02]s propeller without pre-deform design (n = 7 rps).
J T (kg) Q (kg cm) KT 10 * KQ g
0.5 3.793 15.187 0.4758 0.9525 0.3975 deformation of the propeller are the most important factors. The
0.6 3.439 14.141 0.4314 0.8868 0.4645
0.7 3.056 12.943 0.3834 0.8117 0.5262
experiment is performed in a cavitation tank to observe the defor-
0.8 2.655 11.718 0.3331 0.7349 0.5771 mation of the propeller (as presented in Fig. 5). The transparent
0.889 2.337 10.711 0.2932 0.6718 0.6176 window allows close observation. However, the diameter and tor-
1 1.953 9.520 0.2451 0.5971 0.6532 que are limited by the tank size and measurement devices. Hence,
1.1 1.521 8.126 0.1909 0.5096 0.6557
the torque and thrust are not measured in the higher rotational
1.2 1.110 6.890 0.1392 0.4321 0.6154
speed experiments, only deformation is recorded.
 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215 211

1 1
Original Geometry n = 13rps Predeformed Geometry n = 13 rps
[452/902/452/452/452/452/02/02/02/02/02/452]s [452/902/452/452/452/452/02/02/02/02/02/452]s

0.8 0.8

KT, 10KQ, and Efficiency η

KT, 10KQ, and Efficiency η

0.6 0.6

0.4 0.4

10KQ 10KQ

KT 0.2
KT
0.2

η η

0
0
0.4 0.6 0.8 1 1.2
0.4 0.6 0.8 1 1.2
J
J
Fig. 15. Experimental result for [452/902/452/452/452/452/02/02/02/02/02/452]s pro-
Fig. 13. Experimental result for [452/902/452/452/452/452/02/02/02/02/02/452]s pro- peller with pre-deform design (n = 13 rps).
peller without pre-deform design (n = 13 rps).

1 Original Geometry n = 7 rps

[452/902/452/452/452/452/02/02/02/02/02/452]s 1 Predeformed Geometry n = 7 rps
[452/902/452/452/452/452/02/02/02/02/02/452]s

0.8 0.8
KT, 10KQ, and Efficiency η

KT, 10KQ, and Efficiency η

0.6
0.6

0.4
0.4
10KQ
10KQ
0.2
KT
0.2 KT
η
η
0
0
0.4 0.6 0.8 1 1.2
J 0.4 0.6 0.8 1 1.2
J
Fig. 14. Experimental result for [452/902/452/452/452/452/02/02/02/02/02/452]s pro-
peller without pre-deform design (n = 7 rps). Fig. 16. Experimental result for [452/902/452/452/452/452/02/02/02/02/02/452]s pro-
peller with pre-deform design (n = 7 rps).

Two propeller geometries are applied in the experiment: one is

the original DTNSRDC 4498 design; the other is the pre-deformed
DTNSRDC 4498. Table 2 presents the pre-deformed geometric
parameters. Two stacking sequences are applied. One is [452/
902/452/02/452/902/452/02/452/902/452/02]s, which is a quasi-
isotropic sequence. The other is [452/902/452/452/452/452/02/02/
02/02/02/452]s, which is the optimum sequence that was obtained
using a genetic algorithm. The first stacking sequence is denoted
sequence 1 and the second is sequence 2 in the following descrip-
tion. Three propellers are adopted in the experiment: (1) original
shape with sequence 1 (2) original shape with sequence 2 (3)
pre-deformed shape with sequence 2. Each propeller operates at
three rotational speeds, which are n = 20 rps, 13 rps and 7 rps.
The propeller is made from the prepreg of Toho HTA1200 car-
bon fiber/ACD8801 epoxy. Table 3 presents the material constants.
Fig. 6 displays the mold, which is made of aluminum alloy. The pre-
preg is laminated on the lower mold and then the upper mold is
put on. The thickness of the blade determines the number of layers.
Subsequently the mold is heated to 130 °C at a pressure of 30 psi
for 40 min. Then, the blade is taken out and the surface smoothed
using epoxy clay. Fig. 7 displays the finished blade. Finally, the
blade is fixed into a stainless steel hub and the propeller is com-
plete (Fig. 8).
In the cavitation tank (Fig. 5), the water flows from larger cross-
sectional area A1 to the smaller cross-sectional area A2. The cross-
sectional area of A2 is 40 cm  40 cm and A1/A2 is 2.86. The fluid
pressures at A1 and A2 are measured to determine the fluid veloc-
ity. Bernoulli’s equation (Eq. (2)) and the constant flow quantity
equation Eq. (3) are applied to determine the fluid velocity (Eq. (4))
212 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215

0.8 Predeformed Geometry n = 20 rps calculated with n = 7 rps

[452/902/452/452/452/452/02/02/02/02/902/452]s 1
experiment with n = 7 rps
[-452/902/452/02/-452/902/452/02/-452/902/452/02]s

0.8
0.6
KT, 10KQ, and Efficiency η

KQ
0.6

KT, 10KQ
0.4
0.4

KT
10KQ 0.2
0.2
KT
0
η
0.4 0.6 0.8 1 1.2
J
0
Fig. 19. Calculated and experimental results of propeller with sequence one at
0.4 0.6 0.8 1 1.2 n = 7 rps.
J

Fig. 17. Experimental result for [452/902/452/452/452/452/02/02/02/02/02/452]s pro-

peller with pre-deform design (n = 20 rps).

[-452/902/452/02/-452/902/452/02/-452/902/452/02]s
1
calculated with n = 7 rps
(deflection considered)
0.8 calculated by PSF2 (no deflection)

KQ
0.6
KT, 10KQ

0.4

KT
0.2

0
0.4 0.6 0.8 1 1.2
J

Fig. 18. Difference between calculations with and without deflection at n = 7.

1 1
P1 þ qV 21 ¼ P2 þ qV 22
2 2
A1 V 1 ¼ A2 V 2
pffiffiffiffiffiffiffi
DP
V2 ¼
20:95
Measuring the deflection of the blades is the most difficult
problem. Since the propeller rotates and the deflection direction
is unclear, determining the displacement of the blade is difficult.
Therefore, photography is employed to determine the displace-
ment of particular points on the blade. First, the hub is marked,
the propeller rotated and pictures taken every 0.5° in a static state.
Subsequently, pictures of the deflected blade are also taken when
the propeller operates. The pictures of the deflected and the unde-
flected blade are superimposed, to reveal the deflection of the
blade, as presented in Fig. 9. Notably, the marks on the hub must
ð2Þ
ð3Þ Fig. 20. Comparison of numerical [11] and experimental results in [12].
ð4Þ
be exactly superimposed to confirm the position of the blade.
The greatest deflection appears at the tip of the blade. Fig. 101 dis-
plays a partially magnified image; the red marks are at the leading
and trailing edge at 0.95R of the blade (R = radius). The yellow
point indicates the deflected position and the green point indicates
the undeflected point. The displacement can be determined by
counting the pixels between two points. In Fig. 10, each pixel rep-
resents 0.0625 mm.
1
For interpretation of color in Fig. 10, the reader is referred to the web version of
this article.
 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215 213

However, only the axial displacement (dx) can be determined in
this way. Fortunately, only the axial displacement is required to
determine the change of pitch. If the displacement at the leading
edge is smaller than that at trailing edge, then the pitch declines
as revealed by the numerical analysis.
4. Experimental result
Tables 4–10 present the experiment results and Figs. 11–17 plot
the corresponding curves. A small rotational velocity case is con-
sidered to compare the results with those of numerical analysis.
At a small rotational velocity, the pressure on the blade is small,
and the deflection of the blade is therefore tiny. Consequently
the result at small rotational velocity should be close to the result
calculated from PSF2, in which no deflection considered. Fig. 18
plots the calculated result at n = 7 rps using PSF2 neglecting blade
deflection and present method. The figure shows very little differ-
ence between the results with and without propeller deflection.
Fig. 19 plots the results of calculation and experiment at
n = 7 rps. The experimental and calculated KT curves are close to
each other, although the former is higher. An experiment at MIT
yielded similar results. Fig. 20 plots calculated and experimental
results published in another paper [11], it compares the results
of numerical analysis and experiment. The solid curve represents
the experimental result of DTNSRDC [12]; the dashed curve plots

1
[-452/902/452/02/-452/902/452/02/-452/902/452/02]s 1
Predeformed
[452/902/452/452/452/452/02/02/02/02/02/452]s

0.8

0.8

KT, 10KQ
10KQ

0.6

0.6 n = 20 rps
0.4 n = 13 rps
n = 13 rps
n = 7 rps
n = 7rps

0.2
0.4
0.4 0.6 0.8 1 1.2
0.4 0.6 0.8 1 1.2
J
J
Fig. 21. KQ of propeller with stacking sequence one at n = 7 and 13 rps.
Fig. 23. KQ of pre-deformed propeller with stacking sequence two.

1
[452/902/452/452/452/452/02/02/02/02/02/452]s
Predeformed n = 20 rps
1 [452/902/452/452/452/452/02/02/02/02/02/452]s
Non-predeformed n = 7 rps
0.8 [-452/902/452/02/-452/902/452/02/-452/902/452/02]s
0.8
KT, 10KQ

0.6
10KQ

0.6

0.4
n = 13 rps
0.4
n = 7 rps

0.2 0.2

0.4 0.6 0.8 1 1.2 0.4 0.6 0.8 1 1.2

J J

Fig. 22. KQ of propeller with stacking sequence two and without pre-deformed Fig. 24. KQ of optimized propeller and the quasi-isotropic propeller without pre-
design. deformed design.
214 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215

Table 11
Displacement of the tip of the blade at J = 0.6 (n = 20)

Propeller Geometry Stacking sequence Leading edge Trailing edge

Experiment Calculation Experiment Calculation
1 Original 1 0.4375 0.669 0.9375 1.090
2 Original 2 0.3530 0.875 0.8240 2.344
3 Predeform 2 0.3125 0.989 2.1875 2.797

Table 12
Displacement of the tip of the blade at J = 0.889 (n = 20)

Propeller Geometry Stacking sequence Leading edge Trailing edge

Experiment Calculation Experiment Calculation
1 Original 1 0.3125 0.447 0.5625 0.831
2 Original 2 0.1170 0.605 0.4120 1.817
3 Predeform 2 0.0625 0.718 1.1230 2.115

numerical analysis result from MIT [11] and the square points rep-
resent calculations obtained using PSF2. The results demonstrate
that the calculated KQ is smaller than the experimental KQ at
J = 0.6–1.0. The experiments in this study yielded similar results.
Although the calculated results are not very close to the experi-
mental results, this study investigates the variation of deflection
with speed; a small error is acceptable if all calculations are consis-
tent with the same trend.
Fig. 21 compares the performance of the quasi-isotropic propel-
ler at n = 7 and 13 rps to observe the effect of blade deformation
(stacking sequence 1). The rotational speed at n = 13 is 2.6 times
that at n = 7. The fluid pressure on the blade is (2.6)2  6.7 times
that at n = 7, and the deflection is much larger. First, the case of
small deflection is considered with n = 7 rps, the calculated torque
coefficients at J = 0.889 and 0.6 are K 0:889 Q ¼ 0:05201 and
K Q0:6 ¼ 0:07303. For the large deflection with n = 13 rps, the calcu-
lated K 0:889
Q ¼ 0:05072 and K Q0:6 ¼ 0:07169. The calculation shows
that KQ at n = 13 is smaller than that at n = 7. Deformation of the
blade reduces the pitch of the propeller, reducing the torque coef-
ficient. The curve at n = 13 is lower than that at n = 7, as presented
in Fig. 21, which result is consistent with the numerical analysis.
The other two propellers, the propellers with stacking sequence
2, with and without the pre-deformed design, follow the same
trend. Fig. 22 plots the results for the original propeller and Fig.
23 plots those for the pre-deformed propeller. The torque coeffi-
cient curve declines as the rotational speed increases.
The pre-deformed propeller with sequence 2 is the optimized
result in this research and the slope of the torque coefficient curve
should be less than that of the original design. As presented in Fig.
23, the KQ and the slope decrease as the rotational speed increases,
indicating that the pitch of the propeller is actually reduced. In
addition to the reduction of slope, the optimum propeller must
have a KQ at J = 0.889 that equals the originally designed value with
no deflection considered. However, a rigid propeller without
deflection was unavailable in this work, so the optimized propeller
was compared with the original propeller at a very low rotational
speed, which induces very small deflection. Fig. 24 plots the torque
coefficient curves of the optimized propeller at n = 20 rps and the
original propeller at n = 7 rps. The KQ values of both curves at
J = 0.889 are very close to each other, as can be easily seen by com-
parison with Fig. 24. The optimized propeller fulfills both require-
ments of optimization. Unfortunately, however, the measurement
equipment cannot measure the torque at small J when the rota-
tional speed is n = 20. The deflection of the propeller is analyzed
to determine the change of pitch and thus elucidate the perfor-
mance of the optimized propeller at small J.
Table 11 presents the deflection of the propellers at n = 20 and
J = 0.6. The experimental displacements are smaller than the calcu-
lated displacements. The measured points have the same radius
(r = 0.95R) and the change of the pitch can be determined by calcu-
lating the difference between the displacements of the leading and
trailing edges. If the displacement of the trailing edge is larger, then
the pitch is reduced. This table demonstrates that the pitch of the
three propellers decreases as expected. Table 12 presents the deflec-
tions of the propellers at n = 20 and J = 0.889. The change of pitch is
smaller than at J = 0.6. The comparison indicates that the pitch of the
propeller decreases as the J value decreases, indicating that the pro-
peller can automatically change its pitch with the inflow angle.
5. Conclusions
The measured displacements in this study are all smaller than
the calculated values, but both sets of results follow the same
trend. For a given rotational speed, the pitch of the non-optimized
propellers increases at small ship speed (small J), and increasing
the torque of the engine and reducing the efficiency of propulsion.
The pitch of the optimized propeller decreases as the ship speed
decreases, reducing torque and increasing efficiency. The pitch
automatically changes with the inflow angle.
Since the displacement is very small, the measured result is eas-
ily influenced by other factors, such as fluid flow, vibration of blade
and vibration of shaft. Therefore, some differences exist between
experimental and numerical results. The experimental and numer-
ical results can be compared to verify the proposed method. Future
studies should use a larger propeller to yield better results. Large
propellers have more torque and a greater moment on the blade,
so the change in pitch is greater.
Acknowledgement
The authors would like to thank the National Science Council of
the Republic of China, Taiwan, for financially supporting this re-
search under Contract No. NSC 90-2611-E-002-022.
References
[1] Jones RM. Mechanics of composite materials. Washington, DC, USA: Scripta
Book Company; 1975.
[2] Weisshaar TA. Divergence of forward swept composite wings. J Aircraft
1980;17:442–8.
[3] Wilkinson K, Markowitz J, Lerner E, George D, Batill SM. FASTOP: a flutter and
strength optimization program for lifting surfer structure. J Aircraft
1997;14:581–7.
 C.-C. Lin et al. / Composite Structures 89 (2009) 206–215
[4] Weisshaar TA. Aeroelastic tailing of forward swept composite wings. J Aircraft
1981;18:669–76.
[5] Abdul MK. Effects of bend-twist coupling on composite propeller performance.
Mech Compos Mater Struct 2000;7:383–401.
[6] Lee Ya-Jung, Lin Ching-Chieh, Ji Jin-Chih, Chen Jong-Sheng. Optimization of a
composite rotor blade using a genetic algorithm with local search. J Reinf Plast
Compos 2005;24(16):1759–69.
[7] Lee Ya-Jung, Lin Ching-Chieh. Optimization design of composite propeller.
Mech Adv Mater Struct 2004;11(1):17–30.
[8] Lee Ya-Jung, Lin Ching-Chieh. Regression of the response surface of laminated
composite structures. Compos Struct 2003;62:91–105.
215
[9] Lin Ching-Chieh, Lee Ya-Jung. Stacking sequence optimization of composite
laminates using genetic algorithm with local improvement. Compos Struct
2004;63:339–45.
[10] Kerwin JE, Lee CS. Prediction of steady and unsteady marine propeller
performance by numerical lifting-surface theory. SNAME Trans
1978;86:218–53.
[11] Greeley DS, Kerwin JE. Numerical methods for propeller design and analysis in
steady flow. SNAME Trans 1982;90:415–53.
[12] Nelka JJ. Experimental evaluation of a series of skewed propellers with forward
rake: open-water performance, cavitation performance, field-point pressures,
and unsteady propeller loading. DTNSRDC Report 4113, July 1974.