## SUMMARY

Propulsion and maneuvering underwater by flapping foil motion, optimized
through years of evolution, is ubiquitous in nature, yet marine propulsors
inspired by examples of highly maneuverable marine life or aquatic birds are
not widely implemented in engineering. Performance data from flapping foils,
moving in a rolling and pitching motion, are presented at high Reynolds
numbers, Re=*Uc/*ν, or O(10^{4}), where *U* is the
relative inflow velocity, *c* is the chord length of the foil, and ν
is the kinematic viscosity of the fluid, from water tunnel experiments using a
foil actuator module designed after an aquatic penguin or turtle fin. The
average thrust coefficients and efficiency measurements are recorded over a
range of kinematic flapping amplitudes and frequencies. Results reveal a
maximum thrust coefficient of 2.09, and for low values of angle of attack the
thrust generally increases with Strouhal number, without much penalty to
efficiency. Strouhal number is defined as
*St*=2*h*_{0}*f/U*, where *f* is the
frequency of flapping, and 2*h*_{0} is the peak-to-peak
amplitude of flapping. The thrust and efficiency contour plots also present a
useful performance trend where, at low angles of attack, high thrust and
efficiency can be gained at sufficiently high Strouhal numbers. Understanding
the motion of aquatic penguins and turtle wings and emulating these motions
mechanically can yield insight into the hydrodynamics of how these animals
swim and also improve performance of biologically inspired propulsive
devices.

## Introduction

The design of biologically inspired propulsion mechanisms for underwater
vehicles continues to generate significant interest in the hydrodynamics of
fish swimming. Fish and animals, such as aquatic penguins and turtles, are
highly adapted to swimming in the ocean and serve as excellentmodels for
developing novel propulsive devices for underwater vehicles. With the goal of
designing compact, agile autonomous and unmanned underwater vehicles (AUVs and
UUVs), engineers have turned to fish and their aquatic counterparts for
inspiration (e.g. Bandyopadhyay,
2005; Anderson and Chabra,
2002; Triantafyllou et al.,
2000; Anderson et al.,
1998; Bandyopadhyay et al.,
1997*)*. Biologists have studied the kinematics and
morphology of swimming fish in great detail, revealing the superior agility of
these creatures (e.g. Fish and Lauder,
2006; Lauder,
2000; Drucker and Lauder,
1999; Videler,
1993; Fish and Hui,
1991). The undulatory nature of the fish body motion imparts flow
control on the surrounding fluid generating a unique propulsive signature in
the form of a reverse Kármán vortex street
(Triantafyllou and Triantafyllou,
1995). This simplified view of the wake structure does not account
for three-dimensional effects that result from the geometry of the fish body
and tail fin.

Flapping foils have been used as models for swimming fish fins to better understand fish swimming hydrodynamics. Heaving and pitching foils with moderate to high aspect ratios (Anderson, 1996; Triantafyllou et al., 1993), and more three-dimensional motions of foils, modeling pectoral fins with lower aspect ratios, have also been investigated as alternative methods of propulsion for underwater vehicle technology (Bandyopadhyay, 2005; Lang et al., 2006; Kato, 2000). Extensive studies of the two-dimensional wake patterns behind flapping foils and live and robotic swimming fish have been performed both experimentally and numerically. Thorough reviews of this research on flapping foils can be found in the literature (Triantafyllou et al., 2004; Triantafyllou el al., 2005).

Typical wake patterns generated by the two-dimensional flapping foil motion
consist of alternately rotating vortices organized to form a jet-like wake, or
reverse Kármán street, which results in thrust generation over a
range of Strouhal numbers and kinematic parameters. Strouhal number is defined
as *St*=2*h*_{0}*f*/*U*, where
2*h*_{0} is the peak-to-peak heave amplitude of the trailing
edge, *f* is the flapping frequency (in Hz), and *U* is the
incoming flow velocity relative to the foil. Optimal propulsive performance
has been shown to be highly dependent on Strouhal number. Typically, flapping
foils used for propulsion have high efficiencies in a range of Strouhal
numbers 0.2–0.4 (Triantafyllou and
Triantafyllou, 1995); higher thrust values can be achieved at
higher values of Strouhal number but with less efficiency.

The unsteady flapping motion of a foil generates vorticity at the trailing
edge and tips of the foil and also at the leading edge under certain
conditions. These patterns vary with the amplitude and frequency of the motion
as well as the shape of the kinematics employed (e.g.
Hover et al., 2004;
Anderson el al., 1998;
Koochesfahani, 1989;
Katz and Weihs, 1979).
Experiments with heaving and pitching symmetric foils, with a maximum
thickness that is 15% of the chord length (i.e. a NACA 0015 foil), at moderate
Reynolds numbers [5000–12 000
(Freymouth, 1990)] link the
formation of a leading edge vortex, and its successive combination with the
vorticity generated at the trailing edge, with high lift coefficients.
Reynolds number (*Re*) is defined as the ratio of the inertial fluid
forces to the viscous fluid forces:
*Re*=(ρ*U*^{2}*c*^{–1})/(μ*Uc*^{–2})=ρ*Uc*/μ,
where ρ is the density of the fluid, and μ the dynamic viscosity of the
fluid. *Re* is often written in terms of kinematic viscosity:ν
=μ/ρ; for water ν=10^{–6} m^{2}
s^{–1}. This interaction is also important to achieve high
efficiency for flapping foil swimming modes
(Anderson et al., 1998).

A closer look at performance of finite aspect ratio, three-dimensional flapping foils is warranted at higher Reynolds numbers, in order to design vehicles to swim with optimal kinematics and maximizing thrust while optimizing efficiency. In an effort to understand this trade-off, tests were performed using a three-dimensional, linearly tapered hydrofoil forced to move in roll and pitching motions, and the measured force and hydrodynamic efficiency data are presented, discussed and compared with data reported from previous tests (Flores, 2003; McLetchie, 2004; Polidoro, 2003; Read, 2000).

## Experimental methods

### Three-dimensional flapping foil apparatus and kinematics

The three-dimensional (3D) flapping foil motion emulates that of an aquatic
penguin wing or turtle fin. The 3D flapping foil actuator was constructed as a
dual-canister design: two watertight canisters, one each for pitch and roll,
house the motor and chain drive for each motion separately. The two canisters
are coupled together and to the foil (see
Fig. 1A). This apparatus is
mounted in the Massachusetts Institute of Technology recirculating water
tunnel, capable of flow speeds from 0.5 to 5.0 m s^{–1} with
turbulence levels under 3%. The tunnel's square test section is 0.5
m×0.5 m×1.25 m long. Constrained to move in roll ϕ(*t*)
and pitch θ(*t*) only, a diagram of the motion coordinate system
is given in Fig. 1B. This type
of flapping apparatus presents underwater vehicle designers and engineers a
simpler design problem, e.g. fewer sealing issues, than the heaving/pitching
foil. The thesis by Lim (Lim,
2005) presents further details of the housing design and
operation.

The kinematic motion of the foil is measured through potentiometers mounted
to each rotating shaft, which map the pitch and roll motion, and also through
the encoders mounted on each motor. The encoder data are used for redundancy
and kinematics validation. A torque sensor is coupled to the roll motor shaft
to calculate the power input from the motor end during actuation; the power
input can be used to determine the mechanical efficiency of the flapping foil.
The forces and torques acting directly on the foil are measured with a
six-axis, waterproof strain gauge sensor, which measures the three force
components (*F*_{x}, *F*_{y},
*F*_{z}) and the three moment components
(*M*_{x}, *M*_{y}, *M*_{z}).
Hydrodynamic efficiency is calculated using the torque measurements from the
six-axis sensor to calculate power input to the fluid by the foil, thus
bypassing the mechanical losses in the drive mechanism. Cross-coupling between
the six channels is eliminated using the factory supplied matrix, which has
been validated extensively (Lim,
2005).

The average thrust and lift forces are found by applying a force rotation matrix to the force traces and averaging the data over 10–15 cycles. Average force data presented in subsequent sections are given in terms of the non-dimensional force coefficient, based on the planform area of the foil, instead of the swept area as would be used for a propeller. The hydrodynamic efficiency is determined by comparing the power output, calculated from the product of average thrust and velocity, to the sum of the power input, applied to the pitch and roll axes.

The prescribed roll and pitch motions are simple sinusoidal harmonics with
the same circular frequency ω (rad s^{–1}). The roll
motion of the foil is given by:
(1)
where ϕ_{0} is the roll amplitude in radians; the pitch motion of
the foil is defined as:
(2)
where θ_{0} is the pitch amplitude and Ψ is the phase angle
between pitch and roll, both in radians. The static pitch biasθ
_{bias} is an optional parameter that introduces a non-zero
average pitch angle to the foil; for these tests it is set to zero. The phase
angle Ψ for the tests is set to π/2, as recommended
(Read et al., 2003).

The angle of attack profile varies along the span of the three-dimensional
flapping foil as it rolls and pitches. To simplify the kinematics, the motion
can be decomposed to 2D heaving and pitching, *versus* rolling and
pitching, at any span location on the foil. Although McLetchie's results
(McLetchie, 2004) show varying
centers of force, the 70% span location is selected to be consistent with
conventional propeller notations and for easy comparison with past flapping
foils experiments. This location is defined as:
(3)
where *r*_{0} is the distance from the center of roll axis to
the root of the foil, and *S* is the span of the foil, as shown in
Fig. 1A. The heave position is
defined as:
(4)
where *h*_{0} is the amplitude of the heave motion at
*r*_{0.7}; *h*_{0} is defined as:
(5)
The angle of attack at one span location can be found from the instantaneous
pitch position of the foil and the ratio of the heave to forward velocity.
Fig. 2 shows the vector diagram
of the velocity components.

The true angle of attack profile can be calculated mathematically. Since a
foil with a positive pitch produces a smaller angle of attack, one needs to
subtract the angle of attack due to pitch from that due to roll to find the
overall angle of attack profile:
(6)
where *U* is the forward speed of the actuator and
*ḣ* is the heave velocity. From
Eqn 4, we can express the heave
velocity as:
(7)
Substituting Eqn 7 into
Eqn 6, we get the expression for
angle of attack in 3D kinematics:
(8)
The maximum angle of attack, α_{max}, is calculated using
Eqn 8 at
*r*_{0.7}; α_{max} is given in degrees throughout
this paper.

Again, the Strouhal number can be used to describe the foil motion
kinematics based on the roll amplitude and *r*_{0.7}:
(9)
An estimate of the total width of the wake produced by the flapping foil is
2*h*_{0}. This is essentially the peak-to-peak amplitude of the
foil taken at the 0.7 chord for a 90° phase offset between heave and
pitch. The roll amplitude is non-dimensionalized by converting this to 2D
heave amplitude at 0.7 span and dividing by the chord length,
*h*_{0.7}/*c*. Finally, Reynolds number,
*Re=Uc*_{0.7} /ν, is calculated based on the chord at 70% of
the span, *c*_{0.7}, and kinematic viscosity of the fluid,ν
=10^{–6} m^{2} s^{–1}.

The foil used in this experiment is fabricated with a NACA 0012 cross
section. The NACA 0012 is a standard, symmetric foil profile with a maximum
thickness that is 12% of the chord length
(Abbott and von Doenhoff,
1959). The test foil has a span of 24.6 cm from root to tip, and
an average chord length, *c̄*, of 5.5 cm
from leading edge to trailing edge. The foil has a linearly tapering trailing
edge profile. Tests were performed over a range of kinematic parameters:
heave/chord ratio, *h*_{0.7}/*c*={1.0, 1.5, 2.0};
Strouhal number, *St*={0.2:0.1:0.6}; and maximum angle of attack,α
_{max}={15:5:45}°. The experiments were conducted at
Reynolds numbers 27 000 to 55 000. The lower speed was used to achieve higher
Strouhal numbers. The speed was monitored at all times by a Laser Doppler
Velocimetry system and maintained to within ±0.01 m
s^{–1}.

## Results and Discussion

### Three-dimensional flapping foil performance

The 3D flapping foil motion (roll/pitch) presents underwater vehicle
designers with a more straightforward design problem compared to the
heaving/pitching mechanism design and yields further insight into the
performance of swimming animals such as aquatic penguins and turtles. To
evaluate the performance of flapping foils sufficiently, accurate force and
efficiency measurements are paramount. Force data are presented here for the
3D flapping foil mechanism for induced 2D heave-to-chord ratios
(Eqn 4)
*h*_{0.7}/*c*=1.0, 1.5 and 2.0, over a range ofα
_{max} from 15° to 45°. Typical force traces obtained
by the six-axis force sensor are presented in
Fig. 3.

Using data acquired with the six-axis sensor, the average thrust of the
foil is calculated by:
(10)
is the *x*-force component
translated into the reference frame of the tunnel, averaged over one flapping
cycle. By axes convention, *x* is positive upstream. A non-dimensional
thrust coefficient can be defined as:
(11)
where ρ is the fluid density, *U* is the relative flow velocity,
*S* is the foil span and *c̄* is
the average chord length.

The general trend of the thrust results
(Fig. 4) compares well with
those for the 2D case (Read,
2000) and 3D case (McLetchie,
2004). The thrust coefficients decrease with decreasing Strouhal
numbers and α_{max} values, and increases with increasing values
of both Strouhal number and α_{max}. At heave- to- chord ratio
of 1.0 (*h*_{0.7}/*c*=1.0) a peak thrust coefficient of
1.6 is recorded at *St*=0.5 and α_{max}=30°.
Extrapolating up to *St*=0.6, it appears that a higher peak thrust
coefficient could be achieved. Such an increase is substantiated by McLetchie
(McLetchie, 2004).

For the case of *h*_{0.7}/*c*=1.5, peak thrust
coefficients were measured for *St*=0.6 and angle of attacks of 30°
and 35°. At *h*_{0.7}/*c*=2.0, the maximum thrust
coefficients now occur at the same Strouhal numbers (*St*=0.6) but have
lower magnitudes at α_{max}=30° and 35°
(Lim, 2005).
Fig. 4 shows that for a given
Strouhal number the thrust coefficient increases with the maximum angle of
attack, but beyond a critical α_{max} value, the thrust
coefficient is expected to decrease. This is similar to the data obtained for
heaving/pitching foils at *Re*=750 and 1000 (Read, 2006). Increases in
roll amplitude tend to have similar effects. The maximum thrust coefficient
for these runs was recorded at *h*_{0.7}/*c*=1.5. At the
higher roll amplitude, *h*_{0.7}/*c*=2.0, the peak
thrust coefficient is lower, thus it is possible that an optimal roll
amplitude can be found, between the heave-to-chord ratio of 1.5–2.0, for
which the thrust coefficient can be maximized.

The left and bottom `borders' of the parametric space represent the
boundaries for which the angle of attack profiles tend to corrupt, i.e. some
regions of the foil would encounter negative angles of attack, resulting in
drag instead of thrust production. Here, the results at *St*=0.2 show
very low thrust production for all α_{max} values. This low
thrust boundary represent the transition in the wake from drag- to
thrust-producing vortices (Flores,
2003). At high α_{max}, the transition to thrust
does not occur until a Strouhal number of about 0.4. At experimental points
for which low thrust values are measured (*St*⩽0.3;
15°⩽α_{max}⩽25°), there are large errors
associated with the results (>10% in certain cases). A more thorough
analysis of the error from these tests is given in Lim
(Lim, 2005). In addition, the
thrust coefficients presented above were evaluated based on planform area. A
more appropriate normalization, in keeping with that used in propeller
performance analysis, might have been to use the projected swept area. This
formulation would have produced numerically smaller values of
*C*_{T} (Techet el al., 2005).

The hydrodynamic efficiency of the foil is defined as the ratio of power
output *P*_{out} over power input *P*_{in} to
the fluid:
(12)
Power output is the product of the time-averaged thrust and flow velocity:
(13)
The power input *P*_{in} is the power input to the fluid,
calculated from torque measurements with the six-axis sensor mounted to the
foil shaft. While the torque sensor attached to the roll motor could measure
the direct power input from the electric motor, a significant amount of the
power is actually used to move the inertial mass of the pitch canister, and
the losses due to backlash in the motors can be quite large. The motor power
input is found to be in the order of ten greater than the power transmitted to
the flow. Thus a more useful approach is to use the power input measured by
the AMTI sensor to calculate hydrodynamic efficiency of the foil, such that
the hydrodynamic efficiency of the foil can be compared directly with other
foil designs and aquatic animals in future research.

In a reverse fashion to the thrust coefficient results, efficiency peaks at
the lower end of Strouhal and α_{max} values and then decreases
with increasing *St* and α_{max} values
(Fig. 5). This is as expected
since, in low thrust regions, minimal energy is lost as a result of kinetic
energy being imparted to the flow. The maximum efficiency recorded is about
0.7, centered at *St*=0.3 and α_{max}=20° for
*h*_{0.7}/*c*=2.0. Its location corresponds with that
reported by McLetchie (McLetchie,
2004), although the magnitudes of efficiency differ. Read also
shares the same peak location with measured efficiencies not exceeding 0.7,
for 2D flapping foils (Read,
2000).

Higher roll amplitudes result in greater energy expended in moving the foil
through the large oscillations. It is only at higher frequencies where greater
thrust is generated such that the efficiency appears to be improving. Figs
4 and
5 show that for lowα
_{max} values, thrust generally increases with Strouhal number
without significant penalty to efficiency. This is good news with respect to
designing mechanisms for underwater vehicles, as an optimal point can be
identified for relatively high thrust production without sacrificing
efficiency – it is also good news for aquatic swimmers, as they can
produce significant propulsive forces with minimal effort. For example, at
*h*_{0.7}/*c*=1.5, we see relatively high efficiencies
occurring at *St*=0.5, and α_{max} of 15° and
20°. This offers a good design point where efficiencies of more than 0.6
can be achieved with thrust coefficients ranging from 0.7 to 1.2.

The average range of thrust coefficient *C*_{T} is found to
be in the order of ±0.026, or ±0.044 N in terms of absolute
thrust; as such, the percentage error, which is taken as the fraction of
standard deviation over the mean, is naturally higher at the lower thrust runs
(Lim, 2005). In the regions
where the thrust is low (*St*⩽0.3;
15°⩽α_{max}⩽25°), the percentage errors
associated with the measurements still remain significant and thus the
efficiency measurements also require careful consideration and further
validation. Likewise a larger percentage error for efficiency η also
arises for these cases, since the error for η is a combination of errors
from power input and thrust measurements.

## Conclusions

In summary, the 3D flapping foil, similar to a pectoral fin, holds the most
promise for implementation in an underwater vehicle due to the reduced
complexity over the heaving/pitching mechanism. To further extend our
understanding of flapping foil performance at higher Reynolds numbers the 3D
rolling/pitching foil apparatus was studied in a water tunnel at Reynolds
numbers on the order of *Re*∼10^{4}. These tests, using
force sensors to measure thrust and efficiency, with the 3D flapping foils are
highly valuable from a design point of view.

Clearly indicated in the data presented here is that for a fixed Strouhal
number, there is a critical maximum angle of attack value beyond which the
thrust coefficient will start to decrease and performance is reduced. A peak
planform area thrust coefficient of 2.09 was measured at
*h*_{0.7}/*c*=1.5, *St*=0.6 andα
_{max}=30°. Increasing the heave-to-chord ratio from 1.5 to
2.0 does not appear to improve *C*_{T} values, but further
iterations are still warranted. Combined with data from McLetchie
(McLetchie, 2004), the data
presented herein for the 3D foil suggests a useful performance trend where,
for low α_{max}, high thrust and high efficiency can be gained
at sufficiently high Strouhal numbers (*St*=0.6 and possibly higher).
In particular, for higher roll amplitudes, these large oscillations produce
high thrust with relatively little power loss. Higher thrust and efficiencies
can be achieved through pre-shaping algorithms applied to the heaving/pitching
foil motions for 2D flapping (Hover et
al., 2004). Further investigation is warranted to determine
whether pre-shaping would improve the performance of the 3D foil, but data
suggests that this is a reasonable assumption.

Further comprehensive and systematic investigations of foil geometry on flapping foil performance are warranted; for example, it was suggested that an optimal aspect ratio is nearer to 4.0 for flapping wings (Polidoro, 2003). The effect of chordwise or spanwise flexibility on the performance of flapping foils is also still not well understood. Live fish employ active flexure of their fins, especially their pectoral fins, suggesting that they can optimize their fin shape for maximized performance (Fish and Lauder, 2006). Further experiments in tandem with numerical simulations are necessary to further augment our understanding of flapping foil performance as seen in nature.

**List of symbols and abbreviations**

- c
- foil chord
- c̄
- mean chord
*C*_{T}- thrust coefficient
- f
- flapping frequency (Hz)
*F*_{x},*F*_{y,}*F*_{z}- force components
*F̄*_{x}_{0}*x*-force component translated into reference frame of test section- ḣ
- heave velocity
(
*ḣ*=d*h*/d*t*) *h*_{0}- heave amplitude
*h*_{0.7}- heave amplitude at 0.7 span location
- h(t)
- heave position
*M*_{x},*M*_{y,}*M*_{z}- moment components
*r*_{0}- radius from base of foil
*r*_{0.7}- radius from base of foil shaft to 0.7 span
*P*_{in}- power input
*P*_{out}- power output
- Re
- Reynolds number
- S
- foil span
- St
- Strouhal number
- t
- time
- T
- thrust
- U
- free stream velocity
- α
- angle of attack
- αmax
- maximum angle of attack
- Ψ
- phase angle between pitch and roll
- ϕ(t)
- roll angle
- ϕ0
- roll amplitude
- η
- efficiency
- μ
- dynamic viscosity of the fluid (kg m
^{–1}s^{–1}) - ν
- kinematic viscosity of the fluid [ν=μ/ρ (kg
m
^{–1}s^{–1})] - ρ
- fluid density (kg m
^{–3}) - θ(t)
- pitch angle
- θ0
- pitch amplitude
- θbias
- pitch bias angle
- ω
- circular frequency (rad s
^{–1})

## ACKNOWLEDGEMENTS

The author would like to thank K. Lim for his assistance with experiments and Dr Franz Hover for his insightful discussions related to data processing and angle of attack effects. The development of the flapping foil mechanism was made possible through the support of Admiral Paul Sullivan, NAVSEA, and the MIT Sea Grant program (Grant NA16RG2255).

- © The Company of Biologists Limited 2008