Modeling and Testing of a Novel

Matthew Bryant
Laboratory for Intelligent Machine Systems,
Aeroelastic Flutter Energy
Sibley School of Mechanical and Aerospace
Engineering, Harvester
Cornell University,
Ithaca, NY 14853 This paper proposes a novel piezoelectric energy harvesting device driven by aeroelastic
e-mail: mjb377@[Link] flutter vibrations of a simple pin connected flap and beam. The system is subject to a
modal convergence flutter response above a critical wind speed and then oscillates in a
Ephrahim Garcia limit cycle at higher wind speeds. A linearized analytical model of the device is derived
Visiting Scientist, U.S. Department of Homeland to include the effects of the three-way coupling between the structural, unsteady aerody-
Security, namic, and electrical aspects of the system. A stability analysis of this model is presented
Washington, DC 20528; to determine the frequency and wind speed at the onset of the flutter instability, which
Laboratory for Intelligent Machine Systems, dictates the cut-in conditions for energy harvesting. In order to estimate the electrical
Sibley School of Mechanical and Aerospace output of the energy harvester, the amplitude and frequency of the flutter limit cycle are
Engineering, also investigated. The limit cycle behavior is simulated in the time domain with a semi-
Cornell University, empirical nonlinear model that accounts for the effects of the dynamic stall over the flap
Ithaca, NY 14853 at large deflections. Wind tunnel test results are presented to determine the empirical
e-mail: eg84@[Link] aerodynamic model coefficients and to characterize the power output and flutter fre-
quency of the energy harvester as functions of incident wind speed.
关DOI: 10.1115/1.4002788兴

1 Introduction While many other types of flow induced vibration are the result of
vortex shedding effects 关9兴, wing flutter results from the conver-
Several recent studies have explored the possibility of using
gence of two structural modes, giving it the moniker modal con-
piezoelectric materials to harvest energy from flowing fluids using
vergence flutter or coupled mode flutter 关10,11兴. The interaction of
a variety of mechanisms. While traditional piezoelectric energy multiple degrees of freedom also distinguishes modal convergence
harvesting research has predominantly focused on extracting en- flutter from galloping flutter, which is a single degree of freedom
ergy from preexisting vibrating host structures 关1兴, the fluid en- phenomenon that causes vibrations of bluff bodies 关12兴. In modal
ergy harvesting problem must address not only the transduction of convergence flutter, aerodynamic forces drive the natural frequen-
vibration to electrical energy but also the generation of vibrations cies of the two modes, generally a torsion mode and a bending
from the fluid flow as well. Thus the fluid flow energy harvesting mode, to converge near the critical flutter wind speed. Above the
problem introduces a three-way coupling of structural, electrical, flutter wind speed, energy from the fluid flow is transferred to the
and fluid dynamics, all of which must be considered in system structure, causing one of the system poles to become unstable and
modeling and design. A number of groups have investigated pi- leading to growing flutter oscillations 关10,11兴. These oscillations
ezoelectric “eels” or “flags” that are induced to undergo oscillat- grow in amplitude until the deflections become sufficient for sys-
ing bending stresses due to the effects of vortices shed from an tem nonlinearities to create limit cycle behavior or for structural
upstream bluff body. The bluff body periodically sheds vortices failure to occur. It has been noted in the literature that this type of
from either side, producing oscillating surface pressures acting on flutter exhibits less sensitivity to structural damping forces than
the flexible downstream piezoelectric element and causing it to other mechanisms of flow induced vibrations 关8,10兴. This charac-
flap or wave in the flow. This concept has been proposed for use teristic makes modal convergence flutter an ideal candidate to be
in air 关2兴 as well as in water 关3–5兴. A traditional rotary windmill used in energy harvesting; the piezoelectric energy extraction and
design was adapted to explore piezoelectric energy harvesting by charging dynamics of any attached energy storage elements will
using a cam system to induce cyclic bending of a set of radially have a relatively small effect on the flutter vibrations. We propose
arranged piezoelectric bimorphs 关6兴. The continuous rotation of a novel piezoelectric energy harvesting mechanism driven by
the windmill shaft is converted to alternating motion of a series of modal convergence flutter vibrations of a simple cantilevered pi-
stoppers that deflect the tips of the piezoelectric bimorphs. Piezo- ezoelectric beam and a pin connected flap.
electric damping and energy harvesting of vibrations of cantile-
vered pipes carrying flowing fluid have also been the subject of
recent work 关7兴. 2 Experimental Setup
Flow induced vibrations are typically regarded as undesirable The aeroelastic energy harvesting apparatus, as shown in Fig. 1,
and destructive phenomena that occur in a broad range of struc- consists of a rigid flap connected by a ball bearing revolute joint
tures. Such effects occur in systems ranging from heat exchanger to the tip of a flexible 301 stainless steel beam. A pair of MIDE
tube rows, to smokestacks, to aircraft components 关8兴. Aeroelastic Quickpack QP10N piezoelectric patches is laminated to the beam
flutter of aircraft wings and empennage structures is perhaps one at the root end to provide electromechanical transduction, as illus-
of the most dramatic and familiar of these flow induced vibration trated by Fig. 2. The beam root is cantilevered to a streamlined
effects. This type of flutter is distinct from other aeroelastic vibra- sting for mounting in the wind tunnel. This design creates an
tions because of the mechanism responsible for the instability. aeroelastic system with two coupled degrees of freedom, bending
deflection of the beam, and rotation of the flap about the bearing
joint. This system is thus subject to a modal convergence flutter
Contributed by the Technical Committee on Vibration and Sound of ASME for
publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received
response, which provides the driving mechanism for the energy
February 17, 2010; final manuscript received June 23, 2010; published online harvester.
January 5, 2011. Assoc. Editor: Dane Quinn. The physical, mechanical, and electromechanical properties of

Journal of Vibration and Acoustics Copyright © 2011 by ASME FEBRUARY 2011, Vol. 133 / 011012-1

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 Fig. 1 Photograph of the aeroelastic power harvester design

Fig. 3 Section representation of the aeroelastic energy har-

the device and materials used in the modeling of the system are vester describing the airfoil physical parameters
summarized in Table 1 below. Table 2 gives the mass, inertia, and
dimensional parameters for the flap, which was constructed with a
NACA0012 airfoil profile. The NACA0012 profile was selected
because published aerodynamic properties, such as those deter- Wind tunnel testing is conducted in a low speed, open circuit
mined by Sheldahl and Klimas 关13兴, are widely available for this wind tunnel with a 25⫻ 25⫻ 100 cm3 test section with a maxi-
airfoil. mum wind speed of about 9.1 m/s. The freestream air speed dur-
ing wind tunnel testing is measured with an Omega HHF42 probe
hot wire anemometer and the output of the piezoelectric patches
are plotted using an Agilent DSO6014A digital oscilloscope.

3 System Modeling
The aeroelastic energy harvester can be represented by adapting
the wing section model often used in aeroelasticity literature 关11兴.
This method is used to perform flutter analyses on aircraft wing
and empennage structures by examining a typical airfoil section of
Fig. 2 Schematic view of the bender and Quickpack piezoelec-
the structure. The stiffness of the structure is represented by hy-
tric patches with dimensions defined
pothetical compression-extension and torsion springs placed at the
Table 1 Physical, mechanical, and electromechanical param- elastic axis. In order to adapt this representation to the present
eters of the aeroelastic energy harvester study, the linear spring must be replaced with an appropriate beam
model of the piezoelectric bender, and the torsion spring is elimi-
Parameter Symbol Value Units nated because the revolute joint is assumed to have no stiffness.
The typical section for the aeroelastic energy harvester flap is a
Physical uniform thin airfoil, as shown in Fig. 3. A uniform lift distribution
Bender length ᐉ 25.4 cm across the span of the flap is assumed
Bender width w 2.54 cm
Beam substructure thickness ts 0.381 mm
PZT layer thickness tP 0.254 mm
PZT patch length ᐉP 4.60 cm 4 Flutter Boundary Analysis
PZT patch width wP 2.06 cm The flutter boundary, or set of flow conditions at which flutter
Epoxy matrix layer thickness tM 0.0254 mm emerges, is a critical design parameter to the aeroelastic energy
Epoxy matrix length ᐉM 4.84 cm
harvester. For a given air density, the wind speed at the flutter
Epoxy matrix width wM 2.5 cm
boundary will determine the minimum wind speed required for
Beam substructure density ␳S 7850 kg/ m3
PZT density ␳P 7700 kg/ m3
the energy harvester to operate. This value, often referred to as the
Epoxy matrix density ␳m 2150 kg/ m3 “cut-in wind speed” in wind power parlance, can be determined
analytically by combining aircraft flutter analysis models with
Mechanical elastodynamic, piezoelectric coupling models to form the
Beam substructure stiffness cS 212 GPa aeroelastic energy harvester system. In particular, we have fused
PZT stiffness, open circuit cE 67 GPa the linear aeroelastic wing section model of Hodges and Pierce
Epoxy matrix stiffness cM 2.5 GPa 关11兴, the Euler–Bernoulli piezoelectric bender model of Sodano et
al. 关14兴, and the unsteady aerodynamic model of Peters et al. 关15兴
Electromechanical to create a coupled aeroelastic-electromechanical system model.
Piezoelectric constant d31 ⫺190 pm/V
Constant strain permittivity ␧S 15.93 nF/m 4.1 Linear System Model. Hamilton’s principle was em-
ployed to combine both the aeroelastic and electromechanical sys-
tem models. Following the energy formulation, the total system
Table 2 Physical parameters of the NACA0012 profile flap kinetic and potential energies, including the contributions from the
flap assembly, beam, and piezoelectric patches must be defined.
Parameter Symbol Value Units The system kinetic and potential energies, respectively, can be
expressed by

冋
Flap and support mass mT 9.66 g
Flap
Flap
Flap
mass
mass moment of inertia
span
mF
IP
s
6.50
17.2
13.6
g
g cm2
cm
T=
1
2
mTḣ2 + 2mFbx␪ḣ␪˙ + I p␪˙ 2 + 冕 ␳Su̇គ Tu̇គ dVS + 冕 ␳M u̇គ Tu̇គ dVM

册
VS VM

冕
Flap semichord b 2.97 cm
Flap static unbalance x␪ 0.237
+ ␳ Pu̇គ Tu̇គ dV P 共1兲
VP

011012-2 / Vol. 133, FEBRUARY 2011 Transactions of the ASME

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 V=
1
2 冋冕 Sគ TcSSគ dVS + 冕 Sគ Tc M Sគ dV M + 冕 Sគ TcESគ dV P
bender and dissipation in the pin joint. After grouping terms, the
system model is given by

册
VS VM VP

冕 冕 冕
T¨
共M S + M M + M P + mT␾
គ 共ᐉ兲 ␾
T
គ 共ᐉ兲兲r̈គ + mFbx␪␾
គ 共ᐉ兲 ␪ + Chṙគ + 共KS
Sគ Te PTEគ dV P − Eគ Te PSគ dV P − Eគ T␧SEគ dV P 共2兲

冕
− ᐉ
VP VP VP
+ K M + K P兲rគ − ⌰C p−1q = − ␾
គ 共ᐉ兲 L +
T
f␾ T
គ dx 共10兲
where h and ␪ are the flap heave and pitch deflection coordinates, 0
respectively, as depicted in Fig. 3, mF is the mass of the flap, mT
is the total mass attached to the beam tip, and I p is the mass
គ 共ᐉ兲r̈គ + I P␪ + C␪␪ = M 1/4 + b共 2 + a兲L
moment of inertia of the flap about the pin joint. The terms ␳S, ␳m, mFbx␪␾ ¨ ˙ 1
共11兲
and ␳ P are the beam substrate, epoxy matrix, and piezo densities,
respectively, and the variables VS, Vm, and V P are the beam sub-
strate, epoxy matrix, and piezo volumes, respectively. The bender Rq̇ − C P−1⌰rគ + C P−1q = 0 共12兲
displacement in the transverse direction is u, and S and E repre-
where the matrices M S, M M , M P, KS, K M , and K P are the resulting
sent the beam strain and the electric field, respectively. The term
groups of integrals, C P and ⌰ are the piezocapacitance and elec-
cS is the substrate elastic modulus, cE is the piezoelastic modulus
tromechanical coupling matrices, respectively, and the terms Ch
at short circuit, e P is the piezocoupling coefficient, and ␧S is the
piezodielectric or permittivity constant at constant strain 关14兴. Fi- and C␪ are the viscous damping terms for the heave and pitch
nally, x␪ is the flap static unbalance parameter, which can be re- degrees of freedom, L共t兲 is the lift force acting on the flap, and
lated to the parameters shown in Fig. 3 by M 1/4共t兲 is the pitching moment acting at the flap 1/4 chord point.

x␪ = e − a 共3兲 4.2 Linear Aerodynamic Model. The aerodynamic lift force

and pitching moment acting on the structure are modeled using
The beam deflection u共x , t兲 can be expressed in terms of assumed the unsteady flow model of Peters et al. 关15兴. While traditional,
beam mode shapes and a temporal coordinate using the Rayleigh– steady flow aerodynamic expressions would provide a simpler
Ritz modal summation method as model, unsteady effects are generally important in aeroelastic vi-
N brations. The pitching and heaving motions of the flap have sig-
u共x,t兲 = 兺 ␾ 共x兲r 共t兲 = ␾គ 共x兲rគ共t兲
i=1
i i 共4兲 nificant effects on the air flow, prompting vortices to be shed from
the trailing edge, which in turn affect the flow incident on the flap.
In addition, because of the oscillatory motions of the structure, the
where ␾i共x兲 is the beam mode shape, ri共t兲 is the temporal coordi- direction of the wind vector relative to the structure is not fixed.
nate, and N is the number of mode shapes to be considered. The Therefore, a more complex, unsteady aerodynamic model that ac-
flap heave displacement, h共t兲, can also be expressed in this coor- counts for both circulatory and noncirculatory terms in the flow
dinate framework as must be adopted in the analysis of dynamic aeroelastic structures
h共t兲 = u共ᐉ,t兲 = ␾ 关11兴.
គ 共ᐉ兲rគ 共t兲 共5兲
The unsteady aerodynamics models available in the literature
where ᐉ is the length of the bender. The generalized forces acting can be broadly divided into two categories, those that assume
on the system in the heave and pitch directions can be expressed simple harmonic motion of the structure and those that do not.
in terms of the aerodynamic forces as The prior models including that of Theodorsen 关16兴 offer a sim-
pler approach but are only applicable at the flutter stability bound-
Qh = − ␾
គ 共ᐉ兲 L共t兲
T
共6兲 ary condition, not in transient vibrations. The latter category adds
additional complexity but allows analysis of the system behavior
Q␪ = M 1/4共t兲 + b共 2 + a兲L共t兲
1
共7兲 and eigenvalues below the critical flutter speed and in transients.
where L共t兲 is the total lift force and M 1/4共t兲 is the total pitching Among the latter, we implement the finite state, induced flow
moment about the 1/4 chord of the airfoil. In addition to the aero- theory of Peters et al. 关15兴 because it offers a time domain, state
dynamic forces acting on the flap, the beam itself also experiences space representation that allows for convenient simulation of the
aerodynamic loads. The elastic bender is subject to distributed system.
aerodynamic loading as it deflects normal to the flow, effectively This model approximates the effects of the unsteady airloads
changing the bending stiffness and damping as a function of inci- for inviscid, incompressible flows by using induced flow terms to
dent wind speed. The virtual work done by the distributed beam account for the effects of shed vortices on the flow near the airfoil.
aerodynamic forces can be written as The wind velocity near the foil is the free stream velocity plus an
additional local induced flow component 关15兴. The resulting ex-

冕
ᐉ
pressions for the lift and moment acting on the flap can be sum-
␦Wbeam = − ␦rគ 共t兲 f共x,t兲␾
គ 共x兲 dx
T
共8兲 marized as follows:
0

where f共x , t兲 is a function describing the beam aerodynamic loads. 关

L共t兲 = ␲␳⬁sb2共ḧ共t兲 + U␪˙ 共t兲 − ba␪¨ 共t兲兲 + 2␲␳⬁sUb ḣ共t兲 + U␪共t兲
Therefore, the generalized force due to the beam aerodynamic
loading is given by +b 共 1
2 兲
− a ␪˙ 共t兲 − ␭0共t兲 兴 共13兲

冕
ᐉ
f共x,t兲␾
គ 共x兲 dx 共9兲
冋 冉 冊 册
T
Qh,beam = 1 1 a
0 M 1/4共t兲 = − ␲␳⬁sb3 ḧ共t兲 + U␪˙ 共t兲 + b − ␪¨ 共t兲 共14兲
2 8 2
Lagrange’s equations can now be applied to the above relations
with bender deflection rគ 共t兲, flap pitch deflection ␪共t兲, and charge where ␳⬁ is the air density, s is the span of the flap, and ␭0共t兲 is
q共t兲 as the generalized coordinates. The outputs of the piezoelec- the induced flow velocity. The parameters U, h共t兲, and ␪共t兲 are
tric patches are assumed to be wired in parallel and connected defined in Fig. 3 and are the freestream wind velocity, the heave
across a resistive load, R, to facilitate convenient measurement of deflection of the flap, and the rotation angle of the flap, respec-
the output current and power. Viscous damping terms are also tively. The induced flow is then approximated by a series of in-
added to allow for approximation of the structural damping in the duced flow states ␭n共t兲 given by

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011012-3

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 M time, leading to a steady state operating condition for a given set
兺
1
␭0共t兲 ⬇ bn␭n共t兲 共15兲 of incident flow conditions. The characteristics of this limit cycle
2 n=1 behavior and its variation with wind speed are of primary impor-
tance to the operation of the flutter power harvester as a generator.
where M is the number of induced flow states to be included in
The amplitude and frequency of the limit cycle at a given wind
the analysis and bn are a series of least-squares coefficients. The
speed determine the strain and strain rate in the piezoelectric ele-
model states that a system of M differential equations govern ␭n
ments and therefore dictate the electrical output of the system.
as

冋 冉 冊 册
In order to analyze the flutter response and steady state opera-
U 1 tion of the power harvester above the aeroelastic stability bound-
A␭គ̇ 共t兲 + ␭គ 共t兲 = cគ ḧ共t兲 + U␪˙ 共t兲 + b − a ␪¨ 共t兲 共16兲 ary, a more complex model must be adopted. While the coupled
b 2
aeroelastic and electromechanical governing equations derived
where the matrix A and the vector cគ are composed of constants above are adequate for analyzing the cut-in wind speed of the
determined by the number of induced flow states, M, to be in- system, they assume small angles and attached flow to maintain
cluded in the analysis. The matrix A is given by linearity and permit eigenvalue stability analysis. As such, they
1 predict unbounded exponential growth of the oscillation ampli-
A = D + dគ bគ T + cគ dគ T + 2 cគ bគ T 共17兲
tude above the aeroelastic stability boundary. Accurately modeling
where the matrix D and vectors bគ , cគ , and dគ depend on the number the limit cycle behavior of the system requires that the model
of induced flow states. The values of their elements are defined by must account for large deflection amplitudes in the mechanical
the following equations: aspects of the model as well as flow separation effects due to large

冦 冧
flap angles of attack in the aerodynamics. In order to limit the
1 complexity of the analysis, it is assumed that linear Euler–
for i=j+1
2i Bernoulli beam theory remains applicable and beam hyperexten-
Dij = −1 共18兲 sion effects are not significant over the range of wind speeds to be
for i=j−1 considered. This assumption will be verified experimentally in
2i Sec. 6.2 of the paper.
0 for i⫽j⫾1
5.1 Nonlinear System Model. The nonlinear mechanical
model can again be derived using Lagrange equations. The full

冦 冧
共M + i − 1兲! 1 kinetic energy of the flap can be expressed by generalizing the
共− 1兲i−1 for i⫽M simplified expression derived using the small angle approximation
bi = 共M − i − 1兲! 共i!兲2 共19兲
by Hodges and Pierce 关11兴 to include large flap deflection angles.
共− 1兲i−1 for i=M The flap kinetic energy is given by

冦 冧
TF = 2 mFvគ C · vគ C + 2 IC␪˙ 2
1 1
1
for i=1 共24兲
di = 2 共20兲 where the velocity of the flap center of mass point, C, can be
0 for i⫽1 expressed as

2 vគ C = − î2ḣ + ␪˙ 关共1 + a兲b − 共1 + e兲b兴共î1 sin ␪ − î2 cos ␪兲 共25兲

ci = 共21兲
i Substituting and applying trigonometric identities, we obtain
4.3 Linear Beam Aerodynamic Forces. In addition to the TF = 2 mF关ḣ2 − 2共a − e兲b␪˙ ḣ cos ␪ + 共a − e兲2b2␪˙ 2兴 + 2 IC␪˙ 2
1 1
forces acting on the flap, one must also consider the aerodynamic
forces acting on the cantilevered bender as well. When small 共26兲
angles and deflections are assumed to maintain linearity, the aero- The equation can be simplified by introducing the static unbalance
dynamic force acting on a differential element of the beam is parameter, x␪, as given by Eq. 共3兲. Making the substitution yields
given by the lift force acting on the beam strip element as
TF = 2 mF共ḣ2 + 2bx␪␪˙ ḣ cos ␪ + b2x␪2␪˙ 2兲 + 2 IC␪˙ 2
1 1
1
f共x,t兲 = f L,beam = 2 ␳⬁CL,beamwsU2 共22兲 共27兲

where CL,beam is the effective lift coefficient of the beam element. The flap moment of inertia can be expressed about the hinge
Assuming that the flow over the beam is approximately steady, the point, P, using the parallel axis theorem as
beam element lift coefficient can be approximated by a flat plate I P = I C + m Fb 2x ␪2 共28兲
using

冋 册
Substituting Eq. 共28兲 into Eq. 共27兲 simplifies the flap kinetic en-
d ␾គ 共x兲ṙគ 共t兲 ergy to
CL,beam = 2␲ 共␾
គ 共x兲rគ 共t兲兲 + 共23兲
dx U
TF = 2 mF共ḣ2 + 2bx␪␪˙ ḣ cos ␪兲 + 2 I P␪˙ 2
1 1
where the bracketed terms approximate the effective angle of at- 共29兲
tack between the beam element and the incident wind vector in- The beam and tip mass kinetic energy terms remain unchanged as

冋冕 冕 冕 册
cluding the effects of beam slope and the local velocity of the
beam element. In order to apply the aerodynamic model defined 1
by Eqs. 共13兲–共23兲 to the electromechanical system model given by TB = ␳Su̇គ Tu̇គ dVS + ␳M u̇គ Tu̇គ dVM + ␳ Pu̇គ Tu̇គ dV P
2
Eqs. 共10兲–共12兲, the flap position, h共t兲, must be related to the beam VS VM VP

deflection coordinate u共x , t兲. This is accomplished by making the 共30兲

substitution given by Eq. 共5兲. and

5 Analysis of Limit Cycle Behavior 1

Ttip = mtipḣ2 共31兲
2
Above the flutter boundary for the device, experiments show
that stable, nonlinear limit cycle oscillations emerge. The limit respectively. Thus the total kinetic energy of the system can be
cycle holds the flutter frequency and amplitude constant over expressed by the sum of Eqs. 共29兲–共31兲, or

011012-4 / Vol. 133, FEBRUARY 2011 Transactions of the ASME

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 T=
1
2 冋
mtipḣ2 + mF共ḣ2 + 2bx␪␪˙ ḣ cos ␪兲 + I P␪˙ 2 + 冕 ␳Su̇គ Tu̇គ dVS
5.2 Nonlinear Aerodynamic Model. The aerodynamic forces
acting on the flap must be calculated using a model that realisti-
cally accounts for the nonlinear behavior of the aerodynamics at

册
VS

冕 冕
high angles of attack. In addition, as in the linear regime, the flap
+ ␳M u̇គ Tu̇គ dVM + ␳ Pu̇គ Tu̇គ dV P 共32兲 motion causes significant unsteady effects on the flow. One such
VM VP
model that has been extensively applied to aeroelastic flutter
analysis in the literature the ONERA dynamic stall model devel-
The potential energy terms are determined solely by the beam and oped by Tran and Petot 关17兴 and Dat and Tran 关18兴 and later
therefore remain unchanged as refined by Peters 关19兴. This semi-empirical model describes the

V=
1
2 冋冕 Sគ TcSSគ dVS + 冕 Sគ Tc M Sគ dV M + 冕 Sគ TcESគ dV P
aerodynamic forces acting on a fluttering structure subjected to
dynamic stall effects. The dynamic stall phenomenon results from
rapid changes in angle of attack that causes separation and reat-

册
VS VM VP

冕 冕 冕
tachment of the flow to be delayed, effectively altering the aero-
dynamic coefficients 关20兴. These delay effects are modeled with a
− Sគ Te PTEគ dV P − Eគ Te PSគ dV P − Eគ T␧SEគ dV P single lag term incorporated into the linear regime of the aerody-
VP VP VP namic force curve and two lag terms applied to the poststall flow
共33兲 regime. This model represents the aerodynamic force coefficients
as
As in the linearized case, the beam deflection coordinate u共x , t兲
can be expressed as a Rayleigh–Ritz modal summation by substi- Cz = Cz1 + Cz2 共41兲
tuting Eq. 共4兲 into the above relations. Equation 共5兲 is then applied
to express the flap heave coordinate, h共t兲, in terms of the beam ⴱ ⴱⴱ ⴱ

coordinates. Cz1 = sz1␣ + sz2 ␪ + sz3␪ + Cz␥ 共42兲

When large rotations of the flap are considered, the virtual work
ⴱ ⴱ ⴱ ⴱⴱ
due to the flap aerodynamic forces becomes
Cz␥ + ␤1Cz␥ = ␤1a0z共␣ + ␪兲 + ␤2a0z共␣ + ␪ 兲 共43兲
␦Wflap = L共t兲关− ␦h + b共 21 + a兲␦␪ cos ␪兴 + D共t兲关b共 21 + a兲␦␪ sin ␪兴
ⴱⴱ ⴱ ⴱ
+ M 1/4共t兲␦␪ 共34兲 C z2 + r1Cz2 + r2Cz2 = − r2⌬Cz兩␣ − r3⌬Cz兩␣ 共44兲
where D共t兲 is the total aerodynamic drag acting on the flap. In the where the nondimensional time derivative is defined as
above equation forces acting axially along the beam are assumed
to contribute little work and are neglected. After expressing the ⳵共 兲
共ⴱ兲 = 共45兲
flap heave deflection in the beam coordinates, the generalized ⳵␶
aerodynamic forces acting on the flap heave and pitch degrees of
with
freedom can then be identified as
Qh,flap = − ␾ Ut
គ 共ᐉ兲 L共t兲 共35兲
T
␶= 共46兲
b
Q␪ = b共 2 + a兲关L共t兲cos共␪兲 + D共t兲sin共␪兲兴 + M 1/4共t兲
1
共36兲 The coefficients sz1, sz2, sz3, r1, r2, and r3 are empirically deter-
mined constants and are unique to each aerodynamic force coef-
As in the linear case, the generalized force due to the beam aero-
ficient. The effective angle of attack, ␣, includes contributions due
dynamic loading is given by
to pitching and heaving of the flap and is given by

冕 冉冊
ᐉ
Qh,beam = f共x,t兲␾
គ 共x兲 dx
T
共37兲 ḣ ḣ
0
␣ = ␪ + tan−1 ⬇␪+ 共47兲
U U
With the kinetic energy, potential energy, and generalized forces
The nonlinear deviation function, ⌬Cz, expresses the difference
defined, the governing equations of the nonlinear system can now
between the linear static aerodynamic force curve and the nonlin-
be determined. Lagrange’s equation is applied to the above rela-
ear static force curve, with a decrease in force coefficient defined
tions with beam deflection rគ 共t兲, flap pitch deflection ␪共t兲, and
as positive. The full static force curve can therefore be expressed
charge q共t兲 as the generalized coordinates. As in the linear case, by
the electrodes of the piezoelectric patches are connected in paral-
lel across a resistive load, and the damping effects are approxi- Czs共␣兲 = a0z␣ − ⌬Cz共␣兲 共48兲
mated by including proportional damping terms. The resulting
nonlinear system model is given by The static aerodynamic curves used in the current model are fitted
and smoothed curves based on experimental symmetric airfoil

M S + M M + M P + mT␾ ¨ aerodynamic data published by Sheldahl and Klimas 关13兴. The
គ 共ᐉ兲 ␾គ 共ᐉ兲r̈គ + mFbx␪␾
គ 共ᐉ兲 cos共␪兲␪
T T
coefficient curves have also been modified so that there are no
− mFbx␪␾ ˙2 jump discontinuities for computational considerations. The static
គ 共ᐉ兲 sin共␪兲␪ + Chṙគ + 共KS + K M + K P兲rគ − ⌰C p q
T −1
aerodynamic force curves are defined by

冕
ᐉ

冦 冧
=−␾
គ 共ᐉ兲 L +
T
f␾ T
គ dx 共38兲 2␲␣ ␣ ⱕ 7° for
0 − 20.7␣ + 3.252 for 7 ° ⬍ ␣ ⱕ 9°
CL = 共49兲
3.067␣ − 0.495 for 9 ° ⬍ ␣ ⱕ 25°
mFbx␪␾ ˙ ¨ ˙
គ 共ᐉ兲cos共␪兲r̈គ − mFbx␪␾
គ 共ᐉ兲sin共␪兲ṙគ ␪ + I P␪ + C␪␪ 1.1 sin共2␣兲 for ␣ ⬎ 25°

冉 冊 1

冦 冧
= M 1/4 + b + a 关L cos共␪兲 + D sin共␪兲兴 共39兲 0 for ␣ ⱕ 7°
2
C M = − 0.367␣ + 0.045 for 7 ° ⬍ ␣ ⱕ 25° 共50兲
Rq̇ − C P−1⌰Trគ + C P−1q = 0 共40兲 − 0.155␣ − 0.066 for ␣ ⬎ 25°

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011012-5

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 Table 3 Parameters of the nonlinear aerodynamic model †21‡
cantilevered
Parameter Lift Moment
support
a0z 6.28 0
sz1 3.142 ⫺0.786
lift direction
⫺0.589
sz2
sz3
1.571
0 ⫺0.786
iˆ2
␤1 0.15
␤2 0.55 â2
r10 0.700
r12 0.150 iˆ1 drag direction
r20 0.246
r22 0.005 s
r30 ⫺0.024 beam element
r32 0.116 v â1
u(x,t)

CD = 0.3963兩␣4兩 − 2.2584兩␣3兩 + 3.3547兩␣2兩 − 0.0917兩␣兩 + 0.0071 Fig. 4 Cross section schematic showing coordinate directions
for a differential beam strip element
共51兲
where ␣, the angle of attack, is measured in radians. The coeffi-
1
cients r1, r2, and r3 in Eq. 共44兲 are given by the simplified expres- f L,beam = 2 ␳⬁CL,beamwsv2 共60兲
sions of Chen 关21兴 as where CL,beam is the appropriate lift coefficient for the beam sec-
r1 = r10 + r12共⌬CZ兲2 共52兲 tion. The instantaneous wind speed incident on the beam must
account for both the freestream air flow and the motion of the
r2 = 关r20 + r22共⌬CZ兲2兴2 共53兲 beam element normal to the flow. Therefore, the magnitude of the
effective wind speed as a function of time and position along the
r3 = r30 + r32共⌬CZ兲2 共54兲 beam can be expressed by
where the coefficients r10, r12, r20, r22, r30, and r32 are the empiri- v = 冑U2 + 共␾共x兲ṙ共t兲兲2 共61兲
cal constants. These constants were fitted by Chen to match ex-
perimental data for an oscillating NACA0012 airfoil at high The total aerodynamic force vector acting the on a differential
angles of attack published by McAlister et al. 关22兴. The coeffi- beam element in the reference frame oriented relative to the inci-
cients sz1, sz2, sz3, ␤1, ␤2, and a0z govern the linear portion of the dent wind vector is then given by
aerodynamic model. These coefficients were derived from un- f = − f D,beamâ1 + f L,beamâ2 共62兲
steady, incompressible flow theory using a single lag state ap-
proximation of the Theodorsen function 关21兴. Table 3 summarizes with the unit vectors as defined in Fig. 4. This can rotated into the

再冎冋
ground frame by applying the rotation matrix relationship

册再 冎
the parameters used in the present model.
With the necessary aerodynamic coefficients now defined, the î1 cos共␣s − ␪s兲 sin共␣s − ␪s兲 â1
aerodynamic lift, drag, and pitching moment acting on the flap can = 共63兲
now be calculated using the standard expressions as î2 − sin共␣s − ␪s兲 cos共␣s − ␪s兲 â2
L = ␳⬁U2bsCL 共55兲 where ␣s is the angle of attack of the differential beam strip and ␪s
is the angle of the strip relative to horizontal. In order to simplify
D = ␳⬁U2bsCD 共56兲 the analysis, forces acting along the beam axis were neglected and
only the transverse components were considered. This simplifica-
M 1/4 = 2␳⬁U2b2sC M 共57兲 tion is equivalent to considering the forces acting only in the î2
where the flap is assumed to have a uniform, rectangular platform. direction. The simplified beam aerodynamic forces can then be
expressed in beam deflection coordinate direction as
5.3 Nonlinear Beam Aerodynamic Forces. In addition to the
aerodynamic forces acting on the flap, the lift and drag forces f = − f D,beam sin共␣s − ␪s兲 − f L,beam cos共␣s − ␪s兲 共64兲
acting on the beam itself cannot be neglected when large deflec- where the u-coordinate direction is defined positive down, as
tions of the beam normal to the flow are considered. Studies shown in Fig. 4. In order to avoid the additional modeling com-
关23,24兴 have shown that there are two components of the drag plexity of separated flow incident on the flap downstream due to
force that must be considered when a flexible beam vibrates in air. vortex shedding from the beam, the flow over the beam is as-
Specifically, these are a force that is proportional to the instanta- sumed to remain steady and attached.
neous incident wind speed and a force proportional to the square
of the incident wind speed. The total drag force acting on a dif-
6 Comparison of Model and Experimental Results
ferential beam strip can then be expressed as
f D,beam = C1v2 + C2v 共58兲 6.1 Flutter Boundary Analytic and Experimental Results.
The linear system model derived above is applied to the aeroelas-
where v is the instantaneous wind speed incident on the beam. tic energy harvester apparatus to predict the wind speed and flap-
The coefficient C1 has been shown to be 关23兴 ping frequency corresponding to the onset of flutter. The first two
1 mode shapes of a cantilevered beam have been found to be ad-
C1 = 2 ␳⬁CD,beamws 共59兲
equate to model the piezoelectric bender in the frequency range of
where CD,beam is the drag coefficient and ws is the width of the interest and are used as the basis functions for the structural
surface normal to the air flow. The form of the coefficient C2 is model, while four aerodynamic states are included in the unsteady
not known analytically and must be determined empirically. The aerodynamic model. The forms of these aerodynamic states are
lift force acting on the differential beam strip is given by defined by the relations given by Eqs. 共16兲–共21兲. The system ei-

011012-6 / Vol. 133, FEBRUARY 2011 Transactions of the ASME

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 0.5 Table 5 Variation in flutter boundary and first structural bend-

Modal Damping, Re/ 1

ing natural frequency with number of cantilevered beam mode
shapes included in the analysis. All simulations performed with
Stable Flutter M = 4 aerodynamic states.

0 First bending nat. Cut-in wind Cut-in

frequency speed frequency
共Hz兲 共m/s兲 共Hz兲

Experiment 3.8 1.9 3.7

N = 1 mode 4.93 2.83 4.46
-0.5 N = 2 modes 3.83 1.86 3.69
0 1 2 3 4 5
Flow Velocity, m/s N = 3 modes 3.81 1.85 3.68
N = 4 modes 3.81 1.85 3.68
1.2
Modal Freqency, Im/ 1

0.8 structural, and electrical damping terms in the system are over-
Stable Flutter come by destabilizing aerodynamic forces and flutter emerges.
0.6 The imaginary parts of the normalized eigenvalues show a notice-
able convergence trend in the neighborhood of the flutter bound-
0.4 ary, suggesting that the frequencies of two degrees of freedom are
0.2 coalescing as expected, but never fully converge. This behavior is
consistent with other published examples of this unsteady aerody-
0 namic model including those of Hodges and Pierce 关11兴. Turning
0 1 2 3 4 5
to the numerical results of Table 1, the flutter boundary analysis
Flow Velocity, m/s
shows very good agreement with the experimental measurement
Fig. 5 Stability analysis plot as a function of wind speed for
of the wind speed and oscillation frequency at the onset of flutter.
the aeroelastic energy harvester simulated with N = 2 cantilever In fact, considering that the resolution of the hot wire anemometer
beam modes and M = 4 unsteady aerodynamic states. The flut- used for the wind speed measurement is 0.1 m/s, the error in the
ter boundary „dashed vertical lines… represents the cut-in wind predicted cut-in wind speed is effectively negligible.
speed for energy harvesting. The use of the first two cantilevered beam modes in the struc-
tural model was validated by comparing the predicted cut-in wind
speed and flutter frequency for various numbers of modes. Table 5
genvalues are determined as a function of the incident wind speed, lists the results of this analysis. Not surprisingly, the quality of the
as shown in Fig. 5. The flutter boundary is identified as the point model flutter boundary prediction correlates strongly with the pre-
where the real part of the one of the eigenvalues becomes positive, diction of the natural frequency of the first bending mode of the
representing the transition from stable, decaying oscillations to structure. While a single mode approximation significantly over-
growing oscillation. The flutter frequency can then be identified predicts the first bending natural frequency and therefore overpre-
from the imaginary part of the eigenvalue at this wind speed. For dicts the wind speed and flutter frequency at the flutter boundary
both the experimental and model results, the resistive load was set as well, increasing the number of modes to N = 2 matches the
to the experimental optimal resistive load, which is given approxi- experimental result well. Increasing the number of modes further
mately for a weakly coupled piezoelectric system by Guyomar et shows little improvement in the predicted results, indicating that
al. 关25兴 as two structural modes are adequate to model the piezoelectric
bender.
1 As a further validation, the analysis was performed using addi-
Ropt = 共65兲
C P␻ tional aerodynamic states to ensure that convergence of the finite
state, induced flow model had been achieved, as shown in Fig. 6.
where ␻ is the angular frequency of vibration. Table 4 compares The wind speed and frequency of the flutter boundary display
the results of the model with measured wind tunnel data. little sensitivity to the number of aerodynamic states used in the
Figure 5 plots the normalized real and imaginary components model beyond M = 4 states, indicating that four induced flow
of the eigenvalues corresponding to the first bending mode of the states are sufficient to model the unsteady flow over the flap at the
host structure and the wing rotation mode. These eigenvalues cap- onset of flutter.
ture much of the qualitative behavior of the flutter boundary. At
the onset of flutter, the real parts cross the horizontal axis from 6.2 Limit Cycle Experimental and Predicted Results. The
negative to positive, representing the transition from a stable nonlinear system model derived above was simulated in the time
damped system in which any disturbance will produce decaying domain using MATLAB’s ode23t.m numerical integration code to
oscillations, to an unstable system with growing oscillations. predict the structural and electrical responses at wind speeds
Qualitatively, this represents the point at which aerodynamic, above the flutter boundary and fit the model to experimental re-
sults. The unknown empirical beam aerodynamic drag coefficient
from Eq. 共58兲, C2, is empirically determined for each wind speed
Table 4 Comparison of measured and model predicted wind using MATLAB’s fminsearch.m nonlinear optimization tool to fit
speed and flutter frequency at the flutter boundary. Model re- the system response to the corresponding wind tunnel data. Figure
sults generated using N = 2 beam modes and M = 4 aerodynamic 7 shows the variation in the optimized C2 coefficient with the
states. incident wind speed. Figure 8 depicts a typical limit cycle oscil-
lation response simulated at U = 2.6 m / s after the optimization
Experimental Mode Model error
routine has been used to determine the empirical drag coefficient.
Parameter result l prediction 共%兲
As expected, after an initial transient, the simulation predicts the
Cut-in wind speed 1.9 m/s 1.86 m/s 2.1 system settling into a sinusoidal limit cycle and oscillating in a
Cut-in frequency 3.7 Hz 3.69 Hz 0.3 steady state motion.
The transition from a stable, damped system to a flutter limit

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011012-7

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 Fig. 8 Simulated time domain limit cycle oscillation response
for bender tip deflection, flap rotation, and voltage through re-
sistive load at wind speed U = 2.6 m / s and optimized load re-
sistance R = 280 kΩ

speed is the hallmark of a subcritical Hopf bifurcation.

The beam bending and flap rotation amplitudes and frequency
Fig. 6 Variation in predicted wind speed and flutter frequency
of the limit cycle motion vary with the incident wind speed. Fig-
at the flutter boundary with number of unsteady aerodynamic
states used in the model. All simulations performed with N = 2 ure 10 shows the limit cycle trajectory plots for several different
cantilevered beam modes. wind speeds, each calculated beginning from the same initial con-
dition and with the same simulation time. The limit cycle trajec-
tory plots show spiraling elliptical orbit geometry when the flap
rotation angle is plotted against the bender tip deflection. The
cycle as the incident wind speed is increased can be described in thickness of the ellipse is related to the settling time of the system,
nonlinear systems terminology as a Hopf bifurcation. These phe- while orientation of this ellipse reflects the phase relationship be-
nomena describe the loss of stability of the system equilibrium tween the motions of the two degrees of freedom. At lower wind
point and subsequent emergence of an oscillatory limit cycle as a
control parameter is varied and eventually exceeds a critical value,
triggering the bifurcation 关26兴. Experimental observations and the
model simulations both indicate that the nonlinear dynamics of
the aeroelastic energy harvester are governed by a subcritical
Hopf bifurcation. This type of Hopf bifurcation leads to an initial
condition dependent bistability between fluttering and stable be-
havior at low wind speeds. Figure 9 illustrates that at wind speeds
near the flutter boundary, a small initial condition, as shown on
the top, leads to oscillations that decay back to the equilibrium,
while a larger initial condition, as shown on the bottom, leads to
sustained limit cycle oscillations. This coexistence of a stable
equilibrium point and a surrounding limit cycle at the same wind

Fig. 9 Simulated bender tip deflection responses for initial

Fig. 7 Observed variation in empirical beam drag coefficient, conditions of 0.5 cm „top… and 1.0 cm „bottom… initial deflection
C2, with incident wind speed at U = 2.35 m / s incident wind speed

011012-8 / Vol. 133, FEBRUARY 2011 Transactions of the ASME

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 Fig. 10 Trajectory plots showing transient and limit cycle behaviors for several incident
wind speeds simulated over 15 s. From left to right in the top row, the simulated wind
speeds are U = 2.6, 3.6, 4.8 m / s, and from left to right in the bottom row U = 6.0, 6.9, and
7.9 m/s, respectively.

speeds, the plots show a significant phase difference between the Thus this device could provide a useful means of powering sens-
flap rotation and bender deflection, as shown by the incline of the ing electronics placed in areas where hardwired power is not
major ellipse axis. As the wind speed is increased, however, the available and periodically replacing batteries is costly or
major axis of the elliptical orbit approaches the horizontal, imply- impractical.
ing that the two degrees of freedom are deflecting with a 90 deg Figure 12 shows that the power curve exhibits very close agree-
phase difference. ment between the model and the experiment throughout the wind
In order to verify the assumption that the bender deflection speed range with the notable exception that the model fails to
remains in the linear range and that the aerodynamic nonlineari-
ties dominate the system, a static deflection test was performed on
the bender. The results, shown in Fig. 11, show that the piezoelec-
tric bender displays an approximately linear force-displacement
relationship up to static tip deflections of over 7 cm. The simu-
lated limit cycles shown in Fig. 10 never exceed 4 cm tip deflec-
tion amplitude, so the limit cycle tip deflections are safely within
the linear deflection range of the bender and the linear piezoelec-
tric beam model can be reasonably applied to this system.
Figures 12 and 13 compare the modeled system with the ex-
perimental power and frequency data as functions of the incident
wind speed. The power output of the aeroelastic energy harvester
is shown to vary with the incident wind speed between approxi-
mately 0.85 mW and 2.2 mW over the wind speed range tested.
While it is important to note that the experimental setup described
here was built as a proof of concept only and has not been opti-
mized in any way, this power output would be sufficient for pow-
ering low power electronic devices such as wireless sensor nodes.
Fig. 12 Variation in average power through the optimized re-
sistive load with incident wind speed for experimental result
and model prediction

Fig. 11 Static force-deflection experiment results for the pi- Fig. 13 Flutter frequency variation with incident wind speed
ezoelectric bender for experimental result and model prediction

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011012-9

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 Table 6 Effects of piezoelectric energy harvesting on pre- Table 7 Effects of piezoelectric energy harvesting on simu-
dicted wind speed and flutter frequency at the flutter boundary lated steady state flutter limit cycle oscillations at 2.6 m/s

Cut-in wind speed Cut-in frequency Bender tip Flap rotation Flutter
Resistive load 共m/s兲 共Hz兲 deflection amplitude amplitude frequency
Resistive load 共cm兲 共deg兲 共Hz兲
1 ⍀ 1.60 3.74
277 k⍀ 共optimal load兲 1.86 3.69 1 ⍀ 4.11 39.6 3.76
100 M⍀ 1.62 3.80 280 k⍀ 3.24 33.2 3.75
共optimal load兲
100 M⍀ 3.93 38.4 3.83

predict the first data point. This occurs because the nonlinear aero-
dynamic model slightly overpredicts the cut-in wind speed. The
13% and the limit cycle tip deflection amplitude by about 21%
model result captures the shape of the power curve, showing the
when the optimal resistive load and near open circuit conditions
first, lower peak at 2.6 m/s that corresponds to the first bending
are compared. The magnitudes of these shifts are large enough to
resonance of the piezoelectric bender, as well as the wider and
confirm that it is important to account for the full electromechani-
higher peak at around 7.9 m/s. The location of this peak corre-
cal coupling when modeling the system, but not so large that they
sponds to the 90 deg phase difference limit cycle plot at 7.9 m/s,
would pose substantial operational problems in applying the sys-
as shown in Fig. 10, suggesting that this condition represents a
tem to charge more complex circuits with inherent electrical dy-
point of maximum energy transfer from the air flow to the struc-
namics.
ture and therefore maximum available power for energy harvest-
ing. While the frequency curves show less quantitative agreement,
8 Conclusions
the model still captures the general trend of increasing flutter fre-
quency with wind speed and shows values of the appropriate order The problem of designing a piezoelectric energy harvester
of magnitude. The curves differ most significantly in their con- driven by aeroelastic flutter vibrations had been formulated,
cavities; the experimental result shows a concave down trend at solved, and experimentally investigated. The flutter energy har-
lower wind speeds and then transitions to a slightly concave up vester necessarily has a minimum cut-in wind speed below, which
appearance at higher wind speeds, while the model shows concave it cannot operate. This parameter would be an important design
up inflection at across most of the range with the last point dip- point for any potential application and would be constrained by
ping slightly. size, mass, and material limitations in any practical application.
This cut-in wind speed was shown to be predicted by an eigenan-
7 Investigation of Electromechanical Coupling Effects alysis of the coupled linear system model. Simultaneously, the
power generation characteristics of the device at wind speeds
In addition to estimating the flutter boundary characteristics, above cut-in are dictated by nonlinear effects in the system includ-
nonlinear limit cycle behavior, and power output of the aeroelastic ing nonlinearities in the mechanism and the aerodynamic loading.
flutter energy harvester, the model can also be used to investigate A semi-empirical model of the nonlinear electromechanical and
the effects of the electromechanical coupling and energy harvest- aerodynamic system accurately predicted the system power output
ing on the system. The piezoelectric energy harvesting extracts over a range of wind speeds and approximated the variation in
energy from the vibrating structure, essentially creating an elec- oscillation frequency with wind speed. The analysis showed that
tromechanical damping effect on the system. In order to examine the peak power output of the energy harvester occurs when the
the influence of this effect on the flutter boundary, the linear sys- beam and flap are deflecting with a 90 deg phase difference. This
tem model was analyzed with various resistive load values includ- peak power condition adds a second design consideration for the
ing very small and very large resistances in addition to the opti- aeroelastic energy harvester. The design of the energy harvester
mized resistive load value. Table 6 summarizes the results of this would depend largely on the characteristics of the expected flow
analysis. conditions for the intended application. In addition to the cut-in
As expected, the optimized resistive load extracts the most en- wind speed and coupled system resonant peak, the minor peak in
ergy from the system and therefore shows the highest cut-in wind power associated with the first bending resonance of the piezo-
speed and lowest cut-in frequency, both indicative of the largest electric beam can also be leveraged to enhance the power genera-
electromechanical damping effect. In comparison, the 1 ⍀ load tion capacity of the energy harvester depending on the nature of
shows the lowest cut-in wind speed because this load approaches the flow environment. In a varying flow application, the energy
a short circuit condition and creates very little electromechanical harvesting system would need to be tailored to have a cut-in wind
energy extraction and damping effect. When the load resistance is speed low enough for the device to operate throughout the ex-
increased to a very high value, as illustrated by the 100 M⍀ load, pected flow speed range and the bender and system resonance
a significant impedance mismatch occurs between the piezoelec- peaks placed to maximize power output throughout the range of
tric elements and the resistive load, which diminishes the ability expected flow conditions. When the energy harvester is applied to
of energy to flow out of system and into the load resistor, leading a consistent flow condition, however, designing the system to col-
to a lower cut-in wind speed. locate and superpose the bender and system resonances maxi-
The piezoelectric energy extraction also affects the limit cycle mizes power output.
behavior of the system. Table 7 summarizes the results of simu-
lations performed for three different resistive loads at 2.6 m/s Acknowledgment
incident wind speed. In keeping with the results of the flutter
boundary analysis, the optimized resistive load extracts the maxi- This work is supported the National Science Foundation under
mum energy from the motion, resulting in the smallest oscillation Grant No. CMMI-0938870. This grant is monitored by Dr. Shih-
Chi Liu, Program Director, Sensors and Sensing Systems. Addi-
amplitudes for the tip deflection and flap rotation. The 1 ⍀ and
tional funding was provided by the Cornell University Center for
100 M⍀ loads show larger limit cycle amplitudes, again indicat-
a Sustainable Future.
ing that less energy is being extracted from the motion and less
electromechanical damping is present.
The results shown in Tables 6 and 7 show that the electrome- Nomenclature
chanical coupling has a modest effect on the flutter boundary and a ⫽ dimensionless flap joint axis location
limit cycle oscillations, shifting the cut-in wind speed by about b ⫽ flap semichord

011012-10 / Vol. 133, FEBRUARY 2011 Transactions of the ASME

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]
 C ⫽ viscous damping coefficient h ⫽ heave
CD, CL, C M ⫽ drag, lift, and moment coefficients, ␪ ⫽ rotational
respectively
Cz ⫽ general aerodynamic coefficient
C P ⫽ piezoelectric capacitance
cE ⫽ constant electric field PZT stiffness References
c M , cS ⫽ epoxy matrix and beam substructure stiffness, 关1兴 Anton, S. R., and Sodano, H. A., 2007, “A Review of Power Harvesting Using
respectively Piezoelectric Materials 共2003–2006兲,” Smart Mater. Struct., 16, pp. R1–R21.
关2兴 Robbins, W. P., Morris, D., Marusic, I., and Novak, T. O., 2006, “Wind-
D共t兲 ⫽ drag force Generated Electricity Using Flexible Piezoelectric Materials,” Proceedings of
d31 ⫽ piezoelectric constant IMECE 2006, Vol. 71, pp. 581–590.
e ⫽ dimensionless flap center of mass location 关3兴 Allen, J. J., and Smits, A. J., 2001, “Energy Harvesting Eel,” J. Fluids Struct.,
15共3–4兲, pp. 629–640.
f共x , t兲 ⫽ beam aerodynamic load function 关4兴 Taylor, G. W., Burns, J. R., Kammann, S. M., Powers W. B., and Welsh T. R.,
h共t兲 ⫽ flap heave deflection coordinate 2001, “The Energy Harvesting Eel: A Small Subsurface Ocean/River Power
IC, I P ⫽ flap mass moment of inertia about center of Generator,” IEEE J. Ocean. Eng., 26共4兲, pp. 539–547.
关5兴 Pobering, S., and Schwesinger, N., 2004, “A Novel Hydropower Harvesting
mass and joint axis, respectively Device,” Proceedings of the 2004 International Conference on MEMS, NANO
K ⫽ stiffness matrix and Smart Systems, Banff, AL, pp. 480–485.
L共t兲 ⫽ lift force 关6兴 Priya, S., Chen, C. T., Fye, D., and Zahnd, J., 2005, “Piezoelectric Windmill:
ᐉ, ᐉ M , ᐉ P ⫽ length of bender, epoxy layer, and piezo, A Novel Solution to Remote Sensing,” Jpn. J. Appl. Phys., Part 2, 44共3兲, pp.
L104–L107.
respectively 关7兴 Elvin, N. G., and Elvin, A. E., 2009, “The Flutter Response of a Piezoelectri-
mF, mT ⫽ flap mass and total bender tip mass, cally Damped Cantilever Pipe,” J. Intell. Mater. Syst. Struct., 20, pp. 2017–
respectively 2026.
关8兴 Blevins, R. D., 2001, Flow Induced Vibration, Van Nostrand Reinhold Com-
M ⫽ number of aerodynamic modes pany, New York.
M 1/4共t兲 ⫽ quarter chord pitching moment 关9兴 Khalak, A., and Williamson, C. H. K., 1999, “Motions, Forces, and Mode
M M , M P, M S ⫽ mass matrix for epoxy, piezo, and substrate, Transitions in Vortex-Induced Vibrations at Low Mass-Damping,” J. Fluids
respectively Struct., 13共7–8兲, pp. 813–851.
关10兴 Dowell, E. H., Curtiss, Jr., H. C., Scanlan, R. H., and Sisto, F., 1980, A Modern
N ⫽ number of structural modes Course in Aeroelasticity, Sijthoff & Noordhoff, Alphen aan den Rijn, The
Q ⫽ generalized force Netherlands, Chap. 3.
q ⫽ charge 关11兴 Hodges, D. H., and Pierce, G. A., 2002, Introduction to Structural Dynamics
and Aeroelasticity, Cambridge University Press, Cambridge, UK, Chap. 4.
R ⫽ resistive load 关12兴 van Oudheusden, B. W., 1996, “Rotational One-Degree-of-Freedom Galloping
rគ 共t兲 ⫽ bender deflection temporal coordinate in the Presence of Viscous and Frictional Damping,” J. Fluids Struct., 10, pp.
S ⫽ strain 673–689.
关13兴 Sheldahl, R. E., and Klimas, P. C., 1981, “Aerodynamic Characteristics of
s ⫽ flap span Seven Symmetrical Airfoil Sections Through 180-Degree Angle of Attack for
T ⫽ kinetic energy Use in Aerodynamic Analysis of Vertical Axis Wind Turbines,” Sandia Na-
t ⫽ time tional Laboratories Report No. SAND80-2114.
t M , t P, tS ⫽ thickness of epoxy, piezo, and substrate, 关14兴 Sodano, H. A., Park, G., and Inman, D. J., 2004, “Estimation of Electric
Charge Output for Piezoelectric Energy Harvesting,” J. Strain, 40共2兲, pp. 49–
respectively 58.
U ⫽ freestream wind speed 关15兴 Peters, D. A., Karunamoorthy, S., and Cao, W. M., 1995, “Finite State Induced
u共x , t兲 ⫽ beam deflection coordinate Flow Models; Part I; Two Dimensional Thin Airfoil,” J. Aircr., 32共2兲, pp.
313–322.
V ⫽ potential energy 关16兴 Theodorsen, T., 1934, “General Theory of Aerodynamic Instability and the
v ⫽ instantaneous incident wind speed Mechanism of Flutter,” NACA Report No. 496.
V M , V P, VS ⫽ volume of epoxy, piezo, and substrate layers, 关17兴 Tran, C. T., and Petot, D., 1981, “Semi-Empirical Model for the Dynamic Stall
respectively of Airfoils in View of Application to the Calculated Responses of a Helicopter
in Forward Flight,” Vertica, 5共1兲, pp. 35–53.
w ⫽ width 关18兴 Dat, D., and Tran, C. T., 1983, “Investigation of the Stall Flutter of an Airfoil
x ⫽ position along bender length With a Semi-Empirical Model of 2-D Flow,” Vertica, 7共2兲, pp. 73–86.
␣ ⫽ effective angle of attack 关19兴 Peters, D. A., 1985, “Toward a Unified Lift Model for Use in Helicopter Rotor
␦W ⫽ virtual work Blade Stability Analyses,” J. Am. Helicopter Soc., 30共3兲, pp. 32–43.
关20兴 Wickenheiser, A. M., and Garcia, E., 2008, “Optimizing of Perching Maneu-
␧S ⫽ constant strain permittivity vers Through Vehicle Morphing,” J. Guid. Control Dyn., 31共4兲, pp. 815–823.
␾គ 共x兲 ⫽ mode shape 关21兴 Chen, W. C., 1993, “A Formulation of Nonlinear Limit Cycle Oscillation Prob-
lems in Aircraft Flutter,” MS thesis, Massachusetts Institute of Technology,
␭ ⫽ induced flow state Cambridge, MA.
␪ ⫽ angular position 关22兴 McAlister, K. W., Pucci, S. L., McCroskey, W. J., and Carr, L. W., 1982, “An
␳ M , ␳ P, ␳S ⫽ density of epoxy, piezo, and substrate materi- Experimental Study of Dynamic Stall on Advanced Airfoil Sections Volume 2:
als, respectively Pressure and Force Data,” NASA Report No. TM-84245.
关23兴 Baker, W. E., Woolam, W. E., and Young, D., 1967, “Air and Internal Damp-
␳⬁ ⫽ freestream air density ing of Thin Cantilever Beams,” Int. J. Mech. Sci., 9, pp. 743–766.
␶ ⫽ dimensionless time 关24兴 Juang, J.-N., and Horta, L. G., 1987, “Effects of Atmosphere on Slewing
␻ ⫽ angular frequency Control of a Flexible Structure,” J. Guid. Control Dyn., 10共4兲, pp. 387–392.
关25兴 Guyomar, D., Badel, A., Lefeuvre, E., and Richard, C., 2005, “Toward Energy
Subscripts Harvesting Using Active Materials and Conversion Improvement by Nonlinear
T ⫽ tip mass Processing,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, 52共4兲, pp. 584–
595.
B ⫽ bender 关26兴 Strogatz, S. H., 1994, Nonlinear Dynamics and Chaos, Perseus Books, Cam-
F ⫽ flap bridge, MA, Chap. 8.

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011012-11

Downloaded 17 Jan 2011 to [Link]. Redistribution subject to ASME license or copyright; see [Link]