Академический Документы
Профессиональный Документы
Культура Документы
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/281148999
CITATIONS READS
3 123
3 authors:
Abdessattar Abdelkefi
New Mexico State University
126 PUBLICATIONS 1,022 CITATIONS
SEE PROFILE
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Abdessattar Abdelkefi
letting you access and read them immediately. Retrieved on: 01 June 2016
Eur. Phys. J. Special Topics 224, 2929–2948 (2015)
© EDP Sciences, Springer-Verlag 2015 THE EUROPEAN
DOI: 10.1140/epjst/e2015-02599-y PHYSICAL JOURNAL
SPECIAL TOPICS
Regular Article
1 Introduction
During the past few decades, energy harvesters have been deemed as potentially sig-
nificant applications by converting ambient or aeroelastic vibrations into electrical
power in order to operate low-power consumption electronic devices or autonomous
gadgets. The mechanical energy from vibrating bodies can be used to power micro-
electro-mechanical systems (MEMS) or actuators [1, 2], health monitoring and wireless
sensors [3], cameras [4], and cell phones [5]. The application object of energy harvest-
ing is low-power consumption devices, therefore, energy harvesters can be used from
powering equipment in far-flung remote areas to giving energy to high tech equipment
in harsh environments. In addition, the replacement of batteries is quite inconvenient
(e.g. for pacemakers) and hence energy harvesters again become a viable option in
such cases. Ambient vibrations have been always employed as a source of energy har-
vesting for the past two decades [6–12]. Recently, utilizing flow-induced vibrations
for the generation of electrical power has received attention by several researchers
[13–25].
Flow-induced vibrations mainly include flutter in airfoil sections, vortex-induced
vibrations (VIVs) in circular cylinders, and galloping in prismatic structures, which
were employed for energy harvesting through different transduction mechanisms
[13–23]. An experimental investigation was carried out by Bryant and Garcia [13]
in order to study the possibility of harvesting energy from two different wing geome-
tries, namely, a flat plate and a NACA 0012 airfoil. De Marqui et al. [15] developed
a time-domain piezoaeroelastic model for cantilevered plates representing wing-like
structures. Comparing between continuous and segmented electrode configurations,
they demonstrated that using a segmented electrode configuration is beneficial in
terms of enhancing the level of the harvested power and improving the resistive shunt
damping effect. Abdelkefi et al. [26,27] successively investigated the concept of energy
harvesting from flutter of airfoil sections and vortex-induced vibrations of circular
cylinders. Using the concept of vortex-induced vibrations to operate low-power con-
sumption devices, Dai et al. [17,18] constructed a nonlinear distributed-parameter
model for harvesting energy from VIV of a piezoelectric cantilever beam with a cir-
cular cylinder attached to its end, which was validated by the experimental results
of Akaydin et al. [16]. Different galloping piezoelectric energy harvesters consisting of
cantilevers attached to prismatic structures of triangular and D-shape cross-sections
were reported by Sirohi and Mahadik [19,28]. Abdelkefi et al. [29] studied the possi-
bility of harvesting energy from galloping oscillations of a square cylinder attached to
a beam with considering low and high Reynolds numbers configurations. In [30], the
authors developed a nonlinear distributed-parameter model and validated their nu-
merical results with the experimental measurements of Sirohi and Mahadik [28]. The
effects of the tip cross-section geometry on the onset speed of galloping and perfor-
mance of the galloping-based piezoelectric energy harvesters were well-studied using
lumped- and distributed-parameter models by Abdelkefi and his coauthors [20, 31]
and Yang et al. [22]. It was shown that the square cross-section geometry is very
efficient for designing low-wind speed piezoelectric energy harvesters.
It should be noted that the above research studies about energy harvesting from
aeroelastic vibrations have only focused on utilizing the piezoelectric transduction
mechanism. Yet, the electromagnetic induction, as another mechanism for harvesting
energy from unused vibrations, has also been proposed and explored in recent years,
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2931
but only few researchers have considered this transduction mechanism to harvester
energy from aeroelastic oscillations. Jung and Lee [24] performed some experiments to
study the possibility of harvesting energy from wake galloping oscillations with using
an electromagnetic transduction mechanism. The electromagnetic induction was also
used by Zhu et al. [32] to investigate the possibility of harvesting energy from varied
lift force of an airfoil which was placed behind a bluff body. Further investigation on
airfoil-based oscillations for harvesting electrical power by using piezoelectric and/or
electromagnetic induction were carried out by De Marqui and his coauthors [33–35].
Their results were a step forward in understanding the potential of electromagnetic
energy extraction from airfoil-based harvesters. The concept of electromagnetic en-
ergy harvesting from galloping vibrations was first studied by Ali et al. [36] when
considering various cross-section geometries. In their research study, the transverse
galloping self-excited oscillations of the cantilever beam were converted to electri-
cal power through an electromagnetic generator which is composed of a permanent
magnet that is attached to the beam oscillating past a stationary coil. Ali et al. [36]
used a finite element analysis to predict the lift and drag coefficients of the galloping
force and they compared them with available experimental results of these coefficients
for similar flow conditions. In addition to that, they used a finite element model to
solve the electromechanical equations of motion of the harvester. We should note
that in this research study no details were given about the dynamics modeling of
the harvester and only one wind speed value was considered numerically and experi-
mentally. For this considered wind speed and compared to the equilateral triangular
and square cross-section geometries, the authors showed that the D-section is the
best cross-section geometry that gives the highest possible values of the harvested
power. Recently, Vicente-Ludlam et al. [37] and Dai et al. [38] proposed a coupled
electroaeroelastic lumped-parameter model for an electromagnetic energy harvester
using transverse galloping oscillations. Both studies used the quasi-steady approxi-
mation to model the galloping force. Considering a D-section shape of the bluff body,
Vicente-Ludlam et al. [37] indicated that there is an optimum electrical load resis-
tance for each reduced flow velocity in order to maximize the power efficiency. It
should be mentioned that no predictions were addressed in their investigation on the
influences of the electrical load resistance on the onset speed of galloping and asso-
ciated harvested power. Considering a square cross-section geometry, Dai et al. [38]
deeply investigated the effects of the electrical load resistance on the onset speed of
galloping and the levels of the harvested power. For low load resistance values, they
demonstrated that the load resistance has a strong influence on the coupled damping
of the harvester and hence the corresponding onset speed of instability. For each wind
speed value, it was also indicated that there is an optimum value of the electrical load
resistance at which the levels of the power output are maximized.
In the present research study, our focus will be on the modeling and performance
improvement of galloping-based energy harvesting when using both piezoelectric and
electromagnetic transduction mechanisms. To this end, a piezoelectric and an elec-
tromagnetic transducers are attached in the transverse degree of freedom of a pris-
matic cylinder in order to convert galloping vibrations to electrical power. A coupled
lumped-parameter model is developed which relates the motion of the oscillating
cylinder to the generated voltage from the piezoelectric layer(s) and the induced
current from the stationary coil. A quasi-steady approximation is used to model the
aerodynamic force created by the flowing wind over the prismatic structure. The mod-
eling representation is addressed in Sect. 2. We proceed with discussing the effects of
the electrical load resistances on the coupled damping, coupled frequency, and onset
speed of galloping, in Sect. 3. In Sect. 4, the normal form of the Hopf bifurcation is de-
rived in order to determine the effects of hybrid harvester’s parameters on the outputs
power near bifurcation and to identify the type of the ensuing instability. Moreover,
2932 The European Physical Journal Special Topics
(a) (b)
Fig. 1. Sketches of a galloping-based hybrid piezoelectric-electromagnetic energy harvester
(a) physical representation and (b) lumped representation.
the average harvested power generated by each transducer are compared with that
harvested by both transducers. Finally, summary and conclusions are presented in
Sect. 5.
velocity of the cylinder. The empirical values of various cross-section geometries for a1
and a3 are determined by Barrero-Gil et al. [40] which are based on the experimental
results of Parkinson and Smith [41] for a square cross-section geometry.
For the electromagnetic induction, following Faraday’s law, the electromotive force
Vemf is produced on the terminals of the coil can be expressed as [42]:
dii
Vemf − Lc − Rc + Rli ii = 0 (5)
dt
Vp
Cp V̇ p + + θẏ = 0 (6)
Rlp
where m is the total mass per unit length, ζ is the mechanical damping coefficient,
Lb denotes the length of the bluff body, ω0 represents the natural frequency of the pris-
matic structure, and θ denotes the piezoelectric coupling coefficient. The coil winding
across which the vibrating magnet produces change of flux and corresponding current,
is modeled by an electrical internal resistance Rc and an electrical inductance, Lc .
Ẋ1 = X2 (8)
2934 The European Physical Journal Special Topics
ρU Da1 1 ρDa3 3 CBl θ
Ẋ2 = − 2ζω0 − X2 − ω02 X1 + X2 − X3 + X4 (9)
2m 2 mU mLb mLb
CBl Rc + Rli
Ẋ3 = X2 − X3 (10)
Lc Lc
1 θ
Ẋ4 = − X4 − X2 . (11)
Rlp Cp Cp
Dropping the nonlinear term in the equations of motion (Eqs. (4)–(6)) of the hybrid
harvester, Eqs. ((8)–(11)) can be rewritten in the following matricial form:
⎡ ⎤
0 1 0 0
⎢ ⎥
⎢ −ω 2 − 2ζω − ρU Da1 CBl
− mL θ ⎥
⎢ n n 2m mLb ⎥
⎢
B=⎢
b
⎥· (13)
CBl R +Ri ⎥
⎢ 0 − cLc l 0 ⎥
⎣ Lc ⎦
0 − Cθp 0 − Rp1Cp
l
The parameters in matrix B(Rli , Rlp , U ) influence the linear part of the hybrid harvest-
ing system and hence the coupled frequency, damping, and onset speed of galloping.
The matrix B has a set of four eigenvalues λi , i = 1, 2, 3, 4. The first two eigenvalues
λ1 , λ2 exist in the form of complex conjugates. The third and fourth eigenvalues λ3
and λ4 are always real and negative. The physics of the system is governed by the first
pair of eigenvalues that are actually associated with the transverse displacement of
the prismatic cylinder. The real parts of λ1 and λ2 represent the coupled damping of
the system whereas the absolute value of the positive imaginary parts is a representa-
tive of its coupled frequency. Therefore the stability of the trivial solution is entirely
governed by the real part of the first two eigenvalues. The stability of the linearized
system changes its stability from stable to unstable as the real part of the first pair of
eigenvalues λ1 and λ2 switches its value from negative to positive. It should be noted
that a zero real part of this pair of complex conjugates is indicative of the onset
speed of galloping, Ug . In this research study, the values of the physical and empirical
parameters of the hybrid piezoelectric-electromagnetic energy harvesting system are
given in Table 1.
When the external load resistance Rli in the electromagnetic circuit is set to 10Ω,
the impact of Rlp on the coupled damping and frequency of the hybrid harvesting sys-
tem for different values of the wind speed is shown in Fig. 2. During this variation, for
a given wind speed, it is noted that the coupled damping is almost not affected by the
electrical load resistance in the piezoelectric circuit (PC) as its corresponding value
Rlp is lower than 4×103 Ω or higher than 3×107 Ω. While in the range of electrical load
resistances between 4 × 103 Ω and 3 × 107 Ω, the coupled damping varies significantly
and there is an optimal load resistance Rlp at which the value of coupled damping
is lowest. For all considered wind speed values, this optimal value of Rlp is equal to
2.6 × 105 Ω. Moreover, inspecting the plots in Fig. 2(a), it is clear that the coupled
damping is increased when the wind speed is increased over the whole range of the
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2935
Properties Values
Mass, m (kg) 0.44
Diameter, D (m) 0.03
Fluid density, ρ (kg/m3 ) 1.20
Damping ratio, ζ 0.0013
Natural frequency, ω0 (rad/s) 31.42
Electrical resistance, Rc (Ω) 16.8
Inductance, Lc (H) 0.0068
Electromagnetic coupling, CBL (N/A) 1.33
Piezoelectric coupling coefficient, θ (N/V ) 1.55 × 10−3
Capacitance of the piezoelectric layer(s), Cp (F ) 1.2 × 10−7
U=2.5m/s 32.2
U=2.5m/s
U=4m/s U=4m/s
0.4 U=7m/s U=7m/s
32
Coupled frequency
Coupled damping
0.2
31.8
0
31.6
−0.2
−0.4 −2 31.4 −2 0 2 4 6 8
0 2 4 6 8
10 10 10 10 10 10 10 10 10 10 10 10
Rpl (Ω) Rp Ω
l
(a) (b)
Fig. 2. Variations of (a) the coupled damping and (b) coupled frequency as function of Rlp
for different values of the wind speed when Rli is set equal to 10Ω.
load resistance Rlp . Concerning the variations of the coupled frequency as a function
of the PC electrical load resistance Rlp , a significant variation is obtained when Rlp is
increased from 2 × 104 Ω to 2 × 106 Ω for all considered wind speed values. Outside this
range of PC electrical load resistances, the coupled frequency of the hybrid harvester
does not change with Rlp . We should mention here that the wind speed does not effect
the coupled frequency of the piezoelectric-inductive energy harvester.
To determine the effects of the electrical load resistance Rli in the electromagnetic
circuit (EC) on the linear characteristics of the hybrid energy harvester, we plot in
Figs. 3(a) and (b) the variations of the coupled damping and frequency with Rli for
different values of the wind speed when considering Rlp is equal to 2.6×105 Ω. Increas-
ing the external load resistance Rli from the short circuit case (10−2 Ω) to its open
circuit case (105 Ω), we note from Figs. 3(a) and (b) that there is a significant increase
in the coupled damping while a small decrease in the coupled frequency when Rli is
between 1Ω and 1000Ω for all considered wind speed values. Outside this range of Rli ,
the external load resistance Rli in the electromagnetic circuit does not have any influ-
ence on the coupled damping and frequency of the hybrid energy harvester. Clearly,
as the wind speed is increased, the coupled damping is increased over the whole Rli
range while the coupled frequency is decreased. The increase of the coupled damping
when increasing the electrical load resistance in the electromagnetic circuit can be
explained by the presence of the effect of the coil internal resistance when Rli is set
equal to zero which gives an extra resistive shunt damping for the harvester. However,
2936 The European Physical Journal Special Topics
U=2.5m/s U=2.5m/s
−0 31.781 U=4m/s
U=4m/s
U=7m/s U=7m/s
−.1 31.78
Coupled frequency
Coupled damping
−.2 31.779
−.3 31.778
−.4 31.777
−0.5 −2 −1 0 1 2 3 4 5
31.776 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
Ril (Ω) Ri (Ω)
l
(a) (b)
Fig. 3. Variations of (a) the coupled damping and (b) coupled frequency with Rli for different
values of the wind speed when Rlp is set equal to 2.6 × 105 Ω.
for very high values of Rli (open-circuit configuration), the current in the electromag-
netic energy harvester is zero and hence the coupled damping in the harvester is the
same as the piezoelectric coupled damping (without electromagnetic resistive shunt
damping effect). Concerning the coupled frequency in the hybrid energy harvesting
system, it is little bit higher for small values of Rli because an additional frequency
from the electromagnetic equation (Eq. (5)). On the other hand, for high values of
Rli , the coupled frequency of the hybrid harvester is the same as its piezoelectric
counterpart (without any additional frequency from the electromagnetic circuit). It
can be concluded from the plotted curves in Figs. 2 and 3 that both Rli in EC and
Rlp in PC have a strong impact on the coupled damping and frequency of the hybrid
energy harvester and hence they can significantly affect the onset speed of instability
of the hybrid energy harvesting system.
The plotted curves in Figs. 4(a) and (b) show the variations of the onset speed
of galloping as function of Rli for different values of Rlp and Rlp for different values
of Rli , respectively. Inspecting Fig. 4(a), we note that the onset speed of gallop-
ing is decreased when Rli is increased for all considered values of Rlp . In particular,
this decrease in onset speed of instability is obvious when Rli is between 1Ω and
1000Ω. This result is expected because the absolute value of the coupled damping is
clearly decreased in the same range as shown in Fig. 3(a). In addition, it follows from
Fig. 4(a) that when Rlp is set equal to 2.6 × 105 Ω, higher onset speed of galloping are
obtained compared to all other considered values of the external load resistance in
the piezoelectric circuit Rlp . This can be explained due to the fact that the absolute
value of the coupled damping of the piezoelectric-inductive energy harvester is the
highest when Rlp =2.6 × 105 Ω, as shown in Fig. 2(a). As for the variations of the onset
speed of instability as a function of Rlp , it is noted that there is an optimal value of
Rlp at which this onset speed is maximum for all values of Rli which can be clearly
seen in Fig. 4(b). This optimal value is equal to 2.6 × 105 Ω which is also due to the
fact that the absolute coupled damping at this value is the highest. What’s more, as
Rli is increased, the onset speed of galloping is decreased which is also expected from
Fig. 3(a) where the coupled damping gradually increases with increasing Rli in EC.
In this performed comprehensive linear analysis, the effects of the external load re-
sistances in the piezoelectric and electromagnetic circuits on the coupled damping,
frequency, and onset speed of galloping are determined. It is clear that a strong
coupling between the two electrical circuits is observed which will affect the gen-
eral tendency of the hybrid energy harvester compared to its counterpart of separate
energy harvesters (piezoelectric or electromagnetic).
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2937
12
Ril=10 Ω
−2
20 Rp=10−2Ω
l
Ril=10 Ω
0
Rpl =105Ω
9 i 1
p 5
Rl =2.6×10 Ω Rl=10 Ω
15
6
Rpl =10 Ω Ri=102Ω
Ug (m/s)
U (m/s)
l
i 5
8
Rpl =10 Ω 6 R =10 Ω
10 l
g
5 3
0 −2 −1 0 1 2 3 4 5 0 −2 0 2 4 6 8
10 10 10 10 10 10 10 10 10 10 10 10 10 10
i p
R (Ω) Rl (Ω)
l
(a) (b)
Fig. 4. Variations of the onset speed of galloping as a function of (a) Rli in EC for different
values of Rlp and (b) Rlp in PC for different values of Rli .
Applying the center-manifold theorem [43] on Eq. (18) and because σU is small
near bifurcation and only aerodynamic nonlinearities of cubic order are present in the
system, Z3 and Z4 are considered zero to the third approximation. Keeping only the
resonance terms in Eq. (18), one obtains the following complex-valued normal form
of the Hopf bifurcation:
1
ȧ = βr a + αer a3 (21)
4
1
γ̇ = βi + αei a2 (22)
4
where β = σU K11 and the subscripts r and i represent, respectively, the real and
imaginary parts. Equation (20) has three equilibrium solutions which are:
−4βr
a = 0, a = ± · (23)
αer
It is noted that the zero equilibrium solution a = 0 is the trivial solution and the other
two solutions are the nontrivial ones. The origin is asymptotically stable for βr < 0
or βr = 0 and αe r < 0, unstable for βr > 0 or βr = 0 and αe r > 0. When it comes
to the nontrivial solutions, we have βr αe r < 0. The nontrivial solutions are stable
(supercritical Hopf bifurcation) for βr > 0 and αer < 0 and unstable (subcritical
Hopf bifurcation) for βr < 0 and αer > 0. Consequently, to design efficient and
reliable galloping-based aeroelastic energy harvesters, the real part of the effective
nonlinearity αer should be close to zero and negative to ensure maximum values of the
harvested power under supercritical Hopf bifurcation. Considering a square cylinder
cross-section geometry with the linear and nonlinear parameters of the galloping force
stated in Table 1, the real part of the effective nonlinearity for this harvester comes
out to be negative which reflects a supercritical Hopf bifurcation. The transverse
displacement y of the prismatic structure, the induced current in the electromagnetic
circuit ii , and the generated voltage in the piezoelectric circuit V p can be calculated
as follows:
y = a G211r + G211i (24)
ii = a G231r + G231i (25)
V p = a G241r + G241i · (26)
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2939
0.08 0.02
Analytical solution Analytical solution
Numerical solution Numerical solution
0.06 0.015
Pavg(W/m)
y (m)
0.04 0.01
i
0.02 0.005
0 0
1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4
U (m/s) U (m/s)
(a) (b)
−3
x 10
2.5
Analytical solution
Numerical solution
2
Pavg(W/m)
1.5
p
0.5
0
1 1.5 2 2.5 3 3.5 4
U (m/s)
(c)
Fig. 5. Nonlinear normal form (blue dashed lines) and numerical (red solid lines) bifurcation
diagrams of (a) the transverse displacement amplitude; (b) the average harvested power in
EC; and (c) the average harvested power in PC when Rli = 50Ω and Rlp = 103 Ω.
Then, the output average power generated by the electromagnetic induction and
piezoelectric transduction can be expressed, respectively, as:
i 1 i i2
Pavg = Ri (27)
2 l
p 1 V p2
Pavg = · (28)
2 Rlp
To compare the analytical predictions of the nonlinear normal form to the numeri-
cal predictions using Runga-Kutta, we plot in Fig. 5 the bifurcation diagrams of the
transverse displacement of the prismatic structure, the average harvested power in the
electromagnetic circuit, and the average harvested power in the piezoelectric circuit
when Rli = 50Ω and Rlp = 103 Ω. In the considered case of external load resistances,
the onset speed of galloping is equal to 1.516 m/s, which can be seen in Fig. 4(a).
When (U > Ug ), the transverse displacement amplitude and the harvested average
power in both EC and PC show a smooth increase with the increase of the wind
speed which is a characteristic of a supercritical instability. It follows from Fig. 5 that
a very good agreement is obtained between the nonlinear normal form results and the
numerical predictions near bifurcation. As the wind speed U is increased far beyond
the bifurcation, the normal form solutions do not accurately predict the numerical
results. This is predicted because the normal form of the Hopf bifurcation is only
valid near the onset of instability. In terms of design purposes, the nonlinear nor-
mal form can be very useful when other sources of nonlinearities including structural
2940 The European Physical Journal Special Topics
0.25 p
Rl =10−2Ω
p
Rl =10 Ω
−2
0.6
Rp=105Ω Rp=105Ω
l l
0.2
Rp=2.6×105Ω 0.5 p 5
Rl =2.6×10 Ω
l
Pavg (W/m)
Rp=106Ω
p 6
0.15 Rl =10 Ω
l 0.4
y (m)
Rp=108Ω Rp=108Ω
l l
0.3
i
0.1
0.2
0.05
0.1
0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
U (m/s) U (m/s)
(a) (b)
p −2
Rl =10 Ω
4
Rp=105Ω
l
Rp=2.6×105Ω
l
3
Rp=106Ω
(W/m)
l
Rp=108Ω
l
avg
2
Pp
0
0 3 6 9 12 15 18
U (m/s)
(c)
Fig. 6. Bifurcation diagrams of (a) the transverse displacement of the bluff body; (b) the
average harvested power in the EC; and (c) the average harvested power in the PC for
different values of Rlp when Rli is set equal to 16Ω.
To study the influences of the external load resistances Rli and Rlp on the performance
of the hybrid piezoelectric-inductive energy harvester, we plot in Figs. 6 and 7 the bi-
furcation diagrams of the transverse displacement of the prismatic structure and the
average harvested power in PC and EC when considering different scenarios. In Fig. 6,
the external load resistance Rli in the EC is set equal to 16Ω with considering five dif-
ferent values of the external load resistance Rlp in the PC. It follows from these plots
that the onset speed of instability is strongly affected by the electrical load resistance
Rlp . In fact, as the Rlp is increased from 10−2 Ω to 108 Ω, the onset speed of gallop-
ing is first increased and then decreased. When Rlp is 2.6 × 105 Ω, the onset speed
i p −2
is 9.78 m/s which is the highest possible value when R =16Ω. When R = 10 Ω
p 8
l l
or Rl = 10 Ω , the hybrid energy harvester has the lowest onset speed of galloping
which is 2.2 m/s. The results are in total agreement with results of the linear analysis
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2941
0.2
i −2
Rl=10 Ω −2
Ri=10 Ω
l
0.3
0.16 Ri=101Ω i 1
R =10 Ω
l l
Ri=102Ω 0.25 i 2
Rl=10 Ω
l
Ri=105Ω
(W/m)
0.12
l 0.2 Ri=105Ω
l
y (m)
avg
0.08 0.15
Pi
0.1
0.04
0.05
0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
U (m/s) U (m/s)
(a) (b)
Rl=10−2Ω
i
0.04
Rl=101Ω
i
Ri=10 Ω
2
0.03 l
Pavg (W/m)
i 5
R =10 Ω
l
0.02
p
0.01
0
0 2 4 6 8 10 12 14
U (m/s)
(c)
Fig. 7. Bifurcation diagrams of (a) the transverse displacement of the bluff body; (b) the
average harvested power in the EC; and (c) the average harvested power in the PC for
different values of Rli when Rlp is set equal to 103 Ω.
of the bluff body, the average output power in the EC, and the average output power
in the PC when the external load resistance Rlp in the PC is set equal to 103 Ω. In-
specting Fig. 7(a), we note that an increase of Rli is accompanied by a reduction in
the onset speed of instability and an increase in the transverse displacement ampli-
tude of the hybrid energy harvester. This result is in agreement with the findings
of the linear analysis, as presented in Fig. 3(a) which shows that the absolute value
of the coupled electromechanical damping is decreased with the increase of Rli in-
dicating a decrease in the onset speed of galloping and a corresponding rise in the
transverse displacement amplitude of the prismatic structure. As for the harvested
power in EC, it follows from Fig. 7(b) that when Rli = 10−2 Ω and Rli = 105 Ω which
correspond, respectively, to the short- and open-circuit configurations, the levels of
the harvested power are almost zero. This is due to the zero generated voltage in the
electromagnetic circuit in the short-circuit case and zero induced current in the open-
circuit case. Furthermore, it is clear that in the four considered values of the external
load resistance in the EC, maximum levels of the harvested power are obtained when
Rli = 10Ω. It can be concluded from this analysis that there is an optimal value of Rli
at which the harvested power in EC is maximum. Regarding the bifurcation diagrams
of the average harvested power in the PC, it is clear that the output power generated
by Rlp in the piezoelectric circuit is increased with the increase of Rli , as shown in
Fig. 7(c). In fact, this is predicted because the external load resistance Rlp in the PC
is considered equal to 103 Ω and hence the generated voltage and harvested power in
PC have the same tendency as the transverse displacement of the bluff body.
Next, a comprehensive analysis is performed in order to study the impacts of the
external load resistances in both the piezoelectric and electromagnetic circuits on the
hybrid energy harvester’s response. One of the objectives of this investigation is to
determine the optimum values of the external load resistances in the piezoelectric and
electromagnetic circuits and their relations to the operating wind speed. To this end,
four different values of the wind speed, namely, U = 2.5 m/s, U = 5 m/s, U = 7 m/s,
and U = 10 m/s are considered. The variations of the transverse displacement of the
bluff body, average harvested power across Rli in EC (Pavg i
), and average harvested
p p p i
power Rl in PC (Pavg ) as function of Rl when Rl = 16Ω are plotted in Fig. 8(a),
8(b), and 8(c), respectively. It follows from these plots that, when the wind speed is
equal to 2.5 m/s, 5 m/s, or 7 m/s, there is a region of Rlp at which the hybrid energy
harvester does not oscillate. As an example, when U = 5 m/s, the hybrid energy har-
vester exhibits a transverse displacement amplitude of zero, when the external load
resistance Rlp is between 5 × 104 Ω and 106 Ω. This is expected because the onset speed
of instability in this region of Rlp is larger than 5 m/s. When U = 10 m/s, there is a
well-defined value of Rlp at which the transverse displacement of the hybrid energy
harvester is minimum. Generally, in the lower range (Rlp < 103 Ω) or higher range
(Rlp > 107 Ω), the transverse displacement of the prismatic structure does not change
with the external load resistance Rlp for all wind speed values. This is due to the
fact that the effects of the external load resistance Rlp on the coupled electromechan-
ical damping is very negligible in the same range of external load resistances in the
piezoelectric circuit. Concerning the variations of the harvested power in the electro-
magnetic circuit as a function of the external load resistance Rlp , as explained above,
these variations have the same tendency as the ones of the transverse displacement
because the external load resistance in the electromagnetic circuit is constant in these
plots, as shown in Fig. 8(b). However, the variations of the average harvested power
in the piezoelectric circuit is totally different, as presented in Fig. 8(c). Inspecting
this plot, we note that there are two optimum values of Rlp for which more energy can
be harvested. These two external load resistance Rlp peaks strongly depend on the
value of the operating wind speed. It is also noted that the region of Rlp over which
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2943
0.3
U=2.5m/s U=2.5m/s
.16 U=5m/s U=5m/s
U=7m/s 0.25 U=7m/s
U=10m/s U=10m/s
.12 0.2
Pavg (W/m)
y (m)
0.15
.08
i
0.1
.04
0.05
0 −2 0 2 4 6 8 0 −2 0 2 4 6 8
10 10 10 10 10 10 10 10 10 p 10 10 10
p Rl (Ω)
Rl (Ω)
(a) (b)
U=2.5m/s
0.5 U=5m/s
U=7m/s
U=10m/s
0.4
(W/m)
0.3
avg
Pp
0.2
0.15
0 −2 0 2 4 6 8
10 10 10 10 10 10
p
Rl (Ω)
(c)
Fig. 8. Variations of (a) the transverse displacement of the bluff body; (b) the average
harvested power in the EC; (c) the average harvested power in the PC as function of Rlp for
different values of the wind speed when Rli is set equal to 16Ω.
the harvested power is maximum matches the region of Rlp over which the trans-
verse displacement is decreasing for all the considered wind speed values. As shown
in Fig. 6(c), there is a specific wind speed value (> 14.88 m/s) at which only one Rlp
peak exists for a maximum value of the harvested power. These maximum values of
the harvested power which are accompanied by minimum values of the transverse dis-
placement are present due to the resistive Rlp shunt damping effect at which maximum
absolute value of the coupled electromechanical damping takes place. This realization
is very useful from energy harvesting point of view where the hybrid energy harvester
oscillate with smaller amplitudes and maximum levels of the harvested power in the
piezoelectric circuit.
Figure 9 shows the variations of the outputs of the hybrid energy harvester with
the external load resistance Rli for different wind speeds, when Rlp is set equal to 103 Ω.
It follows from Fig. 9(a) that the transverse displacement increases with increasing
the external load resistance Rli . This result can be explained by the decrease of the
absolute value of the coupled electromechanical damping when Rli is increased, as
presented in Fig. 3. When U = 2.5 m/s, it is clear that the hybrid energy harvester
does not oscillate when Rli is below 10Ω. This result is predicted from the performed
linear analysis. Indeed, the onset speed of instability when Rli < 10Ω and Rlp = 103 Ω
is higher than 2.5 m/s. From Fig. 9(b), we find that there is an optimal
value of Rli at
i
which the harvested power in the electromagnetic circuit Pavg is maximum for all
considered wind speed values. Clearly, this optimal Rli value depends on the operating
wind speed value. In fact, an increase in the wind speed value is accompanied by a
decrease in the peak value of Rli . In addition, it can be noted that the region of Rli
2944 The European Physical Journal Special Topics
0.2
U = 2.5 m/s U = 2.5 m/s
0.16 U = 5 m/s U = 5 m/s
U = 7 m/s U = 7 m/s
U = 10 m/s 0.15 U = 10 m/s
0.12
Pavg (W/m)
y (m)
0.1
0.08
i
0.04 0.05
0 −2 −1 0 1 2 3 4 5 0 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
i i
R (Ω) R (Ω)
l l
(a) (b)
.025
U = 2.5 m/s
U = 5 m/s
.02 U = 7 m/s
U = 10 m/s
Ppavg(W/m)
.015
.01
0.005
0 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10
i
Rl (Ω)
(c)
Fig. 9. Variations of (a) the transverse displacement of bluff body; (b) the harvested average
power in EC; (c) the harvested average power in PC as function of Rli for different values of
wind speed when Rlp in PC is set to 103 Ω.
over which the harvested power in EC Pavg i is larger, corresponds to the region of
Rli over which the displacement amplitude significantly increases. It is indicated from
Fig. 9(c) that the variations of the harvested power in the piezoelectric circuit (Pavg p )
have a similar tendency with the ones associated to the transverse displacement of
the bluff body. This result is due to the fact that the generated voltage and hence
the harvested power in the PC are directly related to the transverse displacement of
the hybrid energy harvester when considering a constant value of Rlp .
0.8
i
Power by EM, R = 16 Ω Power by PE, Rpl =104Ω
l
0.7 1.2 p 4 i
Power by EH, R =104Ω,Ri=16Ω
p Power by PH, Rl =10 Ω, Rl=16Ω
l l
0.6 0.025
0.015
.9
P avg (W/m)
P avg (W/m)
0.5 0.02
0.01
0.015
0.4
.6
p
0.01
i
0.3 0.005
0.005
0.2 .3
0
1 1.5 2 2.5 3 3.5 0
1 1.5 2 2.5 3 3.5 4
0.1
0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
U (m/s) U (m/s)
(a) (b)
Fig. 10. Bifurcation diagrams of the average harvested power in (a) the electromagnetic
circuit and (b) the piezoelectric circuit when considering classical (electromagnetic harvester
EM or piezoelectric harvester PE) and hybrid (electromagnetic harvester EH and piezoelec-
tric harvester PH) configurations and when Rlp =104 Ω and Rli =16Ω.
0.8 6
i
Power by EM, Rl = 16 Ω Power by PE, Rp = 2.6×105Ω
l
Power by EH, Rpl =2.6×105Ω,Ril=16Ω 5 Power by PH, Rpl = 2.6×105 Ω, Ril=16Ω
0.6
4
P avg (W/m)
(W/m)
0.4 3
avg
p
i
P
2
0.2
1
0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
U (m/s) U (m/s)
(a) (b)
Fig. 11. Bifurcation diagrams of the average harvested power in (a) the electromagnetic
circuit and (b) the piezoelectric circuit when considering classical (electromagnetic harvester
EM or piezoelectric harvester PE) and hybrid (electromagnetic harvester EH and piezoelec-
tric harvester PH) configurations and when Rlp =2.6 × 105 Ω and Rli =16Ω.
Table 2. Onset speeds of galloping and power outputs for classical and hybrid piezoelectric-
inductive energy harvesters when Rlp = 104 Ω and Rli = 16Ω.
Ug (m/s) P (W/m) P (W/m) P (W/m)
5 m/s 10 m/s 15 m/s
PE 1.4477 0.0362 0.174 0.4158
PH 2.7503 0.0238 0.1484 0.3747
EM 2.1708 0.0316 0.1744 0.4313
EH 2.7503 0.0261 0.1631 0.4103
piezoelectric energy harvester changes smoothly from 8.48 m/s to 9.78 m/s, as pre-
sented in Fig. 11(b) and Table 3. This result is predicted because when Rlp =2.6×105 Ω,
the onset speed of galloping is high (almost 8.5 m/s) compared to the possible onset
speeds of galloping of the classical electromagnetic energy harvester which are in the
order of 1 m/s to 3.5 m/s, as shown in Fig. 4(a). As for the levels of the harvested
power for the classical and hybrid systems, we note that, in addition to the significant
increase in the onset speed of galloping for the hybrid piezoelectric-electromagnetic
energy harvester, a strong decrease is observed in the level of the harvested power in
the electromagnetic circuit, as presented in Fig. 11(a). As for the piezoelectric circuit,
a reduction in the levels of the average harvested power are observed, as shown in
Fig. 11(b). This decrease in the average values of the harvested power in both the
piezoelectric and electromagnetic circuits when using the hybrid energy harvester can
be explained by the increase of the coupled damping of the hybrid harvester due to
the presence of additional resistive shunt damping effects. This increase in the coupled
damping results in a reduction in the transverse displacement of the hybrid energy
harvester. For the same values of the external load resistances (Rlp and Rli ) and be-
cause the induced current in the EC and the generated voltage in the PC are directly
related to the velocity of the prismatic structure, the decrease in the transverse dis-
placement is followed by a decrease in the induced current in the electromagnetic
circuit and a deecrease in the generated voltage in the piezoelectric circuit and hence
their associated outputs power.
It can be concluded from this comparative analysis between classical and hybrid
piezoelectric-electromagnetic energy harvesters that depending on the available wind
speed in the environment and the placement of the harvester, the hybrid piezoelectric-
inductive energy harvester can be a beneficial solution to generate energy from two
Nonlinear and Multiscale Dynamics of Smart Materials in Energy Harvesting 2947
Table 3. Onset speeds of galloping and power outputs for classical and hybrid piezoelectric-
inductive energy harvesters when Rlp = 2.6 × 105 Ω and Rli = 16Ω.
Ug (m/s) P (W/m) P (W/m) P (W/m)
5 m/s 10 m/s 15 m/s
PE 8.4772 0 0.4046 2.6103
PH 9.7797 0 0.0591 2.1031
EM 2.1708 0.0316 0.1744 0.4313
EH 9.7797 0 4.9358 × 10−3 0.1759
possible sources. Although there is an increase in the onset speed of galloping, a good
choice of the external load resistances in both the piezoelectric and electromagnetic
circuits can optimize the performance of the hybrid energy harvester for well-defined
operating wind speeds.
5 Conclusions
The concept of hybrid piezoelectric-inductive energy harvesting from galloping os-
cillations has been deeply studied. In order to convert aeroelastic vibrations into
electrical power, the electromagnetic and piezoelectric transduction mechanisms are
both attached to the transverse degree of freedom of a prismatic structure. Based on
a developed lumped-parameter model which couples the transverse displacement of
the bluff body, induced current, and generated voltage across the external load resis-
tances, linear and nonlinear analyses were performed to investigate the influences of
these electrical load resistances on the onset speed of galloping and the performance
of the harvester. The results showed that the presence of both electromagnetic and
piezoelectric transducers leads to a significant change in the onset speed of galloping.
To characterize the type of the ensuing bifurcation of the hybrid energy harvester
and its performance near bifurcation, the normal form of the Hopf bifurcation was
derived. It was demonstrated that a supercritical instability takes place for all values
of the external load resistances in both piezoelectric and electromagnetic circuits. A
parametric study was also performed to investigate the impacts of the external load
resistances on the performance of the harvester. It was shown that, for a well-defined
wind speed, there exists optimal values of these external load resistances at which
the outputs power in the electromagnetic and piezoelectric circuits are optimum.
Based on a comparative analysis between classical and hybrid piezoelectric-inductive
energy harvesters, it was demonstrated that an additional resistive shunt damping
effect takes place in the hybrid piezoelectric-inductive energy harvester compared to
its classical piezoelectric or electromagnetic energy harvesters which results in an in-
crease in the associated onset speed of instability and a reduction in the levels of
the harvested power. On the other hand, the results showed that with a good choice
of the external load resistances, the hybrid piezoelectric-inductive energy harvester
can be optimized and used in many locations in order to operate multiple low-power
consumption devices.
References
1. P. Muralt, J. Micromech. Microeng. 10, 136 (2000)
2. S.P. Gurav, A. Kasyap, M. Sheplak, L. Cattafesta, R.T. Haftka, J.F.L. Goosen, F.V.
Keulen, Proceedings 10th AIAA/ISSSMO Multidisciplinary Analysis and Optimization
Conference, 3559 (2004)
2948 The European Physical Journal Special Topics
3. D.J. Inman, B.L. Grisso, Smart Struct. Mater. Conf. SPIE 6174, 61740T (2006)
4. A. Abdelkefi, M. Ghommem, Theor. Appl. Mech. Lett. 3, 052001 (2013)
5. N. Sharpes, A. Abdelkefi, S. Priya, Ener. Harv. Syst. 1, 209 (2014)
6. H.A. Sodano, G. Park, D.J. Inman, Shock Vib. Dig. 36, 197 (2004)
7. S. Priya, J. Electrocera. 19, 167 (2007)
8. S.R. Anton, H.A. Sodano, Smart Mater. Struct. 16, 1 (2007)
9. G. Litak, M.I. Friswell, S. Adhikari, Appl. Phys. Lett. 96, 214103 (2010)
10. A. Abdelkefi, F. Najar, A.H. Nayfeh, S. Ben Ayed, Smart Mater. Struct. 20, 115007
(2011)
11. A. Abdelkefi, A.H. Nayfeh, M.R. Hajj, Nonlinear Dyn. 67, 1147 (2011)
12. M.I. Friswell, S.F. Ali, O. Bilgen, S. Adhikari, L.W. Lees, G. Litak, J. Intel. Mater. Syst.
Struct. 23, 1505 (2012)
13. M. Bryant, E. Garcia, Proceedings of SPIE 7493, 74931W (2009)
14. A. Erturk, W.G.R. Vieira, C. De Marqui, D.J. Inman, Appl. Phys. Lett. 96, 184103
(2010)
15. C. De Marqui, A. Erturk, D.J. Inman, J. Intel. Mater. Syst. Struct. 21, 983 (2010)
16. H.D. Akaydin, N. Elvin, Y. Andrepoulos, Smart Mater. Struct. 21, 025007 (2012)
17. H.L. Dai, A. Abdelkefi, L. Wang, J. Intel. Mater. Sys. Struct. 25, 1861 (2014)
18. H.L. Dai, A. Abdelkefi, L. Wang, Nonlinear Dyn. 77, 967 (2014)
19. J. Sirohi, R. Mahadik, J. Intel. Mater. Syst. Struct. 22, 2215 (2011)
20. A. Abdelkefi, M.R. Hajj, A.H. Nayfeh, Smart Mater. Struct. 22, 015014 (2013)
21. A. Abdelkefi, Z. Yan, M.R. Hajj, The Eur. Phys. J. Special Topics 222, 1483 (2013)
22. Y. Yang, L. Zhao, L. Tang, Appl. Phys. Lett. 102, 064105 (2013)
23. A. Bibo, M. Daqaq, Appl. Phys. Lett. 104, 023901 (2014)
24. H.J. Jung, S.W. Lee, Smart Mater. Struct. 20, 055022 (2011)
25. A. Abdelkefi, A. Hasanyan, J. Montgomery, D. Hall, M.R. Hajj, Theor. Appl. Mech.
Lett. 4, 022002 (2014)
26. A. Abdelkefi, A.H. Nayfeh, M.R. Hajj, Nonlinear Dyn. 67, 925 (2011)
27. A. Abdelkefi, M.R. Hajj, A.H. Nayfeh, Nonlinear Dyn. 70, 1377 (2012)
28. J. Sirohi, R. Mahadik, ASME J. Vib. Acoust. 134, 1 (2012)
29. A. Abdelkefi, M.R. Hajj, A.H. Nayfeh, Nonlinear Dyn. 70, 1355 (2012)
30. A. Abdelkefi, Z. Yan, M.R. Hajj, Smart Mater. Struct. 22(2), 025016 (2013)
31. A. Abdelkefi, Z. Yan, M.R. Hajj, J. Intel. Mater. Syst. Struct. 25, 246 (2014)
32. D. Zhu, S. Beeby, J. Tudor, N. White, N. Harris, Proc. IEEE Sens. Kona, HI 1, 1415
(2010)
33. C. De Marqui, A. Erturk, J. Intel. Mater. Syst. Struct. 24, 846 (2012)
34. J.A.C. Dias, C. De Marqui, A. Erturk, AIAA J. (2014)
35. J.A.C. Dias, C. De Marqui, A. Erturk, Appl. Phys. Lett. 102, 044101 (2013)
36. M. Ali, M. Arafa, M. Elaraby, Proc. World Cong. Eng., WCE, London, UK 3, 5 (2013)
37. D. Vicente-Ludlam, A. Barrero-Gil, A. Velazquez, J. Flui. Struct. (2014)
http://dx.doi.org/10.1016/j.jfluidstructs.2014.09.007i
38. H.L. Dai, A. Abdelkefi, U. Javed, L. Wang, Smart Mater. Struct. 24, 045012 (2015)
39. R.D. Belvins, Flow-Induced Vibration Malabar (FL: Krieger, 1990)
40. A. Barrero-Gil, G. Alonso, A. Sanz-Andres, J. Soun. Vib. 329, 2873 (2010)
41. G.V. Parkinson, J.D. Smith, Q. J. Mech. Appl. Math. 17, 225 (1964)
42. J.D. Kraus, Electromagnetics (McGraw-Hill), p. 420
43. A.H. Nayfeh, Method of Normal Forms (Wiley Interscience, Berlin, 2011)