You are on page 1of 18

Accepted Manuscript

Fluttering conditions of an energy harvester for autonomous powering

Stefano Olivieri, Gregorio Boccalero, Andrea Mazzino, Corrado Boragno

PII: S0960-1481(16)31126-0
DOI: 10.1016/j.renene.2016.12.067
Reference: RENE 8404

To appear in: Renewable Energy

Received Date: 5 July 2016


Revised Date: 5 November 2016
Accepted Date: 26 December 2016

Please cite this article as: Olivieri S, Boccalero G, Mazzino A, Boragno C, Fluttering conditions
of an energy harvester for autonomous powering, Renewable Energy (2017), doi: 10.1016/
j.renene.2016.12.067.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT

Fluttering conditions of an energy harvester for autonomous powering

Stefano Olivieria,∗, Gregorio Boccalerob , Andrea Mazzinoa,c,d , Corrado Boragnob


a DICCA, Dipartimento di Ingegneria Civile, Chimica e Ambientale, Via Montallegro 1, 16145 Genova, Università degli
Studi di Genova, Italy
b DIFI, Dipartimento di Fisica, Via Dodecaneso 33, 16146, Genova, Università degli Studi di Genova, Italy
c INFN, Istituto Nazionale di Fisica Nucleare, Sezione di Genova, Via Dodecaneso 33, Genova, 16146, Italy

PT
d Consorzio CINFAI, Sezione di Genova, Via Montallegro 1, Genova, 16145, Italy

RI
Abstract
Flapping states of an energy harvesting device have been investigated by means of experiments, numerical

SC
simulations and a phenomenological model. The main aim is to predict the geometrical/physical properties
of the system allowing sustained flapping limit cycles to emerge. These latter regimes are interesting when
the system is used to harvest energy from flows. The main argument to identify flapping states is based on
a simple resonance condition between the characteristic (elastic) time of the system and the flow time-scale.

U
Similar arguments have been successful in other fields of fluid dynamics and fluid-structure interactions
including turbulent flows of dilute polymer solutions and interactions between the wake originated by bluff
AN
bodies and elastic structures. The predictions of the geometrical/physical properties associated to critical
conditions (i.e. those separating stable stages from flapping regimes) have been compared against the results
of experiments, numerical simulations and a phenomenological model based on a set of ordinary differential
equations. Results clearly confirm the expectations from the resonance condition. Discussions on how to
M

extend our analysis in situations where the extraction stage is taken into account are also provided: this
latter is indeed expected to influence the flapping stage and thus the critical conditions for flapping.
Keywords: Energy harvesting, Autonomous powering, Fully-passive harvesters, Low-speed flutter,
D

Fluid-structure interactions
TE

1 1. Introduction

2 Energy harvesting by fluid-structure interaction (FSI) represents a significant research field for developing
EP

3 innovative solutions for power supply, a topic of well-known importance [1, 2]. Several FSI instabilities can
4 be exploited and many efforts have been made in designing efficient and reliable devices. Examples include
5 systems based on vortex-induced vibrations (VIVs) [3, 4], galloping of bluff bodies [5, 6] and fluttering
6 flags [7, 8, 9, 10], as also other mechanisms at relatively small scales, e.g. Knudsen effects [11, 12].
C

7 A special kind of energy harvesting devices is based on the coupled-mode flutter experienced by flapping
8 foils [13, 14]. Here the foil is able to rotate around a pivot point (pitch) and to translate mainly along the
AC

9 direction transverse to the flow (plunge or heave). Different possibilities are available regarding the system
10 activation: (i) fully-driven devices, in which the motion is prescribed [15, 16], (ii) semi-passive devices, where
11 typically only the pitching mode is driven [17, 18, 19] and (iii) fully-passive devices, where the motion is
12 totally governed by the resulting dynamics [20, 21, 22, 23]. These latter are interesting by virtue of their
13 constructive simplicity, since no control mechanisms and actuators are employed, but pose specific challenges
14 regarding the physical comprehension and modeling [13].

∗ Corresponding author at: DICCA, Dipartimento di Ingegneria Civile, Chimica e Ambientale, Via Montallegro 1, 16145

Genova, Università degli Studi di Genova, Italy. Tel.: +390103532560.


Email address: stefano.olivieri@edu.unige.it (Stefano Olivieri)

Preprint submitted to Renewable Energy December 26, 2016


ACCEPTED MANUSCRIPT

PT
RI
SC
Figure 1: A sketch of the FLEHAP device.

15 In this framework, a novel flow energy harvester has been recently proposed by some of the authors [24, 25]
16 essentially consisting in an elastically bounded flapping wing that is immersed in an incompressible flow.

U
17 The system is fully-passive and has remarkable differences with those based on a pure pitch-and-plunge
18 motion. If the key parameters are properly set, regular and self-sustained limit cycle oscillations (LCOs)
AN
19 are found which may be suitable for energy extraction purposes. However, the dynamics reveals to be rich
20 and complex with different motion regimes including stable, periodic and chaotic states, depending on the
21 main system parameters such as geometrical and mechanical properties.
22 A later extensive investigation was therefore performed [26], considering an idealized model and corrob-
M

23 orated by numerical simulations. Several peculiar features of the system were highlighted, including the
24 dependence on the main parameters (mass density ratio, spring elastic constant, pivot point position) and
25 the existence of flapping instability also in the limiting case of infinite spring stiffness. Nevertheless, further
26 efforts are still required in developing the device design.
D

27 Among the open issues, one is represented by the determination of a critical condition for the emergence
28 of sustained flapping in a framework closer to real situations. The aim of the present work is to assess this
TE

29 point by combining experimental, numerical and analytical investigations.


30 The paper is structured as follows: Sec. 2 reports an up-to-date description of our real device; in Sec. 3
31 we deduce the system natural frequency while in Sec. 4 we identify a proper aerodynamic frequency; Sec. 5
32 reports our predictions based on a resonance condition between these characteristic frequencies; hence,
EP

33 theoretical arguments are verified in Sec. 6 and final remarks are given in Sec. 7.

34 2. Our energy harvesting device


C

35 The device has been named FLEHAP (Fluttering Energy Harvester for Autonomous Powering): a 3D
36 sketch of the most recently adopted configuration is shown in Fig. 1. A freely-turning wing is connected
AC

37 to a rigid frame with four elastomers parallel to the wind direction. The wing is composed by a polyvinyl
38 acetate foil (which can be assumed to be rigid without noticeable deformations during the motion) glued
39 to a polymeric 3D-printed tubular part in where a brass axis can rotate. Other configurations can be set,
40 e.g. the elastomers can be orthogonal to the wind and/or the foil can be fixed to the axis whose rotation is
41 allowed inside two supports attached to the elastomers.
42 The system exploits the aeroelastic fluttering effect: in particular settings, self-excitation is induced
43 due to the coupling between the flow and the structure. Two possibilities are considered to convert the
44 mechanical vibration into electrical power. The first strategy exploits an electromagnetic coupling between
45 a pair of coils on the pivot axis and two bars of magnets placed with alternated polarity in front of these.
46 The second method is to characterise the elastomers as dielectric capacitors leading to an amplified voltage.

2
ACCEPTED MANUSCRIPT

3.41

2.72

PT
(y) [N/m]

1.94
Keff

RI
1.05

SC
0.55

2.2 2.9 3.5 4.2 4.8 5.4 6.1 6.7 7.4 8


U [m/s]

U
AN
Figure 2: Wing motion regimes as a function of flow velocity (horizontal direction) and elastomer equivalent stiffess (vertical
direction; see Sec. 3.2 for its definition). Trailing edge trajectories are acquired by a digital camera with long-time exposure
(more information about the experimental method are given in Sec. 6.3).
M

47 The idea is to integrate both strategies to create an efficient and competitive energy harvester which is able
48 to supply micro-processors, sensors and wireless station transmitters also at very low wind speed ranges.
49 Considering only the aeroelastic aspects, the behaviour of the system is governed by many parameters,
e.g.: fluid properties, type and length of the elastomers, wing geometry, pivot point position and mass
D

50

51 distribution. To give a description of the whole system’s flapping states, Fig. 2 shows the resulting motions
52 for a centimetric size device while varying the incoming flow velocity and the elastomer equivalent stiffness
TE

eff
53 K(y) (that will be introduced in Sec. 3.2). Here, the elastomers are placed orthogonal to the wind; although
54 this arrangement is not the most suitable when employing an electromagnetic coupling, a rich variety of
55 flapping regimes is revealed and looking at the reported trajectories several observations can be made:
1. The aeroelastic instability exists only if the flow velocity is above a critical value (which we will denote
EP

56

57 as Ucr and we aim to identify in the present work), otherwise the wing aligns with the flow in a stable
58 condition. This onset speed clearly appears to depend on the value of the elastic constant: as an
eff
59 example, at 2.9[m/s], for K(y) ≤ 1.05[N/m] flutter is triggered (with asymmetrical trajectories due to
the relative importance of gravity at low wind speeds), while for larger values we are still below the
C

60

61 threshold.
2. If one increases the flow velocity beyond Ucr , for all the reported cases the wing enters a regular
AC

62

63 flapping regime characterised by limit cycle oscillations. The trailing edge trajectory rapidly changes
64 its shape and the amplitude increases with U up to a value which is comparable to the chord size. We
eff
65 highlight how the flow velocity corresponding to this maximum in amplitude varies with K(y) . On the
eff
66 other hand, it can also be noted how the flapping amplitude does not vary significantly on K(y) .
67 3. Although not so evident, the pivot point trajectory is also visible in the figure. Up to a certain
68 wind velocity, the amplitude of this latter is larger than the one of the leading edge, while for U >
69 3.5 ÷ 4.2 [m/s] (depending on the stiffness value) the situation gets opposite. It can be shown that
70 this transitional behaviour corresponds to a significant variation of the phase between the pitch and
71 the plunge motions. Moreover, when the pivot point displacement is maximised, a relative maximum
72 is typically found also for the flapping frequency: even if this aspect is still under investigation, this

3
ACCEPTED MANUSCRIPT

73 condition appears as the most efficient for energy harvesting.


74 4. When increasing further U , the motion regime may vary significantly. At the highest velocities that
75 could be tested (U = 8[m/s]), two particular behaviours are observed: for the smallest stiffness the
76 system goes back to a stable condition, while for larger values the motion becomes chaotic.
77 In order to guarantee a good electrical power extraction, the amplitude of the pivot point motion and
78 the oscillation frequency must be maximised. Currently, there is a lack of analytical models able to describe
accurately the system in the nonlinear regime with large excursions that we want to exploit. Furthermore,

PT
79

80 the correlations between the numerous governing parameters involve a complex framework: e.g., adding
81 mass on the axis increases the amplitude but, at the same time, decreases the flapping frequency, while
82 increasing the elastic tension to recover an higher frequency leads to an higher Ucr . Consequently, predictive

RI
83 scaling laws are still unknown [27].
84 Nevertheless, some interesting configurations have been empirically found: Fig. 3 reports the cycle-
85 averaged dissipated power over a pure resistive load at several flow velocities, for a centimetric size prototype
exploiting the electromagnetic coupling strategy. Clearly, at each wind speed an optimal resistance load

SC
86

87 maximising the output power is found. The magnitude of the obtained power (approximately between
88 1 and 18 [mW]) looks promising for autonomous powering purposes and the aforementioned applications.
89 Detailed results concerning the performance of the device will be presented in upcoming work. Furthermore,
90 a first version of a complete system has been tested using a specialized circuit [28].

3. Device natural frequencies of oscillations


U
AN
91

92 As stated in the introduction, the goal of the present work is to identify the critical condition for sustained
93 flapping, i.e. to investigate the threshold between stable and unstable configurations. With this objective in
94 mind, we first aim at deducing the most appropriate expressions for the natural frequency of the system.
M

95 Confirmed by many of our experimental observations, we assume that the motion of the wing is essentially
96 two-dimensional. Even if our final goal is related to the real device, we will consider two different frameworks.
97 The first one is more idealized but helps for a clearer comprehension of the main features of the problem.
D

98 The second one has some increased complexities, including the proper characterisation of the actual elastic
99 elements, and is closer to the real device.
TE

100 3.1. A two-dimensional model with hookean springs


101 We first consider the following framework (already introduced in Ref. [26]): a homogeneous plate is
102 immersed in a uniform incompressible flow with the pivot point E connected through an ideal spring to an
EP

18
16 U = 3.0[m s−1 ]
U = 3.5[m s−1 ]
14
C

U = 4.0[m s−1 ]
12 U = 4.5[m s−1 ]
P [mW]

10 U = 5.0[m s−1 ]
AC

8
6
4
2
0
103 104 105
R [Ω]

Figure 3: Power extraction measurements as a function of a purely resistive load for different values of the wind speed
(considering a wing of chord c = 35[mm], span s = 85[mm] and using the electromagnetic coupling extraction method).

4
ACCEPTED MANUSCRIPT

(a) (b)

E l E l
y
A1 A2
A
L L
y

PT
x (c)

RI
E θ

Figure 4: (a) The ideal model (already presented in Ref. [26]). The dotted line represents the linear spring connecting the
anchor point A to the wing pivot point E. (b) The real model representative of our energy harvester. Each elastomer connects

SC
the respective anchor point Ai to the wing pivot point E; L is the initial elastomer length while l is the present length during
the wing motion; y represents the vertical oscillation of the pivot point. (c) The wind vane situation. The wing is hinged at
the pivot point E and rotates about it with a pitching angle θ.

U
103 anchor point A (Fig. 4a). Gravity is neglected. The spring follows Hooke’s law and the elastic force modulus
104 is thus expressed by:
AN
105 Fel = k|l − l0 | (1)
106 where k is the spring stiffness, l and l0 are the present length and the rest-length of the spring, respectively.
107 In this case, the well-known expression for the natural frequency holds:
M

r
1 k
108 fn = (2)
2π m
where m is the mass of the homogeneous plate (the spring is considered to be massless and without any
D

109

110 dissipation).
TE

111 3.2. The real model based on polymeric elastomers


112 In order to get closer to our real device, we refer to the situation sketched in Fig. 4b. Here, x and y
113 are the horizontal and vertical displacement of the pivot point (assumed to be small), respectively; l is the
114 present length, L the initial length and l0 is the rest-length of the elastomer; hence, L = l0 +  where  is
EP

115 the initially given pre-stretching. Additionally, let G be the shear modulus of the elastomer material and
116 A0 the initial cross-sectional area of the elastomer. The elastic force modulus is thus assumed to be [29]:
"  2 #
l l0
C

117 Fel = GA0 − . (3)


l0 l
AC

118 The total elastic force along the y-direction can be expressed as:
p !2 
L2 + y2 l
119 Fy = 4 GA0  − p
0 p y (4)
l0 2
L +y 2 L2 + y 2

120 that in the limit of small oscillations (i.e. y  l0 < L) yields:


"  2 #
L l0 y
121 Fy = 4 GA0 − (5)
l0 L L

5
ACCEPTED MANUSCRIPT

eff
122 which can be intended as a linear function of y so that it can also be written as Fy = K(y) y, where we have
123 introduced the equivalent linear spring stiffness, or effective stiffness, with respect to the vertical direction:
"  2 #
eff GA 0 L l0
124 K(y) =4 − . (6)
L l0 L

125 Additionally, for the sake of simplicity we can consider the pre-stretching to be small compared to the
eff
rest-length i.e.   l0 . In this case, K(y) reduces to:

PT
126

eff 
127 K(y) = 12 GA0 . (7)
l0 2

RI
128 The natural frequency associated with the vertical oscillations can thus be expressed as:
s
eff
K(y)
1
129 f(y) = (8)
2π m

SC
130 where m is the mass of the device.
131 Similar considerations can be made for the horizontal displacement. The total elastic force can be
132 expressed now as:

U
" 2 # " 2 #
L−x
 
l0 L+x l0
133 Fx = 2 GA0 − − 2 GA0 − (9)
l0 L−x l0 L+x
AN
eff
134 that for x  l0 yields: K(x) = 4 GA0 /(10 L3 )(L3 + 2 l03 ) and for   l0 further reduces to K(x)
eff
= 12 GA0 /l0 .
135 Under these assumptions, it is easily seen that:
eff
M

K(y) 
136
eff
= 1 (10)
K(x) l0

137 which means that the natural frequency associated with the horizontal oscillations is much higher than the
D

138 one related to the vertical oscillations. For this reason, we will assume the latter to be the first frequency
139 to couple with the hydrodynamic one.
TE

140 4. Aerodynamic (wind vane) frequency


141 After the characterisation of the natural frequencies, the next step is to identify the proper one concerning
the fluid dynamical side. For this purpose, we consider the situation in which the wing is hinged at the pivot
EP

142

143 point with only the pitching motion to be allowed (Fig. 4c). We note that this corresponds to the limit of
144 infinitely large spring stiffness that was already defined as the wind vane situation in Ref. [26].
145 In this case, it is sufficient to refer to the moment equation only, which yields:
C

146 IE θ̈ = MEaero (11)


and assuming the rotation angle to be small (with the wing almost aligned with the unperturbed flow) and
AC

147

148 the lift force to be applied at a distance c/4 from the pivot point, the aerodynamic moment can be expressed
149 as MEaero = π/4 ρf c S θ U 2 , where S is a reference area (usually S = c s, where s is the wing span) and ρf is
150 the fluid density. If we search for harmonic solutions, the following frequency can be derived:
r
ρf c S
151 fv = U (12)
16π IE
152 which we denote as the wind vane frequency. Within the idealized framework (introduced in Sec. 3.1), we
153 can simplify the expression above by taking S = c 1 and IE = mc2 /3, obtaining:
r
3 ρf
154 fv,id = U. (13)
16π m
6
ACCEPTED MANUSCRIPT

155 5. Resonance condition for the emergence of flapping

156 Let us now assert that the critical condition for the emergence of flapping is related to a resonance
157 condition between the natural frequency and the aerodynamic frequency. Arguments of this type have been
158 successful, e.g., to explain symmetry breaking mechanism in fluid-structure interaction [30, 31] as well as the
159 emergence of elastic instabilities [32, 33] and macroscopic spatial scales at which microscopic polymers cause
160 viscoelastic behaviour [34]. Following this line of reasoning, when these two frequencies are comparable, i.e.

PT
161 fv ≈ fn , one can easily derive an expression for the critical velocity for the onset of sustained flapping.
162 For the ideal model, this means to equate Eq. (2) with Eq. (13), from which we derive:
s
2 k
Ucr,id ≈ √ . (14)

RI
163
3 π ρf

164 Interestingly, we note that this quantity depends only on the spring stiffness and the fluid density.
For the real model, as previously motivated, we assume that the natural frequency to be considered is

SC
165

166 the one associated with vertical oscillations (y direction), hence fn = f(y) . By equating Eq. (12) with Eq. (8)
167 we obtain: s
eff
IE K(y)
2
168 Ucr ≈ √ . (15)

U
π ρf c S m
169 that can be intended as a more general form with respect to Eq. (14). It should be said that the moment
AN
170 of inertia IE is a function of the wing mass distribution and geometry. Nevertheless, this form for the
171 expression is useful for our scope, as it will be shown later.

6. Verification of theoretical predictions


M

172

173 In order to corroborate the theoretical arguments exposed in the previous sections, we proceed as follows.
174 For the idealized model, numerical simulations are first performed and a phenomenological model is later
used. Concerning the real model, an experimental wind tunnel investigation is carried out. The following
D

175

176 subsections report the main features and findings of each study.
TE

177 6.1. A numerical study


178 In the present section the two-dimensional idealized model introduced in Sec. 3.1 is studied by means
179 of numerical simulations. Several methods have been developed for the solution of FSI problems [35],
including those based on overlapping grids [36], level set [37] and immersed boundaries [38]. Here we use
EP

180

181 a conforming-mesh, finite-volume approach handled by the open source toolkit OpenFOAM R
[39], written
182 in C++ [40]. A Cartesian-based computational mesh is generated using the cfMesh package [41]. The
183 chosen solver is pimpleDyMFoam which solves the incompressible Navier-Stokes equations for a moving mesh
using the PIMPLE (merged PISO-SIMPLE) algorithm [39, 42]. The rigid body dynamics is handled by the
C

184

185 sixDoFRigidBodyMotion library using a symplectic splitting method [43] while dynamic mesh morphing is
186 performed accordingly [44, 45].
AC

187 We consider a plate with rectangular cross-section with chord c and thickness δ and a rectangular domain
188 ranging from (−5c, −5c) to (10c, 5c) as shown in Fig. 5a. Boundary conditions are set as follows: no-slip
189 (moving wall) on the plate, slip on the top and bottom boundaries, fixed velocity at the inlet and specific
190 mixed condition at the outlet [39]. The plate initially lies aligned with the flow and its center located in the
191 origin.
192 As a baseline, we fix the plate properties to be (unless otherwise specified): chord c = 50[mm], mass
193 m = 0.25[g], chord-to-thickness aspect ratio c/δ = 100, structure-to-fluid density ratio ρs /ρf = 250. The
194 spring has default stiffness k = 0.25[N m−1 ], zero rest-length and is attached at the plate’s leading edge.
195 The Reynolds number can vary but is kept always smaller than 1000. Moreover, a sufficiently large angular
196 momentum is initially imposed in order to reach always the flapping state, if existing. Indeed, the co-
197 existence of both stable and unstable states has been previously observed for certain cases (bi-stability) [26].
7
ACCEPTED MANUSCRIPT

(a)

y
5c

PT
x

5c

RI
SC
5c 10c

(b)

U
AN
M
D
TE

Figure 5: (a) 2D domain used in our numerical investigation. The plate (for clarity, not in scale with the rest of the domain)
is initially aligned along the x-axis with its center in the origin (0, 0). (b) Detail of the 2D mesh near the plate.
EP

198 The used mesh has approximately 50000 cells with the resolution varying from ∆min = c/5 (base resolu-
199 tion) to ∆max = c/160 (near the plate), using five refinement levels (increasing the resolution while getting
200 closer to the plate). A detail showing the region near the body is given in Fig. 5b.
201 Dependency of the solution upon the resolution was investigated: the present mesh satisfies a good
C

202 compromise between accuracy and computational efficiency, with a variation of the flapping amplitude below
203 1% for further refining. Additionally, we successfully reproduced some of the results presented in Ref. [26],
obtaining very good agreement between two solvers with several different features (e.g., overlapping grids
AC

204

205 versus mesh morphing).


206 We are now ready to corroborate the theoretical predictions presented in Sec. 5. To do this, a parametric
207 investigation over the fluid density ρf is conducted. We proceed as follows: we fix the value of ρf and
208 perform simulations for different values of the flow velocity U , distinguishing between stable and unstable
209 cases. Next, we change the value of ρf and repeat the procedure. The overall results are reported in
210 Fig. 6a together with the marginal curve based on the theoretical arguments exposed in Sec. 5 (predicting

211 a scaling law according to Ucr,id ∼ ( ρf )−1 ). Numerical simulations are found to corroborate the analytical
212 prediction: for cases below the curve, the wing tends to align with the flow; for those above the curve, the
213 flapping motion is sustained.
214 Furthermore, a similar study is done over the spring stiffness k. Now, from the theory based on the

8
ACCEPTED MANUSCRIPT

(a) (b)
3 3.5
CFD, stable
CFD, unstable
3
2.5
P.M., stable 2.5
U [m s−1 ]

U [m s−1 ]
2 P.M., unstable 2

PT
1.5 1.5 CFD, stable
1 CFD, unstable
1 P.M. stable
0.5 P.M. unstable

RI
0.5 0
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0 0.2 0.4 0.6 0.8 1 1.2 1.4
ρf [kg m−2 ] k [N m−1 ]

SC
Figure 6: Threshold for sustained flapping (a) in the (ρf , U ) plane and (b) in the (k, U ) plane. Results from numerical
simulations and phenomenological model (this latter will be presented in Sec. 6.2. The dashed line represents the marginal
curve based on Eq. (14).

U
resonance condition we have Ucr,id ∼ k. Results are collected in Fig. 6b, showing also here confirmation
AN
215

216 from the numerical investigation.


217 To deepen our understanding, we look at the plate flapping frequency (computed from the time history
218 of the angular velocity) for the unstable cases √ closest to the marginal line (Fig. 7a). We observe that the
flapping frequency goes almost linearly with k, as expected. It should be pointed out that it is practically
M

219

220 impossible to compute cases lying exactly on the marginal line. Nevertheless, one can notice that the lock-in
221 phenomenon, typical of VIV systems [46], seems not to occur here: this is not surprising since the device is
222 based on a flutter instability, as stated previously.
D

223 Hence, through further computations we explore what happens when the spring stiffness is fixed and the
224 flow velocity is increased over the flapping threshold, showing in Fig. 7b that the resulting flapping frequency
increases with the flow velocity. This evidences how the flow is able to govern the post-critical regime.
TE

225

226 Regarding this aspect, further analyses are certainly deserved in order to characterise more accurately the
227 system behaviour.

6.2. A phenomenological model


EP

228

229 Seeking for an analytical description of the system and to confirm our findings from numerical simula-
230 tions in a complementary way, we adopt and modify the phenomenological, quasi-steady model originally
231 developed for falling plates by Andersen et al. [47] and later used for insect flight [48]. We note that a similar
C

232 attempt has been recently made for an active flapping foil energy harvester [49].
233 The model consists of a system of first order ODEs written in the frame of reference (x0 , y 0 ) co-rotating
with the plate. We modify the equations to be suitable for our idealized model by taking into account the
AC

234

235 elastic force due to the spring and the incoming flow, so that they become:

236 (m + m11 )v̇x0 = (m + m22 )θ̇vy0 + Fxel0 − ρf Γṽy0 − Fxν0 (16)

237

238 (m + m22 )v̇y0 = −(m + m11 )θ̇vx0 + Fyel0 + ρf Γṽx0 − Fyν0 (17)
239

240 (IG + Ia )ω̇ = (m11 − m22 )vx0 vy0 − crFyel0 − lτ ρf Γṽx0 − τ ν (18)
241

242 ẋ0G = vx0 + θ̇yG


0
(19)
9
ACCEPTED MANUSCRIPT

(a) (b)
18 10
CFD
16
P.M.
9.5 CFDP.M.
14
12 9
f [Hz]

f [Hz]
10 8.5

PT
8 8
6 7.5
4
2 7

RI
0 6.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3

k [N1/2 m−1/2 ] U [m s−1 ]

SC

Figure 7: (a) Flapping frequency as a function of k for the unstable cases closest to the marginal line. The dotted line
represents the natural frequency fn (Eq. 2). (b) Flapping frequency as a function of the free-stream velocity. Parameters are set
according to the baseline case (described at beginning of Sec. 6.1). Results from numerical simulations and phenomenological

U
model (introduced in Sec. 6.2).
AN
243
0
244 ẏG = vy0 − θ̇x0G (20)
245
M

246 θ̇ = ω (21)
247 where vx0 and vy0 denote the center of mass velocities, ṽx0 = (vx0 − U cos θ) and ṽy0 = (vy0 + U sin θ) are
248 those relative to the unperturbed flow, m is the plate mass, IG is the moment of inertia with respect to the
D

249 center of mass, r represents the distance between the center of mass and the pivot point normalized with the
250 chord, x0G and yG0
are the coordinates of the plate center of mass, θ is the pitching angle, Fel is the elastic
force and lτ = c/4 cos θ is the moment arm of the circulatory force. The added mass coefficients m11 , m22
TE

251

252 and Ia are expressed for the plate of rectangular cross-section following Huang et al. [50]: m11 = (3π/8)ρf δ 2 ,
253 m22 = (3π/8)ρf c2 , Ia = (5π/256)ρf (c2 − δ 2 )2 .
254 The circulation Γ, the viscous force Fν and the dissipative fluid torque τ ν are expressed in a semi-
255 empirical manner using several free parameters [47]:
EP

ṽx0 ṽy0 1
256 Γ = −CT c q + CR c2 θ̇ (22)
ṽx20 + ṽy20 2
C

257 q
2 2
ν1  2 2
 ṽx0 + ṽy0
F = ρf c CD (0) ṽx0 + CD (π/2) ṽy0 (ṽx0 , ṽy0 ) (23)
AC

258
2 ṽx20 + ṽy20
259 !
3 4
|ṽ y 0 |c | θ̇|c
260 τ ν = Cτ ρf θ̇ + (24)
24 64

261 where CT and CR are the translational and rotational lift coefficients, respectively, CD (0) and CD (π/2) are
262 the drag coefficients while Cτ is the dissipative torque coefficient.
On the numerical side, the evolution of equations (16)–(21) is performed using a third order Adams-
Bashfort integration:
Xn+1 = Xn + cf1 Gn + cf2 Gn−1 + cf3 Gn−2 (25)

10
ACCEPTED MANUSCRIPT

PT
RI
Figure 8: The employed wind tunnel where the real device is tested; (A) nozzle, (B) test chamber, (C) diffuser.

where X is a vector containing the six unknown quantities and G collects the right-hand-side of the equations;

SC
263

264 the superscript denotes that the quantity is evaluated at a certain discrete time tn = n∆t, while expressions
23
265 of the coefficients are: cf1 = 12 ∆t, cf2 = − 34 ∆t and cf3 = 12 5
∆t [51]. The timestep was set equal to
p
266 ∆t = 0.05 m/k, i.e. based on the natural frequency (that is, as previously shown, of the same order of
the flapping frequency), with negligible differences found if choosing smaller values (while increasing the

U
267

268 computational time).


We calibrate the model by tuning the coefficients in order to fit the trajectory of the pivot point and the
AN
269

270 pitching angle time trace with those from one case among the numerical simulations presented in Sec. 6.1.
271 After this operation, the following set is considered to be satisfactory: CT = 1.5, CR = 0.22, CD (0) = 0.05,
272 CD (π/2) = 3 and Cτ = 1.5. The values of the model coefficients are then retained for the whole investigation.
273 We thus proceed to repeat the parameter space exploration already done by means of numerical sim-
M

274 ulations and presented previously. Looking again at Fig. 6, the prediction by the model is found to be in
275 complete agreement with simulations and therefore corroborating the theory, too.
276 Similarly, we look again at the flapping frequency for cases closest to the critical condition (Fig. 7a).
D

277 Good agreement is found with the numerical study, even if the model underestimates this quantity especially
278 at the highest spring stiffness values. Subsequently, we investigate again how the frequency varies with the
flow velocity as shown in Fig. 7b. The prediction from the phenomenological model confirms the monotonic
TE

279

280 increase already found by numerical simulations, with a maximum difference of around 5%, providing a
281 more linear trend. This may be explained by the fact that the model is not able to capture all the fluid
282 dynamical features, in particular those related to the vortex shedding mechanism which occurs in conditions
283 of effective flapping. Despite this, the phenomenological model confirms our findings and represents a useful
EP

284 and complementary tool for the analysis and design of our system.

285 6.3. An experimental wind tunnel study


The wind tunnel used for the experimental measurements is a subsonic aspirating open circuit tunnel
C

286

287 built at the Department of Physics of the University of Genoa (Italy). With a total length of 3[m], its shape
288 is designed to obtain a stable and controlled laminar flow inside the test chamber, the 40 × 40 × 70 [cm3 ]
AC

289 middle part of tunnel (Fig. 8). By exploiting the Venturi effect, the air is aspirated through a honeycomb
290 grid into the nozzle (A), conveyed to the test chamber (B) and through the diffuser (C), at the end of which
291 a three-phase motor of the aspirating fan is collocated.
292 The tunnel calibration was performed using a Cobra probe sampling at 1250[Hz]. A free-stream turbu-
293 lence level of around 0.6% is ensured thanks to the use of the honeycomb grid, the appropriate curvature of
294 the nozzle and the slope of the diffuser. By varying the fan rotational speed, the operational flow velocity in
295 the test chamber ranges from 1.7 to 10.5 [m/s] with an absolute error of 0.15[m/s]. Moreover, visualizations
296 of the free-stream flow and of the flow-structure interaction have been possible using a home-made smoke
297 generator apparatus (Fig. 9).
298 The use of a high definition camera (500[fps]) coupled to an appropriate illumination and video processing
299 system allows to obtain a precise acquisition of the wing motion: a light scatterer placed on the rotational
11
ACCEPTED MANUSCRIPT

(a) (b)

PT
RI
Figure 9: Flow visualizations using smoke generation: (a) free-stream; (b) experiment.

SC
300 axis of the wing is illuminated by a line laser beam and a numerical elaboration of the movie returns the
301 main quantities in time such as the pivot point position and the pitching angle.
302 A hysteresis phenomenon has been observed in what it concerns the evolution of motion whether increas-

U
303 ing or decreasing the wind speed. This is found to be crucial also for the flapping threshold determination,
304 which is the main subject of the present study. As also observed in similar systems [52, 27], this opens
AN
305 the problem to define a convention to evaluate the critical flow velocity Ucr : we will assume this to be the
306 lower value of the range, i.e. the value that, if further decreased, leads the system to stop the existing
307 self-sustained excitation. We mention however that we typically find this lower value to be between 0.8 and
308 0.9 times the upper one (which is measured while increasing the wind speed), similarly to Ref. [27].
In order to verify the resonance condition for the emergence of flapping proposed in Sec. 5, two studies
M

309

310 have been performed. For all measurements, the type of elastomers is fixed such that GA0 = 0.604[N] while
311 the wing geometry is set to be: c = 60[mm], s = 70[mm].
312 In the first study, we investigate the dependence of the critical velocity Ucr on the elastomer rest-length
D

313 l0 (while a constant pre-stretching  is retained) i.e. varying the effective stiffness. Results are shown in
314 Fig. 10a together with the prediction expressed by Eq. (15) where we make use of the simplified expression
eff
, Eq. (7), so that Ucr ∼ l0−1 . The measurements are found to be in good agreement with the expected
TE

315 for K(y)


316 scaling law. For the lowest tested value of l0 = 25[mm] the critical velocity is around 8.5[m s−1 ] while for
317 the highest one (l0 = 120[mm]) it drops to approximately 3[m s−1 ]. This clearly shows how the cut-in speed
318 of the system can be governed by such a simple parameter.
EP

319 Subsequently, we focus on the dependence of Ucr upon the mass of the system m. In this experiment,
320 the mass is varied by placing additional weights on the wing √ rotation axis so that the moment of inertia IE
321 remains constant. Hence, a scaling according to Ucr ∼ ( m)−1 has to be expected. Fig. 10b reports the
322 experimental evidences along with the theoretical prediction, finding again a satisfactory agreement. The
C

323 plot shows that, for the given configuration, Ucr may be reduced of around 33% by placing a 4.5[g] additional
324 mass.
Similarly to what we have found in the ideal framework, the wind tunnel study appears to verify the
AC

325

326 simple resonance condition idea considering the more realistic model treated in Sec. 3.2. We underline the
327 importance of this result in view of the practical application, supplying useful insights for the development
328 of the proposed energy harvesting system.

329 7. Conclusions

330 This work has presented an updated picture of the FLEHAP device, a novel energy harvester exploiting
331 FSI which is under active development. The main focus has been, in particular, on identifying the critical
332 condition for the emergence of sustained flapping.

12
ACCEPTED MANUSCRIPT

(a) (b)
12 4.4
11 4.2
10 4
Ucr [m s−1 ]

9 3.8

Ucr [m s−1 ]
8 3.6

PT
7 3.4
6 3.2
5 3
4 2.8

RI
3 2.6
20 40 60 80 100 120 2.4
l0 [mm] 2 3 4 5 6 7 8
m [g]

SC
Figure 10: Experimental measurements of the critical velocity for the onset of sustained flapping (a) as a function of the
elastomer rest-length l0 (here, m = 1.63[g]) and (b) as a function of the wing mass m (with l0 = 58[mm], L = 66[mm]). The
dashed line represents the marginal curve based on Eq. (15).

U
The current study has been carried out considering two similar but different models. The first one is more
AN
333

334 idealized and has been investigated by means of numerical simulations and applying a phenomenological
335 model based on a set of ordinary differential equations. The expected scaling laws for the critical velocity
336 with respect to the fluid density ρf and the spring stiffness k have been verified. The second model is more
representative of the real device and the corresponding predictions have been successfully compared with
M

337

338 the results from wind tunnel experiments, investigating the dependence of Ucr on the elastomer rest-length
339 l0 and the mass m (retaining the same moment of inertia IE ).
340 These evidences are in agreement with the idea of a resonance condition between the natural frequency
D

341 of the system and the so-called wind vane frequency, the characteristic frequency when a rigid plate is hinged
342 at a pivot point and immersed in an incompressible flow.
We point out that these findings are useful also for practical purposes related to the design of efficient
TE

343

344 energy harvesting devices. Indeed, the critical velocity for sustained flapping represents the lower threshold
345 of the operational flow velocity range.
346 On the other hand, a key point for future developments is the characterisation of the flapping behaviour
within the operational range, i.e. for U > Ucr , with a particular focus on the optimization of amplitude and
EP

347

348 frequency, which are crucial quantities for energy harvesting, leading to highest efficiency or output power.
349 Finally, the present work has not considered the presence of the power extraction: on-going experiments
350 and future analyses are planned to describe how this essential effect modifies the dynamics, including the
flapping threshold. In this regard, we mention that the phenomenological model here used can be modified
C

351

352 to apply to the real model and can be coupled to additional differential equations mimicking the dissipative
353 effect caused by the extraction phase. Moreover, the same concept can be addressed by more complex
AC

354 numerical simulations.

355 Acknowledgements

356 We thank Jacopo Fugardo for his precious contribution to the experiments. We thank the financial
357 support from the PRIN 2012 project n. D38C13000610001 funded by the Italian Ministry of Education.
358 We also thank the Italian flagship project RITMARE for the financial support and the computational
359 infrastructure. Helpful discussions and suggestions with Alessandro Bottaro, Joel Guerrero and Jan Oscar
360 Pralits are warmly acknowledged.

13
ACCEPTED MANUSCRIPT

361 References
362 [1] A. Harb, Energy harvesting: State-of-the-art, Renewable Energy 36 (10) (2011) 2641 – 2654. doi:http://dx.doi.org/
363 10.1016/j.renene.2010.06.014.
364 URL http://www.sciencedirect.com/science/article/pii/S0960148110002703
365 [2] P. Rawat, K. D. Singh, H. Chaouchi, J. M. Bonnin, Wireless sensor networks: a survey on recent developments and
366 potential synergies, The Journal of Supercomputing 68 (1) (2014) 1–48.
367 [3] L. Ding, L. Zhang, M. M. Bernitsas, C.-C. Chang, Numerical simulation and experimental validation for energy harvesting
368 of single-cylinder VIVACE converter with passive turbulence control, Renewable Energy 85 (2016) 1246 – 1259. doi:http:

PT
369 //dx.doi.org/10.1016/j.renene.2015.07.088.
370 URL http://www.sciencedirect.com/science/article/pii/S0960148115301841
371 [4] C. Grouthier, S. Michelin, R. Bourguet, Y. Modarres-Sadeghi, E. de Langre, On the efficiency of energy harvesting using
372 vortex-induced vibrations of cables, Journal of Fluids and Structures 49 (2014) 427 – 440. doi:http://dx.doi.org/10.

RI
373 1016/j.jfluidstructs.2014.05.004.
374 URL http://www.sciencedirect.com/science/article/pii/S0889974614001005
375 [5] A. Barrero-Gil, G. Alonso, A. Sanz-Andres, Energy harvesting from transverse galloping, Journal of Sound and Vibration
376 329 (14) (2010) 2873 – 2883. doi:http://dx.doi.org/10.1016/j.jsv.2010.01.028.
377 URL http://www.sciencedirect.com/science/article/pii/S0022460X10000891

SC
378 [6] H. Dai, A. Abdelkefi, U. Javed, L. Wang, Modeling and performance of electromagnetic energy harvesting from galloping
379 oscillations, Smart Materials and Structures 24 (4) (2015) 045012.
380 URL http://stacks.iop.org/0964-1726/24/i=4/a=045012
381 [7] D. T. Akcabay, Y. L. Young, Hydroelastic response and energy harvesting potential of flexible piezoelectric beams in
382 viscous flow, Physics of Fluids 24 (5) (2012) –. doi:http://dx.doi.org/10.1063/1.4719704.

U
383 URL http://scitation.aip.org/content/aip/journal/pof2/24/5/10.1063/1.4719704
384 [8] S. Michelin, O. Doaré, Energy harvesting efficiency of piezoelectric flags in axial flows, Journal of Fluid Mechanics 714
385 (2013) 489–504. doi:10.1017/jfm.2012.494.
AN
386 URL http://journals.cambridge.org/article_S0022112012004946
387 [9] M. Pineirua, O. Doaré, S. Michelin, Influence and optimization of the electrodes position in a piezoelectric energy harvesting
388 flag, Journal of Sound and Vibration 346 (2015) 200 – 215. doi:http://dx.doi.org/10.1016/j.jsv.2015.01.010.
389 URL http://www.sciencedirect.com/science/article/pii/S0022460X15000371
390 [10] J. Favier, A. Revell, A. Pinelli, Fluid Structure Interaction of Multiple Flapping Filaments Using Lattice Boltz-
M

391 mann and Immersed Boundary Methods, Springer International Publishing, Cham, 2016, pp. 167–178. doi:10.1007/
392 978-3-319-27386-0_10.
393 URL http://dx.doi.org/10.1007/978-3-319-27386-0_10
394 [11] T. Zhu, W. Ye, Origin of Knudsen forces on heated microbeams, Phys. Rev. E 82 (2010) 036308. doi:10.1103/PhysRevE.
D

395 82.036308.
396 URL http://link.aps.org/doi/10.1103/PhysRevE.82.036308
397 [12] T. Zhu, W. Ye, J. Zhang, Negative Knudsen force on heated microbeams, Phys. Rev. E 84 (2011) 056316. doi:10.1103/
TE

398 PhysRevE.84.056316.
399 URL http://link.aps.org/doi/10.1103/PhysRevE.84.056316
400 [13] Q. Xiao, Q. Zhu, A review on flow energy harvesters based on flapping foils, Journal of Fluids and Structures 46 (2014)
401 174 – 191. doi:http://dx.doi.org/10.1016/j.jfluidstructs.2014.01.002.
402 URL http://www.sciencedirect.com/science/article/pii/S0889974614000140
[14] J. Young, J. Lai, M. F. Platzer, A review of progress and challenges in flapping foil power generation, Progress in Aerospace
EP

403
404 Sciences 67 (2014) 2–28.
405 [15] W. McKinney, J. DeLaurier, Wingmill: an oscillating-wing windmill, Journal of energy 5 (2) (1981) 109–115.
406 [16] T. Kinsey, G. Dumas, Parametric study of an oscillating airfoil in a power-extraction regime, AIAA journal 46 (6) (2008)
407 1318–1330.
[17] Q. Zhu, Z. Peng, Mode coupling and flow energy harvesting by a flapping foil, Physics of Fluids 21 (3) (2009) –. doi:
C

408
409 http://dx.doi.org/10.1063/1.3092484.
410 URL http://scitation.aip.org/content/aip/journal/pof2/21/3/10.1063/1.3092484
[18] J. Wu, Y. L. Qiu, C. Shu, N. Zhao, Pitching-motion-activated flapping foil near solid walls for power extraction: A
AC

411
412 numerical investigation, Physics of Fluids 26 (8) (2014) –. doi:http://dx.doi.org/10.1063/1.4892006.
413 URL http://scitation.aip.org/content/aip/journal/pof2/26/8/10.1063/1.4892006
414 [19] J. Deng, L. Teng, D. Pan, X. Shao, Inertial effects of the semi-passive flapping foil on its energy extraction efficiency,
415 Physics of Fluids 27 (5) (2015) –. doi:http://dx.doi.org/10.1063/1.4921384.
416 URL http://scitation.aip.org/content/aip/journal/pof2/27/5/10.1063/1.4921384
417 [20] Z. Peng, Q. Zhu, Energy harvesting through flow-induced oscillations of a foil, Physics of Fluids 21 (12) (2009) –. doi:
418 http://dx.doi.org/10.1063/1.3275852.
419 URL http://scitation.aip.org/content/aip/journal/pof2/21/12/10.1063/1.3275852
420 [21] Q. Zhu, Optimal frequency for flow energy harvesting of a flapping foil, Journal of Fluid Mechanics 675 (2011) 495–517.
421 doi:10.1017/S0022112011000334.
422 URL http://journals.cambridge.org/article_S0022112011000334
423 [22] Q. Zhu, Energy harvesting by a purely passive flapping foil from shear flows, Journal of Fluids and Structures 34 (2012)
424 157–169.

14
ACCEPTED MANUSCRIPT

425 [23] J. Young, M. A. Ashraf, J. C. Lai, M. F. Platzer, Numerical simulation of fully passive flapping foil power generation,
426 AIAA journal 51 (11) (2013) 2727–2739.
427 [24] C. Boragno, R. Festa, A. Mazzino, Elastically bounded flapping wing for energy harvesting, Applied Physics Letters
428 100 (25) (2012) 253906.
429 [25] C. Boragno, G. Boccalero, A new energy harvester for fluids in motion, Proc. SPIE 9431 (2015) 94310G–94310G–6.
430 doi:10.1117/12.2084591.
431 URL http://dx.doi.org/10.1117/12.2084591
432 [26] A. Orchini, A. Mazzino, J. Guerrero, R. Festa, C. Boragno, Flapping states of an elastically anchored plate in a uniform
flow with applications to energy harvesting by fluid-structure interaction, Physics of Fluids 25 (9) (2013) 097105. doi:

PT
433
434 10.1063/1.4821808.
435 [27] X. Amandolese, S. Michelin, M. Choquel, Low speed flutter and limit cycle oscillations of a two-degree-of-freedom flat
436 plate in a wind tunnel, Journal of Fluids and Structures 43 (2013) 244 – 255. doi:http://dx.doi.org/10.1016/j.
437 jfluidstructs.2013.09.002.

RI
438 URL http://www.sciencedirect.com/science/article/pii/S0889974613001874
439 [28] G. Boccalero, C. Boragno, D. D. Caviglia, R. Morasso, Flehap: A wind powered supply for autonomous sensor nodes,
440 Journal of Sensor and Actuator Networks 5 (4) (2016) 15. doi:10.3390/jsan5040015.
441 URL http://www.mdpi.com/2224-2708/5/4/15
442 [29] S. L. Rosen, Fundamental principles of polymeric materials, Wiley, 1982.

SC
443 [30] S. Bagheri, A. Mazzino, A. Bottaro, Spontaneous symmetry breaking of a hinged flapping filament generates lift, Phys.
444 Rev. Lett. 109 (2012) 154502. doi:10.1103/PhysRevLett.109.154502.
445 URL http://link.aps.org/doi/10.1103/PhysRevLett.109.154502
446 [31] U. Lācis, N. Brosse, F. Ingremeau, A. Mazzino, F. Lundell, H. Kellay, S. Bagheri, Passive appendages generate drift
447 through symmetry breaking, Nature Communications 5 (2014) 5310.

U
448 [32] M. Argentina, L. Mahadevan, Fluid-flow-induced flutter of a flag, Proceedings of the National academy of Sciences of the
449 United States of America 102 (6) (2005) 1829–1834.
450 [33] S. Boi, A. Mazzino, J. O. Pralits, Minimal model for zero-inertia instabilities in shear-dominated non-newtonian flows,
AN
451 Phys. Rev. E 88 (2013) 033007. doi:10.1103/PhysRevE.88.033007.
452 URL http://link.aps.org/doi/10.1103/PhysRevE.88.033007
453 [34] I. Procaccia, V. S. L’vov, R. Benzi, Theory of drag reduction by polymers in wall-bounded turbulence, Rev. Mod. Phys.
454 80 (2008) 225–247. doi:10.1103/RevModPhys.80.225.
455 URL http://link.aps.org/doi/10.1103/RevModPhys.80.225
M

456 [35] G. Hou, J. Wang, A. Layton, Numerical Methods for Fluid-Structure Interaction - A Review, Communications in Com-
457 putational Physics 12 (2) (2012) 337–377. doi:10.4208/cicp.291210.290411s.
458 [36] W. D. Henshaw, K. K. Chand, A composite grid solver for conjugate heat transfer in fluid-structure systems, Journal of
459 Computational Physics 228 (10) (2009) 3708 – 3741. doi:http://dx.doi.org/10.1016/j.jcp.2009.02.007.
D

460 URL http://www.sciencedirect.com/science/article/pii/S0021999109000667


461 [37] G.-H. Cottet, E. Maitre, A semi-implicit level set method for multiphase flows and fluid-structure interaction problems,
462 Journal of Computational Physics 314 (2016) 80 – 92. doi:http://dx.doi.org/10.1016/j.jcp.2016.03.004.
URL http://www.sciencedirect.com/science/article/pii/S0021999116001546
TE

463
464 [38] R. Mittal, G. Iaccarino, Immersed boundary methods, Annual Review of Fluid Mechanics 37 (1) (2005) 239–261. arXiv:
465 http://dx.doi.org/10.1146/annurev.fluid.37.061903.175743, doi:10.1146/annurev.fluid.37.061903.175743.
466 URL http://dx.doi.org/10.1146/annurev.fluid.37.061903.175743
467 [39] OpenFOAM. The Open Source CFD Toolbox. User Guide (2015).
URL http://www.openfoam.org
EP

468
469 [40] B. Stroustrup, The C++ programming language, Pearson Education, 2013.
470 [41] F. Juretic, cfMesh v1.0.1. User Guide (2014).
471 URL http://www.c-fields.com
472 [42] H. K. Versteeg, W. Malalasekera, An introduction to computational fluid dynamics: the finite volume method, Pearson
Education, 2007.
C

473
474 [43] A. Dullweber, B. Leimkuhler, R. McLachlan, Symplectic splitting methods for rigid body molecular dynamics, The Journal
475 of Chemical Physics 107 (15) (1997) 5840–5851. doi:10.1063/1.474310.
AC

476 [44] H. Jasak, Z. Tukovic, Automatic mesh motion for the unstructured finite volume method, Transactions of FAMENA 30 (2)
477 (2006) 1–20.
478 [45] H. Jasak, H. Rusche, Dynamic mesh handling in openfoam, in: Proceeding of the 47th Aerospace Sciences Meeting
479 Including the New Horizons Forum and Aerospace Exposition, Orlando, FL, 2009.
480 [46] R. D. Blevins, Flow-induced vibration, New York, NY (USA); Van Nostrand Reinhold Co., Inc., 1990.
481 [47] A. Andersen, U. Pesavento, Z. J. Wang, Unsteady aerodynamics of fluttering and tumbling plates, Journal of Fluid
482 Mechanics 541 (2005) 65–90. doi:10.1017/S002211200500594X.
483 [48] G. J. Berman, Z. J. Wang, Energy-minimizing kinematics in hovering insect flight, Journal of Fluid Mechanics 582 (2007)
484 153–168. doi:10.1017/S0022112007006209.
485 [49] M. Bryant, J. C. Gomez, E. Garcia, Reduced-order aerodynamic modeling of flapping wing energy harvesting at low
486 reynolds number, AIAA journal 51 (12) (2013) 2771–2782. doi:10.2514/1.J052364.
487 [50] W. Huang, H. Liu, F. Wang, J. Wu, H. P. Zhang, Experimetal study of a freely falling plate with an inhomogeneous mass
488 distribution, Phys. Rev. E 88 (2013) 053008. doi:10.1103/PhysRevE.88.053008.
489 [51] R. LeVeque, Finite Difference Methods for Ordinary and Partial Differential Equations: Steady-State and Time-dependent

15
ACCEPTED MANUSCRIPT

490 Problems, Society for Industrial and Applied Mathematics, 2007.


491 [52] N. Razak, T. Andrianne, G. Dimitriadis, Bifurcation analysis of a wing undergoing stall flutter oscillations in a wind
492 tunnel, in: Proceedings of ISMA, 2010, pp. 3857–3602.

PT
RI
U SC
AN
M
D
TE
C EP
AC

16
ACCEPTED MANUSCRIPT

Highlights

• Updated presentation of a novel energy harvester from fluid-structure interactions


• Derivation of the natural frequency of oscillation both for ideal and real framework
• Prediction of the threshold for sustained flapping based on a resonance condition
• Verification by means of experiments, simulations and phenomenological model

PT
RI
U SC
AN
M
D
TE
C EP
AC

You might also like