Journal of Fluids and Structures 72 (2017) 96–113

Contents lists available at ScienceDirect

Journal of Fluids and Structures

journal homepage: www.elsevier.com/locate/jfs

Structural response and energy extraction of a fully passive

flapping foil
Zhuo Wang a , Lin Du a, *, Jisheng Zhao b , Xiaofeng Sun a
a
School of Energy and Power Engineering, Beihang University, Beijing 100191, China
b
Fluids Laboratory for Aeronautical and Industrial Research (FLAIR), Department of Mechanical and Aerospace Engineering,
Monash University, Melbourne, Victoria 3800, Australia

article info a b s t r a c t
Article history: The structural response and energy extraction of a foil undergoing two-degree-of-freedom
Received 26 January 2017 fully passive flapping motions in a two-dimensional flow are numerically investigated
Received in revised form 25 April 2017 at Re = 400. The simulations of the fluid–structure interaction were conducted using
Accepted 5 May 2017
the Immersed Boundary Method (IB Method). In the parametric space of flow reduced
Available online 18 May 2017
velocity and pivot location investigated, five response regimes are identified. This paper
focuses on the stable synchronisation regime, which is characterised by harmonic wake-
Keywords:
Flapping foil body synchronisation with stable large-amplitude oscillations. Correspondingly, a novel
Energy extraction wake pattern composed of a triplet of vortices and a pair of vortices shed per cycle, referred
Flow-induced vibration to as T+P pattern, is encountered. An analysis of the dynamic nonlinearity showed that the
inertia forces can induce perturbations in the form of harmonics to the dynamics of the
system. Furthermore, the highest cycle-averaged power output coefficient and efficiency
were found to be C̄P = 0.95 and η = 0.32, respectively. The present results suggest
that high-efficiency case in fully passive flapping motions is associated with a large pitch–
plunge phase and a 2S wake pattern composed of two strong single LEVs shed per cycle.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction

Flow energy harvester based on flapping foils, as a novel design motivated by rapidly increasing demand for renewable
energy, has been received a great deal of research attention in the last two decades. In contrast to conventional rotary
turbines that require the flow remaining attached to the blades for high aerodynamic performance and efficiency, flapping
foils utilise an aerodynamic mechanism involving leading-edge vortices (LEVs) to augment aerodynamic/hydrodynamic lift
and performance. Such an aerodynamic mechanism has been widely seen in insect/bird flights and aquatic locomotion in
nature (see Ellington et al., 1996; Srygley and Thomas, 2002; Birch et al., 2004). Inspired by bio-applications, flapping foils
are practically considered as an alternative to rotary turbines for energy harvesting (see McKinney and DeLaurier, 1981).
This has motivated extensive investigations that aim to characterise the fluid–structure mechanism and assess the energy
harvesting performance of flapping foils. Comprehensive reviews of this subject have recently been given by Xiao and Zhu
(2014) and Young et al. (2014).
Assuming that two degrees of freedom of plunge (or heave) and pitch are allowed, flapping foil flow energy harvesters
are generally classified by their activation mode into three categories (Xiao and Zhu, 2014; Young et al., 2014): (i) fully forced
system with motions that are fully prescribed in both plunge and pitch, (ii) semi-passive system with motions that usually

* Corresponding author.
E-mail address: lindu@buaa.edu.cn (L. Du).

http://dx.doi.org/10.1016/j.jfluidstructs.2017.05.002
0889-9746/© 2017 Elsevier Ltd. All rights reserved.
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 97

have prescribed pitching motion but free plunging motion, and (iii) fully passive system with free plunging and pitching
motions, fully determined by the fluid–structure interaction.
With its simplicity in modelling, a foil undergoing fully forced motions, which mostly are prescribed as sinusoidal, has
been extensively studied. Previous studies have shown that energy-extraction performance is strongly related to pivot
location, oscillation frequency and plunging amplitude, and the relative phase of pitching and plunging motions (see Kinsey
and Dumas, 2008; Platzer et al., 2010; Ashraf et al., 2011; Zhu, 2012; Xiao and Zhu, 2014). In a systematic study, Davids et al.
(1999) predicted a max efficiency of 0.30 and also found that the optimum phase reduced as the pivot was moved towards
the trailing edge. Kinsey and Dumas (2008) presented a mapping of energy-extraction efficiency in the parametric space of
◦
reduced frequency and pitching amplitude for a NACA0015 foil that had a fixed plunging amplitude, a fixed phase of 90 and
a given pivot location, in which the highest efficiency was found to be 0.34. Further, Zhu (2011) observed that the efficiency
was maximised when the imposed flapping frequency matched the most unstable frequency of the wake that was associated
with multiple LEVs, suggesting the feasibility of high-efficient energy extraction of a fully passive system.
Considerable attention has also been given to semi-passive systems that allow the foil to plunge freely while it undergoes
prescribed pitching oscillations (e.g. Zhu and Peng, 2009; Abiru and Yoshitake, 2011; Wu et al., 2015). Recently, Deng et
al. (2015) investigated the effects of reduced frequency, mass ratio and rotational inertia on the efficiency of a semi-passive
system. They observed a maximum efficiency of 0.34 with an optimum set parameters of reduced frequency and pitching
amplitude, and also found that the efficiency decreased monotonically as the mass ratio was increased.
Compared to fully forced or semi-passive systems, there has been much less work that investigates a fully passive system.
This is partly because it is challenging to experimentally or numerically model such a system that involves with complicated
problems of coupled fluid–structure interaction and nonlinear coupling in two-degree-of-freedom (2-DOF) flapping motions.
Similar to fully forced and semi-passive systems, Peng and Zhu (2009) found that the interaction between the LEV and the
foil can enhance energy harvesting performance in fully passive motions. Moreover, they suggested that the pivot location is
an important factor to the energy harvesting capacity. Simulations of fully passive motions by Platzer et al. (2010) reported
a maximum efficiency 0.30, which was well predicted by prescribed motions. Young et al. (2013) identified a broad region of
high efficiency with a maximum value of 0.294. It should be noted that, in the studies of Platzer et al. (2010) and Young et al.
(2013), their analysis is simplified by reducing the system from two DOFs (plunge and pitch) to one DOF (flywheel angle). The
◦
phase by which pitching motion leads plunge motion was fixed at 90 . Zhu (2012) suggested that full passive systems may
be much more adaptable to various real applications than fully forced or semi-passive systems. More recently, Veilleux and
Dumas (2017) achieved a peak efficiency 0.34 in an optimised fully-passive system at a high Reynolds number of 500 000,
which was enhanced by the adequate synchronisation between both plunging and pitching motions. These results indicate
that fully passive systems can achieve comparable energy extraction performance to fully forced or semi-passive systems.
While most previous studies of flapping foils assumed that the foil mass was negligible, investigations of flow-induced
vibrations (FIVs) of bluff bodies have shown that the mass ratio, defined as the ratio of the total mass of the oscillating system
to that of the displaced fluid, is an important parameter in the fluid-structural mechanism. FIVs of bluff bodies at low mass
ratio can exhibit a rich variety of dynamic characteristics and wake structures as a function of reduced velocity, which is
significantly different to high-mass-ratio case (see Govardhan and Williamson, 2000; Zhao et al., 2014a,b).
A clear gap in the literature is the lack of detailed knowledge of the fluid–structure interaction mechanism of a fully
passive flapping foil, including characteristics of vibration response, dynamic nonlinearity, and wake patterns. Therefore,
this paper specifically examines the following aspects: (i) regimes of the vibration response and corresponding flow patterns
in a parametric space of pivot location and reduced velocity, (ii) effects of the inertia forces on the dynamic nonlinearity, and
(iii) the performance of energy extraction.
The rest of this paper is structured as follows. The fluid–structure system modelling and the numerical method are
described in Section 2. The results and discussion, including the structural response, an analysis of the dynamic nonlinearity,
and the energy extraction performance are presented in Section 3. Finally, conclusions are drawn in Section 4.

2. Numerical approach

The fluid–structure system modelling is described in Section 2.1. Details concerning the dynamics of the foil are given in
Section 2.2. A brief description of the mechanism of power generation of the hydro-elastic system is presented in Section 2.3.
The numerical method employed for the flow dynamics is presented in Section 2.4 and its validation is given in Section 2.5.

2.1. Fluid–structure system

The current paper investigates the dynamic response and the energy transfer of an elastically mounted NACA0012 foil
in a free-stream, as shown in Fig. 1. The fluid density and dynamic viscosity, and the free-stream velocity are denoted by
ρ , µ and U, respectively. The foil is constrained to plunge (or heave) in the cross-flow direction and free to pitch about
a pivot point. The instantaneous plunge displacement is denoted by h(t), which can be non-dimensionalised by the foil
chord (c) as h∗ (t) = h/c. The pitch displacement is denoted by θ (t), and it is normally measured in radian in this study.
The instantaneous transverse fluid force and pitching moment are denoted by Fh (t) and Mθ (t), respectively. The structural
damping coefficient in plunge is designated by ch . The structural stiffness constants are designated by kh and kθ in plunge
98 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Fig. 1. Schematic of the fluid–structure system studied: a NACA0012 foil is allowed to undergo 2-DOF fully passive plunging and pitching motions.

and pitch, respectively. The distance from the pivot point to the leading edge is denoted by xp , and its nondimensional form
is denoted by x (0 ⩽ x ⩽ 1), as given by (1), which is an important parameter in the present study.

x = xp /c . (1)

2.2. Governing equations of motions

The hydro-elastic system studied is considered as a linear mass–spring–damper system, in which the plunging and
pitching motions of the foil are governed by the second-order oscillator Eqs. (2) and (3), respectively, as follows:

mḧ + ch ḣ + kh h − mb cos θ · θ̈ + mb sin θ · θ̇ 2 = Fh (t) (2)

Iθ θ̈ + cθ θ̇ + kθ θ − mb cos θ · ḧ = Mθ (t). (3)
Here, m denotes the mass of the foil. In the present paper, the mass ratio is set equal to 2.0, of the same order of O(2) reported
in previous studies on flow-induced vibration bluff bodies (see Khalak and Williamson, 1997; Zhao et al., 2014b). b denotes
the distance between the pivot point and the mass centre. The nondimensional distance between the mass centre and the
leading edge is 0.4603 for the present NACA0012 foil. Iθ denotes the inertia moment. The structural damping coefficients ch
and cθ are initialised to zero. The stiffness constants in plunge and pitch can be determined by kh = 4π 2 U 2 m/(cU ∗ )2 and
kθ = 4π 2 U 2 Iθ /(cU ∗ )2 , respectively, to achieve the same reduced velocity in both plunge and pitch. The reduced velocity is
defined by U ∗ = U /(fN c), in which fN is the natural frequency of the system in plunge.
The transverse lift and the pitching moment are computed using (4) and (5), and their non-dimensional forms for per
unit foil hereby span can be written as (6) and (7), respectively:
∫
Fh (t) = Fy ds (4)
∫Γ
Mθ (t) = (Fy yr − Fx xr ) ds (5)
Γ
( )
1
CFh = Fh (t)/ ρU 2c (6)
2
( )
1
C Mθ = Mθ (t)/ ρ U 2 c 2 . (7)
2
In order to provide a better understanding of the nonlinear coupled effect of the dynamics, it is of significant interest
to consider the inertial forcing components associated with the plunging and pitching motions. In the left-hand side of (2),
the products of −mb cos θ · θ̈ and mb sin θ · θ̇ 2 represent the inertial forces due to the angular and centripetal accelerations
related to the pitching motion, respectively. Similarly, in the left-hand side of (3), the product of −mb cos θ · ḧ represents
the inertial moment component acting in the pitching direction due to the plunging acceleration. The coefficients of these
three inertial forcing components are defined by (8)–(10) as follows:
( )
1
CIFAθ = −mb cos θ · θ̈/ ρU 2c , (8)
2
( )
1
CIFC θ = mb sin θ · θ̇ 2 / ρU 2c , (9)
2
( )
1
CIMh = mb cos θ · ḧ/
A
ρU c .
2 2
(10)
2
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 99

2.3. Mechanism of power generation

Power extraction of a flapping foil is typically implemented through a damper in plunge with nonzero damping while
the damping in pitching is negligible. As such, the instantaneous power output can be determined by (11) and its coefficient
is defined by (12) below, in which ch denotes the damping coefficient of the damper:

P = ch · ḣ2 , (11)
( )
1
CP = P / ρU 3c , (12)
2
ch∗ = ch /(ρ cU). (13)
The damping coefficient can be normalised by (13). Correspondingly, the coefficients of instantaneous power transmitted to
the foil by the transverse lift and pitching moment are defined by (14) and (15), respectively, as follows:
( )
F 1
CP h = Fh (t) · ḣ/ ρU c ,
3
(14)
2
( )
Mθ 1
CP = Mθ (t) · θ̇ / ρU 3c . (15)
2
Assuming the plunging and pitching motions are periodic, the mean power output and the corresponding contributions
by Fh and Mθ over one cycle time (T ) can thus be non-dimensionally evaluated by (16)–(18):
∫ t +T
1
C̄P = CP (t) dt , (16)
T t
∫ t +T
F 1 F
C̄P h = CP h (t) dt , (17)
T t
∫ t +T
Mθ 1 M
C̄P = CP θ (t) dt . (18)
T t

The efficiency of energy extraction, denoted by η, is measured as the ratio of the mean power output (P̄) to the power
available (Pa ) in the oncoming flow passing through the frontal area swept by the foil (see Kinsey and Dumas, 2008), as in
(19):
P̄ P̄ c
η= = = C̄P , (19)
Pa 1
2
ρU 3d d
where d is the largest distance swept by the foil undergoing both plunging and pitching motions.

2.4. The incompressible flow equations

In the present simulations, the flow remains two-dimensional over the parametric space investigated; therefore, the
flow dynamics is solved by employing the two-dimensional Navier–Stokes equations. The numerical method employed is
the same as in the previous studies of Du et al. (2016a, b) concerning the fluid–structure interaction in turbomachinery. The
numerical method, boundary conditions and numerical constructions have been detailed and validated in these previous
studies, and so will only be briefly described here.
For convenience, the continuity and momentum equations are written in non-dimensional forms:
{
∇ · V⃗ = 0
(20)
∂ V⃗ /∂ t + (V⃗ · ∇ )V⃗ = F⃗ − ∇ p + (∇ · ∇ )V⃗ /Re,
in which Re denotes the Reynolds number, V⃗ = (u, v ) denotes the fluid velocity, and p denotes the pressure. The Reynolds
number, defined by Re = ρ Uc /µ, was set equal to 400 in the present study. The flow equations were discretised on
orthogonal staggered grids and solved by the so-called ‘‘SIMPLE’’ method developed by Patankar and Spalding (1972)
and Patankar (1981).
In order to characterise the interaction of the fluid and the boundary, the body force F⃗ = (Fx , Fy ) was calculated using the
immersed boundary method. For no-slip boundary condition, it can be treated as a process of negative feedback:
∫ t
⃗f (xk , yk , t) = α v⃗ (xk , yk , t ′ ) − v⃗o (xk , yk , t ′ ) dt ′ + β [v⃗ (xk , yk , t) − v⃗o (xk , yk , t)] ,
[ ]
(21)
0

where v⃗f and v⃗o are the velocities of the fluid and the body, respectively, at the kth surface element (xk , yk ), t is the time
instance, and α , β are the feedback factors. It should be noted that great values of these factors can lead to sensitive response
100 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Table 1
Parametric settings of the present case study of a NACA0015 foil undergoing
forced plunging and pitching oscillations.
Parameter Setting value
A∗h 1.0
◦
A∗θ 1.33 (76.3 )
fh∗ 0.14
fθ∗ 0.14
◦
φ 90
x 1/3
Re 1100

Table 2
Mesh resolutions and the R.M.S values of transverse lift coefficients and
pitching moment coefficients in one cycle at Re = 1100, respectively.
Mesh ∆x ∆t CFrms
h
rms
CMθ

A 0.019531 0.0002 1.5707 0.2906

B 0.009466 0.0002 1.4494 0.2611
C 0.009466 0.0001 1.4834 0.2625
D 0.004883 0.0002 1.4342 0.2545

of the foil, which may also easily lead to unexpected divergence in solving the unsteady equations. In order to conveniently
solve the flow equations, the body force described in the Lagrangian form as in (21) was converted into the Euler coordinates
using the Dirac function, as expressed by (22):
∫
F⃗ (x, y, t) = ⃗f (xk , yk , t)δ (x − xk )δ (y − yk )ds, (22)
Γ
where (x, y) represents a point in the Cartesian coordinates and Γ depicts the rigid boundary. The numerical construction
of the δ functions followed that detailed in Peskin (2002), where a continuous and segmented function was used to replace
the singular Dirac function (23):
0 |r | ⩾ 2
⎧
⎪
⎨(
⎪ √ )
5 − 2| r | − − 7 + 12| r | − 4r 2 /8 1 ⩽ |r | ⩽ 2
Φr = (23)
⎪ ( √ )
3 − 2|r | + 1 + 4|r | − 4r 2 /8 0 ⩽ |r | ⩽ 1,
⎪
⎩

r = ∆x/∆h and ∆h being the mesh width.

2.5. Validation of the numerical method

To provide a further validation of the numerical method, a study of forced oscillations was conducted to directly compare
with the previous work of Kinsey and Dumas (2008), where they investigated the power extraction of a NACA0015 foil
undergoing sinusoidal plunging and pitching oscillations with prescribed amplitudes and frequencies over a parameter space
at a pivot location of c /3 and Reynolds number of Re = 1100.
The motion profiles were set exactly matching those of Kinsey and Dumas (2008), where they observed an optimal
efficiency of power extraction of η = 0.34. Table 1 shows the parameter settings for this validation, in which the plunging
amplitude normalised by the foil chord is A∗h = 1.0, the pitching amplitude in radian is A∗θ = 1.33, the reduced frequencies
of plunging and pitching, defined by fh∗ = fh c /U and fθ∗ = fθ c /U, respectively, are equal to 0.14, and the relative phase of the
◦
pitching motion with respect to the plunging motion (for convenience, as the pitch–plunge phase hereafter) is φ = 90 .
Firstly, the convergences of the solutions at four different grid sizes and time steps are validated due to the parameter
setting in Table 1. The mesh resolutions and their root-mean-square(R.M.S) values of transverse lift coefficients and pitching
moment coefficients in one cycle, respectively, are shown in Table 2, and their time traces are shown in Fig. 2. As can be
found from Table 2 and Fig. 2, the difference between the load predictions by mesh B, C and D are small. Considering the
computational accuracy and efficiency, mesh B in Table 2 is chosen for the computation at Re = 1100, and its result is used
to make the comparison between Re = 1100 and Re = 400.
As shown in Fig. 2, the present results agree well with Kinsey and Dumas (2008), with the time traces of CFh and CMθ
matching for one oscillation period. It should be noted that the present simulations were precisely conducted with the
computational domain size of 20c × 40c.
To investigate the effect of Reynolds number on the hydrodynamic characteristics and flow pattern, simulations were
also performed at a lower Reynolds number of Re = 400, with grid sizes and time steps setting in Table 3. By comparing
the rms values of the transverse lift and pitching moment of different mesh resolutions in one cycle, as well as their time
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 101

Fig. 2. Comparisons of the time traces of the transverse lift coefficient in (a) and the pitching moment coefficient in (b) at Re = 1100. The grey dashed lines
represent the results of Kinsey and Dumas (2008) and other lines represent the results in our work with different mesh resolutions. Red dashed lines with
triangles: mesh A. Black dash lines with circles: mesh B. Green dash lines with squares: mesh C. Blue solid line: mesh D. Detail information about mesh
A–D are show in Table 2.

Fig. 3. Comparisons of the time traces of the transverse lift coefficient in (a) and the pitching moment coefficient in (b) at Re = 1100 and Re = 400 of the
present work. The grey dash lines represent the results at Re = 1100, and others lines represent results at Re = 400 with different mesh resolutions. The
black dash lines with triangle: mesh A. The blue solid line: mesh B. The red dash lines with circle: mesh C. The green dash lines with square: mesh D. Detail
information about mesh A–D are show in Table 3.

Table 3
Mesh resolutions and the R.M.S values of transverse lift coefficients and
pitching moment coefficients in one cycle at Re = 400, respectively.
Mesh ∆x ∆t CFrms
h
rms
CMθ

A 0.039063 0.0004 1.5272 0.2398

B 0.019531 0.0004 1.4592 0.2495
C 0.019531 0.0002 1.4956 0.2483
D 0.009766 0.0002 1.4166 0.2461

traces in Fig. 3, the numerical convergence could be validated for Re = 400. For Re = 400, mesh B in Table 3 is able to
provide satisfactory accuracy. Clearly, Fig. 3 also shows that the hydrodynamic characteristics are highly similar, with the
time traces of CFh and CMθ matching closely, for Re = 400 and 1100. This implies that the flow structures (i.e. the LEV
and the near-wake vortices), which are strongly related to the hydrodynamic characteristics, are also similar. These results
indicate that the Reynolds number has little influence on the hydrodynamic characteristics and vortex structures with no
strong turbulence or other three-dimensional features observed in this regime, and thereby the flow can be regarded as
two-dimensional. Simulations were therefore carried out at Re = 400, which ensured a two-dimensional flow over the
parameter space investigated in the present study.

3. Results and discussion

3.1. Regimes of the vibration response

The structural vibration response of the foil undergoing fully passive plunging and pitching oscillations is investigated in
this subsection.
The vibration response of the foil is plotted in the (x, U ∗ ) parameter domain in Fig. 4. Five different vibration regimes,
as illustrated by ①-⑤ in the figure, are identified based on the amplitude responses of plunging and pitching oscillations.
Correspondingly, the representative points labelled by A–E for each regime are examined with their time traces shown in
Fig. 5.
For regime ①, the vibration response is characterised as in a stable state. Initially, the foil was set at zero angle of attack
(AoA), and then a small sinusoidal perturbation with A∗θ less than 0.001 as a ‘‘kick-start’’ was applied in the pitch direction
to speed up the convergence of simulations. However, the kick-start perturbation can only cause extremely-low-amplitude
oscillations in both plunge and pitch at the starting stage, as shown in Fig. 5(a). Then, the oscillations soon dissipate and the
foil remains stationary at its initial position.
102 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Fig. 4. Dynamic response as a function of the reduced velocity and the pivot point location. The regimes labelled by ①–⑤ represent five different vibration
responses of the foil. The black circular (•) markers denote the representative cases of the five zones considered in Fig. 5. The open square (□) markers
represent the cases analysed for energy extraction in Fig. 15.

Fig. 5. Time traces of the plunging and pitching oscillations of the representative locations for the five regimes shown in Fig. 4. The black solid lines and the
red dash lines represent the plunging motion and the pitching motion, respectively: (a) point A at (x, U ∗ ) = (0.33, 1.5), (b) point B at (x, U ∗ ) = (0.43, 1.1),
(c) point C at (x, U ∗ ) = (0.43, 1.4), (d) point D at (x, U ∗ ) = (0.55, 1.6) and (e) point E at (x, U ∗ ) = (0.33, 4.0).

Regime ② is identified as a transition regime between ① and ③. Notably, this regime has a very narrow extended region
bounded between regimes ① and ③ over the range of 0.4 ⩽ x ⩽ 1.0. In this regime, highly periodic oscillations with
low amplitudes but high frequencies are encountered. The oscillations in some cases also show chaotic characteristics. The
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 103

Fig. 6. Offsetting behaviours in the transition regime ②. In (a), the offsets are due to the initial perturbation at (x, U ∗ ) = (0.50, 1.0). In (b), the centre
position of oscillations shifts regardless of the initial perturbation at (x, U ∗ ) = (0.33, 2.5).

maximum amplitudes are observed to be A∗h = 0.48 and A∗θ = 0.33 at (x, U ∗ ) = (0.33, 2.5). However, the vibration response
is sensitive to the initial perturbation and the pivot point location.
As shown in Fig. 6, the oscillations are about a non-zero offset position, while the offsets are observed to be dependent
on the initial perturbation. This is illustrated in Fig. 6(a). In this case, the signs of the offsets are dependent on that of
the perturbation. For instance, a positive initial perturbation applied in the pitch leads to a positive offset in the pitching
oscillations and vice versa; however, the magnitudes of the time–displacement profiles are identical. Furthermore, the centre
position of oscillations, regardless of the initial perturbation, may shift suddenly over time, as illustrated in Fig. 6(b). Similarly,
Peng and Zhu (2009) studied the passive oscillation of a 2-dof foil and it is found that the foil tended to oscillate around an
offset position when the reduced velocity was high or the pivot point was located close to the trailing edge. In the present
study, the offsetting was persistently observed with the pivot point located close to the trailing edge (x > 0.29). The above
results indicate that regime ② is a transition regime, where the vibration response is sensitive to the initial conditions and
the pivot location.
Regime ③ is identified as a stable synchronisation regime, in which the wake-body synchronisation occurs with periodic
large-amplitude plunging and pitching oscillations. To distinguish from the pitch-over regime ④, one criterion of this regime
is defined as the pitching amplitude under the pitch-over threshold of A∗θ = π/2. As can be seen, this regime is bounded by
the transition regime ② and the pitch-over regime ④. Unlike regime ②, the structural response in this regime is observed to be
independent of the initial perturbation. The plunging amplitude response is observed to be in the range of 0.48 ⩽ A∗h ⩽ 1.11
and the pitching amplitude response of 0.33 ⩽ A∗θ < π/2. As illustrated by the representative point C shown in Fig. 5(c),
the body oscillations clearly appear highly periodic but with asymmetric profiles, suggesting the presence of harmonics
and potentially an asymmetric wake pattern in this regime. For further discussion, the corresponding frequency spectra and
harmonic synchronisation are presented in Section 3.1.2; the wake pattern is shown in Section 3.1.3; furthermore, a detailed
analysis of the dynamic non-linearity is given in Section 3.2; the performance of energy extraction of the cases considered
within the regime is given in Section 3.3.1.
Regime ④ is identified as the pitch-over regime, where large pitching amplitudes of A∗θ > π/2 occur intermittently or
persistently. In general, the onset of pitch-over tends to occur at lower U ∗ as x increases, as shown in Fig. 4. Nevertheless,
it is observed that the lowest reduced velocity for the presence of pitch-over is U ∗ = 1.25 with x = 0.708, below which
no pitch-over behaviour is observed for the (x, U ∗ ) space investigated. Like regime ③, as illustrated by the representative
point D shown in Fig. 5(d), the body oscillations also appear highly periodic but asymmetric in this regime, suggesting the
presence of harmonic frequencies and an asymmetric wake pattern. The corresponding frequency spectra and the wake
pattern observed are presented in Section 3.1.2 and Section 3.1.3, respectively. With limited investigations for such a wide
regime, the minimum plunging amplitude response was observed to be A∗h = 0.44 at (x, U ∗ ) = (0.85, 1.4). In this paper,
we do not focus on this regime because it may be difficult to achieve such large-scale oscillations in real applications due
to limitations in mechanism design, although there may exist interesting features. However, further investigations are still
required in the future to quantitatively characterise this regime.
Regime ⑤ is identified as a quasi-steady regime. This narrow regime, bounded by the steady regime ① and the pitch-over
regime ④, tends to occur at high reduced velocity (U ∗ ≳ 4.0) with the pivot point close to the leading edge (x < 0.37). The
structural vibration response is characterised by highly periodic (sinusoidal-like) oscillations with very low amplitudes (i.e.
both A∗h and A∗θ are less than 0.01), as shown in Fig. 5(e). Despite the body oscillating, the flow remains unseparated from the
body surface, as shown in Fig. 9(c), and the system is therefore regarded as in a quasi-steady state.

3.1.1. Effect of pivot location

The plunging and pitching amplitudes play an important role in the performance of energy harvesting, since efficiently
extracting large amount of energy from the flow may require large structural vibrations. In this subsection, the effect of pivot
location on the vibration amplitude responses is investigated.
Fig. 7 shows the maximum plunging and pitching amplitudes about corresponding time-averaged positions as a function
of the reduced velocity over a range of 0 ⩽ U ∗ ⩽ 4.0 for five different pivot locations, x ∈ {0.33, 0.37, 0.50, 0.67, 0.85}.
These pivot locations and the reduced velocity range were chosen such that light can be shed on the large-scale responses
in regime ③.
104 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Fig. 7. The maximum amplitude responses as a function of the reduced velocity at five different pivot locations. In (b), the grey area depicts the pitch-over
zone.

For x = 0.33, the large amplitudes that correspond to regime ③ tend to increase gradually over 2.6 ⩽ U ∗ ⩽ 2.9, with
the peak values of A∗h = 0.68 and A∗θ = 0.56 observed at U ∗ = 2.9. At the lower reduced velocities, 1.8 ⩽ U ∗ ⩽ 2.5, the
vibration response was characterised as in regime ②, with the local peak amplitudes of A∗h = 0.48 and A∗θ = 0.33 observed
at U ∗ = 2.5 before jumping to regime ③. At the high velocities (3.0 ⩽ U ∗ ⩽ 3.5) investigated, on the other hand, the system
jumps sharply to regime ①, where extremely low amplitudes are encountered.
For x = 0.37, the structural response that corresponds to regime ③ is observed at U ∗ = 1.7 and 2.1. Compared to the case
of x = 0.33, the regime transition of ②→③ occurs at a much lower reduced velocity, U ∗ ≈ 1.7. Furthermore, the amplitude
peaks that are within regime ③ are observed to be A∗h = 1.11 and A∗θ = 1.12 at U ∗ = 2.1, which are about 63% and 100%
larger than those of x = 0.33, respectively. With x = 0.37, however, the vibration response is very sensitive to the reduced
velocity: A∗h and A∗θ increase sharply for the range of 1.77 ⩽ U ∗ ⩽ 2.2 investigated here; furthermore, around U ∗ = 2.1, a
slight increase in U ∗ may lead to pitch-over.
For x = 0.50 and 0.60, the regime transition of ② → ③ occurs around U ∗ = 1.0, where A∗θ increases abruptly. The plunging
amplitude peaks that are within regime ③ are observed to be 0.98 at U ∗ = 1.7 for x = 0.50 and 0.41 at U ∗ = 1.2 for x = 0.67.
In these two cases, pitch-over is encountered at U ∗ values much lower than the case of x = 0.37.
For x = 0.85, which is close to the trailing edge, the regime transition of ② → ③ occurs around U ∗ = 1.2, where A∗θ surges
to reach the pitch-over zone at U ∗ = 1.4. The peak plunging amplitude prior to pitch-over is observed to be A∗h = 0.40 at
U ∗ = 1.3.
These results show that the pivot location has a significant impact on the regime transitions of ② → ③ and ③ → ④.
Furthermore, large-scale plunging and pitching amplitudes of regime③ are observed at x = 0.37 and 0.50, suggesting that
this range of pivot location is suitable for efficiency of energy extraction of a fully passive foil.

3.1.2. Synchronisations
Many previous studies (e.g. Govardhan and Williamson, 2000; Zhao et al., 2014b; Bourguet and Jacono, 2014) concerning
free flow-induced vibrations of bluff bodies have shown that transitions in vibration amplitude response are associated
with wake-body synchronisations. Compared to bluff bodies, there has been much less attention paid to free vibrations of
a flapping foil, particularly the fluid–structure mechanism that involves structural response transitions. In order to provide
insights into the regime transitions, a frequency spectrum analysis is given in this subsection.
A striking phenomenon persistently observed in the large-amplitude regimes ③ and ④ is harmonic synchronisation, where
the harmonic frequencies of the body oscillations match those of the fluid forcing. To illustrate this, the frequency spectral
analyses of h(t), θ (t), Fh (t) and Mθ (t) of the corresponding representative points C and D, based on the power spectral density
(PSD) of the time series, are shown in Fig. 8, noting that the PSD is denoted by Pf and the corresponding frequencies are
normalised by the first harmonic of the plunging oscillations (denoted by fh1 ).
For regime ③, the spectra of point C shown in Fig. 8(a-i)–(a-iv) clearly reveal that identical harmonic frequencies appear in
the body oscillations and fluid forcing, with the first harmonic (fh1 ) being the dominant and the second harmonic (2fh1 ) being
the second strongest in power. For regime ④, the spectra of point D shown in Fig. 8(b-i)–(b-iv) also reveal the occurrence
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 105

Fig. 8. Power spectral density of the plunging and pitching oscillations and the corresponding hydrodynamic forcing components for the representative
points of regimes ③ and ④ in the (x, U ∗ ) domain. (a) point C at (x, U ∗ ) = (0.43, 1.4) in regime ③, and (b) point D at (x, U ∗ ) = (0.55, 1.6) in regime ④.

of identical harmonics in the body oscillations and fluid forcing, with the first harmonic being the dominant and the third
harmonic (3fh1 ) being the second strongest in power.
The large-amplitude responses in these regimes can be attributable to the harmonic synchronisations. A recent study
by Zhao et al. (2014b) on the transverse flow-induced vibration (FIV) of a square cylinder has shown that at an asymmetric
◦
angle of attack of 20 , the cylinder experiences a jump to the higher branch of amplitude response with the largest oscillation
amplitudes up to 1.2 times of the frontal projected width of the body, which is attributable to the harmonic synchronisation
of the body oscillations and fluid forces. Similarly, Bourguet and Jacono (2015) observed in their numerical simulations that
the wake-body synchronisation that involved with harmonics matching, as opposed to galloping vibrations, could lead to
large-scale structural vibrations of a rotating circular cylinder. To the authors’ knowledge, such wake-body synchronisations
have not previously been reported for a fully passive flapping foil.

3.1.3. Flow patterns

The flow patterns of an oscillating body in a free-stream are intrinsically coupled with the body dynamics. Previous studies
have shown that a rich variety of vortex shedding patterns can be seen in the near wake of a cylindrical body undergoing
free or forced vibrations (see Williamson and Roshko, 1988; Govardhan and Williamson, 2000; Carberry et al., 2012; Morse
and Williamson, 2009; Du et al., 2014; Zhao et al., 2014b; Bourguet and Jacono, 2014). It is not clear, however, what flow
patterns will appear in free vibrations of a flapping foil. In this subsection, the flow patterns associated with the vibration
response regimes are examined by considering the representative points, with particular attention paid to regimes ③ and ④,
where harmonic synchronisations are encountered.
The flow patterns associated with the low-amplitude regimes, namely ①, ② and ⑤, are shown in Fig. 9. As can be
seen, the flow remains attached smoothly to the foil surface in the steady and quasi-steady regimes where extremely low
106 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Fig. 9. Flow patterns of the representative points of the low-amplitude regimes ①, ② and ⑤: (a) point A at (x, U ∗ ) = (0.33, 1.5) in the steady regime ①, (b)
point E at (x, U ∗ ) = (0.33, 4.0) in the quasi-steady regime ⑤, and (c) point B at (x, U ∗ ) = (0.43, 1.1) in the transition regime ②. The normalised vorticity
range is ω∗ = ωc /U ∈ [−1, 1], in which ω is the out-of-plane vorticity.

Fig. 10. The T+P wake pattern being formed during one cycle at point C, (x, U ∗ ) = (0.43, 1.4). ω∗ ∈ [−1, 1].

amplitudes are encountered; as expected, on the other hand, regime ②, where the body oscillations are sufficient to cause
pressure fluctuations that lead to flow separations, sees a periodic vortex shedding pattern forming a Kármán vortex street
downstream.
For regime ③, the existence of asymmetric wake pattern, as suggested by the frequency spectra, is clearly evident in
the near wake, as shown in Fig. 10. As can be seen, a wake pattern comprising vortices grouped in one triplet (T) and
one additional pair (P) is being formed during one cycle. In the first half cycle shown in Fig. 10(a)–(c), as the foil strokes
with strong plunging acceleration from the top to the lowest position, one pair of vortices composed of one strong positive
(anticlockwise) leading-edge vortex (LEV) and one very weak negative (clockwise) trailing edge vortex (TEV) are being
shed off around the position with highest plunging velocity (ḣ(t)). During this downstroke, one strong negative LEV is
simultaneously being formed around the foil upper surface. Further, in the second half cycle shown in Fig. 10(d)–(f), as
the foil strokes upward, a triplet of vortices, including the previous negative LEV, are being formed and then shed off around
the peak plunge and pitch positions. During this upstroke, the plunging and pitching accelerations are relatively low, due
to the strong interaction of the body and the near-body flow structures, particularly the strong positive LEV that is being
formed around the foil bottom surface. This wake pattern may be referred to as T+P, following the terminology introduced
in Williamson and Roshko (1988) and Jauvtis and Williamson (2004). The pattern of 2T (two triplets) has been observed
by Jauvtis and Williamson (2004) in experiments on 2-DOF VIV of a circular cylinder, and the pattern of T+S (one triplet
plus one single) has been observed by Bourguet and Jacono (2014) in numerical simulations of transverse FIV of a rotating
circular cylinder; however, to the authors’ knowledge, the T+P pattern has not previously been reported. This asymmetric
wake pattern can result in harmonic frequency components, thereby leading to harmonic synchronisation if the frequency
of vortex shedding is locked to that of the body oscillations.
For regime ④, like regime ③, the harmonic synchronisation is also associated with an asymmetric wake pattern, as shown
in Fig. 11. This wake pattern is identified as P+S, comprising vortices grouped in one pair and one single shed per cycle.
As shown in Fig. 11(a)–(c), a negative LEV is being formed as the foil strokes downward from the top position. During
this downstroke, a positive vortex is simultaneously being formed on the upper body surface, strongly interacting with
the negative LEV. Around the lowest position, these two vortices are shed off as a pair, then rolling in the downstream.
Further, during the upstroke, a positive LEV is being formed and shed off around the top position. In previous experiments
of forced vibrations of a circular cylinder by Williamson and Roshko (1988) and Morse and Williamson (2009), P+S pattern
was observed to be associated with amplitudes larger than 0.8D (D being the cylinder diameter). More recently, Bourguet
and Jacono (2014) observed P+S pattern with amplitudes larger than 1.5D in free vibrations of a rotating cylinder. Similarly,
the P+S pattern is associated with the large-scale oscillations in the pitch-over regime.
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 107

Fig. 11. The P+S wake patter being formed during one cycle at point D, (x, U ∗ ) = (0.55, 1.6). ω∗ ∈ [−1, 1].

Fig. 12. Time traces of the coefficients of fluid forcing components and inertia forces of point C in (a) column and point D in column (b). In (ai) and (bi),
the black solid lines represent CFh , the grey dashed lines represent CIFAθ , and the blue dot-dashed lines represent CIFC θ ; in (aii) and (bii), the black solid lines
A
represent CMθ and the grey dashed lines represent CIM .
h

3.2. Analysis of the dynamic nonlinearity

While the system is modelled as a linear mass–spring–damper system that is governed by the linear equations of motion
(2) and (3), the dynamic nonlinearity may be caused by the following two aspects: the coupling between the 2-DOF plunging
and pitching motions, and the hydrodynamic load. This subsection focuses on these two aspects.
As can be seen in (2) and (3), the mismatching distance of the mass centre and the pivot point results in the inertia force
components that directly contribute to the coupling of the plunging and pitching motions, which essentially affects the
dynamic response of the fluid–structure system. To provide insights into the coupling of the 2-DOF motions, the structural
nonlinearity is thus investigated by analysing the resulted inertia force components at the representative points C and D of
regimes ③ and ④.
Fig. 12 shows the corresponding time traces of the fluid forcing and inertia force coefficients. As can be seen, the inertia
forces induced by the pitching motion (CIFAθ and CIFC θ ) are much lower than the transverse lift (CFh ), particularly the inertia
force due to the centripetal acceleration of pitching motion being close to zero (CIFC θ ∼ 0), suggesting that the inertia forces
induced by the pitching motion have a minor impact on the plunging motion; on the other hand, however, the inertia force
A
(CIM ) induced by the plunging acceleration is of the same order of magnitude as the pitching moment (CMθ ), indicating that
h
the pitching moment is significantly affected by the plunging acceleration.
Furthermore, the frequency spectra of these force components are shown in Figs. 13 and 14. As can be seen in Fig. 13,
the inertia forces of CIFAθ and CIFC θ exhibit similar frequency components as CFh , while their power, as expected from the their
magnitudes, are much lower than that of CFh . In particular, the dominant frequency of CIFC θ is its first harmonic as the same
as CFh , whereas the dominant frequency of CIFAθ appears to be the component that is of the second strongest in power of CFh
A
(i.e. the second harmonic at point C, and the third harmonic at point D). On the other hand, the dominant frequency of CIM
h
108 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Fig. 13. Frequency spectra of the hydrodynamic force (CFh ) and the inertial forces (CIFAθ and CIFC θ ) induced by the pitching motion for point C in column (a)
and point D in column (b).

A
Fig. 14. Frequency spectra of the hydrodynamic moment (CMθ ) and the inertial force (CIM ) induced by the plunging acceleration for point C in column (a)
h
and point D in column (b).

appears in the same way as CIFAθ , matching the component that is of the second strongest in power of CMθ (Fig. 14), while the
A
harmonics of CIM exhibit the same order of power magnitude as CMθ .
h
The above results reveal that the inertia forces can induce perturbations in the form of harmonics to the dynamics of the
system. As pointed out by He (1992) that only the low harmonics tend to have strong nonlinear interaction, the present
frequency spectra show that the low harmonics, particularly the dominant frequency, of the inertia forces induced by the
pitching motion have a considerable impact on the dynamics of the plunging motion and vice versa, thereby enhancing the
dynamic nonlinearity.
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 109

Fig. 15. The energy harvesting performance as a function of the normalised damping coefficient for different pivot locations: (a) the time-averaged power
output coefficient, (b) the plunging amplitude response, and (c) the energy harvesting efficiency.

3.3. Energy harvesting performance

In order to find the optimal parameters for efficient energy extraction in regime ③, the energy harvesting performance
of the flapping foil is investigated for different x and U ∗ values over a range of the damping coefficient imposed in plunging
motion. An overall assessment of the harvesting capacity is presented in Section 3.3.1. The effect of the pitch–plunge phasing
is analysed in Section 3.3.2. The efficiency as a function of the pitching frequency and amplitude is examined in Section 3.3.3.

3.3.1. Energy harvesting capacity

In Fig. 15, the plunging amplitude response (A∗h ), the cycle-averaged power output coefficient (C̄P ), and the energy-
extraction efficiency (η) are plotted as a function of the normalised damping coefficient (ch∗ ) for different (x, U ∗ ) values.
These (x, U ∗ ) values were selected in such a way that the assessment can be conducted for a wide range of pivot locations,
while the structural vibration response (with zero damping) can remain within regime ③ but also be considerably close to
the pitch-over regime.
As can be seen in the figure, the case of (x, U ∗ ) = (0.35, 3.0) exhibits the greatest A∗h , C̄P and η values of all the cases. In
this case, while A∗h decreases substantially over the range of 0.3 ⩽ ch∗ ⩽ 1.5, C̄P increases at the lowest ch∗ values to reach a
peak of 0.95 at ch∗ = 0.5 and then declines as ch∗ increases further, whereas η, increases gradually to reach a peak of 0.32 at
ch∗ = 0.8. For ch∗ > 0.8, all of A∗h , C̄P and η decrease gradually as ch∗ increases, except the range 1.5 ⩽ ch∗ ⩽ 1.8, where the
efficiency show a slight increase. Nevertheless, η can remain above 0.25 for the ch∗ range investigated, suggesting that the
pivot location of x = 0.35 provides a robust capability for high power output and efficiency.
The results also show that the energy harvesting performance is sensitive to the pivot location in the vicinity of x = 0.35.
This would be expected from the amplitude responses shown in Fig. 7. As shown in Fig. 15, the peak of C̄P increases from
a very low value to 0.95 as x increases from 0.33 to 0.35, and then decreases considerably with η as x increases to higher
110 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Table 4
Peak values of the cycle-averaged power output coefficients at different pivot locations.
F Mθ F M
Case C̄P C̄P h C̄P C̄P h /C̄P C̄P θ /C̄P
x = 0.33, U ∗
= 2.9, ch∗
= 1.0 0.062 0.064 −0.002 1.032 −0.032
x = 0.37, U ∗ = 2.1, ch∗ = 0.5 0.670 0.684 −0.014 1.021 −0.021
x = 0.50, U ∗ = 1.6, ch∗ = 2.5 0.174 0.172 −0.002 0.989 −0.011
x = 0.67, U ∗ = 1.2, ch∗ = 4.0 0.150 0.148 −0.002 0.987 −0.013

Fig. 16. Time traces of the flapping motions and pitch–plunge phase of high- and low-efficiency cases. In (a), the efficiency is found to be η = 0.30 with
◦ ◦
φ̄ = 64 at (x, U ∗ ) = (0.37, 2.1) and ch∗ = 1.5; in (b), the efficiency is found to be η = 0.08 with φ̄ = 8 at (x, U ∗ ) = (0.50, 1.6) and ch∗ = 1.5. In (ai) and
(bi), the black solid lines and the red dashed lines represent the plunging and pitching motions, respectively.

values (x > 0.37). In a global trend, while A∗h tends to decrease as ch∗ increases, both C̄P and η in each case of (x, U ∗ ) tend to
increase over a certain range of ch∗ and then decrease gradually for higher ch∗ values.
To provide insights into the contributions of the fluid forcing components, their cycle-averaged power output coefficients
are evaluated at different pivot locations. Table 4 shows a detailed evaluation of the power outputs contributed by Fh (t) and
Mθ (t) at x ∈ {0.33, 0.37, 0.50, 0.67}. Note that positive values of the coefficients in the table indicate positive energy transfer
from the fluid to the foil, while negative coefficient values indicate negative energy transfer from the foil to the fluid. From
these cases, it can be observed that the net energy transfer is primarily achieved through the plunging motion, while the
energy transfer associated with the pitching motion is negligible. In practice, however, Mθ (t) is coupled with the plunging
motion, thereby significantly impacting the dynamic characteristics of plunging motion for energy extraction.

3.3.2. The effect of the pitch–plunge phase

The phase between pitching and plunging motions has a crucial impact on the energy extraction efficiency of a flapping
foil. Many previous studies (e.g. Kinsey and Dumas, 2008) of prescribed motions have been conducted on the basis that high
◦
efficiency is achieved with large motions when the pitch–plunge phase is fixed at 90 . In the case of a fully passive system,
however, the flapping motions are induced by the flow, and thereby the phase, determined by the dynamics of the system, is
self-sustained. This subsection examines the relationship of the phase and efficiency for two high- and low-efficiency cases
of a fully passive foil.
Fig. 16 shows time traces of the plunging and pitching motions and their relative phase at (x, U ∗ ) = (0.37, 2.1) as the
high-efficiency case (η = 0.30, the highest efficiency in the present study) and (x, U ∗ ) = (0.50, 1.6) as the low-efficiency
case (η = 0.08). In the high-efficiency case, the foil undergoes highly periodic oscillations with A∗h ≈ 0.5 and A∗θ ≈ 1.2,
while the instantaneous phase, calculated using the Hilbert Transform (HT), fluctuates slightly about the time-averaged
◦
mean of φ̄ = 64 . This fluctuation is due to the fact that the flapping motions contain several harmonic components. In the
low-efficiency case, on the other hand, although the foil oscillations are also highly periodic with A∗h ≈ 0.3 and A∗θ ≈ 1.5,
◦
the instantaneous phase fluctuates largely about the time-averaged mean of φ̄ = 8 .
◦
These results show that the highest efficiency is observed to be associated with a large pitch–plunge phase of φ̄ = 64 .
◦
However, this phase angle is considerably lower than those (∼ 90 ) optimised in forced flapping motions. While the reasons
for this discrepancy are not presently clear, there are two possible reasons. First, it might not be possible to achieve such
◦
a large phase of φ̄ ≈ 90 in a self-sustained system; in other words, the largest phase achievable, determined by the
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 111

Fig. 17. Formation of the 2S wake pattern at (x, U ∗ ) = (0.37, 2.1) and ch∗ = 1.5. The efficiency of energy extraction is found to be η = 0.30 with φ̄ = 64◦ .
ω∗ ∈ [−1, 1].

Fig. 18. Formation of the 2P wake pattern at (x, U ∗ ) = (0.50, 1.6) and ch∗ = 1.5. The efficiency of energy extraction is found to be η = 0.08 with φ̄ = 8◦ .
ω∗ ∈ [−1, 1].

◦
system itself, would be much lower than 90 . Second, the phase associated with high efficiency is a function of the pivot
location. Davids et al. (1999) found in his studies of forced motions that the optimum phase reduced substantially as the
◦ ◦
pivot was moved towards the trailing edge (i.e. φ = 100 at x = 0.0, φ = 80 at x = 0.5). Further investigations to better
understand this are necessary in the future.
Furthermore, significant differences in the near-wake patterns are also observed between these two cases. As shown in
Fig. 17, the highest efficiency case is associated with a well-defined 2S wake pattern (comprising two opposite-sign single
vortices shed per cycle). During the downstroke shown in Fig. 17(a)–(c), a negative LEV is shed off around the lowest pitch
position with very high plunging speed, attaching on the foil and strongly interacting with the shear layer from the bottom
leading edge. Further, during the upstroke shown in Fig. 17(d)–(f), a positive LEV is shed off around the highest plunge
position. Clearly, the pitching motion leads the plunging motion to some degree during the evolution of this 2S wake pattern.
The low-efficiency case, on the other hand, is associated with a 2P wake pattern (comprising two pairs of opposite-sign
vortices shed per cycle). As shown in Fig. 18(a)–(b), a negative LEV and a positive TEV are being formed during the downstroke
and then shed off as one pair around the lowest plunging and pitching positions. Further, a positive LEV and a negative TEV
are being formed during the upstroke (Fig. 18(c)–(f)) and then shed off around the highest positions. As can be seen, each
pair is composed of one LEV and one TEV in opposite signs, and the signs alternate in the following pair. Clearly, the pitching
motion appears as in phase with the plunging motion during this vortex shedding cycle.
The present results show that the highest efficiency case is associated with a 2S wake pattern composed of two opposite-
sign single LEVs, whereas the low-efficiency case is associated with a 2P pattern composed of two pairs of opposite-sign LEV
and TEV. This finding agrees well with the experimental observations of forced flapping motions in Hover et al. (2004) that
the highest efficiency is associated with a 2S wake pattern. However, it should be noted that Simpson (2009) (from the same
group as Hover et al. (2004)) reported an opposite behaviour that the highest efficiency cases were associated with a 2P
pattern, whereas low-efficiency cases were associated with a 2S pattern. Nevertheless, it is clear that the phase is strongly
related to the wake pattern, thereby significantly impacting on the efficiency. Moreover, the strong foil–vortex interaction
has positive effect on the energy harvesting performance. Similar results were also found by Veilleux and Dumas (2017).
112 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113

Fig. 19. The highest efficiency of four pivot location cases plotted on the contours of a NACA0015 foil undergoing prescribed sinusoidal motions by Kinsey
and Dumas (2008). The present cases are represented by the symbols as follows. ◀: η = 0.32 at (x, U ∗ ) = (0.35, 3.0) and ch∗ = 0.5; : η = 0.30 at
(x, U ∗ ) = (0.37, 2.1) and ch∗ = 1.5; ▽: η = 0.28 at (x, U ∗ ) = (0.38, 2.0) and ch∗ = 1.8; : η = 0.26 at (x, U ∗ ) = (0.40, 1.7) and ch∗ = 2.0; □: η = 0.08 at
(x, U ∗ ) = (0.50, 1.6) and ch∗ = 4.0.

3.3.3. The efficiency as a function of (Aθ , fθ∗ )

One aim of this paper is to address the question whether fully passive flapping motions achieve as high efficiency as
forced motions. For this purpose, Fig. 19 shows a direct comparison by plotting the present results on the efficiency contours
in (Aθ , fθ∗ ) space of a forced flapping foil by Kinsey and Dumas (2008).
As shown in the figure, the efficiency of present cases investigated is comparable to the results of Kinsey and Dumas
(2008). The highest efficiency is observed to be η = 0.32 with fθ∗ = 0.15 at (x, U ∗ ) = (0.35, 3.0) in the present study, 0.02
lower than that of η = 0.34 at fθ∗ = 0.14 in Kinsey and Dumas (2008). The induced pitching amplitude at highest efficiency
is slightly larger than 90◦ while the prescribed pitching amplitude for optimal efficiency in Kinsey and Dumas (2008) ranges
in 70◦ –80◦ . This discrepancy could be due to differences in the phase, the Reynolds numbers, etc. However, it should be
noted that the cases of (x, U ∗ ) = (0.37, 2.1) and (x, U ∗ ) = (0.40, 1.7) well match the contour level of the forced motions.
This comparison would be helpful for setting parametric conditions (i.e. the pivot location, the damping coefficient) for a
fully passive foil to achieve high efficiency.

4. Conclusions

The structural response of a foil undergoing two-degree-of-freedom fully passive plunging and pitching motions in a
two-dimensional flow has been numerically investigated over a parametric space of the flow reduced velocity and the
pivot location. The simulations of the fluid–structure interaction were conducted using the Immersed Boundary Method
at Re = 400. In the parametric domain investigated, (0, 0) ⩽ (x, U ∗ ) ⩽ (1.0, 7.0), five regimes of structural vibration were
identified: ➀ as the steady regime, ➁ as the transition between ① and ③, ③ as the stable synchronisation regime, ④ as the
pitch-over regime, and ⑤ as the quasi-steady regime. In particular, harmonic synchronisations occur in regimes ③ and ④,
resulting in large-scale body oscillations. The results showed that the amplitude responses are very sensitive to the pivot
location in the vicinity of x = 0.35. It is clear that the reduced velocity and the pivot location have a significant impact on
the structural response, leading to complex dynamic nonlinearity.
From the frequency spectrum analyses for the representative points in ③ and ④, it is found that the harmonics of the
flapping motions and fluid forcing components synchronised in these regimes. Correspondingly, different wake patterns
were observed in these regimes. For regime ③, a novel wake pattern composed of a triplet of vortices and a pair of vortices
shed per cycle, referred to as T+P pattern, was identified. For regime ④, a P+S wake pattern, composed of a pair of vortices
and a single vortex shed per cycle, was encountered. This suggests that the system exhibits a rich variety of flow patterns in
the parametric space.
Furthermore, the dynamic nonlinearity was also analysed for regimes ③ and ④. It is found that the structural nonlinearity
that arises from the mismatch of the mass centre and the pivot point has a significant impact on the inertia forces
and moment in plunging and pitching motions, respectively. Consequently, these inertia forces and moment can induce
perturbations in the form of harmonics to the dynamics of the system, thereby enhancing the dynamic nonlinearity.
Energy extraction performance was assessed with different pivot locations in the stable synchronisation regime ③. The
assessment showed that the pivot location of x = 0.35, at which large-amplitude oscillations were encountered, had a robust
capability for high power out coefficient and efficiency. The highest cycle-averaged power output coefficient and efficiency
were found to be C̄P = 0.95 and η = 0.32, respectively. It is confirmed that the net power generation is primarily through the
plunging motion; however, the pitching motion is important in the sense that it is coupled with the plunging motion, thereby
impacting the dynamic characteristics of the fluid–structure system. The pitch–plunge phase was analysed using the Hilbert
 Z. Wang et al. / Journal of Fluids and Structures 72 (2017) 96–113 113

Transform. The results showed that in the high-efficiency case of η = 0.30 at (x, U ∗ ) = (0.37, 2.1), the phase fluctuated
◦
slightly about the time-averaged mean of φ̄ = 64 , whereas in the low-efficiency case of η = 0.08 at (x, U ∗ ) = (0.50, 1.6),
◦
the phase largely fluctuated about the time-averaged mean of φ̄ = 8 . Correspondingly, it was observed that the high-
efficiency case was associated with a 2S wake pattern composed of two opposite-sign single LEVs shed per cycle, whereas
the low-efficiency case was associated with a 2P pattern composed of two pairs of opposite-sign LEV and TEV. Consistent
with previous studies, the present results reinforce that LEVs can enhance the performance of energy extraction. A direct
comparison of the present results to a previous study of forced flapping motions presented shows that comparable efficiency
can be achieved for a fully passive flapping foil under certain parametric (i.e. x, U ∗ ) conditions.

Acknowledgement

The authors acknowledge financial support from the National Natural Science Foundation grants 51236001 and
11661141020.

References

Abiru, H., Yoshitake, A., 2011. Study on a flapping wing hydroelectric power generation system. J. Environ. Eng. 6 (1), 178–186.
Ashraf, M., Young, J., S. Lai, J.C., 2011. Numerical analysis of an oscillating-wing wind and hydropower generator. AIAA J. 49 (7), 1374–1386.
Birch, J.M., Dickson, W.B., Dickinson, M.H., 2004. Force production and flow structure of the leading edge vortex on flapping wings at high and low Reynolds
numbers. J. Exp. Biol. 207 (7), 1063–1072.
Bourguet, R., Jacono, D.L., 2014. Flow-induced vibrations of a rotating cylinder. J. Fluid Mech. 740, 342–380.
Bourguet, R., Jacono, D.L., 2015. In-line flow-induced vibrations of a rotating cylinder. J. Fluids Mech. 781, 127–165. http://dx.doi.org/10.1017/jfm.2015.477.
Carberry, J., Sheridan, J., Rockwell, D., 2012. Controlled oscillations of a cylinder: forces and wake. J. Fluids Mech. 538 (2005), . http://dx.doi.org/10.1017/
S0022112005005197.
Davids, S.T., et al., 1999. A computational and experimental investigation of a flutter generator. (Ph.D. thesis), Monterey, California: Naval Postgraduate
School.
Deng, J., Teng, L., Pan, D., Shao, X., 2015. Inertial effects of the semi-passive flapping foil on its energy extraction efficiency. Phys. Fluids 27 (5).
Du, L., Jing, X., Sun, X., 2014. Modes of vortex formation and transition to three-dimensionality in the wake of a freely vibrating cylinder. J. Fluids Struct.
49 (2), 554–573.
Du, L., Sun, X., Yang, V., 2016a. Generation of vortex lift through reduction of rotor/stator gap in turbomachinery. J. Propul. Power 32 (2), 472–485.
Du, L., Sun, X., Yang, V., 2016b. Vortex-lift mechanism in axial turbomachinery with periodically pitched stators. J. Propul. Power 32 (2), 1–14.
Ellington, C.P., van den Berg, C., Willmott, A.P., Thomas, A.L., 1996. Leading-edge vortices in insect flight. Nature 384, 626–630.
Govardhan, R., Williamson, C. H.K., 2000. Modes of vortex formation and frequency response of a freely vibrating cylinder. J. Fluids Mech. 420, 85–130.
He, L., 1992. Method of Simulating Unsteady Turbomachinery Flows with Multiple Perturbations. AIAA J. 30 (11), 2730–2735.
Hover, F.S., Haugsdal, Ø., Triantafyllou, M.S., 2004. Effect of angle of attack profiles in flapping foil propulsion. J. Fluids Struct. 19(1), 37–47. http://dx.doi.
org/10.1016/j.jfluidstructs.2003.10.003.
Jauvtis, N., Williamson, C. H.K., 2004. The effect of two degrees of freedom on vortex-induced vibration at low mass and damping. J. Fluids Mech. 509, 23–62.
http://dx.doi.org/10.1017/S0022112004008778.
Khalak, A., Williamson, C. H.K., 1997. Fluid forces and dynamics of a hydroelastic structure with very low mass and damping. J. Fluids Struct. 11 (8), 973–982.
Kinsey, T., Dumas, G., 2008. Parametric study of an oscillating airfoil in a power-extraction regime. AIAA J. 46 (6), 1318–1330.
McKinney, W., DeLaurier, J., 1981. The wingmill: an oscillating-wing windmill. J. Energy 5 (2), 109–115.
Morse, T.L., Williamson, C.H.K., 2009. Prediction of vortex-induced vibration response by employing controlled motion. J. Fluid Mech. 634, 5.
Patankar, S., Spalding, D., 1972. A calculation procedure for heat, mass and momentum transfer in three-dimensional parabolic flows. Int. J. Heat Mass
Transfer 15 (10), 1787–1806.
Patankar, S.V., 1981. A calculation procedure for two-dimensional elliptic situations. Numer. Heat Transfer 4 (4), 409–425.
Peng, Z., Zhu, Q., 2009. Energy harvesting through flow-induced oscillations of a foil. Phys. Fluids 21 (12), 123602.
Peskin, C.S., 2002. The immersed boundary method. Acta Numer. 11, 479–517.
Platzer, M., Ashraf, M., Young, J., Lai, J., 2010. Extracting Power in Jet Streams: Pushing the Performance of Flapping-Wing Technology. In: 27th International
Congress of the Aeronautical Sciences, Nice, France. pp. 19–24.
Simpson, B.J., 2009. Experimental studies of flapping foils for energy extraction. (Master’s thesis), Massachusetts Institute of Technology.
Srygley, R.B., Thomas, A. L.R., 2002. Unconventional lift-generating mechanisms in free-flying butterflies. Nature 420 (December), 487–489.
Veilleux, J.-C., Dumas, G., 2017. Numerical optimization of a fully-passive flapping-airfoil turbine. J. Fluids Struct. 70, 102–130.
Williamson, C.H.K., Roshko, A., 1988. Vortex formation in the wake of an oscillating cylinder. J. Fluids Struct. 2(4), 355–381.
Wu, J., Chen, Y.L., Zhao, N., 2015. Role of induced vortex interaction in a semi-active flapping foil based energy harvester. Phys. Fluids 27 (9).
Xiao, Q., Zhu, Q., 2014. A review on flow energy harvesters based on flapping foils. J. Fluids Struct. 46, 174–191.
Young, J., Ashraf, M.A., Lai, J. C.S., Platzer, M.F., 2013. Numerical simulation of fully passive flapping foil power generation. AIAA J. 51 (11), 2727–2739.
Young, J., Lai, J.C., Platzer, M.F., 2014. A review of progress and challenges in flapping foil power generation. Prog. Aerosp. Sci. 67, 2–28.
Zhao, J., Leontini, J.S., Jacono, D.L., Sheridan, J., 2014a. Chaotic vortex induced vibrations. Phys. Fluids 26 (12), 121702.
Zhao, J., Leontini, J.S., Lo Jacono, D., Sheridan, J., 2014b. Fluid–structure interaction of a square cylinder at different angles of attack. J. Fluid Mech. 747,
688–721.
Zhu, Q., 2011. Optimal frequency for flow energy harvesting of a flapping foil. J. Fluid Mech. 675, 495–517.
Zhu, Q., 2012. Energy harvesting by a purely passive flapping foil from shear flows. J. Fluids Struct. 34, 157–169.
Zhu, Q., Peng, Z., 2009. Mode coupling and flow energy harvesting by a flapping foil. Phys. Fluids 21 (3).