Experimental Thermal and Fluid Science 74 (2016) 58–72

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science

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

Flow-induced vibration of an elastically mounted airfoil under the

influence of the wake of a circular cylinder
J.F. Derakhshandeh ⇑, M. Arjomandi, B. Dally, B. Cazzolato
School of Mechanical Engineering, University of Adelaide, Adelaide, South Australia 5005, Australia

a r t i c l e i n f o a b s t r a c t

Article history: The effect of vortices generated by a rigidly mounted cylinder on the dynamic response of an airfoil is
Received 17 March 2015 investigated in this study. This work extends previous investigation of vortex interaction with a second
Received in revised form 5 November 2015 cylinder to capture the wake energy (Derakhshandeh et al. 2014). Accordingly, the Flow-Induced
Accepted 5 December 2015
Vibration (FIV) of a symmetric NACA 0012 airfoil positioned in the wake of the rigidly-mounted upstream
Available online 12 December 2015
cylinder is studied. The airfoil mount allows it to move in two degrees-of-freedom; pitch and heave. The
pitching axis is kinematically driven using a brushless permanent-magnet DC servo motor. The heave axis
Keywords:
of the airfoil is coupled with a Virtual Elastic Mechanism (VEM), which is an electro-mechanical device
Flow-induced vibration
Vortex shedding frequency
designed to create any desired impedance. The VEM system replaces the physical damper and spring sys-
Circulation tems and allows the airfoil to oscillate due to the lift force in the normal direction of the flow. The airfoil is
Airfoil set at different positions in the wake of the upstream cylinder in order to characterise the impact of the
Circular cylinder arrangement of a coupled cylinder–airfoil on energy extraction. Special attention was paid to the angle of
attack of the airfoil to explore the optimal performance of the system. Force and displacement measure-
ments of the airfoil were conducted in a closed loop water channel. The experimental work was compli-
mented by a Computational Fluid Dynamics (CFD) modelling study which was aimed at calculating the
vortex frequency shedding and visualize the vortex structure. It is observed that the vortex shedding fre-
quency, the length scale and transverse spacing of the vortices are function of the longitudinal and lateral
distances between the cylinder the airfoil. The results also demonstrate that there is a correlation
between the configuration of cylinder and airfoil and vortex structure. Due to this correlation, the shear
forces acting on the airfoil alters the fluttering response of the airfoil depending on the angle of attack,
which in turn influences the obtained power coefficient of the device. The maximum power coefficient
of FIV is obtained for cases with 3.5 6 x0/D 6 4.5 and 1 6 y0/D 6 1.5 arrangements, which is limited to
the narrower lateral distances as compared to the previous study (1 6 y0/D 6 2) by the authors
(Derakhshandeh et al. 2014) employing a pair of cylinders.
Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction demand of electricity around the world. One such renewable

source is ocean energy which can be harnessed using turbine and
It has been predicted that the growth rate in energy consump- non-turbine convertors [20].
tion over the next 20 years is much higher than the growth rate of Vortex-Induced Vibration (VIV) phenomenon, known as a non-
population due to the incremental increase in demand for electric- turbine convertor [20], has been extensively studied in the past
ity in developing countries [33]. Consequently, the production of [30,36,39,4,19,18,31]. In this phenomenon, the fluid-structure
electricity would increase from 20 Petawatt hours in 2010 to interaction occurs due to the synchronization between structure
31.2 Petawatt hours in 2030 [33]. Considering that electricity pro- and vortex shedding. A previous investigation by Bernitsas et al.
duction using fossil fuel is responsible for producing the largest [6] has shown that the VIV convertor is feasible both technically
amount of carbon dioxide emission (26%) [33], it is essential that and economically. It has been found that the theoretical and exper-
alternative sources of energy to be sought to cater for the increased imental efficiencies of the VIV convertor were estimated to be 37%
and 22%, respectively [5]. In addition, it was shown that the Rey-
nolds number plays a more important role than the mass damping
⇑ Corresponding author. Tel.: +61 8 8313 1124. ratio. In addition, Bernitsas et al. [6] have noted that VIV convertors
E-mail address: [email protected] are easily scalable and can operate under different Reynolds num-
(J.F. Derakhshandeh).

http://dx.doi.org/10.1016/j.expthermflusci.2015.12.003
0894-1777/Ó 2015 Elsevier Inc. All rights reserved.
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 59

ber, which makes them applicable for a wide variety of 2. Dynamic model
applications.
An oscillating airfoil might be considered as a non-turbine con- This section outlines the mathematical model used for an elas-
vertor. This convertor has drawn the attention of scholars to avoid tically mounted airfoil with two degrees of freedom.
the problems of rotary turbines. Although the rotary river turbines
can be employed in shallow waters, the wake interference reduces 2.1. Mathematical model
the performance of an array of the turbines [9]. In addition, the
river turbines generally suffer from drawbacks such as starting tor- To study the oscillation of an airfoil under the effect of oncom-
que and low efficiencies, particularly at low flow speeds [20]. Con- ing vortices, the flutter behaviour of an aero-elastic airfoil is anal-
sequently, employing airfoils with oscillating behaviour, known as ysed based on the schematic model shown in Fig. 1. Point ‘p’ is of
flutter, was proposed [25] to capture kinetic energy of the fluid. interest in this model and refers to the aerodynamic centre of
Harnessing the available kinetic energy from wind using an the system. This point also represents the reference point, where
oscillating airfoil was initially suggested by McKinney and DeLau- the heave of the airfoil ‘h’ is measured and it is located at the
rier [25]. It has been shown that the measured efficiency of a flut- quarter-chord of the airfoil xp. By choosing this point, the effect
tering airfoil is approximately 28%, which is comparable to the of torque is eliminated and the airfoil is only under the influence
experimental efficiencies of the rotary devices which range from of the lift force. The heaving and pitching of the airfoil are
20% to 55% depending on the efficiency of the convertors [35]. restrained by two springs with the constants of kh and ka, respec-
The efficiency of oscillation-based converters using an airfoil is tively, and damped with dampers with damping constants bh and
shown to be higher than the measured efficiency of the VIV with ba.
22% when employing a circular cylinder [6]. Along with the exper- In order to find the total virtual work of the airfoil, initially, the
imental tests of a fluttering airfoil mechanism, the numerical anal- equations of motion for an airfoil with two degrees of freedom and
ysis of Zhu and Peng [28] showed unexpectedly that the efficiency with large displacement are formulated as [32]
of 34% was achievable when the normalized oscillation frequency
mh €  cos a  Sa  a_ 2 sin a þ dh h_ ¼ F L ðtÞ;
€ þ k  h þ Sa a ð1Þ
was around 0.15 (the frequency was normalized by the chord h

length of the airfoil and the free stream velocity). Further investi-
€ cos a þ Ia a
Sa h € þ ka  a þ da  a_ ¼ MðtÞ; ð2Þ
gations have been conducted by Zhu and Peng [28] and Zhu [40]
using numerical modelling to capture kinetic energy from flapping
where h, h_ and h € are the transverse displacement, velocity and
foils in a uniform flow. Zhu and Peng [28] have found that the
acceleration of the airfoil (m, m/s and m/s2), respectively, a is the
behaviour of the airfoil is dependent on the stiffness of the system.
It has also been shown that the location of the pitching axis plays angle of attack of the airfoil (rad), a_ and a
€ represent the angular
an important role in the dynamic response of the airfoil. Depending velocity and acceleration of the airfoil (rad/s and rad/s2), m repre-
on the location of the pitching axis and the rotational spring stiff- sents the mass of the airfoil (kg), kh is the spring stiffness (N/m),
ness, four distinguishable responses have been offered by the Sa is the static moment (kg m), dh denotes the structural damping
authors, namely: static, periodic, irregular and oscillation motions. in bending (kg/s), da is the structural damping in torsion (kg m2/
Among these responses, they postulated that the periodic motion s rad), Ia represents the mass moment (kg m2), ka is the spring stiff-
of the airfoil has the highest feasibility of energy harvesting. ness (N m/rad), FL is the fluid force acting on the airfoil perpendic-
While a few numerical studies have shown the potential of a ular to the flow speed (N), and M represents the moment of the
fluttering airfoil to generate useful power, the lack of examination system (N m).
of the fluttering response of an airfoil in capturing hydrokinetic To achieve the total virtual work, both the kinetic and potential
energy has motivated the current experiments. In practice, two energies as well as the resulting aerodynamic forces acting on the
types of flutter response for an airfoil can be considered: purely surface of the airfoil are considered. The potential and kinetic ener-
passive and semi-active mechanisms. In the purely passive flutter- gies of the airfoil can be written as follow [15,16]:
ing, the airfoil will be entirely excited by Flow-Induced Vibration 1 1
kh h þ ka a2 ;
2
P¼ ð3Þ
(FIV) such as those reported by Sváček et al. [32]. On the other 2 2
hand, in the semi-activated system only one of the dynamic modes,
such as heave or pitch response can be excited by the cyclic acting 1 1
K¼ mV 2c þ Ic a_ 2 ; ð4Þ
forces due to the vortices, while the other mode can be applied to 2 2
the airfoil. In the latter case, the semi-actuated design requires an where Ic is the moment of inertia about C and Vc is the velocity of
activator system to make the specified pitching motion, for the centre of gravity ‘c’ (refer to Fig. 1), which can be found by
instance, virtually. The semi-actuated mechanism enables to opti-
mise desirable displacement amplitude of the airfoil using a con-
trol system, which is suitable for the energy harvester.
In this study, a Semi-Active Virtual Elastic Mechanism (SAVEM)
system is employed instead of a real physical spring damper. The
use of SAVEM system can rapidly open the way for optimisation
and further development of energy utilization from FIV. The
SAVEM system consists of two motors and two controllers, a
belt-pulley transmission, a carriage and an airfoil, which allows
the movement of the airfoil in two degrees of freedom. In order
to identify stable conditions in the wake of the cylinder in current
study, the effect of longitudinal and lateral distances between the
cylinder and the airfoil are analysed. In addition, with combination
of heaving and pitching motions of the airfoil to capture hydroki-
netic energy of oncoming vortices, the effect of wake instability
and the arrangement of the coupled cylinder–airfoil are also Fig. 1. Schematic section of a coupled circular cylinder and elastic airfoil with two
investigated. degrees of freedom including pitch and heave.
60 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

V c ¼ V p þ a_ ^J 3  b^J 1 : ð5Þ with a single cylinder and a coupled cylinder–airfoil. In the current
study, 2D time-dependent, incompressible Navier–Stokes equa-
In Eq. (5), ^J axes represent coordinate system mounted on the tions were solved using a symmetry preserving Finite-Volume
reference point ‘p’ and Vp is the initial velocity of the reference scheme of second order spatial and temporal accuracy. The station-
point ‘p’ and it can be calculated as V p ¼ h_ ^I2 ; hence, the velocity ary cylinder was mounted in a bounded domain with the stream-
of the centre of mass is wise spacing of 25D. Furthermore, 9D spacing was created on each
side of the cylinder to remove the blockage effects. The blockage
V c ¼ h_ ^I2 þ ba_ ^J 2 : ð6Þ ratio is calculated based on the projected area of the cylinder and
cross-section area. The smooth surface was assumed for the cylin-
where ^I axes are the fixed unit vectors (refer to Fig. 1). By substitut-
der. The boundary conditions for all numerical models were set as:
ing the velocity of the mass into Eq. (4), the kinetic energy is sum-
(1) the uniform flow was specified at the inlet with U = U1. (2) No-
marized as
slip conditions were applied on the cylinder and airfoil surfaces. (3)
1 1 Both upper and lower sides of the boundary were set as free a
mðh_ 2 þ b x2a a_ 2 þ 2bxa h_ a_ Þ þ Ic a_ 2 :
2
K¼ ð7Þ
2 2 boundary which means both of them are free from any shear or
pressure forces. (4) The outlet boundary was set with zero differen-
The displacement of the airfoil can be obtained by integrating
tial pressure.
the velocity at point ‘p’ e.g. (V p ¼ h_ ^I2 ); Thus, It is worth noting that, time-dependent, incompressible Navier–
Ddp ¼ Dh^I2 ; ð8Þ Stokes equations were solved using a symmetry preserving Finite-
Volume scheme of second order implicit and temporal accuracy.
where h presents heave of the airfoil based on point ‘p’ (refer to The time step was set to Dt = 0.01 s, which is small enough to sat-
Fig. 1). isfy the Courant–Friedrichs–Lewy (CFL) condition (CFL < 1), and is
In the next step, to calculate the work done by the aerodynamic essential for convergence of the simulations. In addition, the con-
lift, the displacement of the airfoil for point ‘p’ is considered. As vergence criteria were set to a maximum residual of 105. Further-
mentioned earlier, the aerodynamic centre of the airfoil is adapted more, the lift and drag coefficients of the cylinder were monitored
by the elastic axis (point ‘p’) at the quarter of the chord length of during the solution process and the stability of these coefficients
the airfoil. It is worth noting that in the current study, the aerody- was taken into account to accompany the convergence criterion.
namic centre of the airfoil is fixed. However, due to the effect of the Flow around a stationary single cylinder at Re = 10,000, was ini-
upstream wake, the centre of pressure can be different from the tially modelled using ANSYS Fluent Workbench. At this Reynolds
aerodynamic centre of the airfoil which can change the acting number, the flow is classified in intermediate Transition in Shear
moment on the airfoil. Here, the generated moment at this point Layers (TrSL) [39]. At TrSL flow regime, in which all numerical
is assumed to be zero (M1/4  0). models and experiments were performed in this paper, the discrete
Therefore, the virtual work of the lift force during one cycle of periodic vortices are formed and accordingly the Strouhal number
oscillation can be written as remains approximately constant [7]. This facilitates the harnessing
of the energy from the vortices, which are generated at a constant
W ¼ F L ðdp Þ; ð9Þ
frequency, which is close to lock-in. In order to predict the wake
Here, the aerodynamic lift force FL acting in the normal direc- regimes, a Scale Adaptive Simulation (SAS) turbulence model was
tion to the flow with two degrees of freedom can be formulated as used based on the Finite Volume Method (FVM) with a pressure

 based algorithm. In comparison to other turbulence models, the
_ a_
F L ¼ qU 2 blcla a þ h=U þ ðb=2Þ : ð10Þ SAS model is a relatively new model and was developed by Menter
U and Egorov [26]. This model uses the Von Karman length scale,
By integrating the right hand side of Eq. (10) and averaging it which allows it to adapt its behaviour to Scale Resolving Simula-
over a complete cycle of oscillation, the mean generated power tion (SRS) according to the stability parameters of the flow. This
during one cycle of vibration (Tcyl) can be found as also allows the model to provide a balance between the contribu-
tions of the simulated and resolved parts of the turbulence stres-
W ses. Hence, the model can effectively and automatically switch
PFIV ¼ : ð11Þ
T cyl from the Large Eddy Simulation (LES) mode to the Reynolds Aver-
The power coefficient of FIV can be calculated based on the age Navier Stokes (RANS) mode [26]. In previous studies, the
power of the fluid Pfluid, which is the product of the acting force authors have shown that this model is suitable for predicting the
VIV response of a cylinder mounted in a transient flow regime
on the airfoil (12 qU 2 AC L ¼ qU 2 blC L ) and the velocity in the same
[10,11]. In the current study, in order to evaluate the numerical
direction of the force (U). Here, l is the length or span of the airfoil;
model of the flow around a circular cylinder, the method of Grid
A is the area and CL is the lift coefficient. Therefore, the power of
Convergence Index (GCI) recommended by Celik et al. [8] were uti-
the fluid can be written as Pfluid = qU3bl and the power coefficient
lized, which is based on the Richardson extrapolation. Table 1 com-
can be written as
pares the numerical results of three mesh elements, including the
PFIV drag and root mean square lift coefficients as well as the Strouhal
gFIV ¼ : ð12Þ number at Re = 10,000. The density of the mesh elements around
Pfluid
the cylinder is shown in Fig. 2. As summarized in Table 1, no con-
siderable differences were observed for the integral quantities such
3. Numerical analysis of the wake as mean drag and lift coefficients between Test Cases M2 and M3. In
addition, the difference between the Strouhal numbers for these
In order to analyse the vortex structure in the wake area of the two test cases is about 0.5%. As a consequence, the refined quadri-
cylinder, a numerical analysis can be employed. Therefore, with lateral Mesh Case M2 was selected for the study of the wake power
the purpose of visualizing the flow pattern and studying the effect density of the cylinder.
of critical parameters that influence FIV, such as the vortex length Table 1 also compares the obtained numerical results with the
scale, the transverse spacing of the vortices and the vortex fre- available published experimental and numerical parameters. Com-
quency shedding, preliminary numerical analyses were conducted parison between the results of Mesh Case M2 and Direct Numerical
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 61

Table 1
Mesh refinement sensitivity for flow around a circular cylinder at Re = 10,000.

Mesh case Mesh quality Mesh elements CD C Lr:m:s St

M1 Coarse 35,244 1.254 0.642 0.176
M2 Refined 51,666 1.135 0.511 0.198
M3 Highly refined 60,398 1.133 0.502 0.199
Dong and Karniadakis [14] (DNS results at Re = 10,000) 1.143 0.448 0.203
Gopalkrishnan [17] (experimental results at Re = 10,000) 1.186 0.384 0.193

Fig. 2. The density of mesh in proximity of the circular cylinder. The detail of each test case is summarized in Table 1.

Simulation (DNS) obtained by Dong and Karniadakis [14] reveals mental results of Lau [23]. It can be seen from the figure that the
that the drag coefficient and Strouhal number are in a good agree- calculated vorticity distribution in the wake of the cylinder is quite
ment. It is also observed that there is approximately 14% difference similar to that reported experimentally. Consistent with the previ-
between the lift coefficients (based on root mean square). In gen- ous investigations Roshko [29], Unal and Rockwell [34] and Lau
eral, it can be seen that there is a large difference between the lift [23], three regimes in the wake of cylinder are identified: pre-
coefficient examined by Gopalkrishnan [17] and both numerical vortex-formation regime (X1), primary-vortex-formation regime
methods (current study and the results of Dong and Karniadakis (X2), and fully developed-vortex formation regime (X3). It is
[14]). It is worth noting that the reason for this difference can be observed that at X1, the shear layers merge and the initial vortex
related to the wide standard deviation of the lift coefficient even starts to shed. As the longitudinal distance is increased to X2 and
for the same experiment. Gopalkrishnan [17] showed that the X3 regimes, the vortices are formed and become fully developed.
standard deviation of obtaining lift coefficient is large and the lift This means that at X2 and X3, fully developed vortices with larger
coefficient realized in the measurements arranged from 0:2 to length scale and higher circulations of the vortices are observed.
0:55 even from one run to another. Since the length scale and circulation of the developed vortices
Fig. 3 presents the vorticity contours of a single cylinder at are directly proportional to the shear forces, e.g. lift force acting on
Re = 10,000 obtained from the CFD model along with the experi- the cylinder, it is predicted that the primary-vortex-formation (X2)

Fig. 3. Vorticity contour (s1) in the wake of the circular cylinder at Re = 10,000, (a) numerical modelling of the present study, (b) experimental result of Lau et al. [24].
62 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

and fully developed-vortex-formation (X3) regimes are more suit- the primary-vortex-formation regime (X2) is considered as an ideal
able for kinetic energy harnessing by a downstream body. Analys- longitudinal distance between the cylinder and airfoil.
ing the power density generated in the wake of the cylinder
including X2 and X3 can provide further insight for this prediction. 3.2. Strouhal number

3.1. Power density Vortex length scale and vortex shedding frequency (or Strouhal
number) was studied for the cylinder–airfoil case. Models were
Fig. 4 shows the non-dimensionalised power density of the fluid developed for different longitudinal and lateral distances between
in the symmetric plane. For 2D model, the power density is defined the cylinder and the airfoil. Two arrangements were modelled, at
as the force acting on the wetted area of the cylinder in the same y0/D = 0 and y0/D = 1, and the airfoil was positioned at all three
direction of the flow. Therefore, the fluid power equation can be wake regimes (X1, X2 and X3) at zero angle of attack.
written as 12 qU 3 A. The locations with higher power density can Typical vorticity contours around the coupled cylinder–airfoil at
be observed as a zigzag pattern in the wake of the cylinder. It is three wake regimes are shown in Fig. 5. In this figure the longitu-
seen that although at X3 the vortices are fully developed (refer to dinal distance, x0/D, was increased to cover X1 to X3 regimes with
Fig. 3), due to the dissipation of the kinetic energy, the power den- the airfoil mounted along the centre line of the cylinder (y0/
sity of the fluid gradually reduces. Consequently, with the aim of D = 0). The vorticity contours have been plotted at the same time
capturing the wake energy of the upstream cylinder both visual for all numerical models (t = 15 s). This facilitates the comparison
patterns of vorticity contours and the power density contours pro- between the vortex structures. It is observed that due to the small
vide an initial estimation for the optimum location of the down- evolution of the vortices at X1, the wake appears to be affected by
stream body. Therefore, for the rest of the current investigation the airfoil. As a result, at smaller x0/D such as pre-vortex formation

1 2 3 4 5 6 7 8

0 5

Fig. 4. Non-dimensionalised power density in the wake of a stationary circular cylinder ðP f =P f 1 Þ) at Re = 10,000.

2
Mag. of FFT of FL

x 0 / D = 1.25
1.5

X: 0.155
1
Y: 0.6353

0.5

(d) 0
0 0.5 1 1.5 2

1 2 3 4 5 6 7 8 2
Mag. of FFT of FL

(a) X: 0.2083
x 0 / D = 3.25
1.5 Y: 1.896

0.5

0
1 2 3 4 5 6 7 8 (e) 0 0.5 1 1.5 2

2
Mag. of FFT of FL

(b) x 0 / D = 6.5
1.5
X: 0.1944
1 Y: 1.423

0.5
1 2 3 4 5 6 7 8
0
(f) 0 0.5 1 1.5 2
(c) St

Fig. 5. Vorticity contours including positive and negative eddies around a coupled cylinder–airfoil as a function of longitudinal distance at y0 =D = 0, a = 0 and Re = 10,000, (a)
X 1 regime, (b) X 2 regime, (c) X 3 regimes, (d–f) in the right column present the FFT plot of the lift force at each wake regime, respectively.
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 63

regime (X1), the vortices split before they fully develop (Fig. 3a).
0.26
The splitting of the vortices alters the shear forces acting on the X1 X2 X3
airfoil and the frequency of the vortex shedding. At higher longitu- 0.24
dinal distances, e.g. at X2 and X3 regimes (Fig. 3b, and c), the airfoil
is alternately under the effect of either the primary or the fully 0.22
developed vortices regimes.
The spectral analysis of the lift force acting on the airfoil was 0.2
also carried out. The right column of Fig. 5 shows the magnitude

St
0.18
of the lift force versus Strouhal number at three wake regimes from
X1 and X3 (Fig. 5d–f, respectively) obtained using the Fast Fourier
0.16
Transform (FFT). In the FFT profile, the sampling time was 0.01 s,
which is equal to the selected time step used in the numerical 0.14 y0/D = 0, present study

simulations. y0/D = 1, present study

When comparing the FFT plots it becomes readily apparent that 0.12 Single cylinder, Blevins (1990)
when the airfoil is located at X1, the minimum Strouhal number is Ozono (1999)
obtained. The Strouhal number becomes close to the vortex shed- 0.1
0 1 2 3 4 5 6 7
ding frequency of a single cylinder (St  0.2) at higher longitudinal
x /D
distances (at X2 and X3). It is interesting to note that although the 0
Strouhal number reaches 0.2 at X3 regime, the maximum magni-
Fig. 7. Variation of the Strouhal number as a function of longitudinal distance
tude of the lift force is obtained at the X2 regime, which is almost 3 between a coupled cylinder–airfoil at Re = 10,000. The symbol ( ) reflects the
times the magnitude of the lift force at X1 regime (compare the results of a coupled cylinder-plate at Re = 6700 and y0 =D = 0, Ozono [27].
magnitudes of the lift force in Fig. 5d and e).
Fig. 6 also shows instantaneous vorticity contours of the flow in
the wake of the cylinder at t = 15 s, when the airfoil is offset by the airfoil was mounted at y0/D = 0. FFT plot of the lift force demon-
lateral distance of y0/D = 1. It is seen that similar to the centrally strates that the magnitude of the lift force is approximately similar
aligned arrangements, the transverse spacing of the vortices grad- at three regimes and the discrepancy of the Strouhal number is
ually increases when the longitudinal distance between cylinder insignificant (compare Fig. 6d–f).
and airfoil increases; however, at the lateral distance of y0/D = 1, Fig. 7 summarizes the trend of the Strouhal number as a func-
there is no splitting occurs for the vortices. tion of x0/D and y0/D, at Re = 10,000. For the centrally aligned
It is observed that at the X1 regime, the vortices smoothly pass arrangements, it is observed that when the airfoil was mounted
from the upper surface of the airfoil and they approach the leading at X1, the Strouhal number is far away from the vortex frequency
edge of the airfoil at higher x0/D (X2 and X3 regimes). This can be of a single cylinder and by minimizing the gap between the cylin-
more highlighted at X3 regime, where the oncoming vortex is der and airfoil, the vortices are split causing a reduction in the
attached to the leading edge of the airfoil (Fig. 6c). The effect of lon- length scale. Since the Strouhal number is inversely proportional
gitudinal distance on vortex shedding frequency was evaluated to the length scale of the vortices [34], a reduction in the gap size
from the time history of the lift coefficient of the airfoil when the between cylinder and airfoil would increase the Strouhal number.

Fig. 6. Vorticity magnitude contours (s1) around a coupled cylinder–airfoil as a function of longitudinal distance at y0 =D = 1, a = 0 and Re = 10,000, (d–f) in the right column
present the FFT plot of the lift force at each wake regime, respectively.
64 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

When the plate approaches the cylinder at X1 regime, the vortex 3.3. Transverse spacing of vortices
formation region is evacuated and the shear layers are not able
to grow due to the influence of the plate. This phenomenon has The Transverse Spacing (TS) of oncoming vortices can be also
been also examined and confirmed by Ozono [27] in a wide range considered when the coupled cylinder and airfoil is analysed. The
of wind tunnel tests. As a consequence, the Strouhal frequency TS of a vortex is measured from the centre of the vortex to maxi-
asymptotes to higher values when the plate approaches the cylin- mum transverse movement of the vortex over the airfoil using vor-
der. However, the Strouhal number still shows smaller values at X1, ticity contours and it is normalized by the diameter of the cylinder.
compared with the obtained values at X2 and X3. Although compar- The variation in the TS is extracted from Figs. 5 and 6 and the
ison between Strouhal numbers obtained in the present study and results are summarized in Fig. 8, for both centrally aligned and
by Ozono [27] shows some discrepancy, the variations of obtained staggered arrangements. It can be seen that the TS becomes larger
Strouhal number show similar trend. The difference between when the airfoil was positioned at higher longitudinal distances for
numerical results and experimental tests [27] can be related to both centre aligned and staggered arrangements of cylinder and
two main issues: (1) The geometry of structure; (2) the Reynolds airfoil. The smallest transverse spacing is obtained with minimal
number. The experimental results have been utilized here as they longitudinal distance between the cylinder and the airfoil (at X1
are very close to the present study due to the similarity of the regime). The TS then gradually increases at X2 and there is no sig-
objects (a coupled of cylinder-plate versus to the cylinder–airfoil) nificant change can be seen at X3.
in this range of Reynolds number. When the airfoil was mounted The effect of lateral distance is to increase the TS when the air-
at X2 and X3, the results demonstrated the Strouhal number foil was positioned at y0/D = 1. As a result, it is expected that larger
approximately converges to the vortex shedding frequency of a transverse spacing for the oncoming vortices yield stronger circu-
single cylinder. lation of vortices, U. However, it is worth noting that by increasing
It is worth noting that based on the defined Reynolds number, the gap size between the cylinder and the airfoil, in both directions
Re = 10,000 the flow is considered as a transition in shear layers (x/D and y/D), the dissipation of vortices needs be considered.
(TrSL) [39]. In the wake of the cylinder, in particular when the air- Therefore, extra attention is needed to estimate the circulation of
foil is located at X1 regime and due to the pitch of the airfoil, the vortices, U, for each test case.
flow regime is unpredictable as the Strouhal number significantly
changes. Therefore, the wake of the cylinder can be considered as
3.4. Circulation of vortices
a turbulent flow.

The circulation of vortices, U, can be estimated by the integral

around a closed area of the vorticity field as shown in Fig. 9; hence:
2.5
ZZ X
n
X1 X2 X3
C¼ x ds ¼ xi Dxi Dyi : ð13Þ
i¼1
2
In order to ensure that the circulation is independent from the
area of the vorticity, further analyses is required to obtain indepen-
dent circulation values from the selected area. The independent
1.5
TS/ Dv

solution for circulation has been chosen for a typical model when
the airfoil is mounted at X3 due to the maximum obtainable TS
at this regime. The independent results for U are shown in Fig. 10
1 as a function of the chord length of the airfoil. Here, circulation is
normalized by free stream velocity and the chord length of the air-
y0/ D = 0 foil. It is observed that the selected area (ds) with approximately 2C
0.5
y0/ D = 1
(C is the chord length of the airfoil) length yields insignificant
changes of results in terms of the circulation. Therefore, this length
is chosen to calculate the circulation around the airfoil for all test
0 1 2 3 4 5 6 7
x0 /D cases.
Fig. 11 shows the variation of the dimensionless circulation as a
Fig. 8. Dimensionless transverse spacing of the vortices as a function of longitu- function of the longitudinal and lateral distances. In general, it can
dinal and lateral distances of upstream cylinder and downstream airfoil at a = 0° be seen that the vortex circulation is higher when the airfoil is
and Re = 10,000. arranged at y0/D = 1. The results also demonstrate that for the cen-

Fig. 9. Circulation contours around a coupled cylinder and airfoil surrounded by two areas (ds).
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 65

2.2

2.1

2
Γ /UC

1.9

1.8

1.7
0 0.5 1 1.5 2 2.5 3 3.5
x0/C

Fig. 10. Independent variation of the dimensionless circulation as a function of area

for a typical arrangement of cylinder and airfoil located at the X 3 regime. Fig. 12. Configuration of a coupled cylinder–airfoil investigated in this study (with
16 locations for the downstream elastically mounted airfoil in the wake of the
upstream stationary circular cylinder).

2.5
cylinder and airfoil are shown in Fig. 12. For each test case, ten
2.4 angles of attack of the airfoil were considered leading to a total
of 160 test cases.
2.3 Due to the extensive number of test cases, the SAVEM system is
used to run the experiments rapidly by setting the angle of attack
for each arrangement of the coupled cylinder–airfoil. In Section 4,
Γ /UC

2.2
the details of experimental test are explained and the dynamic
2.1 response of the airfoil in capturing the kinetic energy of vortices
is discussed.
2
y0/ D = 0
1.9 4. Experimental apparatus
y0/ D = 1

1.8 A series of experimental tests were conducted in a closed-loop

0 1 2 3 4 5 6 7 water channel located at the Thebarton laboratory of the Univer-
x0 /D sity of Adelaide. The test section of the water channel is 500 mm
wide, 600 mm deep and 2000 mm long. An acrylic circular cylinder
Fig. 11. Effect of longitudinal and lateral distances on the vortex circulation at with diameter of D = 40 mm and 600 mm length was mounted in a
Re = 10,000. fixed frame, while the downstream aluminium NACA 0012 airfoil
with chord length C = 100 mm and the length l = 400 mm was fixed
on a virtual elastic base with two degrees of freedom. Free stream
tre aligned cylinder–airfoil, the maximum circulation can be
velocity of water channel was set at U = 0.25 m/s. Consequently, a
obtained for the studied model with x0/D = 4. Although it was
corresponding Reynolds number of Re = 10,000 was calculated
expected to achieve the highest circulation at fully develop regime
based on the diameter of the upstream cylinder (D) and the mea-
(X3), it can be seen that due to dissipation, the vortex circulation
sured density of the water at 20 °C.
shows a significant drop for both centre aligned and staggered
A 3D rendered model of the system, including the water chan-
arrangements. Since an increase in circulation of incident vortices
nel, and the SAVEM system is shown in Fig. 13. The SAVEM system
acts to increase the lift force on the airfoil and all fluidic and struc-
is comprised of a vertically mounted airfoil, which is attached to a
tural conditions of the cylinder and airfoil are constant, the heave
carriage sliding on four linear bearings. The advantage of this
performance of the airfoil can be affected by the generated vortices
model with a vertically mounted airfoil is that the static weight
or shear forces.
(minus the buoyancy force) is no larger applied to the motor which
provides the linear movement of the airfoil. This approach elimi-
3.5. System configuration nates the effect of the weight and buoyancy forces on the drive
mechanism. The SAVEM system consisted of two Maxon brushless
Analysing different aspects of flow pattern using numerical servomotors (EC-Max 30 with 2000 quad-count encoder) via a
models was helpful in providing insight into the vortical structure gearbox of 51:1 ratio which controlled by two Maxon ESCON
and the optimum configuration of the cylinder and airfoil with 50/5 servo controllers. The force applied to the airfoil is measured
higher potential to harness the kinetic energy in the flow. There- using a pair of strain gauges in a half bridge arrangement. Due to
fore, the longitudinal and lateral distances between the cylinder the normal force, the instantaneous displacement of the airfoil
and the airfoil along with the angle of attack of the airfoil were was measured and controlled employing a Proportional-Integral-
altered in the experiments. Derivative (PID) controller. The pulses produced by the encoder
Sixteen physical arrangements were tested in a water channel. in the servomotor were utilized for measuring the angular position
The downstream airfoil was kept at the centre of the water channel of the motor shaft as well as speed (angular velocity). The angle of
to avoid any blockage effects, while the upstream cylinder was rotation of the shaft was converted to the real linear displacement
moved to different positions. The different arrangements of the of the carriage to which the airfoil was attached. Consequently, the
66 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

1.4
Present study, Re =10000
1.2 Alam et al. (2010), Re = 5300
5. Timing belt Alam et al. (2010), Re = 10050
1
7. Motors, gearboxes
0.8

CL
and encoders
0.6
0.4
6. Pulley
0.2
4. Carriage with
linear bearings 0
3. Strain gauges
0 10 20 30 40 50 60 70 80 90
α (deg.)
8. Test section
Fig. 15. Comparison of the lift coefficient as a function of angle of attack of airfoil
with published experimental data [1].

2. Airfoil NACA 0012

where x represents the circular frequency of rotation of the airfoil.
1. Cylinder Based on the selected reduced frequency, the mechanism is known
as an unstable [38]. The reduced frequency, as a non-dimensional
parameter, is usually utilized for the case of unsteady aero-
elasticity and based on this parameter the flow is divided into four
Flow direction categories: (a) steady state aerodynamics when k = 0; (b) quasi-
steady aerodynamics for 0 6 k 6 0.05, (c) unsteady aerodynamics
Fig. 13. Schematic of 3D model of the test rig including the upstream circular when k > 0.05, and (d) highly unsteady for k > 0.2 [38].
cylinder, the downstream airfoil, the SAVEM system and the test section of water
channel.
4.1. Lift measurement using SAVEM system

variance between measured oscillation of the airfoil and equivalent In order to utilize work conducted by the lift force acting on the
linear displacement of the shaft was calculated as an error value. airfoil, initially it is essential to measure and compare the exerted
The error value was defined as an input signal of the PID controller shear force in the normal direction to the flow as a function of
and the controller is tuned based on Ziegler Nichols method [3]. angle of attack. The lift forces on the mounted airfoil were mea-
The assembled equipment and main parts of the SAVEM system sured and the results were validated against the available pub-
are shown in Fig. 14. More detailed explanation of the VEM system lished data. The angle of attack of the airfoil can be fixed at any
can be found in the previous study of the authors [12]. desired value relative to the direction of the free stream (refer to
In FIV, the Reynolds number, the mass and damping ratios play a in Fig. 1) by changing the angle of the rotor using a real-time
a significant role in the dynamic response of the airfoil. The mass Simulink model. The lift coefficient acting on the airfoil can be
ratio is defined as m⁄ = 4m/qpC2l, based on the mass (m) of the air- quantified as follows
foil, the density (q) of the fluid, the chord length (C), and the length FL
(l) of the airfoil. The mass damping ratio is (m⁄f) where, f repre- CL ¼ 1 ; ð14Þ
2
qU 2 CL
sents the damping ratio of the elastically mounted airfoil. In the
current study, with the intention of study of the effect of the where FL is the lift force measured by strain gauges in the normal
arrangement between the cylinder and airfoil, the mass and damp- direction to the flow stream line.
ing ratios were kept constant at 2.4 and 0.01, respectively, while Fig. 15 shows the dependence of the lift coefficient on the angle
the angle of attack was varied between 0 < a < 45° degrees with of attack of the airfoil along with the previously measured data of
an increment step of 5°. The reduced frequency was also kept con- Alam et al. [1]. It is observed that the lift coefficient initially drops
stant at kr = 0.0875. The reduced frequency is defined as kr = x C/U, at a = 10°, where the static stall occurs. The static stall is generally

1 3

2 4

(a) (b)
Fig. 14. (a) A photograph of the assembled equipment including the test section, (b) main electric elements of the SAVEM system including 1: AC excited bridge amplifier
with two channels, 2: A dSPACE Board (DS 1104) with 16 bit D/A converter, 3: A Maxon ESCON 50/5 servo controller, and 4: A Maxon servo motor (model EC-max 30) via a
gearbox of 51:1 ratio comprising embedded encoder.
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 67

associated with separation of fluid flow at the suction side of the within a stagnant fluid when both motors and controllers were dis-
airfoil [22]. However, in another scenario Laitone [21] reported abled. Based on the dynamic response of the airfoil, the fundamen-
that due to the transition from laminar to turbulence static stall tal natural frequency of the cantilevered airfoil was found to be
can occur. Consequently, it is seen that when the flow around fn = 4.70 Hz, which is outside of the frequency of oscillation.
the airfoil is about half of the Reynolds number used in this study
(e.g. Re = 5300), the lift coefficient monotonically increases until 4.2. Dynamic response of the airfoil
the second stall appears around a = 45°. At this range of Reynolds
numbers, it is assumed that the flow regimes are laminar [1]. Typical time series of the mean lift coefficient for all Test Cases
Therefore, the lift coefficient around a = 10° is far from those found (TC) are plotted in Fig. 16, when the angle of airfoil was set at
by Alam et al. [1] at the lower Reynolds number. In addition, it is a = 10° and Re = 10,000. It is clear that the behaviour of the lift
observed that at Re = 10,000 the trend of the measured data of CL coefficient varies with both longitudinal and lateral distances.
is very close to the previously published data, which is required The amplitude of the lift coefficient is approximately similar for
for the study of the dynamic response of the elastically mounted Test Cases 1 and 2 and the trend of the signal is almost regular.
airfoil. A small discrepancy between measured data with those However, by increasing the longitudinal distance to x0/D = 4 and
reported by Alam et al. [1] can be seen at a higher angle of attack. 5 (for Test Cases 3 and Test Cases 4), the behaviour of the lift coef-
The maximum error of 13% occurs at a higher angle of attack, when ficient becomes more irregular and its amplitude increases. As the
the angle is greater than a > 60°. This can be related to the blockage position of airfoil shifts to larger longitudinal distances, the fully
effects. Although a similar airfoil has been employed in both exper- developed vortices are generated. As a result, the lift force acting
iments, Alam et al. [1] used horizontal NACA 0012 airfoil with on the airfoil can originate from two components known as
600 mm height in a wind tunnel. In the present study, a vertical potential-force due to the ideal flow inertia force and vortex-force
airfoil with 500 mm was employed in a water channel. As a conse- due to the dynamics of vorticity field around the airfoil [37]. Based
quence, the blockage effect, here, is about 4% more than Alam et al. on the numerical-analysis in Section 3, the higher amplitude of the
[1] tests. Accordingly, for the rest of this study, a safe range of angle lift coefficient was expected at primary vortex formation regime
of attack of a 6 45° was chosen to investigate the dynamic (X2) and this expectation is now confirmed for Test cases 3 and 4.
response of the airfoil. For the test cases with lateral distance y0/D = 1, a higher ampli-
To avoid resonance, it is necessary that the natural frequency of tude of the lift coefficient is observed as compared to the centre
the cantilevered airfoil occurs outside of the frequency of oscilla- aligned arrangements. It can be also seen that among all staggered
tion due to vortices. Therefore, the stiffness and mass of system arrangements, a coupled cylinder–airfoil with y0/D = 1 provides the
was kept constant during the experiments and were defined to maximum amplitude of the lift coefficient, which is due to the
be k = 50 N/m and m = 2 kg, respectively. This provided the system higher circulation of oncoming vortices to the downstream airfoil.
with a frequency of oscillation of fosc = 0.79 Hz. The natural fre- Numerical analysis of the staggered arrangements revealed that,
quency of the fundamental bending mode of the cantilevered air- the maximum transverse spacing of the vortices, for staggered
foil was measured using the free-decay tests. Free-decay tests for models with y0/D = 1, is limited to 2. Therefore, it is expected that
the cantilevered airfoil were performed in the water channel outside of this range (such as y0 /D P 2), the downstream airfoil is

0.1 0.1 0.1 0.1

(a) TC = 15, α = 10 (b) TC = 2, α = 10 (c) TC = 3, α = 10 (d) TC = 4, α = 10
0.05 0.05 0.05 0.05
Lα

0 0 0 0
c

-0.05 -0.05 -0.05 -0.05

-0.1 -0.1 -0.1 -0.1

5 10 15 20 25 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

0.1 0.1 0.1 0.1

(e) TC = 5, α = 10 (f) TC = 6, α = 10 (g) TC = 7, α = 10 (h) TC = 8, α = 10
0.05 0.05 0.05 0.05
Lα

0 0 0 0
c

-0.05 -0.05 -0.05 -0.05

-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

0.1 0.1 0.1 0.1

(i) TC = 9, α = 10 (j) TC = 10, α = 10 (k) TC = 11, α = 10 (l) TC = 12, α = 10
0.05 0.05 0.05 0.05
Lα

0 0 0 0
c

-0.05 -0.05 -0.05 -0.05

-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

0.1 0.1 0.1 0.1

(m) TC = 13, α = 10 (n) TC = 14, α = 10 (o) TC = 15, α = 10 (p) TC = 16, α = 10
0.05 0.05 0.05 0.05
Lα

0 0 0 0
c

-0.05 -0.05 -0.05 -0.05

-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time (s) Time (s) Time (s) Time (s)

Fig. 16. Typical time series of the lift coefficient for sixteen arrangements of the coupled cylinder–airfoil (or four groups) at a = 0° and Re = 10,000, (a) TC = 1, (b) TC = 2, (c)
TC = 3, (d) TC = 4, (e) TC = 5, (f) TC = 6, (g) TC = 7, (h) TC = 8, (i) TC = 9, (j) TC = 10, (k) TC = 11, (l) TC = 12, (m) TC = 13, (n) TC = 14, (o) TC = 15, (p) TC = 16.
68 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

less influenced by the generated vortices. This hypothesis is now for all test cases, the magnitude of the force signal is a maximum
confirmed by the experimental results, where the airfoil was posi- for the second group of test cases with y0 /D = 1, and in particular
tioned at y0 /D = 2 or higher lateral distances (refer to Fig. 16i–p). Test Cases 5 to 10 (Fig. 18e–h). Further increasing the lateral dis-
The importance of the staggered arrangement of a coupled tance between the cylinder and airfoil causes a reduction in the
cylinder and airfoil in capturing kinetic energy of vortices can be magnitude of the signal and indicates that the second peak disap-
observed by comparing the heave response of the airfoil for centre pears; however, the frequency of oscillation still remains at f/f osc
aligned and staggered arrangements. Fig. 17 shows the time his- = 1. As a result, it is expected that the heave amplitude of the airfoil
tory of the mean dimensionless (using chord length of airfoil) increases for staggered arrangements with y0 /D = 1.
heave response of all examined test cases at a typical angle of The effect of angle of attack on the heave of the airfoil is exam-
attack a = 10°. The time history of the airfoil undergoes a smaller ined in this section. The difference between the average of maxi-

range of oscillation when the airfoil is positioned at the centre line mum and minimum of the heave amplitude Dh is calculated as
of the upstream cylinder as compared to the staggered arrange- a percentage of discrepancy among all test cases at each angle of
ments. Furthermore, it can be seen that the maximum time series attack. This highlights the role of angle of attack in the heave
of the heave response belongs to the staggered arrangements of y0 / response of the airfoil against the arrangement of the coupled
D = 1. It is also observed that further increasing at the lateral dis- cylinder–airfoil. Fig. 19 shows the percentage of this discrepancy
tance has no significant effect on the amplitude of the heave (refer as a function of the angle of attack with the respective Standard
to Fig. 17i–p). Error of Measurement (SEM). The heave responses reported in
An explanation for this trend in the heave response of the airfoil the figure are in the format mean ± standard error. The SEM is
can be offered here using the spectral analysis of the lift force. The the deviation of the sampling distribution over all measurement
spectra of the measured force applied on the airfoil is plotted in sample tests.
Fig. 18 when the airfoil angle was set to a = 10°. Herein, the fre- It can be seen that at lower angles of attack such as a = 0° or 5°,
quency of the lift force is normalized by the natural frequency of the difference between the average of maximum and minimum of

the structure. For all test cases a dominant peak is observed in the heave amplitude, Dh , is approximately similar. Herein, the
the spectrum. This peak corresponds to the frequency of the oscil- heave is normalized by the chord length of the airfoil. Fig. 18 indi-

lation. When the airfoil was positioned at y0 /D = 1, particularly for cates a significant growth at a = 10°, in which Dh approximately
TC 6, 7 and 8, another frequency appears in the signals, which is 
130%. There is a significant drop for Dh when the angle of attack
close to the frequency of the oncoming vortices. Compared with increases to 15°. Further increasing in the angle of attack causes
the numerical results for staggered arrangement of y0 /D = 1, there a consequent reduction between the maximum and minimum
is evidence that the measured frequency of the vortices is approx- amplitude of the heave. This means that the effect of the angle of
imately St = 0.2 or the frequency of generated vortices in the wake attack at larger angles is weaker and by increasing the angle of
of the cylinder. Based on the measured frequency (f/f osc = 1.30), the attack, the arrangements of the cylinder and airfoil could not influ-
Strouhal number is calculated to be St  0.19. It can be seen that ence the heave response of the airfoil. The reason for this relates to

0.1 (a) TC = 1, α = 10 0.1 (b) TC = 2, α = 10 0.1 (c) TC = 3, α = 10 0.1 (d) TC = 4, α = 10

0.05 0.05 0.05 0.05
h*

0 0 0 0
-0.05 -0.05 -0.05 -0.05
-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

(e) TC = 5, α = 10 (f) TC = 6, α = 10 (g) TC = 7, α = 10 (h) TC = 8, α = 10

0.1 0.1 0.1 0.1
0.05 0.05 0.05 0.05
h*

0 0 0 0
-0.05 -0.05 -0.05 -0.05
-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

0.1 (i) TC = 9, α = 10 0.1 (j) TC = 10, α = 10 0.1 (k) TC = 11, α = 10 0.1 (l) TC = 12, α = 10
0.05 0.05 0.05 0.05
h*

0 0 0 0
-0.05 -0.05 -0.05 -0.05
-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

0.1 (m) TC = 13, α = 10 0.1 (n) TC = 14, α = 10 0.1 (o) TC = 15, α = 10 0.1 (p) TC = 16, α = 10
0.05 0.05 0.05 0.05
h*

0 0 0 0
-0.05 -0.05 -0.05 -0.05
-0.1 -0.1 -0.1 -0.1

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Time (s) Time (s) Time (s) Time (s)

Fig. 17. Typical time series of the heave response for sixteen arrangements of the coupled cylinder–airfoil (or four groups) at a = 10° and Re = 10,000, (a) TC 1, (b) TC 2, (c) TC
3, (d) TC 4, (e) TC 5, (f) TC 6, (g) TC 7, (h) TC 8, (i) TC 9, (j) TC 10, (k) TC 11, (l) TC 12, (m) TC 13, (n) TC 14, (o) TC 15, (p) TC 16.
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 69
Mag. of FFT of lift
0.04 0.04 0.04 0.04
(a) TC = 1, α = 10 (b) TC = 2, α = 10 (c) TC = 3, α = 10 (d) TC = 4, α = 10
0.03 0.03 0.03 0.03

0.02 0.02 0.02 0.02

0.01 0.01 0.01 0.01

0 0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
Mag. of FFT of lift

0.04 0.04 0.04 0.04

(e) TC = 5, α = 10 (f) TC = 6, α = 10 (g) TC = 7, α = 10 (h) TC = 8, α = 10
0.03 0.03 0.03 0.03

0.02 0.02 X: 1.304 0.02 0.02

Y: 0.01171

0.01 0.01 0.01 0.01

0 0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
Mag. of FFT of lift

0.04 0.04 0.04 0.04

(i) TC = 9, α = 10 (j) TC = 10, α = 10 (k) TC = 11, α = 10 (l) TC = 12, α = 10
0.03 0.03 0.03 0.03

0.02 0.02 0.02 0.02

0.01 0.01 0.01 0.01

0 0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
Mag. of FFT of lift

0.04 0.04 0.04 0.04

(m) TC = 13, α = 10 (n) TC = 14, α = 10 (o) TC = 15, α = 10 (p) TC = 16, α = 10
0.03 0.03 0.03 0.03

0.02 0.02 0.02 0.02

0.01 0.01 0.01 0.01

0 0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5

f /f f /f f /f f /f
osc osc osc osc

Fig. 18. Magnitude of FFT of the lift force acting on the airfoil for sixteen arrangements of the coupled cylinder–airfoil at a = 10° and Re = 10,000, (a) TC 1, (b) TC 2, (c) TC 3, (d)
TC 4, (e) TC 5, (f) TC 6, (g) TC 7, (h) TC 8, (i) TC 9, (j) TC 10, (k) TC 11, (l) TC 12, (m) TC 13, (n) TC 14, (o) TC 15, (p) TC 16.

for reduced frequency of the airfoil, the lower angle of attack,

140 a 6 20°, has the strongest effects on the dynamic response of the
airfoil, whereas the large angles a P 25° plays a secondary role.
120 In order to highlight the combined effect of the longitudinal and
lateral distances along with the angle of attack of the airfoil, a con-
100 tour plot diagram is utilized, which summarizes the effect of all the
above parameters. With the aim of visualization and quantification
Δ h* %

80 of the arrangement on the heave amplitude of the airfoil, the inter-

polation of the measured values is developed to the studied test
cases. Furthermore, to aid in the interpretation of the heave values,
60
isolines are added to the contour plot. The maximum dimension-
less heave of the airfoil as a function of x0 /D, y0 /D, and a is pre-
40
sented in Fig. 20. In general, it is observed that the maximum
heave occurs for the staggered arrangements. Most significantly,
20 it is observed that the highest heave of the airfoil appears when
the airfoil was mounted at the primary vortex-formation regime
0 10 20 30 40 50
(X 2 ). In addition, it can be seen that the minimum and maximum
α (deg) bounds of the primary vortex-formation regime, (e.g. x0 /D = 3–5),
Fig. 19. The percentage of variation of heave of airfoil obtained from difference
are very critical at the heave response of the airfoil. When the air-
between the average of the maximum and minimum of the heave at each set of foil was set at smaller angle of attack of a < 25°, the heave response
arrangements as a function of angle of attack at Re = 10,000 including Standard of the airfoil shifts to the lower bound of X 2 or x0 /D = 3. By increas-
Error of Measurement (SEM). ing the angle of attack to a P 25°, the higher heave amplitude of
the airfoil moves to the higher value of the longitudinal distance
the correlation of the selected reduced frequency and angle of of x0 /D = 4 or even more (Fig. 20e–h). As an effect of the lateral dis-
attack of the airfoil. Zhu [40] found that the dynamic response of tance, in general, it can be noted that the maximum heave occur at
the airfoil at different angle of attacks is a function of the reduced the staggered arrangements when the airfoil was mounted at
frequency. Therefore, it is found that under the defined condition 1 6 y0/D 6 1.5 and 3.5 6 x0/D 6 4.5.
70 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

3 3
(a) (e)

04 5
0. 75
0.046 0.17

0. 0.0
α= 5 α = 25

5
04
54 5
5 370.0 42

25
03 .0 .0

03
0.044
0

0.
2.5 2.5
0. 0
0.16

25
0.042

15

16

5
5
03

16
15
0.

0.
0.

0.
0.
5

17
03

03
2 0.04 2

0.
0.

0.
0.15
0.038
5
03

h*
h*
0.

75

1.5 0.036 1.5 0.14

03
0.

25

04
04

0.
0. 0.034

75
1 1 0.13

03
0.032

0.
5
42

17
0

0.
04

5
0.

0.03

0. 5

0. 5
04

0. 5
11

0. 2
12

0. 3
0.

5
13
0.12

0. 4
1

14
0.

0.

1
0.5 0.5
75

0.028
03
0.

0.11

5
0.026

03
0
0.
0
2 2.5 3 3.5 4 4.5 5 2 2.5 3 3.5 4 4.5 5

3 0.075 3 0.23
(b) (f)
07

α = 10 α = 35
0.
0 6

5
05.

06
05

0.07
5
0.

0.22
06
0.

2.5
0.

5
2.5

22
0.
18
0.065

0. 9

22
2

0. 1
0.5

5
0. 5
1

2
19

21
20
0.

0.

0.
0.21
05

0.

22
.
5 0

0.
2 0.06 2
04
0.

0.2
04

0.055

h*
0.

h*
1.5 1.5
0.19
07

0.05
0.

5
22
0.
1 0.045 1 0.18
06

5
0.04

17

5
0.

17
0.17

18
0.
0.

0.
0.5 0.5
5

0.035
06
0.
5

0.16
05

07
0.

0.

0 0.03 0
2 2.5 3 3.5 4 4.5 5 2 2.5 3 3.5 4 4.5 5

3 0.11 3
(c) α = 15 (g) α = 40
35

0.24
10
0.

23
2.5 0.1 2.5

0.
21

0. 5
5

0. 2
21

0. 5
20

0. 3

5
2

5
0.

22
2

23
23
0.23
0.
0.
75

0.
06

2 0.09 2
0.

0.22
h*

h*
1.5 0.08 1.5

5
24
0. 0.21
1 0.07 1

24
0.
0.2
0. 765

0. 55

45
0.072

9
0. 81

5
08

0. 9
09

09

19
0

2
0
0.

0.

0.

0.

0.5 0.06 0.5

24

0.19
0.
35
10
0.

0 0.05 0
2 2.5 3 3.5 4 4.5 5 2 2.5 3 3.5 4 4.5 5

3 3 0.265
(d) 0.14 (h) α = 45
4
23

α = 20
95

0.26
0.
22
0.
14

2.5 2.5 0.255

0.

0.13
0. 243

1
25 5

26
2
0. 47
0.

0.
2

0.25
2 2
05

0.12 0.245
22
0.

0.24
h*
h*

1.5 0.11 1.5

14
0.

0.235
22
0.
5
22

0.23
0.

1 0.1 1
1
26

0.225
0.
0. 5

0. 5

0. 5

5
09

10

0. 1

12

13
11

12

0.09
65
1

85
0.1

13

0.22
0.

0.
0.

0.5
25

0.5
23
0.

0.
0.
23 5

0.215
9
22

0.08
0.
0.

0 0
2 2.5 3 3.5 4 4.5 5 2 2.5 3 3.5 4 4.5 5

Fig. 20. Isoline contours of the heave of the airfoil as a function of longitudinal and lateral distances at different angle of attack at Re = 10,000. (a) a = 5°, (b) a = 10°, (c) a = 15°,
(d) a = 20°, (e) a = 25°, (f) a = 35°, (g) a = 40°, (h) a = 45°.

4.3. Power coefficient of the FIV cient is shown. The contour plot has been chosen for the maximum

Dh when the airfoil was at a = 10°. The plotted color bar of the fig-
The effect of the arrangements of the coupled cylinder–airfoil ure shows the range of efficiency within a range of 0–30%. The con-
and angle of attack of the airfoil impacting on the power coefficient tour of the efficiency is plotted as a function of longitudinal and
can be considered. In Fig. 21 the contour plot of the power coeffi- lateral distance (x0 =D and y0 =D respectively). It is observed that
 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72 71

3 30 dynamic response the downstream cylinder Derakhshandeh et al.

26
[11]. This is related to the effect of generated upstream vortices,

28

29
28 which are suppressed between the upstream cylinder and the wall

28
2.5

29
and imposed on the downstream cylinder. The presence of the wall

27
20

26
forces the generated upstream vortices to move forward. There-
19

2
fore, the downstream cylinder will be under the effect of the jet
24
18

of flow. This causes an alteration on shear and pressure forces

and finally affects the dynamic response of the cylinder which
17

22

η%
1.5 alters the efficiency of the convertor. A detailed explanation about
20 the dynamic response of the downstream cylinder near a rigid wall
can be found in Derakhshandeh et al. [11].
30

1
16

18 The effect of non-circular structure such as airfoil is considered

in this study. Similar arrangements with the same structural
16 parameters have been chosen here as compared with the previous
21
15

22

0.5
23
24
25
26

study [10]. The results reveal that the angle attack affects the
27

29

14
14

dynamic response of the airfoil which increases the efficiency of

28 the convertor from 28% (using coupled circular cylinders) to 30%
28

0
2 2.5 3 3.5 4 4.5 5
(employing a coupled cylinder–airfoil). However, a comparison
between results demonstrate that although the optimum efficiency
Fig. 21. Variation of the power coefficient as a function of the longitudinal and can be captured at lateral distance 1 6 y0/D 6 2 for two cylinders,
lateral distances between the coupled cylinder–airfoil at a = 10° and Re = 10,000. for a coupled cylinder and airfoil these arrangements alters to
1 6 y0/D 6 1.5. Admittedly, further investigations can be per-
the maximum FIV efficiency of 30% is obtained for cases with 3.5 formed to optimum the efficiency of the convertor by changing
6 x0/D 6 4.5 and 1 6 y0/D 6 1.5. It is seen that the increment of the structural parameters such as mass ratio, damping ratio, type
the lateral distance has no considerable influence on the power of the airfoil and location of the pitch of the airfoil.
coefficient of the FIV as compared to the increase of the longitudi-
nal distance. 5. Conclusions
It is worth noting that the dynamic response of the airfoil stud-
ied in the current paper is comparable with the previous numerical The wake flow patterns around the cylinder and airfoil was
and experimental investigations conducted by Derakhshandeh studied using CFD along with the water channel tests at a transient
et al. [10], Derakhshandeh et al. [11] and Derakhshandeh et al. flow regime (Re = 10,000). The results of the numerical modelling
[12,13]. It has been shown that VIV energy can be harnessed using and experimental tests support the following observations
a single circular cylinder [13], unbounded two circular cylinders accordingly:
[10,12], two circular cylinders under the effect of a rigid wall
[11]. Although it has been shown that VIV energy can be captured 1. The locations with higher power density were determined as a
by a single cylinder [13], the efficiency of VIV converter is approx- zigzag pattern in the wake of the cylinder. It is found that
imately 10%. This efficiency can be optimised employing two circu- although at fully developed-vortex formation regime (X 3 ) the
lar cylinders and using WIV mechanism instead of VIV. It was vortices are fully developed, due to the dissipation of vortices;
shown that in the VIV mechanism the upper branch of amplitude the power density of the fluid gradually reduces. As a result,
as an effective parameter on the efficiency occurs at a limit range the primary-vortex-formation regime (X 2 ), is identified as an
of frequencies or reduced velocities in which the shedding fre- ideal longitudinal distance between the cylinder and airfoil.
quency is very close to the resonance frequency of the structure. 2. The vortex shedding frequency of the wake is a function of the
However, in the WIV mechanism using two cylinders, the maxi- longitudinal distance and lateral distances between coupled
mum amplitude of oscillation can occur at the frequencies outside cylinder and airfoil and it was varied from pre-vortex-
the range in which VIV is observed [2,12]. In the WIV mechanism, formation regime (X 1 ) to primary-vortex-formation regime
the total lift acting on the downstream cylinder can be divided into (X 2 ) and it ultimately converged to the Strouhal number of a
two components [37]: (1) a potential-force component (F potential ) due single cylinder at fully developed-vortex formation regime
to the ideal flow inertia force and (2) a vortex-force (F v ortex ), which is (X 3 ). Consequently, pre-vortex-formation regime (X 1 ) is not
only produced by the dynamics of the vorticity field around the appropriate for harnessing the vortices energy with irregular
downstream cylinder. Here, the potential-force is a function of the frequency of vortex shedding.
cylinder’s acceleration and its magnitude is proportional to the 3. The length scale and transverse spacing of vortices behind a
product of the displaced fluid mass and the acceleration of the cylinder becomes larger at primary-vortex-formation regime
cylinder. Meanwhile, the vortex-force is dependent on the dynam- (X 2 ) and fully developed-vortex formation regime (X 3 ). A larger
ics of generated vortices [2]. Consequently, in the WIV mechanism, vortex provides a stronger circulation and consequently
the maximum lift coefficient acting on the downstream cylinder is increases the lift force on the downstream airfoil. The variation
accessible when both the potential lift coefficient and vortex lift of the length scale of vortices and transverse spacing of the vor-
coefficient become a maximum and the phase angle remains very tices can be seen in both centre aligned and staggered arrange-
close to zero. It was shown that, in the WIV mechanism, the total ments when the longitudinal distance increases. However, this
phase angle is approximately zero when the reduced velocity is variation was higher for the centre aligned arrangements.
roughly 4.6 and both potential and vortex forces are approximately 4. The effect of the arrangements of the coupled cylinder–airfoil
equal [2]. on the aerodynamic force and the heave response of the airfoil
Apart from this, for two cylinders in which the upstream is were examined. It was shown that both longitudinal and lateral
mounted in vicinity of a rigid wall, it has been shown that the distances play an important role in the heave response of the
gap between wall and upstream can significantly affect the airfoil to capture the energy of the vortices.
72 J.F. Derakhshandeh et al. / Experimental Thermal and Fluid Science 74 (2016) 58–72

5. In regard to the angle of attack in capturing vortices energy it [16] D.H. Hodges, G.A. Pierce, Introduction to Structural Dynamics and Aero-
Elasticity, vol. 15, Cambridge University Press, 2011.
was observed that the optimum angle of attack was found to
[17] R. Gopalkrishnan, Vortex-Induced Forces on Oscillating Bluff Cylinders, Ph.D.
be a = 10°. For this angle of attack, the maximum FIV efficiency Thesis, Department of Ocean Engineering, MIT, USA, 1993.
of 30% is obtained for cases with 3.5 6 x0/D 6 4.5 and 1 6 y0/ [18] R. Govardhan, C. Williamson, Critical mass in vortex-induced vibration of a
D 6 1.5 arrangements, which is limited to the narrower lateral cylinder, Eur. J. Mech.-B/Fluids 23 (1) (2004) 17–27.
[19] A. Khalak, C. Williamson, Dynamics of a hydro-elastic cylinder with very low
distances as compare to the previous study of the authors mass and damping, J. Fluids Struct. 10 (5) (1996) 455–472.
[12] employing a pair of cylinders. [20] M. Khan, M. Iqbal, J. Quaicoe, River current energy conversion systems:
progress, prospects and challenges, J. Renew. Sustain. Energy Rev. 12 (8)
(2009) 2177–2193.
[21] E. Laitone, Wind tunnel tests of wings at Reynolds numbers below 70,000, J.
References Exp. Fluids 23 (5) (1997) 405–409.
[22] J.W. Larsen, S.R. Nielsen, S. Krenk, Dynamic stall model for wind turbine
[1] M.M. Alam, Y. Zhou, H. Yang, H. Guo, J. Mi, The ultra-low Reynolds number airfoils, J. Fluids Struct. 23 (7) (2007) 959–982.
airfoil wake, J. Exp. Fluids 48 (1) (2010) 81–103. [23] Y. Lau, Experimental and Numerical Studies of Fluid-Structure Interaction in
[2] G. Assi, Mechanisms for flow-induced vibration of interfering bluff bodies PhD Flow-Induced Vibration Problems. PhD thesis, The Hong Kong Polytechnic
thesis, Imperial College London, UK, 2009. University, 2003.
[3] K.J. Astrom, T. Hagglund, Advanced PID Control, ISA, 2006. [24] Y.L. Lau, R.M.C. So, R.C.K. Leung, Flow-induced vibration of elastic slender
[4] P.W. Bearman, Vortex shedding from oscillating bluff bodies, J. Fluid Mech. 16 structures in a cylinder wake, J. Fluids Struct. 10 (9) (2007) 1061–1083.
(1984) 195–222. [25] W. McKinney, J. DeLaurier, The wing mill: an oscillating-wing windmill, J.
[5] M.M. Bernitsas, K. Raghavan, Converter of Current/Tide/Wave Energy, Energy 5 (2) (1981) 109–115.
Provisional Patent Application, United States Patent and Trademark Office [26] F. Menter, Y. Egorov, The scale-adaptive simulation method for unsteady
Serial, No. 60/628, 2004, p. 252. turbulent flow predictions. Part 1: Theory and model description, Flow,
[6] M.M. Bernitsas, Y. Ben-Simon, E. Garcia, VIVACE (vortex induced vibration Turbulence Combust. 85 (1) (2010) 113–138.
aquatic clean energy): a new concept in generation of clean and renewable [27] S. Ozono, Flow control of vortex shedding by a short splitter plate
energy from fluid flow, J. Offshore Mech. Arct. Eng. 130 (2008) 1–15. asymmetrically arranged downstream of a cylinder, J. Phys. Fluids 11 (1999)
[7] R.D. Blevins, Flow-induced vibration, Krieger Publishing Company, Malabar, 2928–2934.
Florida, USA, 1990. [28] Z. Peng, Q. Zhu, Energy harvesting through flow-induced oscillations of a foil, J.
[8] B.I. Celik, U. Ghia, P.G. Roache, C.J. Freitas, H. Coleman, P.E. Raad, Procedure for Phys. Fluids 21 (12) (2009) 123602–123609.
Estimation and Reporting of Uncertainty Due to Discretization in CFD [29] A. Roshko, On the development of turbulent wakes from vortex streets,
Applications, Journal of Fluids Engineering 130 (7) (2008) 078001, http://dx. National Advisory Committee for Aeronautics, NACA Technical Report 1191,
doi.org/10.1115/1.2960953. 1954.
[9] L.P. Chamorro, D.R. Troolin, S.J. Lee, R. Arndt, F. Sotiropoulos, Three- [30] T. Sarpkaya, Fluid forces on oscillating cylinders, ASCE J. Waterway Port
dimensional flow visualization in the wake of a miniature axial-flow Coastal Ocean Div. 104 (1978) 275–290.
hydrokinetic turbine, J. Exp. Fluids 54 (2) (2013) 1–12. [31] T. Sarpkaya, A critical review of the intrinsic nature of vortex-induced
[10] J.F. Derakhshandeh, M. Arjomandi, B. Cazzolato, B. Dally, The effect of vibrations, J. Fluids Struct. 19 (4) (2004) 389–447.
arrangements of two circular cylinders on the maximum efficiency of [32] P. Sváček, M. Feistauer, J. Horáček, Numerical simulation of flow induced airfoil
Vortex-Induced Vibration power using a Scale-Adaptive Simulation model, J. vibrations with large amplitudes, J. Fluids Struct. 23 (3) (2007) 391–411.
Fluids Struct. 49 (2014) 654–666. [33] L. Trevor, Future Energy: Improved, Sustainable and Clean Options for our
[11] J.F. Derakhshandeh, M. Arjomandi, B. Dally, B. Cazzolato, Effect of a rigid wall Planet, Elsevier Science, Amsterdam, 2013.
on the vortex induced vibration of two staggered circular cylinders, J. Renew. [34] M. Unal, D. Rockwell, On vortex formation from a cylinder, Part 2, Control by
Sustain. Energy 6 (2014) 033114. splitter-plate interference, J. Fluid Mech. 190 (1988) 513–529.
[12] J.F. Derakhshandeh, M. Arjomandi, B. Dally, B. Cazzolato, Harnessing hydro- [35] O. Vries, On the theory of the horizontal-axis wind turbine, Annu. Rev. Fluid
kinetic energy from wake-induced vibration using virtual mass spring damper Mech. 15 (1) (1983) 77–96.
system, J. Ocean Eng. (2014), http://dx.doi.org/10.1016/j. [36] C. Williamson, A. Roshko, Vortex formation in the wake of an oscillating
oceaning.2015.08.003. cylinder, J. Fluids Struct. 2 (4) (1988) 355–381.
[13] J.F. Derakhshandeh, M. Arjomandi, B. Dally, B. Cazzolato, A study of a vortex- [37] C. Williamson, R. Govardhan, Vortex-induced vibrations, Annu. Rev. Fluid
induce vibration mechanism for harnessing hydrokinetic energy of eddies Mech. 36 (2004) 413–455.
using an elastically circular cylinder, J. Appl. Math. Modell., 2015, under [38] JR. Wright, J.E. Copper, Introduction to Aircraft Aeroelasticity and Loads, vol.
review. 20, John Wiley & Sons, 2008.
[14] S. Dong, G.E. Karniadakis, DNS of flow past a stationary and oscillating cylinder [39] M. Zdravkovich, Flow Around Circular Cylinders, 1st ed., vol. 1, Oxford
at Re = 10,000, J. Fluids Struct. 4 (20) (2005) 519–531. University Press Inc., New York, 1997.
[15] Y.C. Fung, An Introduction to the Theory of Aero-Elasticity, Courier [40] Q. Zhu, Optimal frequency for flow energy harvesting of a flapping foil, J. Fluid
Corporation, 2002. Mech. 675 (2011) 495–517.