Journal of Fluids and Structures 74 (2017) 5363

Contents lists available at ScienceDirect

Journal of Fluids and Structures

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

Experimental sensitivity of vortex-induced vibrations to

localized wake perturbations
J.I. Jimnez-Gonzlez a, *, F.J. Huera-Huarte b
a
Departamento de Ingeniera Mecnica y Minera, Universidad de Jan, Campus de las Lagunillas, 23071 Jan, Spain
b
Department of Mechanical Engineering, Universitat Rovira i Virgili, 43007 Tarragona, Spain

article info a b s t r a c t
Article history: We present the experimental sensitivity of vortex-induced vibration (VIV) to localized
Received 20 January 2017 perturbations, in the wake of a low mass-damping one-degree-of-freedom circular cylinder
Received in revised form 29 May 2017 subject to uniform cross-flow. Regions of VIV sensitivity have been identified, clearly
Accepted 10 July 2017
indicating positions in the wake where control systems should be placed in order to
attenuate VIV amplitudes.
As a validation of the sensitivity maps, we demonstrate how by using control cylinders
Keywords:
Sensitivity analysis with diameters of only 12% of the main cylinder diameter, reductions of VIV response of
Vortex-induced vibrations more than 65%, can be reached. The use of Digital Particle Image Velocimetry (DPIV) has
Bluff body wakes allowed us to identify the physical mechanisms underlying the VIV response modifications
Localized perturbations induced by the control cylinders.
2017 Elsevier Ltd. All rights reserved.

1. Introduction

The flow around circular cylinders and the existence of vortex-induced vibrations (VIV) under particular flow and struc-
tural conditions, constitutes a widespread problem which is encountered in many engineering applications. In particular,
in ocean and offshore engineering systems (e.g. riser pipes, undersea cables or other slender structural components), the
amplitude of vibrations can be large. There has been a wide scientific interest that has given rise to a large amount of work
devoted to VIV characterization and control (see e.g. Zdravkovich, 1981; Bearman, 1984; Williamson and Roshko, 1988;
Sarpkaya, 2004; Williamson and Govardhan, 2004; Bearman, 2011). In that sense, to reduce or completely suppress cylinder
vibrations, different approaches have been proposed, which can be generally classified into passive, active open-loop and
active closed-loop methods (Choi et al., 2008). Passive methods offer generally simple and efficient solutions for vibration
attenuation in real configurations, and have traditionally received more attention (Zdravkovich, 1981). These strategies are
mainly based on geometry modifications of the body (e.g. Bearman and Brankovic, 2004; Snchez-Sanz and Velazquez,
2009; Zhou et al., 2011; Lee et al., 2014) and perturbations of the flow around it, using, for instance, splitter plates (Assi et
al., 2009) or smaller control cylinders (Korkischko and Meneghini, 2012; Zhu et al., 2015; Zhu and Yao, 2015), whose aim is
to alter the vortex shedding process. The performance of such control devices might be improved by previously identifying
the wake regions that contribute most to VIV reduction, when the flow is perturbed locally by means of infinitesimal forcing.
This approach stems from the concept of sensitivity to localized perturbations, which, to the best of our knowledge, has not
been yet applied to VIV.
The use of sensitivity to localized perturbations in the wake has been widely investigated in the recent years for rigid
stationary cylinders and other two-dimensional bodies. The seminal experimental study from Strykowski and Sreenivasan

* Corresponding author.
E-mail addresses: jignacio@ujaen.es (J.I. Jimnez-Gonzlez), francisco.huera@urv.cat (F.J. Huera-Huarte).

http://dx.doi.org/10.1016/j.jfluidstructs.2017.07.010
0889-9746/ 2017 Elsevier Ltd. All rights reserved.
54 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363

(1990) demonstrated that the vortex-shedding behind a cylinder can be effectively suppressed at low values of Reynolds
numbers, through the placement of a small cylinder in a particular region of the wake, leading to a shift on the onset of
the instability, towards a higher critical value of Reynolds number well above the value of Re = 46. These observations
on the wake sensitivity were later reproduced using adjoint linear stability formalisms (Hill, 1992; Giannetti and Luchini,
2007; Marquet et al., 2008), identifying similar regions in the flow where an infinitesimal localized perturbation, e.g. control
cylinders, might be able to reduce the growth rate and amplitude of the unstable global mode. The effect of such localized
perturbations is related to wake stabilization mechanisms that reduce the shear or the flow recirculation, which are known to
be a source of global instability. More interestingly, the presence of these perturbations in the wake affects also the frequency
of shedding and the pressure distribution, and therefore, alters the forcing acting on the body. For instance, Sakamoto
and Haniu (1994) showed experimentally that mean and fluctuating values of force coefficients acting on a cylinder can
be considerably reduced by placing a small perturbing cylinder either along the boundary layer or within the separation
region. Moreover, Parezanovi and Cadot (2012) performed experiments in the wake of a D-shaped cylinder at turbulent
regimes, showing that important variations of shedding frequency can be achieved when the cylinder is placed along the
shear layers, modifying the interaction between them and therefore, vortex formation. The effects of localized perturbations
on the global frequency have also been characterized by means of adjoint sensitivity analyses in recent works by (Meliga
et al., 2012; Mettot et al., 2014). Similarly, important variations of drag and lift magnitudes can also be obtained when the
disturbing cylinders are spatially arranged according to the areas with the largest sensitivity (Meliga et al., 2014, 2016).
In view of such results on wakes behind rigid stationary bluff bodies, it seems evident that sensitivity analyses may
provide valuable information on the optimal placement of wake perturbing devices in elastically mounted rigid cylinders. In
that sense, a first theoretical step to understand and derive the flow sensitivity are the works from Meliga and Chomaz
(2011) and Meliga et al. (2011), where the main features of VIV, at low Reynolds numbers, were investigated using
adjoint-based formulation and asymptotic expansions, identifying fluid and structure modes and its role on lock-in and
hysteretic phenomena. However, the lack of results at high Reynolds numbers and the complex system behaviour, makes
the experimental analysis an interesting approach. Consequently, in order to set a base for improved devices, we will apply
the concept of sensitivity maps to chose appropriate locations of perturbing cylinders, prior experimental determination of
VIV response sensitivity.
We aim at obtaining sensitivity maps of VIV amplitude and frequency to localized perturbations in the flow around the
cylinder. A symmetric pair of tinny perturbation cylinders will be placed at different wake positions, covering a fine grid of
wide radial and azimuthal ranges, relative to the origin of the vibrating cylinder. Note that, unlike other previous works on
rigid bodies, symmetric configurations for perturbations cylinders will be employed to obtain dynamic responses without
any mean lift component, which is known to appear if an asymmetric scheme is adopted (see e.g. Assi et al., 2009), and might
be undesirable in terms of control for practical applications. These tests will be conducted for selected values of Reynolds
numbers or reduced velocities. The usefulness of the obtained sensitivity maps will be validated by placing, only in several
locations with the highest sensitivity, pairs of control cylinders with diameters larger that those of the previous perturbation
ones. The VIV response of the new controlled system will be studied in depth using Digital Particle Image Velocimetry (DPIV).

2. Experimental set-up

The experiments were performed in the free surface water channel (FSWC) of the Department of Mechanical Engineering
at Universitat Rovira i Virgili (URV) in Tarragona. The experimental set-up and the layout of the experiment appear in Fig. 1.
The water channel has a cross-section of 1 1.1 m2 , and with a water height of 1 m it can reach flow velocities well over to
0.5 m/s. The flow profile is characterized by a very low flow velocity variability, with a maximum deviation of axial velocity
of 1.58% at the working section.
The circular cylinder model consisted of a rigid acrylic tube of immersed length L = 0.55 m and diameter D = 0.05 m,
i.e. of aspect ratio L/D = 11, and had an end-plate attached to his bottom base, with the aim at suppressing three-dimensional
effect close to the cylinder end. The system, elastically connected to a fixed supporting structure by means of a pair of springs,
was hung from a one-degree-of-freedom air bearing rig, which allowed the cylindrical model to oscillate in the cross-flow,
y-direction (see Fig. 1(a)) with very low damping.
The VIV sensitivity study was performed by placing in the wake a symmetric pair of perturbation cylinders, whose centres
are located at positions (r , ), relative to the coordinate system defined by the oscillating cylinder, as shown in Fig. 1(b).
Perturbation cylinders consisted of stainless steel rods with a diameter d = 0.032D.
The dynamic response validation carried out at the points in which sensitivity was found to be the maximum, was
performed with larger cylinders, consisting of aluminium tubing with an external diameter d = 0.12D. Holding devices
were made of slim, streamlined, 3D printed parts of polylactic acid (PLA), that allowed precise placement of cylinders at any
radius r and angle of interest.
For the sensitivity analysis, a discrete fine grid of points was designed, covering the ranges 0.6D r 1.54D and
90 90 , with r = 0.06D and = 10 , configuring a grid of 13 19 (247) nodes. At the second stage of
the work, several specific control locations were picked for the control cylinders, based on the areas of highest sensitivity
identified in the previous analysis.
The structural damping coefficient in air after decay tests, was found to be = 0.0117. The mass ratio was m =
4m/( LD2 ) = 1.94, with m being the structural mass of the system including the mass of the perturbation cylinders
 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363 55

Fig. 1. (a) Experimental set-up in the free surface water channel (FSWC) and (b) cross-sectional schematic of the experimental arrangement.

and the holding devices. The combined mass-damping ended up being m = 0.023. The Reynolds number Re = u D/ ,
where u and are respectively the free-stream velocity and kinematic viscosity, ranged from 9200 to 26 000. Similarly, the
reduced velocity, defined as Ur = u /fn D, where fn is the natural frequency in still water (0.86 Hz approximately), covered
the interval 4.2 to 11.8.
The y-displacement was measured by means of a laser position sensor, and the velocity in the wake was obtained using
a high-resolution acoustic Doppler velocimeter. Data acquisition was performed with a sampling frequency of Fs = 1000 Hz
sampling during at least 60 s, ensuring temporal resolution enough to capture the system dynamics properly. Considering
that the minimum measured oscillation frequency was 0.67 Hz, at least 40 oscillation cycles were recorded for each test.
Once the VIV sensitivity was obtained, the flow field for certain cases showing high potential for VIV reduction, was
measured using Time-Resolved Digital Particle Image Velocimetry (TR-DPIV). Images were recorded at 500 fps using a CCD-
sensor with 12 bit resolution, a pixel size of 10 m and 1280 1024 pixels, in combination with a lens with 20 mm focal
length. The flow was seeded using 20 m Polyamide particles and the illumination was generated using a Continuous Wave
(CW) Diode-Pumped solid State (DPSS) green laser able to deliver up to 5 W. This set-up allowed measurements of the
velocity field in a region of approximately 8 5 diameters around the cylinder model. Image pre-processing using in-
house codes was performed in order to mask regions without particles in the flow and to identify in each DPIV snapshot
the position of the cylinders. Following, a Fast Fourier Transform (FFT) cross-correlation algorithm (Willert and Gharib,
1991) was applied to the pre-processed images, using an interrogation area of 24 24 pixels with a 50% of area overlapping.
Spurious vectors produced by the cross-correlation scheme, were thresholded and replaced by new values obtained from
neighbour interpolation.

3. Dynamic response sensitivity to localized perturbations

In this section we present the results of the sensitivity analysis to local perturbations. First we show the dynamic response
of the plain cylinder model isolated, without perturbation cylinders (Section 3.1), and then we show the response of the
system after including the localized perturbation cylinders (Section 3.2).
The VIV dynamic response is characterized by the oscillation amplitude A and frequency f , made dimensionless with the
diameter D and the natural frequency in still water fn , so that A = A/D and f = f /fn . Likewise, 1/fn is used for the time
scaling yielding t = tfn . Remaining variables are also in dimensionless form using the same scalings.
The amplitude is defined here using two methods, either by computing the mean value of the signal envelope obtained
through Hilbert transform; or by means of integration of the peak in the Power Spectral Density distribution, PSD(F ),
associated with the fundamental frequency f , in such a way that
f +fu
A 2
=2 PSD(F ) dF , (1)
f fd

where fd and fu stand for the interval around f for which the energy drops to 1% the peak value, and F represents
the frequency values spanning the interval [0, Fs /2]. Both methods resulted in nearly identical estimations of amplitude A .
56 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363

Fig. 2. Amplitude and frequency response of an isolated flexibly-mounted rigid cylinder versus reduced velocity Ur and Reynolds number Re , with
m = 0.023.

The second approach has been used throughout the text as it takes into account only the energy at the response frequency
(including modulations of amplitude at F = f ).

3.1. VIV of the reference plain cylinder

The VIV response of the reference model without perturbations appears in Fig. 2 where the amplitudes and frequencies
are plotted versus the reduced velocity Ur , and the Reynolds number Re. As in other one-degree-of-freedom VIV studies
with flexibly mounted rigid cylinders in water (see e.g. Khalak and Williamson, 1999; Williamson and Govardhan, 2004;
Blevins and Coughran, 2009), results show the initial, upper and lower branches. The maximum amplitude corresponds to
A 0.91, and was reached at Ur 6.1. Besides, as expected, the oscillating frequency follows initially the Strouhal law
(represented by the continuous line in Fig. 2), which has been measured to be St 0.185, up to the beginning of the lock-in
range, where the oscillating frequency increases slowly with Ur . At the right limit of the lower branch, when the amplitude
decreases abruptly, f approaches again the Strouhal law.

3.2. Amplitude and frequency sensitivities to localized perturbations

In this section we present results of VIV response sensitivity analyses after placing a pair of small perturbation cylinders
in a symmetric arrangement, in the near wake. The size of the perturbation cylinders has been chosen to be of the order
of magnitude provided by the boundary
layer thickness, , within the range of Re evaluated, in such a way that, according
to Schlichting (1979), O(D/ Re) 0.028D at Ur = 6.1 where the maximum amplitude is reached (the local Re
provided by the perturbation cylinder diameter under that flow condition is approximately 43).
In view of Fig. 2, two representative values of reduced velocity have been selected to study the dynamic response when
the wake is perturbed, namely Ur = 6.1 and Ur = 9.3, corresponding, respectively, to points in the upper branch (with
the largest amplitude response A 0.91) and in the lower branch (A 0.59). The sensitivity of the dynamic response
has been computed after measuring the amplitudes and frequencies of the system modified with the perturbation cylinders,
located at different positions of the grid described in Section 2. Fig. 3 shows maps of variations in amplitude and frequency,
as a function of the perturbation cylinders location, (xc , yc ), for the selected values of Ur . The change in the dynamic response
is quantified through the variables A = (A A0 )/A0 and f = (f f0 )/f0 , where the subscript 0 denotes in the following,
the corresponding reference values of dynamic response without perturbation cylinders.
Note that the effect of the perturbation cylinders supports was found to be negligible, since the differences between
values of A0 and f0 obtained with and without the holding arrangement were always below 5%.
The maps in Fig. 3 show that the dynamic response sensitivity is highly dependent on the wake regime and response
branch. Interestingly, reductions of amplitude reach values of 0.25 regardless of the Ur considered.
Regions of amplitude reduction are much more easily identifiable for the case with Ur = 6.1. In this case, three major
zones can be distinguished, as Fig. 4(a) shows. First sensitive region, R1, is located along the body surface for a r 0.6 D and
angles in the range 40 90 . A second well defined region, R2, is found slightly further away from the body surface, at
a r 0.9D, with sensitivity values especially large for 5 25 . Finally, the third radial region, R3, is located at r 1.4,
also for small angular values.
The first region, R1, near the surface of the body, with values of angular position extending in the last part of the first
quadrant suggests that the perturbation acts on the boundary layer and its separation region. The other regions of large
sensitivity, R2 and R3, are located closer to the symmetry axis at y = 0, suggesting the perturbation acts in the zones where
 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363 57

Fig. 3. Sensitivity of dynamic response to localized perturbation, as a function of the perturbation cylinders position (xc , yc ): (left column) amplitude
sensitivity A (xc , yc ), and (right column) frequency sensitivity f (xc , yc ), for (a) Ur = 6.1 (Re 13 400) and (b) Ur = 9.3 (Re 20 500). Maps are
obtained using a perturbation cylinder of d = 0.032D.

Fig. 4. (a) Identification of regions of largest amplitude reduction for Ur = 6.1, and (b) selected control cylinders location, (xc , yc ), employed in Section 4.

the shear layers at both sides of the wake interact, and therefore the mechanism is related to the modification of the vortex
formation region, with a length scale usually of order O(Lf ) D, typically shorter than that for stationary cylinders (Bearman,
1984; Williamson, 1996). Interestingly, regions R1 and R2 fairly covers the angular range identified by Sakamoto and Haniu
(1994) for which the presence of a control cylinder reduces mean and fluctuating values of forces coefficients. In that sense,
especially remarkable is the coincidence in the loci of largest amplitude reduction along the cylinder surface, found here at
= 90 and = 25 (see Fig. 4(a)), and their observations for maximum variations of time-averaged lift coefficient (see
Fig. 6 therein), which they indeed interpreted, respectively, in terms of modifications in the separation point location and
the shear layer interaction process.
Regarding the map of A at Ur = 9.3 in row (b) of the figure, there are no clear coherent areas as in the previous
case. This fact might be related to a more complex wake topology and flow three-dimensionality at larger Re and Ur , as
strong modulations of amplitude were observed. Notwithstanding, there is a global stabilizing effect of vibrations when
perturbation cylinders are placed downstream for angles < 80 , with the most sensitive areas located close to r 1.5,
as happened in the case at Ur = 6.1.
Conversely, there is a major region around = 90 whose effect is destabilizing. This is more evident for the case with
Ur = 9.3, where a relative growth of 0.1 in amplitude is observed. Interestingly, the growth of amplitude in this region is
roughly concomitant with reduction in the oscillation frequency, as the map of f for Ur = 9.3 depicts, where there is a fair
overlapping between areas of amplitude and frequency variations of opposite sign, indicating that a slower response motion
arises, when the characteristic oscillations length scale grows. However, this correlation does not apply generally for the rest
of the domain investigated, nor seems clear for Ur = 6.1, suggesting that the modification of the shedding process induced
by wake perturbation is affecting strongly the oscillation period.
The amplitude response is shown to be quantitatively more sensitive to the presence of perturbation cylinders than
the frequency response. Temporal periods of 60 s have been employed to obtain the PSD(F ) distributions, providing with a
58 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363

Table 1
Locations selected for control cylinders of d = 0.12D, following sensitivity
maps in Fig. 3.
Point r Region
#1 0.605 90 R1
#2 0.605 40 R1
#3 0.905 20 R2
#4 1.305 20 R3
#5 1.305 0 R3

Fig. 5. Temporal amplitude response, A (t ), (left column) and corresponding Power Spectral Density, PSD(F ), (right column), of the flexibly-mounted
system at Ur = 6.1 (Re 13 400), when control cylinders of d = 0.12D are placed at positions listed in Table 1: (a) #1, (b) #2, (c) #3, (d) #4 and (e) #5.
Dashed lines represent the value of uncontrolled amplitude A0 and its corresponding PSD.

minimum frequency resolution of F = 0.017. These facts lead to frequency maps that are less resolved than the amplitude
maps, especially at the wake of Ur = 6.1, where values of f are smaller.

4. Control of VIV based on sensitivity maps

The previous results of VIV sensitivity (Fig. 3) obtained with perturbation cylinders, although significant, might be limited
if thinking in practical problems. This was a first step in order evaluate regions of high VIV sensitivity, with potential for VIV
reduction and attenuation, but outside the laboratory such small perturbations are impractical. Here, we evaluate the effect
of different control cylinder configurations using cylinders of larger diameter, with d = 0.12D. Moreover, the controllability
of the oscillating response at the upper branch has been shown to be more localized around coherent regions, that is why
we have decided to choose five specific locations with large sensitivity on the map of A at Ur = 6.1. For these cases, DPIV
interrogations were carried out, so the mechanisms underlying the sensitivity can be better inferred. The positions are listed
in Table 1, and correspond to loci of maximum amplitude reduction within the three coherent regions previously identified,
as Fig. 4(b) displays.
Fig. 5 depicts temporal series and Power Spectral Density distributions of dynamic response for the system with the
control cylinders, when they are placed, respectively, at points #1 (Fig. 5(a)), #2 (Fig. 5(b)), #3 (Fig. 5(c)), #4 (Fig. 5(d)) and
#5 (Fig. 5(e)). The reference value, obtained with the plain cylinder A0 , is also shown for comparison using dashed lines. As
suggested by the time series of A , there are remarkable reductions of oscillation for all configurations, although they are
especially strong when control cylinders are placed at points #2 and #3, coincident with the highest sensitivity in the map
of Fig. 3(a). This confirms previous results on VIV reduction obtained by Zhu et al. (2015). The reduction of vibrations is also
observable in the PSD, where all peaks are smaller than the one corresponding to the isolated cylinder shown with the dashed
lines. Variations of frequency are also identified and, in particular, there is a noteworthy growth in f for the configuration
#1, as the frequency sensitivity map shows (Fig. 3(a), right). A good agreement is also found for the rest of configurations,
including points #2, where the oscillation frequency slightly decreases (note that in the vicinity of such location, there is a
slight positive sensitivity contribution to frequency that is counterbalanced by a stronger negative sensitivity, affected as
well by the presence of control cylinders).
The quantification of the amplitude and frequency variations obtained by means of control cylinders is shown in Fig. 6, for
both Ur = 6.1 and Ur = 9.3, along with their corresponding reference values from the isolated cylinder test. For the lowest
 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363 59

a b

Fig. 6. Amplitude and frequency responses of a flexibly-mounted cylinder with and without control cylinders of d/D = 0.12, located at positions listed in
Table 1, for: (a) Ur = 6.1 (Re 13 400) and (b) Ur = 9.3 (Re 20 500).

value of Ur (Fig. 6(a)), configuration #2 is clearly the most efficient one, since a 66% of amplitude reduction is achieved,
accompanied with a slight decrease in f . Similarly, configuration #3 nearly halves the value of A , while the rest of control
arrangements provide with reductions of approximately 30% or larger. The least efficient option is however #1, with control
cylinders placed at = 90 , where there is an increase of response frequency. A similar result was found by Korkischko
and Meneghini (2012), which, in spite of using a slightly different arrangement, i.e. d = 0.06D and a gap g between main
and control cylinders g = 0.07d (here g = 0.04d in case #1); obtained likewise an increase in f , which seemed to be
linked to a shorter recirculation bubble and formation length induced by control cylinders. As predicted by sensitivity maps
in Section 3, at Ur = 9.3 (Fig. 6(b)), configuration #1 leads to a 13% increase in VIV amplitude, while frequency decreases.
Remaining configurations give rise to a considerable decrease of oscillations, with a remarkable weakening of approximately
56% for the case #2. Response frequency is also significantly reduced in cases #2 and #3, while #4 and #5 are indeed less
affected, as predicted by the sensitivity map of Fig. 3(b).
In view of these results, increasing the size of the cylinders in the wake may contributes positively to the control
of the dynamic response which, in the present case, is certainly attenuated with respect to that obtained through the
sensitivity analysis. However, this positive impact of increasing diameters might be eventually limited by non-local,
nonlinear mechanisms that may be triggered if the cylinder size becomes too large. In any case, the study of the effect of the
control cylinders diameter is beyond the scope of the study presented here, which was obviously focused on the sensitivity
analysis of VIV response.
The proposed control arrangements have been shown to efficiently reduce the cylinder oscillations amplitude, regardless
of the response branch. An exception is however configuration #1, whose stabilizing performance seems highly dependent
on the value of Ur , and therefore, on the flow regime.
In order to better understand the mechanisms that lead to the changes in the VIV response and to gain a deeper insight
into the resulting wake topologies, we next present and discuss results based on DPIV wake interrogations.

4.1. Wake visualizations

With the objective to evaluate the wake topology modifications induced by the control cylinders upon the oscillating
system, we conducted several DPIV measurements for cases with reduced velocities Ur = 6.1 (Re 13 400), where the
maximum value of A was reached. We have limited our study to control positions #2 and #3 (see Table 1), where the
largest amplitude reductions were observed, as shown in Fig. 6. Spanwise vorticity distributions have been computed from
measured velocity fields, and made nondimensional using the free-stream velocity and the cylinder diameter, z = z D/u .
To set a reference for comparison, we display first vorticity distributions corresponding to the isolated plain oscillating
cylinder. Fig. 7 shows a time sequence of phase-averaged maps corresponding to eight different times t (phases ) of an
oscillation period, i.e. from t = 0 to t = 1, taken every eighth of the period. The temporal evolution of the amplitude
appears in the top plot of Fig. 7, and was obtained by image processing the DPIV images, tracking the cylinder displacement.
Each phase-averaged vorticity map is obtained using at least 7 instantaneous fields matching the same phase in different
oscillation cycles. Fig. 7 displays a typical shedding process with a single vortex at each side of the wake, being convected
downstream while the cylinder oscillates. The formation of vortices occurs in the very near wake, barely spanning one
60 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363

Fig. 7. Time sequence of a vibration cycle at Ur = 6.1 (Re 13 400): phase-averaged vorticity fields around a flexibly-mounted isolated cylinder, for
different instants t (phase ) of one oscillation period: (a) t = 0 ( = 0), (b) t = 1/8 ( = /4), (c) t = 1/4 ( = /2), (d) t = 3/8 ( = 3/4), (e)
t = 1/2 ( = ), (f) t = 5/8 ( = 5/4), (g) t = 3/4 ( = 3/2), (h) t = 7/8 ( = 7/4). Top figure depicts temporal signal of A and corresponding
identification of phase values shown in (ah).

diameter downstream of the body, where both shear layers interact. At t = 0 ( = 0, Fig. 7(a)), where the cylinder is
at its equilibrium point, the bottom shear layer, with counter clockwise vorticity, is folded upwards due to the cylinders
downward displacement. Besides, a clockwise vortex has already been shed from the upper shear layer and travels
downstream. The interaction between opposite sign vorticity, breaks the bottom shear layer and an incipient vortex forms,
while the cylinder descends (at t = 1/8 and 1/4 in Fig. 7(b)(c)), being detached at t = 3/8 (Fig. 7(d)), time at which the
cylinder begins to rise again. At this point, the top shear layer is deflected downwards and is trapped between the cylinder
and the bottom vortex. Similarly, at t = 1/2 ( = , Fig. 7(e)) the vortex formation process begins at the upper shear
layer, giving birth to the alternate shedding mode in the wake. In general, vortices diffuse in the wake to very low levels of
vorticity after one oscillation cycle, and at a distance of x 6 vortices have practically vanished.
When the control cylinders are placed in the wake, the vortex shedding process is highly altered. Fig. 8 shows the time
sequence of an oscillation period for the system with control cylinders at location #2, i.e. (r , ) = (0.605, 40) , along with
the temporal evolution of A (Fig. 8 top), where the strong reduction of amplitude is evidenced, if compared with the data
shown in Fig. 7 for the reference cylinder. Control cylinders alter the boundary layer separation point and modify the angle
of separation forming the shear layers in the wake. Moreover, conversely to what was observed for the plain cylinder, the
near wake appears elongated and extends almost two diameters downstream of the body, with the shear layer interaction
region shifted away from the model, to a position x 1.5. Interestingly, shear layers do not fold and wrap around the
cylinder anymore. The cylinder now appears more streamlined and the visualization suggests that the oscillating dynamics
 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363 61

Fig. 8. Time sequence of a vibration cycle at Ur = 6.1 (Re 13 400): phase-averaged vorticity fields around a flexibly-mounted system with control
cylinders at (r , ) = (0.605, 40) , i.e. position #2 in Table 1, for different instants t (phase ) of one oscillation period: (a) t = 0 ( = 0), (b)
t = 1/8 ( = /4), (c) t = 1/4 ( = /2), (d) t = 3/8 ( = 3/4), (e) t = 1/2 ( = ), (f) t = 5/8 ( = 5/4), (g) t = 3/4 ( = 3/2), (h)
t = 7/8 ( = 7/4). Top figure depicts temporal signal of A and corresponding identification of phase values shown in (ah).

are weaker. Besides, once the vortices are shed, the spanwise extent of the wake reduces, and the magnitude of vorticity
becomes lower and diffuses more rapidly if compared to the uncontrolled case.
The shedding mechanism is more complex when control cylinders are placed at position #3, i.e. (r , ) = (0.905, 20) ,
as Fig. 9 shows. As the cylinder descends (i.e. t = 0 and 1/8 in Fig. 9(a)-(b)), the bottom shear layer is split by the presence of
the lower control cylinder (whose local wake is clearly observable) and inclined towards the cylinder base, getting trapped
between the two control cylinders, forcing its angle of inclination. At t = 1/4 (Fig. 9(c)), the cylinder is decelerating and
the inertia displaces the bottom shear layer downwards, while the top control cylinder provokes the detachment of an
isolated vortex, which is subsequently shed into the wake as the cylinder begins its rising motion (t = 3/8, Fig. 9(d)).
Meanwhile, the upper shear layer is split by the top control cylinder as it leans toward the rear of the main cylinder, starting
the corresponding formation and shedding of clockwise vorticity that spans the next half period (Fig. 9(e)-(h)). Although the
shedding process holds similarity with that described in Fig. 7 for the plain cylinder, the level of vorticity in the near wake
and along the rear surface of cylinder is generally lower (at least for an important interval of the oscillation period). That
may diminish the magnitude of forcing, giving rise to a weaker oscillating motion (Fig. 9 top).
In view of the DPIV results, it seems evident that stabilizing mechanisms are very different in the two cases of study. The
data suggests that the control cylinders in position #2 act by modifying the separation points and the angle of separation
of the boundary layers, creating a vorticity-dead region close to the cylinder, presumably modifying the lift magnitude and
phase between forcing and response, in a similar manner as tripping rods (Hover et al., 2001; Rodrigues Quarante and
62 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363

Fig. 9. Time sequence of a vibration cycle at Ur = 6.1 (Re 13 400): phase-averaged vorticity fields around a flexibly-mounted system with control
cylinders at (r , ) = (0.905, 20 ), i.e. position #3 in Table 1, for different instants t (phase ) of one oscillation period: (a) t = 0 ( = 0), (b)
t = 1/8 ( = /4), (c) t = 1/4 ( = /2), (d) t = 3/8 ( = 3/4), (e) t = 1/2 ( = ), (f) t = 5/8 ( = 5/4), (g) t = 3/4 ( = 3/2), (h)
t = 7/8 ( = 7/4). Top figure depicts temporal signal of A and corresponding identification of phase values shown in (ah).

Nishi, 2014). The configuration used for case #3 splits the shear layers and strongly alters the vortex formation process
stemming from the oscillating motion, while conditioning vortex formation length because of the control cylinders location.
These overall near wake modifications might lead to important cylinder forcing changes (in amplitude and phase), and
consequently, to a weaker dynamic response. For instance, in the case of a fixed cylinder, the measurements from Sakamoto
and Haniu (1994) showed that the presence of a control cylinder in the near wake, i.e. 20, leads to weaker oscillations
of drag and lift coefficients. However, due to the complexity induced by the cylinder motion, that hypothesis can be
only confirmed by means of force measurements, which were not available for the present experimental arrangement.
Notwithstanding, visualizations have allowed the identification of distinct physical mechanisms, that help to explain the
existence and the origin of the different coherent zones in the amplitude sensitivity maps presented in Section 3.

5. Conclusions

We have obtained experimentally the sensitivity of vortex-induced vibrations (VIV) to localized perturbations, for a low
mass-damping one-degree-of-freedom circular cylinder (of diameter D) subject to uniform cross-flow, by parametrically
modifying the position of smaller perturbations cylinders in the wake of the flexibly mounted system. Sensitivity maps have
been obtained at two zones of the VIV response of the cylinder, in which amplitudes are known to be the largest, i.e. upper and
lower branches (Fig. 2). Amplitude and frequency maps (Fig. 3) show clearly coherent regions with high sensitivity where the
 J.I. Jimnez-Gonzlez, F.J. Huera-Huarte / Journal of Fluids and Structures 74 (2017) 5363 63

VIV response is largely reduced by placing a localized perturbation, whose size d 0.03D, is similar to the boundary layer
thickness around the cylinder at those Re investigated. That coherence and magnitude reduction are especially remarkable
at the upper branch region of the VIV response, and in particular, for the amplitude sensitivity (as evidenced in Fig. 3(a)).
Following the sensitivity analysis, we have explored the use of control cylinders with a larger size (d = 0.12D), in order to
validate sensitivity maps and explore their potential as basis for designing passive control strategies of VIV. We demonstrate
how by placing control cylinders in the regions identified with the highest VIV response sensitivity, important reductions of
VIV amplitude up to more than 65% can be achieved (especially for Ur = 6.1, which corresponds to the condition with the
largest uncontrolled amplitude response in the upper branch).
By using Digital Particle Image Velocimetry, we have elucidated the physical mechanisms that lead to the VIV amplitude
reduction with the control cylinders, that are dependent of the sensitivity region. The mechanisms are related to alterations
in either the boundary layers or shear layer detachment, or to strong modifications of the vortex formation region. Controlled
configurations give rise to weaker and narrower wakes with lower values of vorticity, that dissipate faster. Besides, these
changes in the vortex shedding and formation processes led to small variations of the cylinders oscillation frequency,
suggesting that the localized perturbations alter the hydrodynamic forcing frequency and phase.

Acknowledgements

This work has been supported by the Spanish MINECO (Subdireccin General de Gestin de Ayudas a la Investigacin),
Universidad de Jan and European Funds under Projects DPI2014-59292-C3-3 and DPI2015-71645-P and UJA2015/06/14
respectively.

References

Assi, G.R.S., Bearman, P.W., Kitney, N., 2009. Low drag solutions for suppressing vortex-induced vibration of circular cylinders. J. Fluids Struct. 25, 666675.
Bearman, P.W., 1984. Vortex shedding from oscillating bodies. Annu. Rev. Fluid Mech. 16, 195222.
Bearman, P.W., 2011. Circular cylinder wakes and vortex-induced vibrations. J. Fluids Struct. 27, 648658.
Bearman, P., Brankovic, M., 2004. Experimental studies of passive control of vortex-induced vibration. Eur. J. Mech. B Fluids 23, 915.
Blevins, R.D., Coughran, C.S., 2009. Experimental investigation of vortex-induced vibration in one and two dimensions with variable mass, damping, and
Reynolds number. J. Fluids Eng. 131, 101202101207.
Choi, H., Jeon, W.-P., Kim, J., 2008. Control of flow over a bluff body. Annu. Rev. Fluid Mech. 40, 113139.
Giannetti, F., Luchini, P., 2007. Structural sensitivity of the first instability of the cylinder wake. J. Fluid Mech. 581 (1), 167197.
Hill, D.C., 1992. A theoretical approach for analyzing the restabilization of wakes. AIAA Paper 92, 0067.
Hover, F.S., Tvedt, H., Triantafyllou, M.S., 2001. Vortex-induced vibrations of a cylinder with tripping wires. J. Fluid Mech. 448, 175195.
Khalak, A., Williamson, C.H.K., 1999. Motions, forces and mode transitions in VIV at low mass-damping. J. Fluids Struct. 13, 813851.
Korkischko, I., Meneghini, J.R., 2012. Suppression of vortex-induced vibration using moving surface boundary-layer control. J. Fluids Struct. 34, 259270.
Lee, K.Q., Aminudin, A., Naomi, K., Pauziah, M., Asnizah, S., Hanida, A., 2014. Investigation on the effectiveness of helical strakes in suppressing VIV of flexible
riser. Appl. Ocean Res. 44, 8291.
Marquet, O., Sipp, D., Jacquin, L., 2008. Sensitivity analysis and passive control of cylinder flow. J. Fluid Mech. 615, 221252.
Meliga, P., Boujo, E., Pujals, G., Gallaire, F., 2014. Sensitivity of aerodynamic forces in laminar and turbulent flow past a square cylinder. Phys. Fluids (1994-
Present) 26 (10), 104101.
Meliga, P., Cadot, O., Serre, E., 2016. Experimental and theoretical sensitivity analysis of turbulent flow past a square cylinder. Flow Turbul. Combust. 97,
9871015.
Meliga, P., Chomaz, J.-M., 2011. An asymptotic expansion for the vortex-induced vibrations of a circular cylinder. J. Fluid Mech. 671, 137167.
Meliga, P., Chomaz, J.-M., Gallaire, F., 2011. Extracting energy from a flow: An asymptotic approach using vortex-induced vibrations and feedback control.
J. Fluids Struct. 27, 861874.
Meliga, P., Pujals, G., Serre, E., 2012. Sensitivity of 2-d turbulent flow past a d-shaped cylinder using global stability. Phys. Fluids (1994-Present) 24 (6),
061701.
Mettot, C., Sipp, D., Bzard, H., 2014. Quasi-laminar stability and sensitivity analyses for turbulent flows: Prediction of low-frequency unsteadiness and
passive control. Phys. Fluids 26, 045112.
Parezanovi, .V., Cadot, O., 2012. Experimental sensitivity analysis of the global properties of a two-dimensional turbulent wake. J. Fluid Mech. 693, 115149.
Rodrigues Quarante, L.A., Nishi, Y., 2014. Amplification/suppression of flow-induced motions of an elastically mounted circular cylinder by attaching tripping
wires. J. Fluids Struct. 48, 93102.
Sakamoto, H., Haniu, H., 1994. Optimum suppresion of fluid forces acting in a circular cylinder. J. Fluids Eng. 116, 221227.
Snchez-Sanz, M., Velazquez, A., 2009. Passive control of vortex induced vibration in internal flow using body shape. J. Fluids Struct. 27, 976985.
Sarpkaya, T., 2004. A critical review of the intrinsic nature of vortex-induced vibrations. J. Fluids Struct. 19, 389447.
Schlichting, G., 1979. Boundary Layer Theory, seventh ed. McGraw-Hill Book Company.
Strykowski, P., Sreenivasan, K., 1990. On the formation and suppression of vortex shedding at low Reynolds numbers. J. Fluid Mech. 218, 71107.
Willert, C., Gharib, M., 1991. Digital particle image velocimetry. Exp. Fluids 10, 181193.
Williamson, C.H.K., 1996. Vortex dynamics in the cylinder wake. Annu. Rev. Fluid Mech. 28, 477539.
Williamson, C.H.K., Govardhan, R.N., 2004. Vortex-induced vibrations. Annu. Rev. Fluid Mech. 36, 413455.
Williamson, C.H.K., Roshko, A., 1988. Vortex formation in the wake of an oscillation cylinder. J. Fluids Struct. 2, 355381.
Zdravkovich, M.M., 1981. Review and classification of various aerodynamic and hydrodynamics means for suppressing vortex shedding. J. Wind Eng. Ind.
Aerodyn. 7, 145189.
Zhou, T., Razali, M., Hao, S.F., Cheng, L.Z., 2011. On the study of vortex-induced vibration of a cylinder with helical strakes. J. Fluids Struct. 27, 903917.
Zhu, H., Yao, J., 2015. Numerical evaluation of passive control of VIV by small control rods. Appl. Ocean Res. 51, 93116.
Zhu, H., Yao, J., Ma, Y., Zhao, H., Tang, Y., 2015. Simultaneous CFD evaluation of VIV suppression using smaller control cylinders. J. Fluids Struct. 57, 6680.