Hindawi Publishing Corporation

Advances in Mechanical Engineering

Volume 2013, Article ID 204974, 12 pages
http://dx.doi.org/10.1155/2013/204974

Research Article
Vortex-Induced Vibrations of a Square Cylinder with
Damped Free-End Conditions

S. Manzoor,1 J. Khawar,2 and N. A. Sheikh3

1
Department of Mechanical Engineering, University of Engineering and Technology, Taxila, Pakistan
2
Beijing University of Aeronautics and Astronautics (BUAA), Beijing, China
3
Department of Mechanical Engineering, Muhammad Ali Jinnah University, Islamabad, Pakistan

Correspondence should be addressed to N. A. Sheikh; [email protected]

Received 28 November 2012; Accepted 28 February 2013

Academic Editor: Yi Wang

Copyright © 2013 S. Manzoor et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The authors report the results of vortex-induced vibrations of a square cylinder in a wind tunnel. This constitutes a high mass ratio
environment. The square cylinder is mounted in the wind tunnel in such a fashion that it only performs rigid body oscillations
perpendicular to the flow direction with damped free-end conditions. This physical situation allows a direct evaluation for analytical
models relying on simplified 2D assumptions. The results are also compared with two-dimensional fluid-structure (CFD-CSD)
numerical simulations. The comparison shows that despite having one-dimensional motion, the analytical model does not predict
the VIV region with correctness. Results show that the numerical simulations and experimental results differ from the analytical
model for the prediction of reduced velocity corresponding to peak amplitude. Also the analytical reduced velocity envelope is
underpredicted compared to both numerical simulations and experimental data despite the structure being lightly damped. The
findings are significant as the experimental results for freely oscillating high mass ratio body show differences from the low mass
ratio especially in the transition between VIV and galloping regions. However the numerical simulations show comparatively close
agreement.

1. Introduction on both fixed and forced oscillating square cylinders. They

determined that the amplification of the fluctuating lift
Vortex shedding excitation of a cylinder is probably one of coefficient for a square cylinder at lock-in was much less than
the most studied problems in flow-induced vibrations. This that of a circular cylinder subjected to similar conditions.
mechanism, referred to as vortex-induced vibration (VIV), Ongoren and Rockwell [6] have studied the effects of cylinder
occurs when the vortices, developed in the wake, can couple inclination with respect to the mean free stream, using a
with the dynamics of the cylinder. It can be seen roughly as a forced circular cylinder in a water channel. Ongoren and
resonance mechanism appearing when the frequency of the Rockwell [7] repeated some experiments with square and
vortex shedding, controlled by the fluid flow, is close to the triangular cross-section cylinders. Their experimental setup
natural frequency of the cylinder. However, the physics of involved vertically arranged cylinders subjected to low mass
the interaction between the flow and the cylinder transverse ratio environment. Williamson and Roshko [8] provided the
motion is not simply linear. Mathematical modelling for mechanism of vortex formation and the underlying physics
predicting the amplitude of the cylinder motion has become a for mode shifts. Parkinson [9] resumed the phenomenology
widely studied problem in engineering. Wilkinson [1], Otsuki and the theoretical modelling tools available to understand
et al. [2], and Nakamura and Mizota [3] presented some the vortex-induced oscillations and the galloping instability
experimental data on the forced oscillations of square section in case of flow past bluff bodies. Brika and Laneville [10]
cylinders. Sarpkaya [4] presented a selective review of the studied a hollow slender cylinder in a wind tunnel and
then existing know-how about vortex-induced oscillations. showed that the cylinder’s steady response was hysteretic.
Bearman and Obasaju [5] studied the pressure fluctuations Each branch in the hysteresis loop is associated to either the
2 Advances in Mechanical Engineering

2S or the 2P mode of vortex shedding. Abrupt change in detailed discussion on the comparison of numerical simula-
the amplitude curve is attributed to the sudden mode shift. tions and experimental results is also included. At the end we
Govardhan and Williamson [11] presented the transverse conclude the findings of this research paper.
vortex-induced oscillations of an elastically mounted rigid
cylinder in a fluid flow. The authors point out that in a classical
high mass ratio system the initial and lower amplitude 2. Experimental Methods
branches can be distinctly identified due to a discontinuous
2.1. Wind Tunnel and Flexibly Mounted Cylinder. The square
mode transition. In case of lower mass ratio systems a further
cylinder has a cross-section, 𝐷 = 0.02 m, and a span, 𝑏 =
upper amplitude branch is clearly identifiable attributed to a
0.15 m. The cylinder is put in place in the test section using
second instance of mode transition. Hémon and Santi [12]
four linear springs mounted outside the test section. Specific
submitted experimental and numerical results on the aero-
chord wiring is used to restrain the cylinder such that it
elastic behavior of slender rectangular and square cylinders
oscillates only transverse to the air flow executing rigid body
subjected to a cross flow. Morse and Williamson [13] discov-
motion. The use of linear spring mountings allows a damped
ered the 2Poverlap mode using high-resolution data from
free-end condition for the cylinder rather than conventional
a forced oscillating cylinder at a fixed Reynolds number.
fixed and/or cantilever-type boundary conditions. Special
Manzoor et al. [14] presented their experimental results
attention is paid to keep the structural damping as low as
based on the free oscillations of a square cylinder in a high
possible. The experiment is conducted in a vertical axis Eiffel
mass ratio environment. Vortex shedding mode switch at the
Wind Tunnel which has a closed circular test section of a
resonant velocity was identified. Hysteresis in the reduced
diameter of 0.20 m. A centrifugal fan, downstream of the test
amplitude curve was observed for one of the experimental
section, is used to produce the wind stream. This free stream
cases. The same author has presented a possible theoretical
flow velocity can be safely assumed to be uniform over the
explanation of the experimental results in Manzoor et al.
cylinder span given the comparison between the cylinder
[14]. The reduced-order van der Pol model proposed by
span and the test section diameter. Mean velocity in the test
Facchinetti et al. [15] was used. Distinct limitations of the van
section is varied from 2 m/sec to ∼7 m/sec. Turbulence level
der Pol model in predicting the VIV behavior have been iden-
of the upstream airflow is measured and found less than
tified. Sen and Mittal [16] recently submitted their numerical
1% over the velocity range during the course of this study,
investigation of the two dimensional behavior of a square
Amandolèse and Hémon [19].
cylinder oscillating freely in a low mass ratio environment.
For this experimental setup no endplates are attached
The authors have neglected the effect of structural damping in
to the test structure. Due to the high aspect ratio of the
their study while varying the Reynolds number in the lower
cylinder used in this experiment (𝑏/𝐷 = 7.5), flow around the
range.
longitudinal ends can have a significant effect on the vortex
As catalogued above, almost all of the work being
dynamics, the correlation of the induced fluid forces on the
reported is largely focused on forced oscillations of cir-
body, and thus the vibrations. However the proximity of both
cular cylinders in setups with lower mass ratios generally.
the ends of the cylinder with the test section wall reduces
Experimental data on noncircular bluff bodies is obtained
the effect of the end conditions. As reported by Morse et al.
either with lower mass ratio configurations or with forced
[20], for a circular cylinder the vortex-induced vibrations for
oscillations mechanisms. Mostly, the experimental setups
attached and unattached endplates are nearly the same.
demonstrate bending modes of vibration rather than rigid
body motion. In this study we use an experimental setup
which allows only the rigid body modes of oscillations 2.2. Measurement System. At low flow velocity accurate
with damped-free ends, thereby providing ground for direct measurement is always challenging. In the present case, a
testing of analytical and numerical models relying on planar nozzle is mounted downstream of the test section. Using
kinematics. This work contributes also in the sense that we Bernoulli’s theorem we calculate the airflow velocity in the
use experimental results for a more realistic configuration test section and compare it with the flow velocity in the
in comparison with the numerical simulations and analytical nozzle section. Pressure readings in the test section and the
predictions for a practical real-life problem, civil engineering nozzle section are obtained by using pairs of static pressure
projects for example. For comparison with experimental data, taps in each section. A thermocouple is also employed for
analytical as well as numerical models are used to simulate temperature correction. This technique allows us to measure
the free vibration of lightly damped square section with free low flow velocities with accuracy better than 1%, Manzoor et
ends. To best of our knowledge this work is the first attempt al. [14].
to simulate and understand the physics of square section Measurements of the transversely displacement of the
oscillation with free-end conditions. cylinder are obtained by a laser displacement sensor. The
The organization of this paper is as follows. First we intro- measurement resolution is 40 𝜇m and the accuracy is 1%
duce the experimental setup and the details of measurement of the full scale range. Signals from the laser displacement
system with details of system parameters. This is followed sensors are transmitted to an acquisition system named
with experimental results along with comparison of analytical PAK provided by Mueller-BBM. It consists of an 8-channel
model predictions for the setup. In the next section the details acquisition card of 24 bit with a signal processing software.
of the numerical simulation model and the parameters for the Sampling resolution is 1024 Hz and the typical duration of
coupled fluid structure interaction simulations are detailed. A data accumulation is 60 s. However this duration can be
Advances in Mechanical Engineering 3

Table 1: Physical parameters of the experiments.

Diameter of the cylinder 𝐷 20 Mm

𝑈 𝑈 Length of the cylinder 𝐿 150 Mm
Stiffness of the setup 𝐾 597.6 ± 35 N/m
200 Cable Spring
Mass of the cylinder 𝑀 0.0654 ± 0.004 Kg
150 Critical damping 𝑐𝑐 12.5 ± 0.75 N⋅s/m
Structural damping 𝑐 0.0104 ± 0.0008 N⋅s/m
𝐷 𝑓𝑜 15.21875 ± 0.01563 Hz
Square Spring Natural frequency
cylinder Wind tunnel velocity 𝑈 1.5–6.0 m/s
𝑧(𝑡)
Air density 𝜌 1.205 kg/m3
Front view Side view Kinematic viscosity 𝜈 15𝑒−6 m2 /s

Figure 1: Sketch showing the principles of the experimental setup

Table 2: Nondimensional parameters.
[14].
Reynolds
Re 𝑈 𝐷/𝜈 2000–8000
number
increased to 300 s to enable the frequency measurements of Mass ratio 𝑚∗ 𝑚/𝜌 𝐷2 𝐿 905
the limit cycle oscillation (LCO) of the cylinder. Increasing Damping ratio 𝜂 𝑐/𝑐𝑐 0.000828 ± 0.000014
the time duration also enabled us to obtain a better frequency
Scruton number Sc 2𝜂𝑚∗ 1.498
resolution of the measurements. The physical degree of
Strouhal
freedom “𝑧” is obtained through the recombination of the St 𝑓𝑤 𝐷/𝑈 0.127
number
measured signals using the system kinematics, Manzoor et al.
Skop-Griffin
[14]. 𝑆𝐺 4𝜋2 St2 Sc 0.954
parameter
Reduced
2.3. Structural Parameters of the Experiments. The structure 𝑈𝑟 𝑈/𝑓0 𝐷 5–20
velocity
is a rigid square cylinder with the length of the side “𝐷”. The
external attachments of the cylinder makes it constrained to
oscillate in a direction normal to the mean flow direction as 905 associated with a very low damping ratio 𝜂 ≈ 0.0828%.
depicted in Figure 2. Very low damping leads to a relatively small Scruton number,
Flow is assumed to be uniform all along the cylinder Sc ≈ 1.5 which is the key parameter in the observation of
length. The cross flow displacement “𝑧” of the cylinder can vortex shedding vibrations.
be described by the standard damped linear oscillator: Unsteady wake measurements have also been performed
in order to measure the Strouhal number of the cylinder at
𝑚𝑧̈ + 𝑟𝑧̇ + 𝑘𝑧 = 𝑆. (1)
rest. Measurements were obtained using a single-component
The undamped natural frequency of the system is defined hot wire anemometer installed downstream of the square
as cylinder at a distance equal to one length of the cylinder side.
For the system under consideration, the Strouhal number
1 √𝑘 was found to be 0.127 over the velocity range of the vortex
𝑓𝑜 = . (2) shedding oscillation regime. This is in accordance with
2𝜋 𝑚
Norberg’s data for low Reynolds number, Norberg [21].
Here structural parameters are obtained under zero cross
wind velocity. The natural frequency 𝑓𝑧 and the system 2.4. Experimental Results. Experimental results presented in
damping 𝑟𝑠 are obtained using spectral analysis; Manzoor Figure 3 show limit cycle oscillations (LCO) of a square
et al. [14]. A static weight calibration technique is used to cylinder measured for reduced velocity ranging from 𝑈𝑟 ≈
calculate the stiffness 𝑘. Mass 𝑚 is then calculated using: 6 ≈ 16. Different symbols in the figure signify different
experimental runs conducted at different times to ensure
𝑘 repeatability of the experimental procedure. Apart from the
𝑚= 2
. (3)
(2𝜋𝑓𝑜 ) obvious dispersion of experimental points at higher reduced
velocities we can safely assume that the resonant frequencies
The reduced velocity, 𝑈𝑟 , is given as 𝑈𝑟 = 𝑈/𝑓𝑜 𝐷 and lie approximately in the same reduced velocity range for
the RMS amplitude, 𝑍∗ , is given as 𝑍∗ = 𝑧/𝐷. Structural each experimental run. The result can be explained by the
supports are assumed to behave linearly throughout the type-II frequency coalescence mechanism, Shehryar [17]. As
amplitude envelope. Important physical parameters of the obvious from the figure, no significant oscillations take place
experiment are resumed in Table 1. Pertinent nondimen- at lower reduced mean free stream velocities, 𝑈𝑟 < 6, when
sional parameters are also reported in Table 2. From the two the two natural frequencies of the system are far apart. The
tables, it is clear that the system has a high mass ratio 𝑚∗ ≈ fluid structure interaction mechanism is dominated by the
4 Advances in Mechanical Engineering

𝑞(𝑡)
∙

Figure 2: Schematic for square cylinder coupled wake oscillator for 2D vortex-induced vibrations in a vertical wind tunnel [17].

inertial effects due to the incoming flow. As the 𝑈𝑟 increases, galloping instability sets in; interestingly it starts immediately
the vortex shedding frequency downstream of the cylinder after the VIV region.
gets increasingly synchronized with the cylinder oscillating
frequency (6 ≤ 𝑈𝑟 ≤ 13). This results in an increasing 3. Comparison with the Analytical Model
energy transfer from the wake to the oscillating cylinder.
Facchinetti et al. [15] have stated that for a positive energy For a one-dimensional case the equation of motion of a rigid
transfer of this sort, 0 < (−𝜑/𝜋) < 1, where 𝜙 is the cylinder oscillating in the transverse 𝑧-direction (normal to
phase angle between the vortex shedding frequency and the flow) reads as (4), where “𝑚” is the mass of the cylinder,
the cylinder oscillating frequency. Increased energy transfer “𝑟” is a viscous damping coefficient, “𝑘” is the stiffness of the
shows as higher cylinder oscillating amplitude. The cylinder setup, and “𝑆” is the aerodynamic force resulting from the
exhibits maximum oscillation amplitude for a critical flow fluid structure interaction in this case, mainly due to vortex
velocity where the two frequencies of the system match shedding.
exactly. Generally, this point of maximum amplitude for a The mass “𝑚” takes into account the cylinder mass and
circular cylinder with higher mass ratio is achieved for a the fluid added mass 𝑚𝑓 which models the inviscid inertial
reduced upstream velocity close to (1/St), 𝑈𝑟 ≈ 8. For the effects. Following Blevins [18] we can define
square cylinder, however, it occurs at 𝑈𝑟 ≈ 10 which is slightly
more than (1/St) in our case. As the mean free stream velocity 𝑚 = 𝑚𝑠 + 𝑚𝑓 ;
is increased further, (𝑈𝑟 > 10), a sudden phase shift results
𝜋
in an abrupt change in the wake mode pattern, Khalak and 𝑚𝑓 = 𝐶𝑀𝑎 𝜌𝐷2 ;
Williamson [22] and Brika and Laneville [10]. This point can 4 (4)
be identified on the amplitude curves where there is a sudden 𝑚
𝜇=
change in amplitude. Shehryar [17] observed that for such 𝜌𝐷2
type-II fluid structure interaction systems, the phase between
the structure and wake oscillators showed an overall jump of
𝜋 passing through the lock-in range. 𝑟 = 𝑟𝑠 + 𝑟𝑓 ; 𝑟𝑓 = 𝛾Ω𝜌𝐷2 ; (5)
As the mean free stream velocity increases, a nega-
where “𝛾” is a stall parameter as Skop and Balasubramanian
tive energy transfer takes place accredited to the sudden
[23] and “Ω” is a reference angular frequency. Assuming that
phase shift. Consequently, the cylinder oscillations amplitude
the mean free stream velocity “𝑈” is zero, “Ω” is the angular
decreases, Figure 3, until a second critical value of the
frequency of the structure oscillation and “𝛾” is a function of
reduced mean free stream velocity is reached (𝑈𝑟 ≈ 14).
the amplitude of oscillations related to the mean structural
Experimental data presented in Figure 3 depicts amplitude
drag coefficient 𝐶𝐷, Blevins [18]. Blevins [18] defines “𝛾” for
values for a velocity just below this second critical value.
a circular cylinder as:
Beyond this second critical value of 𝑈𝑟 , the two system
frequencies bifurcate, Shehryar [17]. If the mean free stream 1
velocity is increased further, (𝑈𝑟 > 14), the secondary 𝛾= 𝐶 . (6)
4𝜋St 𝐷
Advances in Mechanical Engineering 5

0.2

0.15

0.1
𝑍∗

0.05

0
6 8 10 12 14 16 18
𝑈𝑟

Figure 3: Reduced RMS amplitude of the limit cycle oscillations

versus reduced velocity [17].

In our case, however, “Ω” is the vortex shedding angular

frequency, Ω = Ω𝑓 = 2𝜋St(𝑈/𝐷). “St” is the Strouhal Figure 4: Single degree of freedom galloping model. reproduced as
number. Following Skop and Balasubramanian [23], “𝛾” will blevins [18].
be assumed to remain constant for the sake of simplicity.
Govardhan and Williamson [11] used two distinct formula-
tions of the equation of motion for the “total force” and the study however is based on the vortex-induced vibrations of a
“vortex force”. Following the same approach the fluid effects, square section cylinder. It is imperative to remark here that
namely, the added mass and added damping, have been this basic difference in geometry has to be taken into account
included directly in the structural oscillator in (4) through (1) to be able to compare theoretical findings and experimental
and (2). The vorticity effects are modelled by the forcing term results later. Blevins [18] suggested a simplified system to be
at the right hand side in (4). Structural angular frequency Ω𝑠 considered, Figure 4.
is typically defined as Ω𝑠 = √𝑘/𝑚 and the reduced structural The angle of attack “𝛼” can be calculated as 𝛼 =
damping 𝜉 = 𝑟𝑠 /2𝑚Ω𝑠 . Equation (1) may be re written as −1
̇
tan (𝑧/𝑈) such that if 𝛼 = 0, 𝑧 = 0 implying that the
cylinder is at its equilibrium position. Also, if “𝛼” is very
𝛾 𝑆 ̇
𝑧̈ + (2𝜉Ω𝑠 + Ω ) 𝑧̇ + Ω2𝑠 𝑧 = . (7) small, 𝛼 ≈ 𝑧/𝑈 and 𝑈rel ≈ 𝑈. Now from Figure 4, the
𝜇 𝑓 𝑚 resultant vector is assumed to be positive downwards:
Facchinetti et al. [15] and de Langre [24] have used the 1
𝐹𝑧 = −𝐹𝐿 cos 𝛼 − 𝐹𝐷 sin 𝛼 = 𝜌𝑈2 𝐷𝐿𝐶𝑧 , (10)
standard van der Pol equation to model the fluctuating nature 2
of the vortex street downstream of the oscillating square where 𝐶𝑧 is the vertical force coefficient:
cylinder:
2
𝑈rel
2
𝑞 ̈ + (𝜀Ω𝑓 (𝑞 − 1)) 𝑞 ̇ + Ω2𝑓 𝑞 = 𝐹. (8) 𝐶𝑧 = − [𝐶𝐿 cos 𝛼 + 𝐶𝐷 sin 𝛼] . (11)
𝑈2
The fluid variable “𝑞” is defined as the reduced vortex Now because “𝛼” is very small, advancing the last equa-
lift coefficient. Mathematically, 𝑞 = 2(𝐶𝐿 /𝐶𝐿 𝑜 ). 𝐶𝐿 𝑜 is tion
the reference lift coefficient measured on a fixed cylinder 𝜕𝐶𝐿
subjected to vortex shedding. “𝐹” shall be “𝐴𝑧” for displace- 𝐶𝑧 (𝛼) = −𝐶𝐿 − (− + 𝐶𝐷) 𝛼. (12)
𝜕𝛼
ment coupling, “𝐴𝑧”̇ for velocity coupling, and “𝐴𝑧”̈ for
acceleration coupling. Facchinetti et al. [15] have presented a In the last equation if 𝜕𝐶𝑧 /𝜕𝛼 < 0 or equiva-
detailed study of the dynamics of such a coupled system. They lently (𝜕𝐶𝐿 /𝜕𝛼) + 𝐶𝐷 > 0, total damping of the system
estimated the near wake van der Pol parameter “𝜀” and the remains positive. The system remains stable. This is known
coupling force scaling “𝐴” from the available experimental as the Den Hartog’s Stability Criteria. Equation (12) at 𝛼 ≈ 0
data. The oscillation amplitude at lock-in, “𝑧𝑀”, is obtained gives 𝐶𝑧 = 𝐶𝐿 . Trivial algebraic manipulation of (10) and (12)
from yields
1 1 𝜕𝐶
𝐶𝐿 𝑜 /2 𝐴 𝐶𝐿 𝑜 /4 𝐹𝑧 = − 𝜌𝑈2 𝐷𝐿𝐶𝐿 + 𝜌𝑈𝐷𝐿 𝑧 𝑧.̇ (13)
𝑧𝑀 = √1 + . (9) 2 2 𝜕𝛼
𝑆𝐺 + 4𝜋2 St2 𝛾 𝜀 𝑆𝐺 + 4𝜋2 St2 𝛾
Comparing (1) and (13), retaining only lower-order terms
of 𝛼, we arrive at
3.1. Den Hartog’s Instability Criteria. It should be kept in
mind here that the model proposed by Facchinetti et al. [15] 1 𝜕𝐶𝑧
𝛾= , (14)
as discussed above is valid for circular cylinders. The present 8𝜋St 𝜕𝛼
6 Advances in Mechanical Engineering

0.18
0.16 2.35
0.14 2.3
0.12 2.25
0.1 2.2
𝑍∗

0.08 𝑞 2.15
0.06 2.1
0.04 2.05
0.02 2
0 1.95
6 7 8 9 10 11 12 13 6 7 8 9 10 11 12 13
𝑈𝑟 𝑈𝑟
(a) (b)

Figure 5: Hysteresis standard wake oscillator model solved numerically using velocity coupling as in Facchinetti et al. [15]. Solid line:
Increasing Velocity, dashed line: decreasing velocity. 𝑍∗ is the amplitude normalized with respect to the dimension “𝐷” of the square cylinder,
Figure 1, and 𝑞 is the reduced vortex lift coefficient.

0.2 suitable for our high mass ratio experimental setup. This
study therefore focuses on the velocity coupling model.
0.15 Following the same approach as Facchinetti et al. [15] and
the references there in, we solved the model numerically as
0.1 presented in (7) and (8) using the Fox and Goodwin iteration
𝑍∗

scheme. Details of the iteration scheme are described in

Shehryar [17]. System parameters are described in Tables 1
0.05
and 2, Figure 5.
We shall now compare our experimental results as pre-
0 sented with the theoretical wake oscillator model. As can be
6 8 10 12 14 16 18
observed from Figure 6, the wake oscillator model predicts
𝑈𝑟
the location of maximum limit cycle amplitude closer to the
Figure 6: Reduced RMS amplitude of the limit cycle ocillations expected value of (1/St) for this case. However, the extent of
starting from rest configuration. solid line: velocity coupling sim- frequency lock-in domain does not match the experimental
ulations of the wake ocillator model as presented in Facchinetti et result. We can safely state that only the lock-in amplitude
al. [15]. model in (9) provides acceptable results given that the added
aerodynamic damping effect due to bluff body cross section
(14) has been taken into account. de Langre [24] states that
the extent of the lock-in domain is in fact controlled by the
product AM where “𝐴” is the coupling force scaling in the
where (𝜕𝐶𝑧 /𝜕𝛼) is 3 for a square cylinder in smooth flow and wake oscillator model. Following Facchinetti et al. [15], we
3.5 for a square cylinder in turbulent flow, Blevins [18]. determine 𝐴 = 2.7 for our case. “𝑀” is a mass parameter
Investigations by Facchinetti et al. [15] revealed that such that
using acceleration coupling in the forcing term of the wake
oscillator best matches with the then available experimental 𝐶𝐿 𝑜 1
𝑀= . (15)
results. The coupling terms were formulated to essentially 2 2𝜋 St2 𝜇
2

allow only the linear functions of the wake variable “𝑞”

and displacement “𝑧” and their temporal derivatives. In 4. Comparison with Numerical Simulations
order to set the values of parameters “𝐴” and “𝜀”, the total
lift coefficient was assumed to be the same as the vortex For the current case we also carried out coupled fluid-
lift coefficient. Although Facchinetti et al. [15] recommend structure interaction of the square cylinder to resolve the
acceleration coupling as the most suitable, it is important physics of the flow and its effect on the rigid body oscillations.
to keep in mind that the theoretical model was devised for The flow physics shall enable us to visualize the dynamics
systems with lower mass ratios. Since this experimental work of wake structure and its influence on the structure through
involves a higher mass ratio system we refer to the lock- coupled simulation. For simplicity, the coupled model is
in domain calculations as a function of mass ratio by the reduced to 2D interaction only because of the nature of the
same authors. Given the seemingly asymptotic relationship vibration (one dimensional oscillation) and low turbulence.
between the upper and lower bounds of the lock-in region This section details the coupled computational fluid dynam-
in the velocity coupling model at lower mass ratios, we can ics (CFD) - computational structural dynamics (CSD) model
conclude that it is the velocity coupling model which is best and the simulation results.
Advances in Mechanical Engineering 7

4.1. Coupled CFD-CSD Model. The coupled CFD-CSD sim- is induced by the fluid itself without the requirement of
ulations are carried out using Fluent through its user an external plug or trigger. The asymmetry leads to the
defined function (UDF) feature. An in-house developed formation of an alternating vortex which is continuously
UDF developed using C-Language was linked with Fluent shed in the wake of square cylinder as shown in Figures
through Dynamic Link Library (DLL). The coupled CFD- 10 and 11. The resultant is the formation of a continuous
CSD system is modelled using loosely coupled partition vortex street behind the square section as seen in Figure 12.
procedure Matthies and Steindorf [25]. The time step size This asymmetric vortex shedding introduces asymmetric
chosen for unsteady CFD-CSD simulation is 2𝑒 − 4 sec based pressure distribution around the cylinder which introduces
on the vortex shedding frequency of the structure measured fluctuating lift force. This lift forces causes the square cylinder
experimentally. to vibrate. As the incoming velocity is varied the amplitude of
the lift force increases; this results in the higher amplitudes of
vibrations. However this does not continue for all incoming
4.2. Computational Fluid Dynamics Model. For computa-
velocities. At a certain critical velocity, the amplitude maxi-
tional fluid dynamics simulations the chosen domain is two
mizes to its highest value and after than the peak amplitude
dimensional containing 60000 cells and 60600 nodes as
decreases.
shown in Figure 7. For the square cylinder a solid boundary
A comparison of the peak amplitudes of oscillations pre-
nonslip wall condition was prescribed while pressure-far-
dicted by the coupled CFD-CSD simulations and experimen-
field boundary condition was used on the outer domain
tal data is shown in Figure 13. For illustration, comparison of
boundaries. The distance between the solid wall boundary
results predicted by the analytical model is already shown in
and the outer boundary of the domain was set at more than
Figure 6. It is evident that the analytical model is by far the
10 times the size of the square cylinder as shown in the
least demonstrative of the trend seen in the experiments. On
Figure 7. The grid size is selected after carrying out a grid
the other hand the coupled CFD-CSD simulations show good
independence study.
comparison with the peak amplitude prediction as well as the
The two-dimensional flow was modelled using viscous
width of the VIV region.
two-dimensional Navier strokes equations along with shear
stress transport (SST) kw turbulence model, Menter [26].
This particular turbulence model was used because of its good
performance for the flows with adverse-pressure gradients, 5. Discussion
Dina et al. [27]. The initial turbulence intensity was set
Different authors have published their work in attempts to
to be equal to one percent because the Reynolds number
enhance the wake oscillator model since it was proposed
based on the size of the square cylinder is in low Reynolds
by Currie and Hartlen [28]. Such models are based on the
number regime (2000∼6000). The numerical simulations are
reduced vortex lift coefficient manipulated in the form of
performed for various flow velocities ranging from 1.5 m/sec
a standard van der Pol equation to model the effects of
to 6 m/sec. Value for the air density was kept constant at
the vortex shedding on the solid object. The solid object in
1.205 kg/m3 with static air pressure of 101325 Pa to match the
turn is modeled in the form of a standard damped spring
experimental test conditions.
mass system. Forcing terms in both the equations impose
the necessary coupling. An important aspect associated with
4.3. Computational Structural Dynamics Model. The compu- using this type of theoretical models is the fact that the ratio
tational structural dynamics (CSD) model used for the square between the coupling force scaling “𝐴” and the near wake van
cylinder has a single degree of freedom in plunge direction der Pol parameter “𝜀” has to be fixed independently for each
“𝑧”. The mass “𝑚” of cylinder is 0.0654 kg with the stiffness case of a bluff body executing vortex-induced vibrations. This
“𝑘” of the cylinder is modeled using a lineal spring attached ratio governs the maximum limit cycle amplitude at lock-in.
to the center of the cylinder with stiffness of 597.6 N⋅m. The The extent of the lock-in domain is controlled by the product
value of structural damping “𝜉” coefficient was taken to be of the coupling force scaling “𝐴” and the mass parameter “𝑀”,
around 0.0104 to match the parameters of experiments. de Langre [24]. While our experimental results validate the
For the single degree of freedom a mass spring-damper model’s ability to predict the maximum lock-in amplitude,
system is modeled (see (1)) and solved numerically using the comparison reflects an important drawback regarding
fourth-step Rung Kutta (RK4) method with the lift force “𝑆” the extent of the lock-in range. Especially the model fails
as the input to the system. The lift force, generated due to to predict the complete VIV envelope in the post peak
airflow, was calculated by numerically integrating pressure, amplitude region as observed in Figure 3. It can also be
calculated through CFD, around the cylinder. noticed that the van der Pol model predicts a sudden decrease
The time-resolved coupled CFD-CSD simulation show in the amplitude of vibration after the point of peak vibration
that initially a separation bubble is formed downstream amplitude. This suggests that the force scaling parameter “𝐴”
the cylinder as shown in the velocity magnitude plot of in this region may not remain constant for the entire envelope
Figure 8. This velocity magnitude contour plot is shown for and requires further tuning. As described earlier, the change
free-stream flow of 5.5 m/sec at time 0.14 sec. As shown, in the amplitude curve can be linked with the mode shift
initially the contour plot is symmetric about the cylinder, in the vortex shedding pattern, Govardhan and Williamson
but after a certain time the separation bubble starts to [11]. However the van der Pol model does not account for the
become asymmetric as shown in Figure 9. The asymmetry change in the lift force scaling in the region.
8 Advances in Mechanical Engineering

Pressure far field Pressure far field

Solid wall

Figure 7: Illustrating the mesh around the square cylinder.

6.94𝑒+00
6.66𝑒+00
6.38𝑒+00
6.1𝑒+00
5.83𝑒+00
5.55𝑒+00
5.27𝑒+00
4.99𝑒+00
4.72𝑒+00
4.44𝑒+00
4.16𝑒+00
3.88𝑒+00
3.61𝑒+00
3.33𝑒+00
3.05𝑒+00
2.77𝑒+00
2.5𝑒+00
2.22𝑒+00
1.94𝑒+00
1.66𝑒+00
1.39𝑒+00
1.11𝑒+00
8.32𝑒−01
5.55𝑒−01
2.77𝑒−01
0𝑒+00
Contours of velocity magnitude (m/s) (time = 1.4𝑒−01) Mar 05, 2013
ANSYS FLUENT 12.1 (2d, pbns, dynamesh, sstkw, transient)

Figure 8: Velocity magnitude contour plot for free-stream velocity of 5.5 m/sec at time 0.14 sec.

Also, a closer inspection of our experimental results in the simplest of the structural model which only accounts for
Figure 3 shall reveal that the maximum lock-in amplitude the structural damping. Despite the nonlinearity, the model
lies at a slightly higher value of the reduced velocity than has been able to predict the maximum lock-in amplitude as
the expected (1/St). The wake oscillator model proposes well as the VIV envelope reasonably. The structural model can
that this maximum amplitude should lie indeed at reduced be upgraded to include other forms of damping and forcing
velocity closer to (1/St). This discrepancy in the location of terms to improve the comparison. Significantly the only
the maximum lock-in amplitude and the extent of the lock- difference is the coupled numeric simulation of the flow field
in domain may be connected. A detailed investigation of that links the wake dynamics with the oscillations. The wake
the blockage effects in this case may yield new experimental physics is highly complex especially with the introduction
evidence. At the same time extending the model to consider of turbulence. The wake oscillator model (Section 4) cannot
higher-order terms in the forcing functions could improve account for such nonlinearity.
the accuracy of theoretical curves. Another important tool As we know that the value of the van der pol parameter
which could be used is a visualization study. The ability to is tuned to obtain the range of reduced velocity, we suspect
simply see the vortex shedding patterns in real time could that the parameter is highly nonlinear and cannot be fixed
shed some more light on any existing phenomena near res- for the complete reduced velocity envelope. Rather we think
onance. One easier approach is to simulate the coupled fluid- that the parameter should evolve into a form at different
structure interaction. Here, we solved 2D Navier-Stokes with reduced velocities and mass ratios so that it accounts for the
Advances in Mechanical Engineering 9

6.91𝑒+00
6.64𝑒+00
6.36𝑒+00
6.08𝑒+00
5.81𝑒+00
5.53𝑒+00
5.25𝑒+00
4.98𝑒+00
4.7𝑒+00
4.42𝑒+00
4.15𝑒+00
3.87𝑒+00
3.59𝑒+00
3.32𝑒+00
3.04𝑒+00
2.77𝑒+00
2.49𝑒+00
2.21𝑒+00
1.94𝑒+00
1.66𝑒+00
1.38𝑒+00
1.11𝑒+00
8.3𝑒−01
5.53𝑒−01
2.77𝑒−01
0𝑒+00
Contours of velocity magnitude (m/s) (time = 1.6𝑒−01) Mar 05, 2013
ANSYS FLUENT 12.1 (2d, pbns, dynamesh, sstkw, transient)

Figure 9: Velocity magnitude contour plot for free-stream velocity of 5.5 m/sec at time 0.16 sec.

7.07𝑒+00
6.78𝑒+00
6.5𝑒+00
6.22𝑒+00
5.94𝑒+00
5.65𝑒+00
5.37𝑒+00
5.09𝑒+00
4.81𝑒+00
4.52𝑒+00
4.24𝑒+00
3.96𝑒+00
3.68𝑒+00
3.39𝑒+00
3.11𝑒+00
2.83𝑒+00
2.54𝑒+00
2.26𝑒+00
1.98𝑒+00
1.7𝑒+00
1.41𝑒+00
1.13𝑒+00
8.48𝑒−01
5.65𝑒−01
2.83𝑒−01
0𝑒+00
Contours of velocity magnitude (m/s) (time = 2𝑒−01) Mar 05, 2013
ANSYS FLUENT 12.1 (2d, pbns, dynamesh, sstkw, transient)

Figure 10: Velocity magnitude contour plot for free-stream velocity of 5.5 m/sec at time 0.2 sec.

wake dynamics especially when transition from laminar to held with the damped free-end conditions. The experimental
turbulent occurs. It is recommended that further theoretical findings show lock-in peak amplitude away from region of
investigation is conducted so as to find a better approximation reduced velocity predicted by the analytical models. On the
of the van der Pol parameter especially for the free vortex- other hand numerical simulations show peak amplitude and
induced oscillations of noncircular cross sections. In this VIV envelope corresponding well with the experimental data.
regard we suggest that reduced-order modelling technique However, the predictions of the analytical model show lesser
can be used. spread of VIV region with the peak lock-in amplitude in
accordance with the experimental findings. The comparison
6. Conclusion of analytical and numerical results suggests that the wake
model used for analytical lift force predictions needs revision.
The research paper summarizes the vortex-induced vibra- Especially the van der Pol parameter needs correction to
tions of a square cylinder subject to cross flow. The cylinder is incorporate the effects of turbulence experienced in the wake.
10 Advances in Mechanical Engineering

7.17𝑒+00
6.88𝑒+00
6.6𝑒+00
6.31𝑒+00
6.02𝑒+00
5.74𝑒+00
5.45𝑒+00
5.16𝑒+00
4.87𝑒+00
4.59𝑒+00
4.3𝑒+00
4.01𝑒+00
3.73𝑒+00
3.44𝑒+00
3.15𝑒+00
2.87𝑒+00
2.58𝑒+00
2.29𝑒+00
2.01𝑒+00
1.72𝑒+00
1.43𝑒+00
1.15𝑒+00
8.6𝑒−01
5.74𝑒−01
2.87𝑒−01
0𝑒+00
Contours of velocity magnitude (m/s) (time = 2.2𝑒−01) Mar 05, 2013
ANSYS FLUENT 12.1 (2d, pbns, dynamesh, sstkw, transient)

Figure 11: Velocity magnitude contour plot for free-stream velocity of 5.5 m/sec at time 0.22 sec.

8.52𝑒+00
8.18𝑒+00
7.84𝑒+00
7.5𝑒+00
7.16𝑒+00
6.82𝑒+00
6.48𝑒+00
6.14𝑒+00
5.8𝑒+00
5.45𝑒+00
5.11𝑒+00
4.77𝑒+00
4.43𝑒+00
4.09𝑒+00
3.75𝑒+00
3.41𝑒+00
3.07𝑒+00
2.73𝑒+00
2.39𝑒+00
2.05𝑒+00
1.7𝑒+00
1.36𝑒+00
1.02𝑒+00
6.82𝑒−01
3.41𝑒−01
0𝑒+00
Contours of velocity magnitude (m/s) (time = 3.2𝑒−01) Mar 05, 2013
ANSYS FLUENT 12.1 (2d, pbns, dynamesh, sstkw, transient)

Figure 12: Velocity magnitude contour plot for free-stream velocity of 5.5 m/sec at time 0.32 sec.

0.2

0.15
𝑍∗

0.1

0.05

0
6 8 10 12 14 16 18
𝑈𝑟

Figure 13: Comparison of coupled CFD-CSD simulations (solid line) with experimental data (symbols).
Advances in Mechanical Engineering 11

At the moment the van der pol parameter is tuned to adjust [7] A. Ongoren and D. Rockwell, “Flow structure from an oscil-
the spread of the region, but it needs better understanding lating cylinder. Part 2: mode competition in the near wake.,”
using first principles. We expect that in the near future Journal of Fluid Mechanics, vol. 191, pp. 225–245, 1988.
further numerical simulations and experimental results along [8] C. H. K. Williamson and A. Roshko, “Vortex formation in the
with reduced-order modelling may help in improving the wake of an oscillating cylinder,” Journal of Fluids and Structures,
formulation of the van der Pol parameter. vol. 2, no. 4, pp. 355–381, 1988.
[9] G. Parkinson, “Phenomena and modelling of flow-induced
vibrations of bluff bodies,” Progress in Aerospace Sciences, vol.
Nomenclature 26, no. 2, pp. 169–224, 1989.
𝑐: Structural damping [10] D. Brika and A. Laneville, “Vortex-induced vibrations of a long
𝑓𝑜 : Undampednatural frequency (Hz) flexible circular cylinder,” Journal of Fluid Mechanics, vol. 250,
𝑓: Frequency of the LCO (Hz) pp. 481–508, 1993.
𝑓𝑤 : Vortex shedding frequency (𝑓𝑤 = 𝑓𝑜 St𝑈𝑟 ) [11] R. Govardhan and C. H. K. Williamson, “Modes of vortex for-
St: Strouhal number mation and frequency response of a freely vibrating cylinder,”
𝐹: Lift force (N) Journal of Fluid Mechanics, vol. 420, pp. 85–130, 2000.
𝑓∗ : Reduced frequency of the LCO (𝑓∗ = 𝑓/𝑓𝑜 ) [12] P. Hémon and F. Santi, “On the aeroelastic behaviour of rectan-
gular cylinders in cross-flow,” Journal of Fluids and Structures,
𝐿, 𝐷:Span and chord of the square cylinder (m)
vol. 16, no. 7, pp. 855–889, 2002.
𝑘: Stiffness (N⋅m/rad) and (N/m)
[13] T. L. Morse and C. H. K. Williamson, “Fluid forcing, wake
𝑚: Mass (kg) modes, and transitions for a cylinder undergoing controlled
𝑈: Mean wind velocity (𝑚/𝑠) oscillations,” Journal of Fluids and Structures, vol. 25, no. 4, pp.
𝑈𝑟 : Reduced velocity (𝑈𝑟 = 𝑈/𝐷𝑓o ) 697–712, 2009.
̂
𝑍: Amplitude of the LCO [14] S. Manzoor, P. Hémon, and X. Amandolèsse, “Vortex induced
̂ vibrations of a square cylinder in a wind tunnel,” in Proceedings
𝑍∗ : Reduced RMS amplitude of LCO (𝑍∗ = 𝑍/𝐷) of the 3rd ASME 2010 Fluids Engineering Summer Meeting
𝜂: √
Damping ratio 𝜂 = 𝑐/2 𝑘𝑚. (FEDSM ’10), Montreal, Canada, 2010.
[15] M. L. Facchinetti, E. de Langre, and F. Biolley, “Coupling of
Acknowledgments structure and wake oscillators in vortex-induced vibrations,”
Journal of Fluids and Structures, vol. 19, no. 2, pp. 123–140, 2004.
The experimental part of this work was done at the Labo- [16] S. Sen and S. Mittal, “Free vibration of a square cylinder at low
ratoire d’Hydrodynamique (LadHyX) under the supervision Reynolds numbers,” Journal of Fluids and Structures, vol. 27, no.
of Dr. Pascal Hèmon and in collaboration with Dr. Xavier 5-6, pp. 875–884, 2011.
Amandolèse. Portions of the work have already been pub- [17] M. Shehryar, Transient instability mechanisms by frequency coa-
lished in ASME 3rd FEDSM and in Shehryar [17]. This is lescence in fluid structure systems [Ph.D. thesis], Hydrodynamics
an attempt to further the ongoing discussion in the light Lab (LadHyX), Ecole Polytechnique, Palaiseau, France, 2010.
of computational fluid and structure dynamics. The authors [18] R. D. Blevins, Flow Induced Vibrations, Van Nostrand Reinhold,
acknowledge the contribution by the team at LadHyX in New York, NY, USA, 1990.
designing the experiment and comprehending the experi- [19] X. Amandolèse and P. Hémon, “Vortex-induced vibration of a
mental data. square cylinder in wind tunnel,” Comptes Rendus Mécanique,
vol. 338, no. 1, pp. 12–17, 2010.
References [20] T. L. Morse, R. N. Govardhan, and C. H. K. Williamson, “The
effect of end conditions on the vortex-induced vibration of
[1] R. H. Wilkinson, On the vortex induced loading on long bluff cylinders,” Journal of Fluids and Structures, vol. 24, no. 8, pp.
cylinders [Ph.D. thesis], Faculty of Engineering, University of 1227–1239, 2008.
Bristol, Bristol, UK, 1974. [21] C. Norberg, “Flow around rectangular cylinders: pressure forces
[2] Y. Otsuki, K. Washizu, H. Tomizawa, and A. Ohya, “A note on and wake frequencies,” Journal of Wind Engineering and Indus-
the aeroelastic instability of a prismatic bar with square section,” trial Aerodynamics, vol. 49, no. 1-3, pp. 187–196, 1993.
Journal of Sound and Vibration, vol. 34, no. 2, pp. 233–248, 1974. [22] A. Khalak and C. H. K. Williamson, “Motions, forces and mode
[3] Y. Nakamura and T. Mizota, “Unsteady lifts and wakes of oscil- transitions in vortex-Induced vibrations at low mass-damping,”
lating rectangular prisms,” Journal of Engineering Mechanics, Journal of Fluids and Structures, vol. 13, no. 7-8, pp. 813–851,
vol. 101, no. 6, pp. 855–871, 1975. 1999.
[4] T. Sarpkaya, “Vortex induced oscillations, a selective review,” [23] R. A. Skop and S. Balasubramanian, “A new twist on an old
Journal of Applied Mechanics, vol. 46, pp. 241–258, 1979. model for vortex-excited vibrations,” Journal of Fluids and
[5] P. W. Bearman and E. D. Obasaju, “An experimental study of Structures, vol. 11, no. 4, pp. 395–412, 1997.
pressure fluctuations on fixed and oscillating square-section [24] E. de Langre, “Frequency lock-in is caused by coupled-mode
cylinders,” Journal of Fluid Mechanics, vol. 119, pp. 297–321, 1982. flutter,” Journal of Fluids and Structures, vol. 22, no. 6-7, pp. 783–
[6] A. Ongoren and D. Rockwell, “Flow structure from an oscillat- 791, 2006.
ing cylinder. Part 1: mechanisms of phase shift and recovery in [25] H. G. Matthies and J. Steindorf, “Partitioned strong cou-
the near wake.,” Journal of Fluid Mechanics, vol. 191, pp. 197–223, pling algorithms for fluid-structure interaction,” Computers and
1966. Structures, vol. 81, no. 8-11, pp. 805–812, 2003.
12 Advances in Mechanical Engineering

[26] F. R. Menter, “Two-equation eddy-viscosity turbulence models

for engineering applications,” AIAA journal, vol. 32, no. 8, pp.
1598–1605, 1994.
[27] J. E. Dina, P. G. Huang, and T. H. Coakley, Turbulence Modeling
Validation, Testing and Development, NASA Technical Memo-
randum, 110446, 1997.
[28] I. G. Currie and R. T. Hartlen, “Lift oscillator model of vortex
induced vibration,” Journal of Engineering Mechanics, vol. 96,
no. 5, pp. 577–591, 1970.