You are on page 1of 16

China Ocean Eng., Vol. 28, No. 4, pp.

541 556
2014 Chinese Ocean Engineering Society and Springer-Verlag Berlin Heidelberg
DOI 10.1007/s13344-014-0044-1 ISSN 0890-5487

Simulation of Vortex-Induced Vibrations of A Cylinder Using ANSYS CFX

Abu Bakar IZHARa, 1, Arshad Hussain QURESHIa and Shahab KHUSHNOODb


a
University of Engineering & Technology, Lahore, Pakistan
b
University of Engineering & Technology, Taxila, Pakistan

(Received 17 December 2012; received revised form 27 May 2013; accepted 23 July 2013)

ABSTRACT
In this paper, vortex-induced vibrations of a cylinder are simulated by use of ANSYS CFX simulation code. The
cylinder is treated as a rigid body and transverse displacements are obtained by use of a one degree of freedom spring
damper system. 2-D as well as 3-D analysis is performed using air as the fluid. Reynolds number is varied from 40 to
16000 approx., covering the laminar and turbulent regimes of flow. The experimental results of (Khalak and Williamson,
1997) and other researchers are used for validation purposes. The results obtained are comparable.

Key words: vortex-induced vibrations; ANSYS CFX; flow induced vibrations; fluid structure interaction

1. Introduction
Vortex-induced vibrations (VIVs) occur when fluid crosses over a cylinder due to transverse and
inline forces that are generated due to vortex formation. Vortices are formed at a certain frequency and
if the vortex shedding frequency coincides with the natural frequency of the cylinder, resonance will
occur. VIVs occur in different engineering structures such as heat exchangers which are subjected to
VIV as a result of cross flow induced vibrations over the exchanger tubes. Offshore risers in petroleum
exploration also suffer from vortex-induced vibrations. Different researchers have contributed to the
prediction and analysis of VIV in literature like Brika and Laneville (1993), Khalak and Williamson
(1997) and Bearman (2000). Khalak and Williamson (1997) obtained the response of an elastic
cylinder with a low mass ratio of 2.4. They identified two separate regimes of high amplitude
resonance labeled as upper and lower branches of response in contrast to a single amplitude response
curve which was previously supposed by Sarpkaya (1979). Bearman (2000) gave a detailed review of
experimental work on VIV and addressed the VIVs from bluff bodies and showed that vortex shedding
frequency coincides with natural frequency of the structure over a wider band of reduced velocities if
displacement is high. Bearman (2000) conducted study on free and forced vibrations and stated that
forced vibration experiments allow the individual control of reduced velocities and displacement ratios.
VIVs depend upon several parameters including mass ratio (ms), added mass and mass damping
parameter (ms). Khalak and Williamson (1997) obtained data for three different mass ratios from low
to high and concluded that the range of reduced velocity is narrowed with the increasing mass ratio.
The damping effect was also studied with different damping ratios, and high upper branch amplitude
was observed with lower damping ratio. Khalak and Williamson (1997) also studied the effect of
combined mass damping parameters (ms) on the response and observed that upper branch amplitude

1 Corresponding author. E-mail: bakarizr@gmail.com


542 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

has a larger dependence upon the combined mass damping parameter while lower branch amplitude
does not show much dependence. The distinction between the lower and upper branch of Khalak and
Williamson (1997) is attributed to different modes of oscillation as given by the study of Williamson
and Roshko (1979). Khalak and Williamson (1997) observed from the plot of non-dimensional
frequency response versus reduced velocity that for higher damping ratios the frequency ratio (ratio of
frequency of oscillation to the natural frequency of cylinder) was close to 1 but for lower damping
ratios (e.g 2.4) the ratio was roughly around 1.5. Williamson and Roshko (1979) identified two modes
of oscillation in synchronization region i.e 2P and 2S modes. In the 2S mode, one vortex is formed in
each half cycle of vibration and occurs during initial branch of the response and is associated with
large amplitudes whereas for 2P mode two vortices are formed at each half cycle and occurs at upper
and lower branch. Khalak and Williamson (1997) showed that for low (ms) type transition from initial
to upper branch is hysteretic and from upper to lower branch is intermittent. For high ms the transition
from initial branch to lower branch is hysteretic. These transitions are associated with jump in
displacement and frequency.
Wanderley et al. (2008) performed a numerical simulation of vortex-induced vibrations using a
two-dimensional numerical scheme by solving RANS (Reynolds Averaged Navier Stokes) equations.
Their results agreed well with the results of previous researchers at the initial and upper branch of
displacement response, but not so close to the lower branch level, and they attributed this disagreement
to the three-dimensional effects, the addition of artificial dissipation term, and the turbulence scheme
used. In a subsequent work, Wanderley et al. (2008) solved RANS equations using an upwind and
TVD (Total Variation Diminishing) scheme of Roe (1984) and Sweby (1984). They verified the code
results by duplicating the data and flow conditions of Khalak and Williamson (1997), by taking a
damping ratio of 0.00542 and mass ratio of 2.4. The reduced velocity was varied from 2 to 15
corresponding to a Reynolds number varying from 2000 to 12000. The present work is an attempt to
simulate VIV for flow over an elastic rigid cylinder by taking the same flow parameters as that of
Khalak and Williamson (1997) using ANSYS CFX. These results are validated to the experimental and
numerical results of previous researchers. ANSYS CFX is a commercially available CFD (Computational
Fluid Dynamics) simulation package having advanced turbulence schemes and numerical algorithms
and can accurately predict the experimental data by employing the latest CFD and FSI (Fluid Structure
Interaction) techniques. Previously, not many papers are found in literature in which CFX is used to
validate VIV data. Kuehlert et al. (2008) validated the data for a single tube with 2-DOF (Degree of
Freedom) flow induced vibration with experimental results of Hover et al. (1998) by employing RANS
turbulence scheme in 2D. Kuehlert et al. (2008) varied the fluid properties and free stream velocity
while increasing the reduced velocity and keeping Reynolds number constant at 3800. Spring and
damper combination was used to support the tube. Their simulated data for the displacement and
frequency response agreed well to that of Hover et al. (1998).

2. Non-Dimensional Parameters
The following non-dimensional parameters used in this simulation are listed in Table 1.
Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 543

Table 1 Dimensionless parameters used in simulation


Description Formulae
Obtained by taking the structural frequency and U
Reduced velocity Ur
diameter of cylinder into account for generalization fn d
4m
Specific mass ratio of mass of the cylinder to that of displaced fluid ms
d 2 l
Ud
Reynolds number Ratio of inertia force to viscous force Re

fs d
Strouhal number Measure of the predominant shedding frequency fs St
U
Ratio of damping coefficient to critical damping c
Damping ratio
coefficient ccrit
Dimensionless Ratio of y displacement of the cylinder to the diameter y
A
displacement of the cylinder d
Parameter that quantifies drag or resistance of an object Fx
Drag coefficient Cd
in a fluid medium 0.5 U 2 dl

Parameter that quantifies the lift developed in body due Fy


1Lift coefficient Cl
to dynamic pressure in flowing fluid 0.5 U 2 dl
Note: U= free stream velocity; fn= natural frequency of cylinder; d= diameter of cylinder; m= mass of cylinder;
= density of fluid; l = length of cylinder; = dynamic viscosity of fluid; fs= vortex shedding frequency
stationary cylinder; c= damping coefficient; ccrit= critical damping coefficient; y= transverse displacement
of cylinder; Fx=drag force; Fy=lift force.

3. 2D Simulation
The laminar case was setup in CFX with an extruded 2D mesh. CFX meshing was used to mesh
the domain. The domain and the mesh are shown in Fig. 1. The mesh is refined in the wake region while
coarser mesh is used near the outer walls of the domain. The mesh is extruded one element in z
direction. ANSYS CFX is intrinsically a 3D solver and therefore 2D planar geometries cannot be
imported into ANSYS CFX. This can be called a quasi 2D case. For this simulation, the length in the
extruded direction is almost equal to the diameter of the cylinder. This convention provides reasonable
accuracy. This should be noted that 2D simulation is intrinsically a 3D simulation. 2D simulation
should be referred here as quasi-2D simulation, because certain length of cylinder is considered in z
direction which can calculate the drag and lift coefficients using the projected area (diameter of
cylinder multiplied by length of cylinder) and the projected force at the fluid structure boundary using
CFX force functions. The data in Table 2 is obtained using the quasi-2D simulation not planar 2D
simulation. Inflation layers were used at the cylinder wall and the outer walls to resolve the boundary
layer at the walls. No slip condition at cylinder wall is used while free slip condition is used at the
upper and lower boundaries. Front and back sides of the domain are given symmetry conditions which
make this setup two dimensional. The left side was specified as velocity inlet while the right side as the
outlet pressure. An initial velocity profile was also specified for the domain that helps in achieving the
steady state efficiently. The fluid properties and velocity were varied to keep the Reynolds number
constant. This setup was run for five cases with Reynolds numbers equal to 40, 100, 200 and 1000.
544 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

The lift and drag coefficients were obtained using the intrinsic force function in CFX.

Fig. 1. Isometric view of the setup showing the boundaries.

Fig. 2 shows the contours of streamlines over the cylinder for Re=40. The two vortices are formed
due to boundary layer separation at Re=40 while pressure gradients are clearly visible at the leading
and trailing edge of the cylinder. Fig. 2 shows the development of laminar vortices pair at Re=40.
These elliptic vortices have clearly defined the boundary with characteristic dimensions. a is the
distance from the trailing edge of the cylinder to the center of the vortex, l is the recirculation length of
vortex, and b is the center to center distance of the vortex pair. The characteristic dimensions of the
vortices have much significance in the dynamics of the vortex. These parameters are used in the
calculation of average velocity of the vortex ring, as this is not the focus of the current research topic, a
comprehensive description is avoided. The parameters are quoted for the validation of CFD results.
The reader can consult Akhmetov (2009) and Green (1995) for further details. These dimensions are
measured from the present simulation and they are comparable to the results of Wanderley et al. (2008)
and that of previous researchers as given in Table 2, which gives a comparison of these dimensions
with data in literature and it can be seen that data compares well with those of previous researchers.
Fig. 3 shows a plot of lift and drag coefficients obtained for Re=200 from the present study. Fig. 4
shows the contours of velocity for Re=100. The convergences from one of the cases for Re=2000 and
3D are shown in Fig. 20 and Fig. 21 in Appendix B.
Table 2 Characteristic dimensions of two vortices formed at Re=40 from previous researchers and present study
Reference Cd l/d a/d b/d () Comments
Sphaier and Rengel (1999) 1.610 2.23 0.7200 0.579 54.06 FVM
Wanderley et al. (2008) 1.560 2.29 0.7300 0.600 53.79 FDM
Present 1.546 2.20 0.7147 0.590 52.40 CFX

Table 3 shows a comparison of drag (Cd) and lift (Cl) coefficients from the present study and
previous literature for Reynolds numbers 100, 200 and 1000. It can be seen from Table 3 that results
for Cl for Re=100 and 200 are closer to the results of Li et al. (2011) while Cd and Cl at Re=100 agrees
well with the data of Wanderley et al. (2008). Values of Cd obtained from the present simulation are a
little larger than those of previous researches (Wanderley et al., 2008; Norberg, 2003).
Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 545

Fig. 2. Streamlines over the cylinder for Re=40 (present simulation).

Fig. 3. Dynamic plot of drag and lift coefficients for Re=100 (present simulation).

Fig. 4. Cylinder wake at Re=100.

From the plots of Cd and Cl, we can perform a discrete FFT analysis to find the frequencies. The
frequencies of Cd and Cl can be observed from the frequency amplitude plot at Re=100 shown in Fig. 5.
The peak of Cl corresponds to the amplitude axes on the left while Cd peak corresponds to the
amplitude axes on the right and it can be observed from the plot that the frequency for Cd signal is
546 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

about twice that of Cl signal. The Strouhal frequency at Re=2000 is determined in a similar manner by
using DFT (Discrete Fourier Transform) for velocity signal and it can be seen from Fig. 6 that a peak
is formed at the frequency of 4.66 Hz.
Table 3 Data comparison from previous researches at Re=100, 200 and 1000
Reynolds
Reference Cd Cl Comments
number
100 1.30 0.25
Wanderley et al.
200 1.27 0.51 FDM
(2008)
1000 0.96 0.22
100 0.32
Norberg (2003) 200 0.53 Experimental
1000 0.08
100 1.36 0.32
Sphaier and Rengel FEM
200 1.35 0.67
(1999) (10079 nodes)
1000 1.60 1.70
100 1.350 0.363
Li et al. (2011)
200 1.342 0.690
100 1.49 0.20
Present study 200 1.34 0.46 CFX
1000 1.04 0.52

Fig. 5. Frequency amplitude plot from DFT analysis of Cd and Cl at Re=100.

Fig. 6. Frequency amplitude plot for velocity signal at Re=2000.

The Reynolds number for the flow over cylinder in our case is defined as:
Ud
Re , (1)

Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 547

where is the fluid density (kg/m3), d is the diameter (m) of the cylinder over which the fluid flows
externally, U is the incoming free stream velocity of the fluid (m/s), and is the dynamic viscosity of
the fluid (Pa-s).
For the case of Re=2000, the free stream inlet velocity was taken as 22 m/s. The diameter of the
cylinder was taken as 2 m. Therefore, substituting these values into Eq. (2) yields the Strouhal number
fd 4.66 1
St 0.2119 , (2)
U 22
where f is the Strouhal frequency.

4. 3D Simulation
The 3D model representation is clearly shown in Fig. 7. One way unsteady coupling was
implemented between fluid and structure fields to calculate the displacements. The structural solver
was implemented in Mathematics. Runge K-5 solution routine for ordinary differential equations was
developed. The details are given in Appendix A. The fluid force data at each step was imported from
CFX to Mathematica where structural field was calculated using Eq. (16). The step size for the
structural case was calculated by dividing the period of oscillation of the cylinder into 20 steps for
numerical accuracy. This should be clear that flexibly rigid cylinder was assumed with an equivalent
system stiffness and damping and hence deflection profile along the span of cylinder was not
considered.
This should be cleared here that 2D simulation is intrinsically a 3D simulation. 2D simulation
(Section 3) should be referred here as quasi-2D simulation, because certain length of cylinder is
considered in z direction which can calculate the drag and lift coefficients using the projected area
(diameter of cylinder multiplied by length of cylinder) and the projected force at the fluid structure
boundary using equations in Table 1. The data in Table 2 are obtained using the quasi-2D simulation
not planar 2D simulation. Similar is the case of 3D simulation in Section 4, the mass of the cylinder
was calculated by use of the length and mass ratio of the cylinder. Therefore, 3D and 2D simulations
are not differentiated in geometric sense.

4.1 Spring Damper System Parameters


A three-dimensional case was setup using the flow and mass-damping parameters of Khalak and
Williamson (1997) for validation purpose. Rigid tube was simulated using a mass-spring-damper
system for one-degree-of-freedom in the transverse direction of the flow. Fig. 7 shows the representation
of the distributed spring damper system that is implemented in three-dimensional simulation.
The natural frequency of the cylinder is calculated by treating the cylinder as a free-free beam, as
shown in Fig. 8.
Free-free beam conditions can be applied to the cylinder. The natural frequency of the cylinder
using the above beam analogy is calculated. For more details of the formulation of natural frequency,
reader can consult Irvine (2012). The solution of Euler Bernoulli beam equation as represented by Eq. (3)
is employed as:
548 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

4 yL 2 yL
EI A 0, (3)
xL4 t 2
where xL is the direction along the length of the beam, yL is the direction of displacement perpendicular
to xL, and A is dimensionless displacement. From the separation of variables, the solution to Eq. (3) can
be written as:
yL ( xL , t ) X ( xL )T (t ) , (4)
where t is temporal variable.

Fig. 7. Distributed spring damper system for a free-free beam for 3D analysis.

Fig. 8. Euler beam representation.

The wavenumber k and the circular frequency can be represented as:


EI 4
2 k . (5)
A
The general solution for xL function is represented by Eq. (6):
yL xL L1 sin kxL L2 cos kxL L3 sinh kxL L4 cosh kxL . (6)
For free condition, the shear force and the bending is taken as zero, hence by applying these
conditions on each boundary and some arrangement, the following equation holds
cosh kl cos kl 1 , (7)
where L1, L2, L3, and L4 in Eq. (6) are constants of integration and l in Eq. (7) is the length of the beam
under consideration.
The transcendental equation represented by Eq. (7) can be solved to find the values of knl, where n
Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 549

is the mode number. For the first mode,


kl 4.73 . (8)
By substituting the length of cylinder l into Eq. (8), k can be calculated and hence from Eq. (5) the
circular frequency is found which then yields natural frequency fn in Hz.
The specific mass (ms) of the cylinder i.e. mass per unit length of the cylinder is represented as:
4m
ms . (9)
d 2 l
The mass of the cylinder (m) can be calculated from Eq. (9) with different values of specific masses
such as 2.4 and 10.3.
The added mass is calculated from Eq. (10) below
ma r 2 . (10)
The natural frequency (fn) and spring constant (ksp) of the system are related as given by
1 ksp
fn . (11)
2 m ma
By rearranging Eq. (11) and substituting for fn , the spring constant is calculated as:
ksp 4 2 f n2 (m ma ) . (12)
Therefore, by substituting the values of natural frequency, the mass of the cylinder and the added
mass of the spring constant of the system can be calculated.
The damping ratio was taken approx. equal to 0.00542. From this damping ratio, the damping
constant c is calculated as follows:
c 2 ksp (m ma ) . (13)
Therefore, we have calculated the values of damping constant and spring stiffness of the spring
damper system.
The forces on the cylinder wall are obtained for a given reduced velocity by using the intrinsic
force function in CFX. The displacement of the rigid elastic tube is obtained by numerically integrating
Eq. (14).
my cy ksp y Fy . (14)
This should be clear here from Fig. 9 that X is the direction of flow i.e. positive X direction. Y is
perpendicular to X, Z is perpendicular to X and Y and is directed into the paper. We are taking Fy as the
lift force acting in Y direction and y is the displacement in Y direction due to force Fy. This should be
noted that displacement y is a small letter while the direction Y is a capital letter. The fluid forces were
obtained at each step and were transferred to structural side to determine the displacement of the
cylinder. The displacement of the cylinder was not transferred to the fluid side.

Fig. 9. Directions for analysis.


550 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

4.2 Simulation Setup


The 3D simulation was run for different reduced velocities from 2 to 15 with Reynolds numbers
varying from 2000 to 12000. SAS-SST turbulence scheme was employed in CFX. It is a class of
URANS (Unsteady Reynolds Averaged Navier Stokes) method and it can provide LES (Large Eddy
Simulation) like behavior in detached flow regions. SAS is based on including the von Karman length
scale into the turbulence scale equation. Von Karman length scale allows SAS technique to
dynamically adapt to the resolved structures in a URANS simulation which provides an LES like the
behavior in unsteady regions of flow and simultaneously gives standard RANS capabilities in steady
flow field. SAS-SST is particularly suited to flows with large separation zones and vortex formations
as described in ANSYS (2012).

4.3 Results and Discussion (3D Simulation)


Fig. 10 shows a comparison of the normalized displacement A versus the reduced velocity from
the present study and previous literature. It can be observed from Fig. 10 that initial branch is close to
the response from previous authors like Khalak and Williamson (1997) who used experimental
technique and Iwan and Blevins (1974) who employed analytical methods. The upper branch is not as
clearly visible as in Khalak and Williamson (1997) or Iwan and Blevins (1974). It is also noted from
Fig. 10 that the maximum amplitude from different researchers is varying. The maximum amplitude
obtained through the present simulation is close to that of Iwan and Blevins (1974) and Wanderley et
al. (2008). The reduced velocity (Ur) value at which the maximum amplitude occurs is approx. 4.6 in
the present case while that of Wanderley et al. (2008) and Khalak and Williamson (1997) is around 5.5.
The lower branch of response is a little lower than that of Khalak and Williamson (1997) and Parra
(1996). The desynchronization branch from the present simulation, Khalak and Williamson (1997),
Meneghini et al. (1997) and Parra (1996) is a little non uniform while that of Wanderley et al. (2008)
and (Iwan and Blevins, 1974) are more uniform.
Figs. 11 and 12 show the dynamic behavior of non-dimensional displacement and lift coefficient
for Ur=4 and 6, respectively. It is evident from these figures that the response for Ur=4 is much more
periodic than the response for Ur=6. Fig. 12 shows presence of different frequencies. It is also clear
from these plots that as the reduced velocity increases, the phase difference between the lift and
displacement responses also increases as shown in Fig. 13 which shows a comparison of phase lag
between the lift and displacement responses of the current simulation and the results of Wanderley et
al. (2008) and Guilmineau and Queutey (2004). Fourier function in Mathematica was used which finds
the discrete Fourier transform. The argument of the transformed data at the peak position of each
signal was found in radians. The difference in these two argument values was converted into degrees to
find the phase difference between the lift and displacement signals. A sudden increase in the phase
difference is seen near Ur=4 and this sudden increase is attributed to a switching of vortex formation
mode from 2S to 2P. Brika and Laneville (1993) first showed the evidence of 2P mode from free
vibration of a cable and confirmed the results of Williamson and Roshko (1979). The initial branch
therefore corresponds to 2S mode while the upper and lower branches are associated with 2P mode as
is obvious from Fig. 14 which shows the 2P vortex shedding pattern at Ur=7.6 from the present
Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 551

simulation. As the reduced velocity increases, the speed reaches a value at which fs (vortex shedding
frequency) becomes closer to fn (natural frequency) of the structure which causes the frequencies to
synchronize. The vortex shedding frequency (fs) and the tube oscillation frequency (f) becomes close to
the natural frequency (fn) and the ratio f*=f/fn becomes close to unity which is known as lock-in or
synchronization phenomenon.

Fig. 10. Comparison of present calculated data with previous researches found in literature.

Fig. 11. Dimensionless displacement and lift coefficient at Ur=4.

Fig. 12. Dimensionless displacement and lift coefficient at Ur=6.


552 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

Fig. 13. Phase lag between the lift and displacement as function of reduced velocity (Ur).

Fig. 14. Velocity profile showing vortices at Ur=6.

This behavior can be seen in Fig. 15 which shows a plot of frequency ratio versus the reduced
velocity. With reference to Fig. 15, the line with triangles is the frequency ratio of vortex shedding of
stationary cylinder normalized with the natural frequency of cylinder in air, the line with diamonds is
the frequency ratio for vibrating cylinder of specific mass of 2.4 and the line with squares is the
frequency ratio of vibrating cylinder of specific mass of 10.3 normalized with the natural frequency in
air. Similarly, Fig. 16 shows the plot of frequency ratios obtained by normalizing the frequencies with
the natural frequency of cylinder in water (fnw). It is obvious from Fig. 15 and Fig. 16 that as the
specific mass increases, the frequency response becomes closer to the natural frequency and the ratio
gets closer to unity. This behavior is evident from the plots of Fig. 17 and Fig. 18 taken from Khalak
and Williamson (1997). Fig. 19 shows the amplitude profile as a function of the reduced velocity for a
specific mass of 10.3 from the present study and Khalak and Williamson (1997). It can be seen that
upper and lower branches from the present simulation and Khalak and Williamson (1997) are not
much close whereas the initial branch and the desynchronization branches of the response are
comparable and a generalized trend of amplitude response can be seen. This is also obvious from Fig.
19 that upon increasing the specific mass the range of the reduced velocity over which the response
(upper and lower branches) extends is clearly reduced when compared to amplitude response at
specific mass of 2.4 in Fig. 10.
Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 553

Fig. 15. Frequency ratio (normalized with fna) for ms=2.4 from the present simulation.

Fig. 16. Frequency ratio (normalized with fnw) for ms=2.4 from the present simulation.

Fig. 17. Frequency ratio from Khalak and Williamson (1997) for ms=2.4.

Fig. 18. Frequency ratio from Khalak and Williamson (1997) for ms=10.3 and 20.4.
554 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

Fig. 19. Comparison of amplitude profile against reduced velocity for ms=10.3.

5. Concluding Remarks
It can be concluded after observing the data obtained from this simulation that ANSYS CFX can
predict the vortex-induced vibrations behavior considerably well. The SAS-SST turbulence scheme
available in CFX used for this simulation is well suited to this type of case. It can be seen from this
study that a better amplitude response could be achieved if small increments in the reduced velocity
were used along with a finer grid but these could be used at the case of computationally expensive cost.
Future work can be extended to fluid structure interaction simulation available in CFX using rigid
body solver along with a grid refinement study.

References
Akhmetov, D. G., 2009. Vortex Rings, Springer-Verlag Berlin Heidelberg.
ANSYS, 2012. ANSYS CFX Documentation, ANSYS Incorporated.
Bearman, P. W., 2000. Developments in vortex shedding research, Workshop on Vortex-Induced Vbrations of
OffShore Structures, Sao Paulo, Brazil.
Brika, D. and Laneville, A., 1993. Vortex-induced vibrations of a long flexible circular cylinder, J. Fluid Mech.,
250, 481508. doi:10.1017/S0022112093001533.
Green, S. I., 1995. Fluid Vortices: Fluid Mechanics and Its Applications, Kluwer Academic Print on Demand.
Guilmineau, E. and Queutey, P., 2004. Numerical simulation of vortex-induced vibration of a circular cylinder
with low mass-damping in a turbulent flow, J. Fluid. Struct., 19(4): 449466. doi:10.1016/j.jfluidstructs.
2004.02.004.
Hover, F. S., Techet, A. H. and Triantafyllou, M. S., 1998. Forces on oscillating uniform and tapered cylinders in
cross flow, J. Fluid Mech., 363, 97114. doi:10.1017/S0022112098001074.
Irvine, T., 2012. Bending Frequencies of Beams, Rods, and Pipes, http://www.vibrationdata.com/tutorials2/
beam.pdf.
Iwan, W. D. and Blevins, R. D., 1974. A model for vortex-induced oscillation of structures, J. Appl. Mech.,
ASME, 41(3): 581586.
Khalak, A. and Williamson, C. H. K., 1997. Investigation of relative effects of mass and damping in vortex-
induced vibration of a circular cylinder, Journal of Wind Engineering and Industrial Aerodynamics, 6971,
341350. doi: 10.1016/S0167-6105(97)00167-0.
Kuehlert, K., Webb, S., Schowalter, D., Holmes, W., Chilka, A. and Reuss, S., 2008. Simulation of the fluid
structure-interaction of steam generator tubes and bluff bodies, Proc. 14th International Conference on
Energy, 238(8): 20482054. doi:10.1016/j.nucengdes.2007.11.017.
Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556 555

Li, T., Zhang, J. Y. and Zhang, W. H., 2011. Nonlinear characteristics of vortex-induced vibration at low
Reynolds number, Communications in Nonlinear Science and Numerical Simulation, 16(7): 27532771.
Meneghini, J. R., Sahara, F. and Bearman, P. W., 1997. Numerical simulation of vortex shedding from an
oscillating circular cylinder, Computational Methods and Experimental Measurements, Computational
Mechanics Publications, 409418. doi:10.2495/CMEM970401.
Norberg, C., 2003. Fluctuating lift on a circular cylinder: Review and new measurements, J. Fluids Struct., 17(1):
5796. doi:10.1016/S0889-9746(02)00099-3.
Parra, P. H. C. C., 1996. Semi Empirical Model for Vortex-Induced Vibrations: Theoretical and Experimental
Analysis, Msc. Thesis, Sao Paulo, Brazil.
Roe, P. L., 1984. Generalized Formulation of TVD Lax-Wendroff Scheme, ICASE Report.
Sarpkaya, T., 1979. Vortex-induced oscillations: A selective review, J. Appl. Mech., 46(2): 241258.
Sphaier, S. H. and Rengel, J. E., 1999. A projection method for unsteady Navier Stokes equation with volume
method and collocated grid, Hybrid Methods in Heat and Mass Transfer 1, No. 4.
Sweby, P. K., 1984. High resolution scheme using flux limter for hyperbolic conservation laws, SIAM Journal on
Numerical Analysis, 21(5): 9951011.
Wanderley, J. B. and Levi, C., 2005. Vortex induced loads on marine risers, Ocean Eng., 32(11-12): 12811295.
doi:10.1016/j.oceaneng.2004.12.007
Wanderley, J. B. V., Souza, G. H. B., Sphaier, S. H. and Levi, C., 2008. Vortex-induced vibration of an elastically
mounted circular cylinder using an upwind TVD two-dimensional numerical scheme, Ocean Eng., 35(14-15):
15331544. doi:10.1016/j.oceaneng.2008.06.007.
Williamson, C. H. K. and Roshko, A., 1979. Vortex formation in the wake of an oscillating cylider, J. Fluid.
Struct., 2(4): 355381.

Appendix A
The structural side was solved by employing Runge K5 method for the system of the first order
ordinary differential equations. Eq. (14) was represented by a system of the first order ODE. The
system is as follows:
F cv ku
f1 x, u , v v; f 2 x, u , v . (15)
m
The displacement u and velocity v are calculated as:
7 k11 32k31 12k41 32k51 7k61
ui1 ui 90
(16)
v v 7 k12
32 k 32
12 k42 32k52 7k62
i1 i
90
where,
k1i hf i x, u , v ;
h k k
k2 i hf i x , u 11 , v 12 ;
2 2 2
h 3k11 k21 3k12 k22
k3i hf i x , u ,v ;
4 16 16
h k k
k4 i hf i x , u 31 , v 32 ;
2 2 2
556 Abu Bakar IZHAR et al. / China Ocean Eng., 28(4), 2014, 541 556

3h 3k21 6k31 9k41 3k22 6k32 9k42


k5i hf i x , u ,v ;
4 16 16
k 4k21 6k31 12k41 8k51 k12 4k22 6k32 12k42 8k52
k6 i hf i x h, u 11 ,v , (17)
7 7
where h is the step size.
Appendix B

Fig. 20. Momentum and turbulence convergences for 2D case (Re=2000).

Fig. 21. Turbulence and wall scale convergence for Ur=3. (3D case).

You might also like