Professional Documents
Culture Documents
Coupling Between Structural and Fluid Dynamic Problems Applied To Vortex
Coupling Between Structural and Fluid Dynamic Problems Applied To Vortex
a r t i c l e i n f o a b s t r a c t
Article history: A simplified fluid–structure interaction approach is used to study the dynamic behaviour of a particular
Received 2 December 2010 90 m steel chimney under vortex-induced vibrations. Navier–Stokes equations for incompressible flow
Received in revised form are solved in 2D in several transverse planes of the line-like structure. The resultant pressure field is
13 October 2011
introduced using standard FEM interpolation techniques, together with the dynamical behaviour of the
Accepted 15 October 2011
structure and its boundary conditions. A fractional step scheme is used to solve the fluid field. In each
Available online 22 November 2011
fluid plane, the displacements are taken into account considering an Arbitrary Lagrangian Eulerian
Keywords: approach. The stabilisation of incompressibility and convection is achieved through orthogonal quasi-
Steel chimney static subscales, an approach that is believed to provide a first step towards turbulence modelling. In
Fluid–structure interaction
order to solve the structural problem, a special one-dimensional element for thin walled cross-section
Vortex induced vibrations
beams is implemented. The standard second-order Bossak method is used for the time integration of
Across-wind vibration
the dynamic problem. The wind is modelled as an incompressible fluid acting on the structure in a
series of planes, transverse to the structure that are supposed to be independent among them. For each
period of time, the fluid problem is solved, the aeroelastic analysis is carried out and the geometry of
the mesh of each fluid plane is updated according to the structure displacements.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction Research carried out in latest years about analytical and semi-
empirical models to study the dynamic response in vertical
Multidisciplinary methods for fluid–structure interaction (FSI) structures due to the vortex shedding have followed two distinct
can be used to describe a wide range of interesting problems in lines. The first one was aimed at determining the maximum
engineering. Essentially, FSI phenomena are caused by the pres- effects due to along-wind response, whereas the second extends
sure forces from the flow around a structure resulting in a original methods from along-wind response to cross-wind and
deformation of the structure, which in turn alters the flow in a torsional responses.
dynamic way. In particular, vibrations generated by vortex shed- The first research line was focused on along-wind response and
ding are of practical interest to civil and industrial engineering derived from the observation that equivalent static force, as
and slender structures such as bridges, tall buildings, industrial defined by Davenport (1967), usually produces correct mean
chimneys or cables are designed according to rules for cross-wind maximum displacements but may give rise to other effects (above
vibrations. all bending moments and shear forces) marked by noteworthy
The significance of vortex-induced vibrations (VIVs) and the errors. Efforts were made by Kasperski (1992) and Holmes (1994)
complexity of the cross-wind forcing mechanisms have led to a to develop new analytical methods for defining the equivalent
considerable amount of research work, either experimental or static force in a more realistic and simplified manner. Analogous
theoretical, the latter through nonlinear-oscillator modelling methods were applied by several authors to different structural
approaches and, recently, also through Computational Fluid types (Holmes, 2002), sometimes arriving at a closed form solution
Dynamics (CFD). In the review by Williamson and Govardhan (Dyrbye and Hansen, 1997). The progress of analytical methods for
(2008) fundamental results and discoveries concerning VIV are studying the wind-induced response of structures, with special
discussed. regard to cantilever vertical structures, was presented by Solari
(2002).
The second research line was aimed at determining the three-
dimensional wind-induced response of structures. With reference to
n
Corresponding author. Tel.: þ34 983546504. slender structures and structural elements in the atmospheric
E-mail address: alivas@cartif.es (A.V. Belver). boundary layer, Piccardo and Solari (1996) schematised along-wind,
0167-6105/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2011.10.007
A.V. Belver et al. / J. Wind Eng. Ind. Aerodyn. 100 (2012) 30–37 31
y
jy
n
ix x s
O
ds j
ds
dz O' dz
s i
k
s z
kz
z
a Love–Kirchhoff shell model to relate the stresses and strains in equations can be expressed as
the shell. In these relations, the generalised beam deformations @v 2
corresponding to the Navier–Bernoulli and Vlasov models are þðvUrÞvnr v þ rp ¼ b ð7Þ
@t
introduced through geometric considerations but deformations of
the cross-section as ovalization are neglected, according to beam rUv ¼ 0 ð8Þ
element modelling.
The solution procedure is based on a fractional step scheme
Fig. 2a shows the geometry and coordinate system for a thin-
(Codina, 2001), namely, a second order algorithm based on the
walled beam with an arbitrary cross-section, while Fig. 2b shows
Crank–Nicolson discretization for the viscous and convective
similar information for any shell portion composing the beam. z is
terms and a second order pressure splitting, leaving the pressure
the longitudinal coordinate for the beam, which is the line-like
gradient at a given time level in the first step and computing its
structure, and no curvature is considered in this direction.
increment in the second one (Dadvand et al., 2010).
To describe the kinematics, Uz, Ux and Uy will be considered to
The fractional step scheme is described briefly below, in its
be the displacement of the beam along Oz, Ox and Oy global axes,
basic form, without the introduction of any stabilisation for the
respectively; bx and by will be the rotations of the cross-section
pressure term. The starting point is the spatial discretization of
beam about the Ox and Oy axes, respectively, and f will be the
the Navier–Stokes Eqs. (7) and (8). The matrix form of the
twist of each section around the Oz axis. The local shell displace-
problem using the Galerkin approach is given by
ments in the O0zns system will be indicated as u, vn and vt. Merging
the Love–Kirchhoff theory for shells together with the Navier– Vn þ 1 Vn
M þ KðVn þ y ÞVn þ y þ GPn þ 1 ¼ Fn þ y ð9Þ
Bernoulli and Vlasov hypothesis for beams (Monleón, 2001; Jung dt
et al., 2002; Jung and Park, 2005 ), a matrix relation between
external forces, Fext ðz,tÞ, and generalised displacements, x, at both DVn þ 1 ¼ 0 ð10Þ
ends of the beam element can be achieved in the form: where V and P are the arrays of nodal velocities and pressures,
M x€ þ C x_ þKx ¼ Fext ðz,tÞ ð4Þ respectively. If the node indexes are denoted with superscripts a,
b, the space indexes with subscripts i, j, and the standard shape
where K is the stiffness matrix, M is the mass matrix and C is the function of node a by Na, the components of the arrays involved in
damping matrix. The generalised displacement vector is these equations are
n ot
x ¼ U z U x U y f bx by f,z ð5Þ Mab a b
ij ¼ ðN ,N Þdij ðdij is the Kronecker dÞ,
1
where f,z is the axial warping and the generalised force vector is KðVn þ y Þab a nþy
ij ¼ ðN ,v UrN b Þdij þ ðN a ,ðrUvn þ y ÞN b Þdij þ nðrN a , rN b Þdij ,
2
given by
n ot Gab a b
i ¼ ð@i N ,N Þ,
Fext ðz,tÞ ¼ N V x V y T s M x M y M o ð6Þ Fai ¼ /N a ,f i S,
where N is the axial force; Vx and Vy are the transverse shear Dab a b
j ¼ ðN ,@j N Þ:
forces; Mx and My are the bending moments about the x and the y The fractional step scheme is based on the introduction of an
directions, respectively; Ts is the St. Venant twisting moment and nþ1
auxiliary variable V~ , which allows to rewrite the time deriva-
Mo is the Vlasov bimoment. tive as
Once defined the governing equations in the local coordinate
nþ1 nþ1
system, standard assembly procedures are used and boundary Vn þ 1 Vn Vn þ 1 V~ V~ Vn
¼ þ ð11Þ
conditions imposed. The resulting system of differential equations dt dt dt
is solved using the second-order Bossak scheme. and the fractional step scheme can be presented as
nþ1
V~ Vn
M þKðVn þ y ÞVn þ y þGPn ¼ Fn þ y ð12Þ
4. Computational fluid dynamics dt
nþ1
Assuming that the air is an incompressible fluid, typically Vn þ 1 V~
M þ GðPn þ 1 Pn Þ ¼ 0 ð13Þ
accepted in the civil engineering applications, the Navier–Stokes dt
A.V. Belver et al. / J. Wind Eng. Ind. Aerodyn. 100 (2012) 30–37 33
DVn þ 1 ¼ 0 ð14Þ the friction resistance, where a time-accurate viscous flow solver is
needed. In our case, the interest focuses on the behaviour of a beam
Using the following approximation:
subjected to a flow orthogonal to the beam axis. Under this
nþy nþy assumption, the fluid domain was modelled with a number of
KðVn þ y ÞVn þ y KðV~ ÞV~ ð15Þ
independent planes of fluid where the problem can be solved
where separately. Conceptually, the CFD solution on each fluid plane
nþy nþy provides a force density, acting on the cross-section beam, which is
V~ ¼ yV~ þ ð1yÞVn ð16Þ
obtained by integrating the pressure of the fluid over the boundary of
nþ1
and expressing V in terms of V~
nþ1
using Eq. (13) and inserting the cross-section. So, a time-varying distributed load over the
the result in Eq. (14), the set of equations to be solved is structure is obtained by interpolation between consecutive fluid
nþ1 planes. The deformation of the structure (and the necessary move-
V~ Vn nþy nþy
ment of the fluid mesh) provides a kind of correlation between the
M þ KðV~ ÞV~ þGPn ¼ Fn þ y ð17Þ
dt motion of the different sections and fluid planes.
nþ1 It is interesting to make a conceptual consideration on the
dtDM1 GðPn þ 1 Pn Þ ¼ DV~ ð18Þ nature of the coupling. The beam formulation makes use of the
nþ1
small strain hypothesis, which implies that the reference and
Vn þ 1 V~ deformed configuration are considered to coalesce in writing the
M þ GðPn þ 1 Pn Þ ¼ 0 ð19Þ
dt equilibrium. On the other hand, the motion of the cross-section is
Note that DM 1G can be replaced by the Laplacian operator L and obtained exactly once given the motion of the corresponding axis,
hence the fractional step scheme takes the form: without taking advantage of the small strain hypothesis. This
implies that the loads acting on the structure will be allowed
Step A depending in a non-linear way on the motion of the beam. This
nþ1
feature may become important for the cases in which the model
V~ Vn nþy nþy moves at the limits of the small strain formulation, which is not
M þKðV~ ÞV~ þ GPn ¼ Fn þ y ð20Þ
dt the case for the example to be shown.
Finally, it is needed to choose a suitable fluid–structure
Step B coupling algorithm. In this work, the aeroelastic problem is
nþ1 characterised by large Reynolds numbers and flow separation
dtLðPn þ 1 Pn Þ ¼ DV~ ð21Þ around the circular cross-section. It is well known that loose
coupling procedures (Dadvand et al., 2010) are very efficient for
Step C such a problem. In the coupling scheme used to solve the fluid–
nþ1
structure interaction problem, the structural solution is inte-
Vn þ 1 V~ grated in the fractional step procedure and the Arbitrary Lagran-
M þ GðPn þ 1 Pn Þ ¼ 0 ð22Þ
dt gian-Eulerian (ALE) formulation is used to take into account the
movement of the fluid mesh.
The coupling scheme is based on the following steps
The fluid solver used in this work is based on the use of
Orthogonal Sub-Grid Scale (OSS) stabilisation, as described for 1. Solve the structural problem for an initial guess pressure,
example in Codina (2002). Such stabilisation falls into the Xnn þ 1s ¼ Xnn þ 1s ðPnn þ 1 Þ.
category of Variational Multi-Scale (VMS) techniques, which have 2. Move the fluid domain according to the structure motion,
been deeply investigated during the last decade. An important Xnn þ 1 ¼ Xnn þ 1 ðPnn þ 1 Þ.
nþ1
property of such techniques is that, aside of allowing a stable 3. Step A, V~ .
Finite Element solution, they tend to introduce a dissipation into 4. Step B, Pn þ 1. Impose pressure boundary conditions.
the solution that acts similarly to a turbulence model. A mathe- 5. Solve the structural problem for the calculated pressure,
matical discussion on the effectiveness of such models in model- Xn þ 1s ¼ Xn þ 1s ðPn þ 1 Þ.
ling the turbulence can be found for example in the work of 6. Step C, Vn þ 1. Impose velocity boundary conditions.
Codina et al. (2010), Principe et al. (2010) or Oñate et al. (2007),
while some empirical evidence of their success in simulating high The structural model, the CFD abilities and the former numerical
Reynolds number flows around bluff bodies can be found in the methods have been implemented in Kratos (Dadvand et al., 2010,
work of Rossi and Oñate (2010). Belver et al., 2010) (an object-oriented framework for developing
finite element codes for multi-disciplinary applications).
5. Fluid-structure coupling
6. Steel chimney under wind action
To perform the interaction between the structural and fluid
problems, it is necessary to identify the geometry of the cross- Across-wind vibrations caused by vortex shedding in a parti-
section of the beam and the distribution of pressure or force cular 90 m steel chimney are studied using the former approach.
density on its boundary. Under the usual assumptions, the motion For low damping ratios, the stack is likely to oscillate strongly
of the beam axis describes the motion of the whole cross-section, (Areemit and Warnitchai, 2001) in a plane perpendicular to the
which can be imagined as a rigid surface, which follows the mean wind direction for speeds around 9 m/s. The diameter of the
translation and the rotation of the beam axis. Any section of finite chimney varies from 2.20 m at the top to 5.50 m at the base and
dimensions in the xy plane is associated to the corresponding the thickness of the shell from 22 mm to 12 mm, as shown in
point of the one-dimensional beam element, assumed to be Fig. 3. The average mass per unit height is 1916 kg/m and the
oriented in the Oz axis. diameter dm ¼4.0 m is adopted as reference to get dimensionless
Current work focuses on bluff bodies, in particular a circular cross- values for displacements. The elastic properties are Young’s
section body, with a clear predominance of the shape resistance over modulus E¼2.1 1011 Pa, Poisson’s ratio n ¼0.3 and density
34 A.V. Belver et al. / J. Wind Eng. Ind. Aerodyn. 100 (2012) 30–37
Table 1
Inflow velocities and Reynolds numbers.
10 m 2.20 m; 12 mm
h [m] vx [m/s] Re L1 [m] L2 [m] W [m]
90 m 30 m 4.10 m; 19 mm
v
S
u d W
30 m 5.20 m; 22 mm W/2
y
L1 L2
Fig. 3. (a) 90 m steel chimney in Rayong, (b) basic dimensions of the chimney
x
(Areemit and Warnitchai, 2001).
Fig. 5. Boundary conditions’ definition.
0.16
0.12
disply / dm x 102
0.08
0.04
0
-0.04
-0.08
-0.12
-0.16
0 5 10 15 20 25 30 35 40 45 50
t [s]
0.16 0.004
ns1= 0.55Hz
0.14 0.0035
ns2 = 0.89Hz ne1 = 0.91Hz
0.12 0.003
ns3 = 1.13Hz
0.1 0.0025
Rm
Rm
0.08 0.002
0.06 0.0015
0.04 0.001
ns4 = 1.62Hz ne2 = 3.6Hz
0.02 0.0005
0 0
0 1 2 3 4 5 0 2 4 6 8 10
fr [Hz] fr [Hz]
Fig. 8. FFT analysis for the force vibration, 0 ot o 40 s (a) and free vibration, 40 ot o 50 s (b).
0.65
0.5
0.35
0.2
CL
0.05
-0.1
-0.25
-0.4
-0.55
0 5 10 15 20 25 30 35 40 45 50
t [s]
Fig. 9. Lift coefficient.
3500 Table 2
Vortex shedding frequencies and frequency response of the chimney.
ns =1.62Hz
3000 Diameter [m] Velocity [m/s] Theoretical freq. [Hz] Frequency resp. [Hz]
1500
As the frequency of the oscillation of the lift coefficient coin-
1000 cides with the vortex shedding frequency, the Strouhal number can
be evaluated considering the Strouhal relation, Eq. (3):
500 d 2:2
St ¼ ns ¼ 1:62 ¼ 0:284 ð24Þ
U 12:52
0
The Root Mean Squared (RMS) value of the lift coefficient
0 1 2 3 4 5
(Fig. 9) is 0.3. Both this value and the Strouhal number, St¼ 0.284,
fr [Hz]
agree with the experimental results (see Fig. 1) for the corre-
Fig. 10. FFT analysis for the lift coefficient. sponding Re (1.84 106)
36 A.V. Belver et al. / J. Wind Eng. Ind. Aerodyn. 100 (2012) 30–37
Principe, J., Codina, R., Henke, F., 2010. The dissipative structure of variational Solari, G., 2002. The role of analytical methods for evaluating the wind-induced
multiscale methods for incompressible flows. Computer Methods in Applied response of structures. Journal of Wind Engineering and Industrial Aerody-
Mechanics and Engineering 199, 791–801. namics 90, 1453–1477.
Repetto, M.P., Solari, G., 2004. Directional wind-Induced fatigue of slender vertical. Son, J.S., Hanratty, T.J., 1969. Numerical solution for the flow around a cylinder at
structures. Journal of Structural Engineering, ASCE 130, 10032–11040. Reynolds number of 40, 200, 500. Journal of Fluid Mechanics 35, 369–386.
Rodi, W., 1997. Comparison of LES and RANS calculations of the flow around bluff Techet, A.H., 2005. 13.42 Lecture: Vortex Induced Vibrations, /http://web.mit.
bodies. Journal of Wind Engineering and Industrial Aerodynamics 69–71, eduS.
55–75. Verboom, G.K., Koten, H., 2010. Vortex excitation: three design rules tested on 13
Rossi, R., Oñate, E., 2010. Analysis of some partitioned algorithms for fluid– industrial chimneys. Journal of Wind Engineering and Industrial Aerodynamics
structure interaction. Engineering with Computers 27 (1), 20–56. 98 (3), 145–154.
Ruscheweyh, H., 2009. Experience with vortex-induced vibrations, CICIND Inter- Williamson, C.H.K., Govardhan, R., 2008. A brief review of recent results in vortex
national Committee on Industrial Chimneys. Technical Report 26 (2), 49–57. induced vibrations. Journal of Wind Engineering and Industrial Aerodynamics
Sampaio, P.A.B., Coutinho, A.L.G.A., 2000. Simulating vortex shedding at high 96 (6–7), 713–735.
Reynolds numbers. In: Proceedings of the Tenth International Offshore and Zhou, Y., Kareem, A., Gu, M., 2002. Mode shape corrections for wind load effects.
Polar Conference, Seattle, USA. Journal of Engineering Mechanics 128 (1), 15–23.