Anda di halaman 1dari 8

The Fifth International Symposium on Computational Wind Engineering (CWE2010)

Chapel Hill, North Carolina, USA May 23-27, 2010

Three-dimensional numerical flutter simulation


Gergely Szaboa, Gergely Kristof b
a
Phd Student, Budapest, Hungary, mr.gergely.szabo@gmail.com
b
Associate professor, Budapest, Hungary, kristof@ara.bme.hu

ABSTRACT: In this paper a numerical flutter simulation with the ANSYS software is presented.
A fully aeroelastic wind tunnel model was constructed in order to make a validation possible.
First the dynamic properties of the bridge were determined. Second the critical wind speed for the
flutter instability was captured. Taking into account the dynamic properties and the contour of the
bridge, the CSD and the CFD models of that were built. By using the fluid-structure interaction
solver of the ANSYS, the three-dimensional coupled simulation was performed. The critical wind
speed was calculated by monitoring the structural motion at different inlet velocities. Comparing
the measured and calculated critical wind speeds, good agreement was found considering the
complexity of the problem. Alternatively the critical wind speed was evaluated by using the flut-
ter derivatives extracted from a two-dimensional CFD simulation made with the FLUENT soft-
ware.

1 INTRODUCTION

Nowadays, with a strong computational background, CFD (Computational Fluid Dynamics)


simulations appear to be powerful rivals of the wind tunnel tests. Recently a number of numerical
simulations have been made aiming at determining the aerodynamic performance of an ordinary
bridge deck section. These simulations are two-dimensional mainly for the sake of time effective-
ness. The main shortcoming of the two-dimensional approach is that it cannot capture the rather
complex three-dimensional coupling of the bridge deck motion and the fluid flow around it. Our
main goal is to substitute the costly wind tunnel test with advanced three-dimensional coupled
simulations. In bridge design practice such investigation is rare but for an airfoil Wang et al
(2008) have profound studies.

2 THE WIND TUNNEL MODEL


2.1 Model dimensions
As our main purpose was to validate the numerical solution, a wind tunnel model as simple as
possible was designed. The general dimensions are shown in Figure 1.

Figure 1. Main dimensions of the aeroelastic wind tunnel model.


The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

The bridge is made up using an aluminum core beam on which balsa elements are fixed. The
cross section can be seen in Figure 2. The contour of the model represents an idealized box bridge
cross section with a wind nose on both sides. This streamlined section was chosen because it was
thought to be handled properly with simpler turbulence models.

Figure 2. Cross section of the bridge model.

The core beam is fully constrained on both sides. There are four steel wires working like stay ca-
bles. These wires were softened by introducing spring elements in order to make the bridge more
flexible. The side view and a view from the bottom can be seen in Figure 3.
The elastic modulus of the aluminum is 2.06E10 N/m², the density is 2500 kg/m³. The den-
sity of the balsa is 123 kg/m³. The stiffness of the above mentioned spring is 230 N/m.

Figure 3. Bridge model mounted in the wind tunnel.

2.2 Dynamic parameters


At the middle balsa element of the bridge, on both sides of the segment two piezoelectric acceler-
ometers were fixed for monitoring the vertical motion. For the wind-off case the natural frequen-
cies and logarithmic decrement values were measured. The first vibration mode is a symmetrical
heave motion of the bridge deck with a natural frequency of 2.02 Hz. For the flutter analysis the
first torsion mode is dominant, which was the fifth one with a natural frequency of 5.55 Hz. The
logarithmic decrement for this mode was δ=0.012. These parameters were used for tuning the
CSD model. During the measurement the wind speed was incrementally increased. The high fre-
quency vibrations caused by the vortex shedding were ignored and the flutter phenomenon was
only considered. The tests were done under low turbulence level.
The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

3 TWO-DIMENSIONAL STUDY
3.1 Setting up a numerical solution
Before applying the three-dimensional coupled simulation, a two-dimensional CFD calculation
was performed using the FLUENT software in order to determine the flutter derivatives for this
cross section. The numerical mesh around the contour can be seen in Figure 4.

Figure 4. Two-dimensional numerical mesh.

The flutter derivatives were extracted by means of forced vibration method. The oscillation
frequency was constant while the inflow velocity was incrementally increased. The maximum
flow velocity was 14.4m/s when the motion frequency was 6.00 Hz. The lift motion amplitude
was set to 20 mm, the rotation angle was 0.1744 rad. The applied cell number is 12590. The k-ε
turbulence model was chosen. The motion of the wall domain of the bridge needs the deformation
of the numerical mesh, which can be assured using the dynamic mesh option of the FLUENT.
The mesh deformation is relatively low, so there is no need to implement re-meshing technique
but a deforming mesh is enough to handle the bridge deck motion. In the URANS simulation the
time step was 0.00002 s. Thus, the Courant-number was set below 1 for the whole computational
domain.

3.2 Results
By performing URANS simulation for both vertical and rotational motion, the flutter derivatives
were extracted within a certain reduced velocity range (0-12). The velocity contour plot can be
seen in Figure 5 in case of rotational motion. Compared with the flat plate derivatives there were
no significant differences as expected. Once the derivatives are known, a simplified critical wind
speed calculation can be performed.

Figure 5. Velocity contour plot around the dislocated bridge deck.

The previously introduced two-dimensional mesh was extruded, and a three-dimensional


mesh was acquired. The cell number now is 503600. For a certain reduced velocity the same cal-
culation was done with this updated model.
The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

The main questions were whether the two-dimensional approach was adequate for extracting the
flutter derivatives and the k-ε turbulence model worked well for this problem. In Figure 6 the
streamlines with LES and URANS shows the differences in the flow field but in Figure 7 it can
be seen that the small fluctuations evaluated using LES do not alter the derivatives significantly
in this case. Comparison of the 2D and 3D results with the k-ε model and the 3D LES showed
that the 2D approach gave fairly good results.

Figure 6. Streamlines around three-dimensional bridge deck section using LES (left picture) and URANS

0,04
0,03 2D_URANS
0,02 3D_URANS
0,01 3D_LES
cL [-]

0,00
-0,010,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80

-0,02

-0,03
-0,04
t [s]
Figure 7. Lift force coefficients on the bridge deck section with different meshes and turbulence models.

4 THREE-DIMENDSIONAL STUDY
4.1 FEM model of the bridge
So far two- and three dimensional section model studies were shown with CFD simulations. The
three-dimensional aero-elastic modeling demands a mechanical model first. The FEM model with
ANSYS for the coupled calculation can be seen in Figure 8.

Figure 8. FEM representation of the wind tunnel model.


The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

The proper arrangement of the FEM model is essential for the coupled solver. As the interaction
between the flow field and the structure happens on the boundary of the bridge, the geometrical
similarity should be assured first. The main problem in this case is that the wind tunnel model is
made up using a core beam bounded with balsa segments, which is not entirely continuous along
the bridge axis. This system is to be transformed into a fully continuous one in order to avoid any
unpleasant distortion of the CFD mesh. Thus, the mechanical properties of the aluminum beam
were converted into a boxed shell element configuration being on the bridge contour. The thick-
ness and elastic modulus were adjusted so that the torsion and bending stiffness of the bridge
deck be the same as that of the aero-elastic model. It was necessary to introduce special mass dis-
tribution for adjusting the natural frequencies as well.
As the bending stiffness of these shell elements is low, the box was filled with solid ele-
ments that provide the cross section with enough stiffness without interfering into the global dy-
namic behavior. With this method the boundary shell elements do not suffer large local deforma-
tion due to the pressure load from the fluid flow. In Figure 9 the first heave and first torsion
modes shape are shown.

Figure 9. Mode shapes of the FEM model.

4.2 Numerical mesh around the bridge


Once the shape of the bridge is set the computational domain for the CFD simulation can be de-
signed, see Figure 10. The boundary conditions are shown. The length of the domain in z direc-
tion is 2800 mm in accordance with the wind tunnel model.

Figure 10. Main dimensions of the computational domain.


The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

In Figure 11 the details of the numerical mesh are shown. The main concept was to apply as many
cells as necessary but as few as possible in order to reduce the computational costs. A boundary
layer mesh was created around the bridge surface in 2D, which was extruded along the bridge
axis providing a structured boundary layer mesh. In the farther regions of the domain unstructured
mesh was applied using tetrahedral cells. This meshing concept kept the cell number below
200.000. Naturally this mesh is coarse enough to capture vortices in the wake of the body but
adequate to calculate forces during vertical and rotational motion of the bridge.
The k-ε turbulence model was used. The time step was 0.0008 sec. The inlet velocity was
constant. The key moment in the coupled analysis is the force-displacement transfer from one
physic to the other. The MFX solver involved in the ANSYS handles the problem automatically.
The CFD mesh needs to be deformed during the FSI simulation, which is handled by a mesh dif-
fusion method. This technique involved in the ANSYS makes the displacement on the bridge
deck boundary diffused to the inner regions of the domain. The boundary layer region can be
maintained of good quality by introducing higher stiffness factors for the smaller cells and lower
for the larger ones so the large displacement is handled by the farther regions of the mesh without
unwanted distortion.

Figure 11. Numerical mesh for the CFD simulation.

4.3 Critical wind speed calculation with updated spatial eigenvalue analysis
Considering the complexity of the three-dimensional coupled simulation, a check simulation was
made in order to compare the results from different approaches. The main idea is to consider a
method for determining the critical wind speed for an NDOF system. The method proposed by
Starossek (1998) was used (Equation 1) with some modifications.
M&x& + (1 + iγ )Kx = q (1)
where M=mass matrix, i=complex unit, γ=damping parameter, K=stiffness matrix,
x=displacement vector, q=force vector.
The Equation 1 can be rewritten to an eigenvalue problem in Equation 2.
{(1 + iγ)K − ω 2 [M + L(U )]}~x = 0
red (2)
where L=loading matrix that is compiled of the flutter derivatives from the two-dimensional CFD
simulations and is a function of the reduced velocity. In (2) the displacement vector is complex.
The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

The NDOF system equation of motion (Equation 2) with the complex notation can be simplified
using the modal analysis technique (Equation 3).
V T MV = E , V T KV = ω , V T LV = A , ~
r x = V~y (3)
where V=matrix compiled from the mode shape vector extracted from the FEM model, ωr=r-th
natural frequency, ~ y =a newly introduced unknown complex vector.
The final form of the system to be solved is the Equation 4.
{(1 + iγ ) ω − ω 2 [E + A(U )]}~y = 0
r red (4)
In this method the critical wind speed can be calculated in the same way as in case of a 2D sys-
tem; first the total damping is determined, and when the damping is zero the critical wind speed is
found. On this wise there is no need neither to reduce the bridge into a 2DOF mechanical system
nor to build up the FEM model of the bridge but the mode shapes and the corresponding frequen-
cies from the ANSYS FEM model can be used.

5 CRITICAL WIND SPEED


5.1 Comparison of the critical wind speed from different methods
In a flutter analysis the target parameter is the critical wind speed. The wind tunnel test showed
the flutter motion of the bridge deck with remarkable amplitude at the wind speed of 9.2 m/s. The
deformed shape at two stages can be seen in Figure 12.

Figure 12. Flutter motion of the wind tunnel model.

The same motion could be observed during the FSI simulation as well. The calculation was re-
peated at larger and larger inflow wind velocities in order to find the critical wind speed. It was
observed that at 10 m/s the motion was damped while at 12 m/s the motion grew (see Figure 14).
Thus, the critical wind speed is between 10 and 12 m/s according to the coupled three dimen-
sional simulation. The flutter motion can be seen in Figure 13.

Figure 13. Streamlines around the deformed FEM model.


The Fifth International Symposium on Computational Wind Engineering (CWE2010)
Chapel Hill, North Carolina, USA May 23-27, 2010

In Figure 14 the rotation of the middle section of the bridge can be seen at 10 and 12 m/s prompt-
ing a critical wind speed value of 10-12 m/s. The spatial eigenvalue analysis gave a critical wind
speed of 8.8 m/s, while the wind tunnel model showed flutter motion at 9.2 m/s. These values are
relatively close together considering the great differences in the methods.

1,00

U=10m/s
0,50
U=12m/s
rot [rad];-

0,00
0,00 0,15 0,30 0,45

-0,50

-1,00
t [s]
Figure 14. Rotation of the middle section of the bridge at two different wind velocities.

5.2 Conclusions
In this paper a fully aero-elastic bridge model was considered. The main question was the critical
wind speed for this system. Our main goal was to apply an advanced numerical method for bridge
flutter assessment therefore a validation was essential. The critical wind speed was calculated in
three different ways. In the wind tunnel the flutter speed was determined. In the FSI analysis with
the ANSYS software the wind speed was increased step-by-step until the critical wind speed was
found. Alternatively, considering the complexity of the problem, an approximation was also ap-
plied. First the aerodynamic derivatives were extracted from a 2D CFD simulation and the critical
wind speed was evaluated using an updated method afterwards.
The critical wind speeds from the three different methods are in line well. The FSI simula-
tion gave a slightly higher value but it seems good coincidence in this really complex simulation.
Naturally the inflow velocity steps should be more accurate in order to capture the flutter speed
more precisely. In addition to that the 3D CFD mesh should be denser.
In our further work the CFD will be updated in order to investigate the flow field in detail.
Furthermore the method will be applied to real bridge structures in the near future.

6 ACKNOWLEDGEMENTS

The authors are grateful for the support of the Department of Fluid Mechanics (University of
Technology and Economics, Budapest), the CFD.hu Ltd. (Hungary), and the Pont-Terv Ltd.
(Hungary).

7 REFERENCES

Starossek, U. (1998): Complex notation in flutter analysis, ASCE, Journal of Structural Engineering, Vol. 124,
No. 8.
Wang Y., Lin Y. (2008): Combination of CFD and CSD packages for fluid-structure interaction, Journal of Hydro-
dynamics, 2008, 20(6):756-761

Anda mungkin juga menyukai