Agard 445.6 Wing Flutter Validation
Agard 445.6 Wing Flutter Validation
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
FLUID STRUCTURE INTERACTION ON AGARD 445.6 WING AT MACH 0.9
1 Introduction
The high velocity on aircraft lead to thin wings which are more likely to a mechanical failures. Consequently,
a flutter study on airplane wings and turbomachinery blades is a critical issue to asses the reliability of the aircraft.
In this phenomenon, aerodynamics forces acts on the structure deforming it. In this sense, flutter is a result of the
Fluid-Structure Interaction (FSI) and is commonly associated with complex physics such as shock wave, boundary
layer collapse and oscillations. In this sense, an accurate prediction of flutter is a challenging problem which
requires a powerful software and computational resources.
Over the last years advances in computer technology and developments in numerical methods turned a FSI
problem problems able to be simulated. As this issue deals with both fluid and solid, the governing equations have
to be couple by efficient algorithms. Therefore, it requires the simultaneous application of Computational Fluid
Dynamics (CFD) code so as to solve the aerodynamic forces and Computational Structural Dynamics (CSD) code
to determine the solid deformation. Therefore, a fluid motion generates forces on structures, which deform due
these forces and then the fluid movement is also affected by the structural displacement which results in new forces
an this loop goes on. Traditionally, each of these computational code have been developed in different numerical
technique. While most of CFD solvers are based on Finite Volume Method (FVM), CSD solvers uses a Finite
Element Method (FEM). As the former uses a Eulerian or spatially fixed-coordinate system and the latter uses a
lagragian or material fixed-coordinate system some approaches must be done so as to get the interaction between
the two simulation modules. Therefore, an interfacing technique is one of the key factor since these two codes
must communicate efficiently.
In general, there are two methods used to couple a FSI problem in the time: fully and loosely. In the former
model, the structural response delays behind the fluid solution. In addition, this method is limited to first-order
accuracy in time anyhow of the temporal precision of the solvers (Melville & Morton 1998). On other hand, fully
coupled communicate simultaneously by exchanging the aerodynamic loads and structural displacement within
each interaction. Consequently, it also brings a computational difficulties such as instability during the solution
process (Kamakoti & Shyy 2004).
In this work, a FSI problem will be present using the AGARD 445.6 which is a benchmark experiment
Yates Jr (1987) of flutter phenomena. Then a extensive research have been done using this wing (Lee-Rausch &
Batina 1993, 1996, Lesoinne & Farhat 1998, Liu et al. 2001, Geuzaine et al. 2003, Thirifay & Geuzaine 2008,
Chen et al. 2007). More recently, (Goud et al. 2014) performed a simulation of FSI problem in this wing at Mach
0.9 and discussed the results concerning the pressure, turbulence kinetic energy and temperature over the wing.
The objective of this paper is to perform a validation of flutter in Agard 445.6 wing at 0.9 mach using a commercial
platform Ansys in a personal computer (intel i-7 1.8GHZ - 16Gb RAM).
The AGARD 445.6 wing was tested by Yates Jr (1987) in both air and Freon-12 in the 16 x 16 foot NASA
Langley Research Center. The first numerical digit named in the wing is associated with the aspect ratio. then,
the following two digits refer to the quarter-chord sweep angle (45°) and the last digit indicates the taper ratio
(0,66). As it is shown in This wing was based in the Naca 65A004 airfoil with no twist or curvature along its
length. The root chord of this model was 0,55m and the semi-span was 0,76m. The semi-span model was attached
directly to the wind tunnel wall, therefore, the wing root is into the wall boundary layer. The test model used in
the experiments (?) was constructed using laminated mahogany, which has different material properties in each
direction. Also, the model used in tests was weakened by drilled hole
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
SILVA. P.A.S., MORAIS, M.V.G.,
0,368m
e
lin
o rd
ch
r
rte
0,762m
ua Naca 65A004
Q
45°
0,559m
The test model used in the experiments(Yates Jr 1987) was constructed by laminated mahogany, which has
different material properties in each direction. Also, the model used in tests was weakened by drilled holes which
were filled by foam in order to obtain easily the flutter conditions. The material properties in each direction for this
weakened AGARD 445.6 wing is given in table x, where E id the Elasticity Modulus, G is Shear modulus and ρ
is density.
Table 1: Material Properties for weakened AGARD 445.6 wing Yates Jr (1987)
In order to represent the deformation problem, the governing equations related to a structural dynamics is
given by a stress strain relationship for a linear material as follows:
σ = D~ (1)
where σ is the stress tensor, D is the elasticity matrix and ~ is the train vector. Since Ansys workbench environment
has a multi-physics platform, it provides a much easier for a user perform a fluid structure interaction. In this sense,
so as to facilitate the communication between CFD and CSD, the Ansys transient structural solver was employed
to deal with the solid governing equations.
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
FLUID STRUCTURE INTERACTION ON AGARD 445.6 WING AT MACH 0.9
The simulation process on Ansys passes through many steps from the geometry preparation to the solution.
First, the Agard 445.6 wing was built by a CAD software, such as Solidworks, and imported into Ansys. In the
next step, the mesh was generated in the wing, which reached a 11727 nodes. Then, the material properties of the
mahogany were set according to data presented by Yates Jr (1987), which also is exposed in table 1. Last but not
least, the boundary conditions were applied as shown in figure 2. Besides the gravitational forces in the y axis,
the root wing was considered as fixed and in others wing surfaces, such as top,tip and bottom, were applied fluid
structure interface. Also, the solver was set to perform a one sub-step per time-step, which is controlled defined by
the CFD solver.
A similar procedure was done in the modal simulation. The main difference from the transient structural
solver was the boundary conditions. In the modal part the wing root surface still fixed, however, the fluid-structure
interfaces were not taking into account. In this matter, this solver were decoupled from the CFD part.
FLUID SOLID
INTERFACE
FIXED
Y
Z X
The fluid flow is assumed as compressible and fully turbulent. Consequently, velocity and pressure fields are
governed by the Navier-Stokes equations. Due to the computational resource available, the turbulence phenomena
was considered without numerically solving all the eddy scales. Then, the Unsteady Reynolds Averaged Navier-
Stokes (URANS) approach was adopted. Hence, the governing equations of the flow are the continuity, momentum
and energy conservation equations that can be described, respectively, as
∂ρ ∂ui
+ =0 (2)
∂t ∂xi
∂ρui ∂ρuj ui ∂p ∂
+ =− + (2µSij − ρui uj ) + pfi (3)
∂t ∂xj ∂xi ∂xi
" !# " !#
∂ 1 ∂ 1 ∂ ∂qj
ρ e + ui ui + ρuj h + ui ui = [uj (2µSij − ρui uj )] − (4)
∂t 2 ∂xj 2 ∂xj ∂xj
where ui are the components of the velocity, ρ is the density, p is the mechanical pressure, µ is the viscosity, fj is a
force per unit of volume such as gravitational force, e is specific internal energy, h = e + p/ρ is specific enthalpy,
and qj is the heat flux vector. In the present work, the so-called κ − ω Shear-Stress Transport (SST) model,
in the form as described by Menter (1993, 1994) Traditionally, in Reynolds Averaged Navier-Stokes (RANS)
calculations, the steady-state solution is main objective. However, as the time interval used for the time averaging
is reduces, and the transport equations are solved as dependent on time, an URANS scheme can be obtained.
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
SILVA. P.A.S., MORAIS, M.V.G.,
The simulation were carried out by Ansys CFX, which is based on finite-volume method with an unstructured
parallelized coupled algebric multigrid solver with a second order advection scheme. Moreover, the unsteady
terms in the governing equations are solved by an implicit second order backward time-stepping algorithm.
The computational domain has a shape and dimensions as depicted in figure 3. The domain were assumed
as big enough to represent the physics of the problem. The mesh generated for this domain achieved 505364
nodes. Then, the boundary conditions were applied in this domain in order to accurately represent the experiment
as follows:
• Inlet velocity: a Dirichlet condition which is applied on the curved surface set a uniform velocity of
269, 69ms− 1 in the x direction with turbulence intensity of 5% and temperature of 269, 86°K. In this
manner, the pressure is determined in a way to satisfy transport equations.
• Outlet pressure: in the upwind face another Dirichlet boundary condition was applied so as to define the
pressure gauge as 0 Pa. Therefore, the velocity field is evaluated in order to satisfy transport equations.
• wall: a non-slip condition were applied to the wing surfaces, implying that the relative velocity of the fluid
particle to the mesh motion is zero. Also, these surfaces wee set to has a interface with the CSD solver being
able to receive data concerning the mesh displacement.
• Symmetry: applied on the lateral surface. In this condition, zero normal velocity at symmetry plane.
• Time-step: on the unsteady simulation was set a timestep of 0.0025s and a total time of 0.5s.
OUTLET
SYMMETRY
INLET
WALL
Y
m
10
Z X
Ansys workbench software performs FSI by connecting the coupling participants to a component system
named as System Coupling in a loosely-coupled method. In this sense, the participant, CSD and CFD, either feed or
receives data into the coupled analysis. Firstly, the system collects information from the participants so as to adjust
the system to start the simulation and then the information to be exchanged are given to the respective participant.
Afterwards, the work process is organizing the sequence of transfer of information. Lastly, the convergence of
coupling step is evaluated at end of every coupling iteration. In this case, as any time-step stars, CFD solver obtain
the convergence in its own criteria and transfers the fluid forces to CSD solver. Hence, the displacement result of
a CSD is solved using the solution provided by CFD for the same time-steps, as it is shown in figure 4 and 5.
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
FLUID STRUCTURE INTERACTION ON AGARD 445.6 WING AT MACH 0.9
Steps
1 5
CFD
2 6
4 8
3 7
CSD
t t + Δt t + 2Δt
Figure 4: System coupling steps
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
SILVA. P.A.S., MORAIS, M.V.G.,
At each time step showed in figure 4, four phase are done as follows:
A Solving the flow field as the boundary conditions are applied originally, shown in step 1
B Transferring data from CFD to CSD and converting the surface pressure to the general aerodynamic forces
imposed in structural motion equations as showed in step 2.
C Performing structural motion equations, and updating the generalized displacement and velocity vector as
showed in the step 3.
D Transferring data from CSD to CFD, and the aerodynamic mesh on the wing boundary is deformed in
agreement with the structural distortion, shown in step 4.
E Solving the Navier Stokes equations again with the new boundary conditions and updating the variables of
the flow field, shown in step 5.
Hence, by doing these steps successively, the time history resolution of both fluid and structure field are acquired.
The modal data for the Agard 445.6 wing, in the form of the frequencies of the many natural nodes and its
mode shapes were provided by Yates Jr (1987). Then, the numerical results of the first four frequencies were
compared with the experimental data and other works as it is shown in table 2. It is noted that the results achieved
here are well agreed with the experiment of Yates Jr (1987). In addition, as shown in figure 6 the contour of the
mode frequencies also are similar to the experiment.
Before the flutter analysis of the wing takes place, the steady state response of the wing was numerically
solved with the freestream conditions set to ux = 269.69 m/s, uy = 0.26969 m/s and T = 269.86°K. A steady
state solution was first computed to a undeformed wing. Then, this perturbation fo the steady state solution was
used as an initial condition for the unsteady state. In this part, uy was set equal to zero and ux remained at the same
value. As mentioned before, it was set a time step ∆t = 0.0025s and a total time of 0.5s and the displacement
of both tip leading and trailing edge was monitored in time as show in figure 8. As it is clearly noted there are
a damping in the displacement over the time. In addition, it is noted that even though the velocity far away from
the wing is set in 0.8 Mach, at the leading edge it reaches 0.9 Mach as can be seen in figure 7. In this sense, for
the flutter analysis on the leading edge, the velocity at 0.9 Mach was considered to compare against experimental
results.
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
FLUID STRUCTURE INTERACTION ON AGARD 445.6 WING AT MACH 0.9
NUMERICAL EXPERIMENTAL
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
SILVA. P.A.S., MORAIS, M.V.G.,
4
Displacement [m]
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time [s]
So as to obtain the flutter frequency, a Fast Fourier Transform (FFT) is applied on the tip leading edge
displacement data as shown in figure 9. The small bump around 7Hz is associated to data corruption that may
be occurred due to the initial perturbation in the system. Hence, the first peak at 14Hz is the flutter frequency
concerning the first bending mode (9.6Hz). Also, the small bump around 50Hz is likely to be the first torsion
mode (38.10Hz), which requires smaller timesteps to be solved accurately. Thereby, to reach further modal modes
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
FLUID STRUCTURE INTERACTION ON AGARD 445.6 WING AT MACH 0.9
it is required a finer mesh and smaller timesteps. In this sense, as the experimental flutter frequency is 20.3Hz
(Yates Jr 1987), the numerical result achieved here presents an error of 31%.
7 Conclusions
This work was largely aimed at giving a basic understanding and an overview of the fundamental FSI prob-
lem using a benchmark experiment such as AGARD 445.6 wing and complex phenomena like mesh motion and
compressible flow in a personal computational resource. As previously mentioned, FSI plays a crucial role in
many engineering fields which in many cases neglected due to lack of resources. In this sense, the present study
performed a classic FSI flutter problem by a intel i-7 1.8GHZ - 16Gb RAM computer using a commercial soft-
ware, which ran for 23 hours using only one core. It is important to mention that concepts and results achieved
in studies like this one can be easily used as educational practices, as the limitations of mesh used here fit in an
student license provided by Ansys. In addition, a finer mesh and smaller time-step would provide a better accuracy
of results, especially in flutter frequency. All in all, the content presented here provides a great exposure of Fluid
Structure Interaction flutter problem of a wing at Mach 0.9.
ACKNOWLEDGEMENTS
The authors would like to thanks University of Brasilia - UnB and Laboratory of Energy and Environment
for all support.
CILAMCE 2016
Proceedings of the XXXVII Iberian Latin-American Congress on Computational Methods in Engineering
Suzana Moreira Ávila (Editor), ABMEC, Braslia, DF, Brazil, November 6-9, 2016
References
Ansys, A. F., 2011. ‘14.0 theory guide’, ANSYS inc pp. 218–221.
Ayabakan, S., 2008. Loosely-coupled flutter analysis of agard wing 445.6, in MpCCi, ed., ‘9th MpCCI User
Forum’, pp. 53–61.
Chen, X., Zha, G.-C. & Yang, M.-T., 2007. ‘Numerical simulation of 3-d wing flutter with fully coupled fluid–
structural interaction’, Computers & fluids pp. 36,5,856–867.
Geuzaine, P., Grandmont, C. & Farhat, C., 2003. ‘Design and analysis of ale schemes with provable second-order
time-accuracy for inviscid and viscous flow simulations’, Journal of Computational Physics pp. 191,1,206–227.
Goud, T. S. K., Sai Kumar, A. & Prasad, S. S., 2014. ‘Analysis of fluid-structure interaction on an aircraft wing’,
Analysis 3(9).
Kamakoti, R. & Shyy, W., 2004. ‘Fluid–structure interaction for aeroelastic applications’, Progress in Aerospace
Sciences pp. 40,8,535–558.
Lee-Rausch, E. & Batina, J. T., 1996. ‘Wing flutter computations using an aerodynamic model based on the navier-
stokes equations’, Journal of Aircraft pp. 33,6,1139–1147.
Lee-Rausch, E. M. & Batina, J. T., 1993. ‘Calculation of agard wing 445.6 flutter using navier-stokes aerodynam-
ics’, AIAA paper pp. 3476,9–11.
Lesoinne, M. & Farhat, C., 1998. ‘Higher-order subiteration-free staggered algorithm for nonlinear transient aeroe-
lastic problems’, AIAA journal pp. 36,9,1754–1757.
Liu, F., Cai, J., Zhu, Y., Tsai, H. & F. Wong, A., 2001. ‘Calculation of wing flutter by a coupled fluid-structure
method’, Journal of Aircraft pp. 38,2,334–342.
Melville, R. B. & Morton, S., 1998. ‘Fully-implicit aeroelasticity on overset grid systems’, AIAA Paper-98-0521 .
Menter, F. R., 1993. ‘Zonal two equation k-turbulence models for aerodynamic flows’, AIAA paper pp. 2906,1993.
Menter, F. R., 1994. ‘Two-equation eddy-viscosity turbulence models for engineering applications’, AIAA journal
pp. 32,8,1598–1605.
Thirifay, F. & Geuzaine, P., 2008. ‘c’, URL http://citeseerx. ist. psu. edu/viewdoc/summary .
Yates Jr, E. C., 1987. ‘Agard standard aeroelastic configurations for dynamic response. candidate configuration
i.-wing 445.6’.
Zuijlen, A. V., 2007. High-order time integration throuh smooth mesh deformation for 3-d fluid-structure interation
simulations, in MpCCi, ed., ‘Journal of Computational Physics’, pp. 224,414–4430.
11