You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228737999

Time and frequency domain flutter solutions for the AGARD 445.6 wing

Article · January 2005

CITATIONS READS
22 2,832

3 authors, including:

F. Nitzsche Daniel Feszty


Carleton University Carleton University
147 PUBLICATIONS   1,036 CITATIONS    36 PUBLICATIONS   551 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Smart Spring View project

All content following this page was uploaded by F. Nitzsche on 19 May 2014.

The user has requested enhancement of the downloaded file.


TIME AND FREQUENCY DOMAIN FLUTTER SOLUTIONS
FOR THE AGARD 445.6 WING
Ryan J. Beaubien1, Fred Nitzsche1, and Daniel Feszty1
1
Department of Mechanical and Aerospace Engineering, Carleton University
1125 Colonel By Drive, Ottawa, Ontario, Canada
e-mail: fnitzsch@mae.carleton.ca

Key words: Aeroelasticity, CFD, Flutter, Time Domain.

Abstract. Time marching simulations using the Euler and Reynolds Averaged Navier-Stokes
equations have been performed for the AGARD 445.6 wing in transonic flow. These
simulations have been compared to experimental results as well as a frequency domain
solution using the Doublet-Lattice Method. The time marching simulations show good
comparison to the experimental results unlike the frequency domain solution, in accordance
with previous works.

1 INTRODUCTION
The most widely used method for flutter certification is based on linearized aerodynamic
potential theory, specifically, the Doublet-Lattice Method (DLM)1. This method is available
in various commercial software packages such as the Aeroelastic module for
MSC/NASTRAN. However, a major disadvantage of the DLM is that it fails to capture the
location and magnitude of local shock waves and the associated shock wave-boundary layer
interactions on the wing surface in the transonic regime. The prediction of these phenomena
are crucial for assessing the aeroelastic behaviour of a wing in transonic flow as they are the
source for several nonlinear aeroelastic effects, such as the transonic dip and limit cycle
oscillations.

A solution to this problem is to complete a time marching analysis where a Computational


Fluid Dynamics (CFD) solver is coupled with a Computational Structural Dynamics (CSD)
solver. In these simulations, the structure is given an initial velocity in one of the dominant
modes and the subsequent time evolution of the modal response is calculated to see whether it
grows or decays2. The flow may be simulated by solving the Euler or Reynolds Averaged
Navier-Stokes (RANS) equations. The computational cost is not unreasonable when the
intention is to examine behaviour at previously identified problem conditions3. However, for
unknown problem conditions, i.e., defining the flutter boundary of a new wing, this method is
extremely expensive thus negating its use in a production environment.

The purpose of this paper is to compare the flutter boundary obtained from using DLM
aerodynamics in a frequency domain analysis with the time marching analysis using Euler and
RANS equations for a wing in the transonic regime.

2 TEST CASE
The AGARD 445.6 wing was selected as the test case since flutter measurements are
available for a wide range of Mach numbers4. Results have also been published in various
papers on computational aeroelasticity, including Goura2, Badcock3 and Melville5. The wing
has a quarter chord sweep of 45°, an aspect ratio of 1.65, a taper ratio of 0.66 and a constant
NACA 65A004 symmetric profile. The experiment was conducted in the NASA Langley

1
Transonic Dynamics Tunnel and the results were published in 1963. Various wing models
were tested; the case most commonly used in computational aeroelasticity papers is the
weakened wing model at zero angle of attack in air. This model was constructed of laminated
mahogany and had holes drilled through the wing to reduce its stiffness. Flutter speed
coefficients, U, for Mach numbers in the range of 0.338 to 1.141 were reported4. The flutter
speed coefficient is expressed as:
U = U∞ / ( bs ωα μ1/2 ) (1)
where U∞ is the freestream velocity at flutter, bs is the semispan, ωα = 39.44 Hz is the
frequency of the first torsional mode (39.44 Hz) and μ = m/( ρ ∞ V) where m = 1.693 kg is the
mass of the wing, V = 0.130 m2 is related to the volume of the wing and ρ ∞ is the freestream
density at flutter.

Time marching simulations will also be run for an angle of attack of 5° in order to test the
ability of the simulations to capture significant nonlinear effects. Note that no experimental
data is available for this particular case.

3 FREQUENCY DOMAIN SOLUTIONS USING MSC/NASTRAN


The flutter analysis in MSC/NASTRAN is conducted in the frequency domain. The solution
involves a series of complex eigenvalue solutions. MSC/NASTRAN provides three methods
of analysis: the American (K) method, a restricted but more efficient American (KE) method
and the British (PK) method. The PK-method was selected as it also provides approximate
estimates of system damping at subcritical speeds which is useful for monitoring flight tests6.

3.1 Structural Model


The linear structural model for the 445.6 wing was created in MSC/NASTRAN using the
model parameters in the aeroelastic optimization study by Kolonay7. This model was selected
for comparative purposes as it was also used by Goura2, Badcock3 and Melville4 for time
marching studies. The wing is modelled with plate elements as a single layer orthotropic
material. The model consisted of 231 nodes and 200 elements (Fig. 1). The thickness
distribution was governed by the airfoil shape. The material properties used were
E1 = 3.1511 GPa, E2 = 0.4162 GPa, ν = 0.31, G = 0.4392 GPa and ρ = 381.98 kg/m3 where E1
and E2 are the moduli of elasticity in the longitudinal and lateral directions, ν is Poisson's
ratio, G is the shear modulus in each plane and ρ is the wing density7. Table 1 compares the
measured and calculated first four fundamental modes and Fig. 2 shows the calculated mode
shapes.

Figure 1: AGARD 445.6 structural model.

2
Mode 1 [Hz] Mode 2 [Hz] Mode 3 [Hz] Mode 4 [Hz]
4
Experiment 9.60 38.10 50.70 98.50
7
Kolonay 9.63 37.12 50.50 89.94
2
Goura 9.67 36.87 50.26 90.00
Calculated 9.46 39.44 49.71 94.39

Table 1: Comparison of modal frequencies for AGARD 445.6 wing.

Mode 1 (1-B) Mode 2 (1-T)

Mode 3 (2-B) Mode 4 (2-T)


Figure 2: AGARD 445.6 structural model.

Using the density and model definition of Kolonay7 resulted in a wing mass of 1.693 kg, 9%
lighter than the Kolonay model. The mass of the Kolonay model was equal to the
experimental model. The modes and mode shapes showed good comparison between all three
models, thus the density was not adjusted to match the wing mass.

The linear structural model was used for both the Euler and RANS time marching as well as
the MSC/NASTRAN simulations. No structural damping is possible for the time marching
solutions. Thus, for comparative purposes, the structural damping for the NASTRAN model
was set to zero.

3.2 Aerodynamic Model


A linear aerodynamic model was created for the DLM. The wing is modelled by panels and
the lifting surfaces are assumed to lie nearly parallel to the flow. DLM does not account for
thickness effects of the lifting surfaces and assumes the undisturbed flow is uniform or
varying harmonically. The model consisted of 25 points in the spanwise direction and 28
points in the chordwise direction. The spanwise spacing was set similar to a CFD grid
distribution, i.e., concentrated near the tip region. The chordwise point distribution was
concentrated at the leading edge distribution while maintaining the recommended three-to-one
box ratio6. This model was adequate for the Mach numbers below 1.0; however, it grossly
under predicted the flutter speed for higher Mach numbers. A second model was created with
box ratios close to unity as recommended by Ref. 6. This model possessed the same number
of boxes as the previous model.

3
4 TIME MARCHING SIMULATIONS
The time marching simulations were performed using the PMB (Parallel Multi-block) code
developed at the University of Glasgow. This code features a finite-volume Euler and RANS
CFD solver with a proved capability of accurately capturing transonic effects2,3.

4.1 Mesh Generation


For the Euler calculations, an O-O grid was used whereas a C-H grid was generated for the
RANS calculations. The spanwise grid densities of the structured, multiblocked grids were
increased towards the tip based on the grid convergence study by Badcock3. Medium and
coarse grids were created for the O-O and C-H grids. The medium and coarse O-O grids had
190,000 and 27,000 nodes respectively, with 4453 and 1131 points on the wing surface. The
medium and coarse C-H grids had 324,000 and 45,000 points respectively, with 2862 and 979
points on the wing surface. A zoom of the root airfoil section for each medium grid is shown
in Fig. 3.

Figure 3: C-H and O-O grid topologies for the RANS and Euler simulations, respectively.

4.2 Numerical Method


PMB uses a cell centred finite volume technique to solve the Euler and RANS equations. The
diffusive terms are discretized using a central differencing scheme and the convective terms
use Roe's scheme with MUSCL interpolation offering third-order accuracy. Steady flow
calculations proceed in two parts, initially running an explicit scheme to smooth out the flow
solution, then switching to an implicit scheme to obtain faster convergence. The
preconditioning is based on block incomplete lower-upper surface factorization and is also
decoupled between blocks to increase the parallel performance. The linear system arising at
each implicit step is solved using a generalized conjugate gradient method8.

There are several turbulent models available in the code, including the Spalart-Allmaras, k-ω,
k-ε and SST models8, although in the present work only laminar calculations were performed.

For aeroelastic simulations, the flow domain is deforming which is achieved by interpolating
the boundary displacements to interior points. Grid speeds and transformation Jacobians are
calculated by finite differencing. Cell volumes are recalculated using a global conservation
law by considering volume fluxes through cell sides, the so-called Geometric Conservation
Law (GCL). The internal volume of the wing is kept constant by the Constant Volume
Tetrahedron (CVT) method. Full details can be found in Ref. 8.

4
4.3 Verification of the Numerical Method
A grid convergence study was conducted for a steady case at zero angle of attack, Mach
number of 0.96 and a Reynolds' Number of 4.51x105 (based on mean aerodynamic chord).
The results for the Euler and RANS solutions are shown in Fig. 4 for four wing stations. Both
the medium and coarse grids show good agreement for the respective topologies. The Euler
and RANS solutions show disagreement near the trailing edge of the wing.
Span = 0.0% Span = 25.0%
-0.2 -0.2

0
0

0.2
Cp

Cp
0.2
0.4

0.4
0.6

0.8 0.6
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

Span = 90.0% Span = 95.0%


-0.2 -0.2

0 0
Cp

Cp

0.2 0.2

RANS Coarse
0.4 0.4 RANS Medium
Euler Coarse
Euler Medium
0.6 0.6
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

Figure 4: Grid convergence study for AGARD 445.6 wing (M = 0.96, Re = 4.51x105, α = 0°).
Pressure coefficients shown for the upper wing surface.

5 COMPARISION OF MEASURED AND CALCULATED RESULTS

5.1 Grid Density and Fluid Model Effect on Flutter Speed


The comparison of the time marching solutions at Mach 0.96 (near the bottom of the transonic
dip) to various published grid densities and fluid models is shown in Table 2.

Flutter Speed
Reference Grid Volume Fluid Model
Coefficient
Medium 0.308
Current RANS
Coarse 0.327
Medium 0.317
Current Euler
Coarse 0.330
Fine 0.175
Badcock3 Medium Euler 0.192
Coarse 0.227
Fine 0.314
Melville5 Medium RANS 0.304
Coarse 0.285
Table 2: Comparison of flutter speed coefficients at Mach 0.96 for various grids.

5
The results of Melville5 show a 6% downward trend in the flutter speed between the medium
and coarse grids. The current results show an upwards trend (6% for RANS, 4% for Euler),
similar to the results of Badcock3. The Euler and RANS solutions show good agreement for
each grid density.

5.2 Flutter Boundary for Zero Angle of Attack


The flutter speed coefficients for Mach 0.499 to 1.072 for the current methods, previously
published results by Goura2 and the experimental results are shown in Fig. 5. The coarse grid
time domain Euler and RANS simulations produced similar flutter boundaries for the
subsonic and transonic regime. These boundaries also exhibit a shape similar to the
experimental results. The difference may be attributed to the lack of structural damping in the
time domain solutions. Previous results have shown using a value of structural damping of
0.5% will shift the boundary towards the experimental results3. The transonic dip is more
pronounced in Goura's2 results than the experimental and time domain Euler and RANS
solutions.

The MSC/NASTRAN solution using DLM aerodynamics produced a significantly higher


flutter boundary. This disagreement was expected as current industry practice is to alter the
aerodynamic data by incorporating wind tunnel data or steady state CFD results9,10. The
velocity-damping (V-g) and velocity-frequency (V-f) plots for Mach 0.499 are shown in
Figs. 6 and 7. The mechanism of flutter for all tested Mach numbers is between modes 1
(first bending) and 2 (first torsion).

0.50

0.48

0.46

0.44
Flutter Speed Coefficient

0.42

0.40 Experiment
Time Marching, Euler
Time Marching, RANS
0.38
Nastran DLM
Goura, Euler
0.36

0.34

0.32

0.30

0.28
0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10
Mach Number

Figure 5: Flutter boundary for the AGARD 445.6 wing (α = 0°).

6
0.5
Mode 1 (1-B)
Mode 2 (1-T)
0.4 Mode 3 (2-B)
Mode 4 (2-T)

0.3

0.2

0.1
Damping

-0.1

-0.2

-0.3

-0.4

-0.5
50 100 150 200 250
Velocity (m/s)

Figure 6: V-g plot for the AGARD 445.6 wing using MSC/NASTRAN (M = 0.499).

100

90

80

70

60
Frequency (Hz)

50

40

30

20

Mode 1 (1-B)
10 Mode 2 (1-T)
Mode 3 (2-B)
Mode 4 (2-T)
0
50 100 150 200 250
Velocity (m/s)

Figure 7: V-f plot for the AGARD 445.6 wing using MSC/NASTRAN (M = 0.499).

7
It is interesting to note that the runtimes for the time domain simulations were quite
acceptable. On a cluster of four 3.2 GHz machines, the average simulation time required to
calculate each flutter data point was 40 minutes and 50 minutes, using Euler and RANS
equations, respectively. The entire flutter boundary required approximately 2.5 hours and 3.5
hours to complete.

5.3 Flutter Boundary for α = 5°


The time domain solutions were also ran for an angle of attack of 5° where nonlinear effects
are even more pronounced. The flutter speed coefficients for Mach 0.678 to 1.072 are shown
in Fig. 8. The Euler and RANS solutions diverge after Mach 0.901. At this angle of attack,
the shock-boundary layer interaction becomes significant and the Euler solution is unable to
capture these effects.

0.42

0.40

0.38

0.36
Flutter Speed Coefficient

0.34

0.32

Time Marching, Euler


0.30 Time Marching, RANS

0.28

0.26

0.24

0.22

0.20
0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10
Mach Number

Figure 8: Time marching flutter boundary for the AGARD 445.6 wing (α = 5°).

6 CONCLUSION
Flutter results were obtained for the AGARD 445.6 wing in the time domain, using the Euler
and RANS equations, as well as in the frequency domain, using DLM aerodynamics and
MSC/NASTRAN. The following conclusions can be drawn from the study: (1) for transonic
flow conditions with insignificant nonlinear effects, the time domain Euler and RANS
simulations produce similar flutter boundaries; (2) Euler solutions are unable to produce
accurate flutter boundaries when the nonlinear flow effects are significant, thus the RANS
equations must be used for these cases; and, (3) unreliable transonic flutter boundaries are
produced from frequency based solvers using unaltered DLM aerodynamics.

8
7 REFERENCES
[1] Yurkovich, R., "Status of unsteady aerodynamic prediction for flutter of high-
performance aircraft", Journal of Aircraft, Vol. 40, No. 5, 2003, pp. 832-842.
[2] Goura, G. S. L., "Time marching analysis of flutter using Computational Fluid
Dynamics", Ph. D. thesis, University of Glasgow, 2001.
[3] Badcock, K. J., Woodgate, M. A. and Richards, B. E., "Direct aeroelastic bifurcation
analysis of a symmetric wing based on the Euler equations", Technical Report 0315,
Department of Aerospace Engineering, University of Glasgow, 2003.
[4] Yates, E. C., "AGARD standard aeroelastic configurations for dynamic response I-wing
445.6", AGARD Report 765, North Atlantic Treaty Organization, Group for Aerospace
Research and Development, 1988.
[5] Gordnier, R.E. and Melville, R.B., "Transonic flutter simulations using an implicit
aeroelastic solver", Journal of Aircraft, Vol. 37, No. 5, 2000, pp. 872-879.
[6] Rodden, W. P. and Johnson, E. H., MSC/NASTRAN Aeroelastic Analysis User's Guide,
Version 68, The MacNeal-Schwendler Corp., 1994.
[7] Kolonay, R. M., "Unsteady aeroelastic optimization in the transonic regime", Ph. D.
thesis, Purdue University, 1996.
[8] Goura, G. S. L., Badcock, K. J., Woodgate, M. A. and Richards, B. E., "Implicit methods
for the time marching analysis of flutter", Aeronautical Journal, Vol. 105, 2001, pp. 199-
215.
[9] McCain, W. E., "Measured and calculated airloads on a transport wing model", Journal
of Aircraft, Vol. 22, No. 4, 1985, pp. 336-342.
[10] Pitt, D. M. and Goodman, C. E., "Flutter calculations using doublet lattice aerodynamics
modified by the full potential equations", Paper No. AIAA-87-0882, American Institute
of Aeronautics and Astronautics, Inc., 1987, pp. 506-512.

View publication stats

You might also like