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

net/publication/313457160

Aeroelastic Simulations Using ANSYS Multiphysics Software

Conference Paper · January 2017

DOI: 10.2514/6.2017-0192

CITATION READS
1 1,912

3 authors, including:

Balasubramanyam Sasanapuri Krishna Zore

ANSYS ANSYS Inc. Pune India
16 PUBLICATIONS 129 CITATIONS 22 PUBLICATIONS 114 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Krishna Zore on 11 September 2019.

The user has requested enhancement of the downloaded file.

 AIAA 2017-0192
AIAA SciTech Forum
9 - 13 January 2017, Grapevine, Texas
58th AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference

Aeroelastic Simulations Using ANSYS Multiphysics

Software

Krishna Zore1 and Balasubramanyam Sasanapuri2

ANSYS Software Pvt Ltd, Plot no. 34/1, Hinjewadi, Pune - 411057, Maharashtra, India

Eric Bish3
ANSYS Inc, 10 Cavendish Court, Centerra Resource Park Lebanon NH 03766

This paper presents results from simulations performed for the 2nd AIAA Aeroelastic Prediction Workshop
using ANSYS multiphysics software. Two cases are studied. The first case is a forced oscillation simulation to
model the Oscillating Turn Table (OTT) experiment. The second is a flutter simulation to model experiments
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

done with the Pitch and Plunge Apparatus (PAPA). The results obtained by ANSYS multiphysics solvers
CFX and Fluent compare well with experiments of the Benchmark Supercritical Wing (BSCW) for steady
wing pressure coefficient and unsteady pressure frequency response function (FRF). A slight sensitivity in
pressure coefficient with grid refinement was observed on the suction surface just downstream of the leading
edge. To accurately capture the suction peak pressure coefficient further grid refinement is recommended.
BSCW simulations performed at experimental on-set flutter dynamic pressure using ANSYS CFX show good
agreement with experimentally measured flutter frequency and pressure coefficient magnitude and phase.
The on-set wing flutter is predicted at a lower dynamic pressure than the experimental on-set flutter dynamic
pressure.

Nomenclature
AePW = Aeroelastic Prediction Workshop
DPW = Drag Prediction Workshop
HiLiftPW = High Lift Prediction Workshop
TDT = Transonic Dynamic Tunnel
OTT = Oscillation Turntable
PAPA = Pitch and plunge Apparatus
FRF = Frequency Response Function
FFT = Fast Fourier Transform
BSCW = Benchmark Supercritical Wing
Cd = Drag coefficient
Cl = Lift coefficient
Cp = Pressure coefficient
c = Chord length
M, Mach = Mach Number
AOA = Angle of Attack
q = Dynamic pressure
RANS = Reynolds Averaged Navier-Stokes
URANS = Unsteady Reynolds Averaged Navier-Stokes
SA = Spalart-Allmaras
SST = Shear Stress Transport
SST k-ω = Shear Stress Transport k-ω
f = Frequency
tt = Time step per period
CAD = Computer Aided Design
XC = Streamwise position normalized by chord length

1
Application Software Developer
2
Lead Technology Specialist, Senior AIAA Member
3
Senior Manager, Software Development
1
American Institute of Aeronautics and Astronautics

Copyright © 2017 by ANSYS Inc.. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 I. Introduction
The Aeroelastic Prediction Workshop (AePW) provides an opportunity to assess state-of-the-art computational
methods and tools for predicting aeroelastic phenomena. The test cases proposed for AePW-2 are Case-1: Steady
and forced oscillation tested in the NASA Langley Transonic Dynamic Tunnel (TDT) using the Oscillation
Turntable (OTT)5,6 facility, and Case-2: Wing flutter tested in the TDT on a flexible mount pitch and plunge
apparatus (PAPA)13,15 which provides two degree-of-freedom dynamic motion. For Case-1 experimental
measurements of steady pressure and pressure frequency response function (FRF) magnitude and phase at 60% wing
span are provided for comparison. For Case-2, the onset flutter dynamic pressure and FRF pressure magnitude and
phase at 60% and 95% wing span are provided.
ANSYS unstructured Reynolds Averaged Navier-Stokes (RANS) CFD solvers Fluent and CFX are used to
obtain the computational results. Case-1 computations are performed on supplied coarse, medium and fine grids
using the Spalart-Allmaras (SA), Shear stress transport (SST) and SST k-ω turbulence closure models. Modified
meshes were generated based on a validated strategy14 to incorporate pitch and plunge dynamic motion and were
used for Case-2 flutter analyses. Case-1 results by both the ANSYS solvers were presented in the 2nd Aeroelastic
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Prediction Workshop (AePW-2) held in January 2016.

II. Geometry and Computational Grids

A. Geometry
The Benchmark Supercritical Wing (BSCW) airfoil is a NASA SC (2)-0414 2nd generation supercritical airfoil
with design normal force coefficient of 0.4 and 14% thickness to chord ratio (see Figures 1 and 2). The planform is
rectangular with wing tip cap shape as a tip of revolution. The pitching axis is at 30% and 50% chord for the OTT
and PAPA tests, respectively. For the OTT7 test, the boundary layer transition was fixed at 7.5% chord using a size
30 grit flow transition strip. The PAPA13 test was conducted with several flow transition strip configurations; only
data using 35 grit was used for these comparisons. All cases have transition grits on both upper and lower wing
surfaces. Pressure measuring transducers12 were placed at 60% for OTT test and at 60% and 95% for the PAPA test.
Engineers at NASA constructed CAD geometry referring to the measured data shown in Table 1. The CAD
IGES file includes a splitter plate geometry, which isolates the wing from the wind tunnel wall, thus preventing flow
disturbances arising from the wall hitting the wing1. However, the splitter plate is not included in the provided
computational meshes and hence the CFD computations.

Figure 1. Cross-sectional view of the SC(2)-0414 airfoil, with BSCW instrumentation.

2
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

(a) Photograph of the BSCW model mounted on the OTT in (b) BSCW Geometry.
the TDT.
Figure 2. BSCW model.

Table 1. BSCW geometric reference parameters.

Description Symbol Value

Reference chord cref 16 inches
Model span b 32 inches
Area A 512 inch2
Moment reference point relative to axis system def. x 4.8 inches, 30%
y 0.0 inches
z 0.0 inches
Frequency Response Function reference quantity FRF Pitch angle

B. Provided Grids
In order to minimize the solution variability due to the use of customized grids, the organizers of AePW-2
provided three sets of common grids to all participants. These grids were generated following the gridding
guidelines developed for the Drag Prediction Workshop (DPW)17 and High Lift Prediction Workshop (HiLiftPW)18.
The grid metrics are shown in Table 2 and representative grid images in Figure 3. These grids are consistent with
grids used in AePW-1.

Table 2. Supplied mesh metrics.

Coarse (Grid 1) Medium (Grid 2) Fine (Grid 3)

Number of nodes 1,347,650 4,041,006 11,703,071
Number of elements 3,607,995 11,572,143 36,028,725
Minimum grid angle 12.33o 12.71o 10.16o
Maximum aspect ratio 5,317 9,223 4,419
First grid node @ wall, 0.000188” 0.000126” 0.000084”
inch

3
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

(a) Coarse (Grid 1) Mesh. (b) Medium (Grid 2) Mesh. (c) Fine (Grid 3) Mesh.
Figure 3. Supplied grids.

C. Generated Grids
A new validated mesh approach14 was used, in which a non-conformal cylindrical interface accommodates
rotational wing pitch motion, and layering plus a non-conformal interface accommodates translational wing plunge
motion. The meshes were generated using the ANSYS Fluent mesher (see Figure 4). The meshing tool provides
flexibility to use existing surface meshes from supplied grids and create required geometry shapes such as the
cylinder and rectangle for generating non-conformal interfaces.

4
American Institute of Aeronautics and Astronautics
 (b) Non-conformal interface imprints on (c) Close view of cylindrical non-
(a) Cylindrical and rectangle shapes for conformal interface.
symmetry plane.
creating non-conformal interfaces.
Figure 4. Generated grids.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

III. Test Cases for the 2ndAeroelastic Prediction Workshop

Five test cases have been proposed for the AePW-2; two are compulsory, and three are optional (see Table 3).
ANSYS analyses were carried out on the compulsory cases 1 and 2. Case-1 is a forced oscillation with frequency, f,
of 10 Hz and amplitude, |Ɵ|, of 1 degree. Experimental results were obtained in the NASA Langley Transonic
Dynamics Tunnel (TDT) Oscillating Turntable (OTT) facility using refrigerant R-134a as a test medium. Case-2
onset flutter predictions were obtained in the TDT on a flexible mount Pitch and Plunge Apparatus (PAPA), which
provides two degree-of-freedom dynamic motion. R-12 refrigerant was used as a test medium.

Table 3. AePW-2 workshop cases.

Case 1 Case 2 Optional Case 3A Optional Case 3B Optional Case

3C
Mach 0.7 0.74 0.85 0.85 0.85
AOA 30 00 50 50 50
Dynamic Forced Flutter Unforced Forced Flutter
Data Type oscillation Unsteady oscillation
f = 10Hz, |Ɵ| =1o f = 10Hz, |Ɵ| =1o
Notes: -Attached flow -Attached flow -Separated flow -Separated flow -Separated flow
-OTT exp. data -PAPA exp. data -OTT exp. data -OTT exp. data -No exp. data
-R-134a -R-12 -R-134a -R-134a -R-134a

IV. Results and Discussions

A. Case-1
The analysis matrix (see Table 4) provides insight into the detailed simulations carried out for Case-1. Unsteady,
forced oscillation solutions were initialized using converged, steady-state RANS solution. The SA, SST k-ω and
SST turbulence closure models were used for the analysis. Committee provided coarse, medium and fine grids were
used to perform the grid refinement study. Temporal convergence was studied for the unsteady, forced oscillations
by varying the time-step size from 32 to 128 time-steps per period (see Table 5).

Table 4. ANSYS CFD Case 1 Analysis matrix.

ANSYS CFD Codes CFX Fluent

Turbulence Models SST SA SST k-ω SA
Unstructured Coarse (Grid 1) RANS RANS RANS RANS
(VGRID-Mixed- URANS URANS URANS URANS
Cell Center) Medium (Grid 2) RANS RANS RANS RANS
URANS URANS URANS URANS
Fine (Grid 3) RANS RANS RANS RANS
URANS URANS URANS URANS
5
American Institute of Aeronautics and Astronautics
 Table 5. Forced oscillation, unsteady time-step per period.

Grid Coarse Medium Fine

Time (sec) dt/dn/N dt/dn/N dt/dn/N
1 0.003125/32/5 0.003125/32/5 0.003125/32/5
2 0.0015625/64/5 0.0015625/64/5 0.0015625/64/5
3 0.00078125/128/5 0.00078125/128/5 0.00078125/128/5

1. ANSYS Fluent
RANS force coefficient predictions with respect to grid refinement are shown in Figure 5. Lift and moment
coefficients obtained from the SST k-ω turbulence model are higher than SA, whereas, drag coefficient is lower, and
are consistent over all the grids. Monotonic variation in force coefficient is observed over grid refinement.
However, grid independent solution was not obtained on the provided grids.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 6 shows force coefficient monitor convergence. Fluent converges within 1000 iterations, and both the SA
and SST k-ω turbulence models show similar behavior. Both the models predict same lift coefficient. However, the
drag predicted by the SA model is higher than SST k-ω, and the moment coefficient predicted by the SST k-ω model
is higher than SA. All the force coefficient monitors convergence completely after 800 iterations.

(a) Lift Coefficient. (b) Drag Coefficient. (c) Moment Coefficient.

Figure 5. ANSYS Fluent grid convergence.

Figure 6. ANSYS Fluent, RANS, Force coefficients convergence with SA and SST k-ω turbulence models on supplied
grids.

6
American Institute of Aeronautics and Astronautics
 Figure 7 shows the comparison between computed and experiment pressure coefficient at 60% wing span. Both
SA and SST k-ω models show excellent match with the experimental pressure coefficients. However, the suction
peak on the upper surface near wing leading edge is better predicted by the SST k-ω than the SA turbulence model.
Also, finer grids predict better suction peak for both the models. In order to better capture the exact experimental
suction peak further grid refinement is recommended.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 7. ANSYS Fluent, Case-1, RANS, Pressure coefficient comparison between computations and experiment at 60%
wing span.

URANS forced oscillation predictions of force coefficient show good temporal convergence with respect to
varying time-step-size of 32, 64 and128time steps per period on given grid size (see Figure 8). There is slight
sensitivity seen during the first cycle, but the variations die out subsequently and the results show complete
convergence within five oscillation cycles.

7
American Institute of Aeronautics and Astronautics
 Figure 8. ANSYS Fluent, URANS, Force coefficients, Temporal and spatial convergence.

Spatial convergence with fixed finer time-step-size of 128tt and varying grid size show sensitivity to force
coefficients. Lift coefficient variations even though very small are seen with respect to grid refinement. Drag and
moment coefficient shows greater sensitivity with grid refinement, and the variations are monotonic.
The experimental FRFs of pressure magnitude and phase are compared with the computationally obtained FRFs
at 60% wing-span location. Computational FRFs did not show temporal sensitivity at a given grid size, and hence
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

the results are compared at 64 time-steps per period. However, a spatial sensitivity is clearly seen from Figure 9. The
SA turbulence model prediction of pressure magnitude FRFs on upper surface at 60% wingspan shows slightly
lower value at suction peak compared to experiments. However, there is an increment in magnitude with grid
refinement. Lower surface pressure magnitude FRFs at wingspan location 60% shows excellent match with
experiments, with no sensitivity to grid refinement. Pressure phase FRFs on the upper surface exactly matches with
experiments at the critical suction peak region, whereas, the prediction is slightly off near trailing edge. Lower
surface phase shows good over all comparison with experiments, except near trailing edge. The phase sensitivity to
grid refinement on both the surfaces at 60% wingspan are smaller compared to magnitude.

Figure 9. ANSYS Fluent, Case-1, URANS, SA turbulence model, Pressure magnitude and phase FRFs comparison with
experiment at 60% wing span.

8
American Institute of Aeronautics and Astronautics
 SST k-ω turbulence model predictions of pressure magnitude FRFs on the upper surface at 60% wing span
shows higher value at suction peak compare to experiments (see Figure 10). The magnitude FRFs show sensitivity
to grid refinement at the critical suction peak region, with medium and fine grids predict closer results. Lower
surface pressure magnitude FRF at 60% wing span shows excellent match with experiments, with no grid
sensitivity. Pressure phase FRFs on upper surface closely matches with experiments at critical suction peak region,
whereas, slightly off near trailing edge. Lower surface phase shows good over all comparison with experiments,
except near trailing edge. The phase sensitivity to grid refinement on both the surfaces at 60% wingspan are smaller
compared to the magnitude.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 10. ANSYS Fluent, Case-1, URANS, SST k-ω turbulence model, Pressure magnitude and phase FRFs comparison
with experiment at 60% wing span.

Figure 11 shows experimentally measured FRFs of pressure magnitude and phase compared to computational
results from ANSYS Fluent SA and SST k-ω turbulence models on the fine grid and 64 time-steps per period. At the
critical suction peak region on upper surface at 60% wing span, pressure magnitude FRFs show some variation
between the ANSYS Fluent turbulence models. The SST k-ω model predicts a higher magnitude, whereas, the SA
model predicts a lower magnitude than experiments. Away from critical region there is no sensitivity observed with
turbulence models. Pressure magnitude FRFs at 60% wing span on the lower surface show very small sensitivity
near the leading edge. The SST k-ω model predicted slightly higher magnitude than SA. However, over all
magnitude comparison on lower surface shows minimal sensitivity to turbulence models and matches very well with
experiments. Pressure phase FRFs at 60% wing span on the upper surface show sensitivity to turbulence model near
trailing edge, with SA model showing heigher value than SST k-ω, which is closer to experiments. Lower surface
phase does not show turbulence model sensitivity and both the models closely match with experiments with small
discrepancies near the trailing edge.

9
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 11. ANSYS Fluent, Case-1, URANS, Fine grid, 64tt, SST k-ω and SA turbulence models, Pressure magnitude and
phase FRFs comparison with experiment at 60% wing span.

2. ANSYS CFX
RANS force coefficient predictions with respect to grid refinement are shown in Figure 12. Lift and drag
coefficients obtained from the SA turbulence model are slightly higher than SST. The difference, in lift and drag
coefficient reduces with grid refinement, and for the finest grid the drag coefficient predicted by both the turbulence
models is the same. Moment coefficients predicted by SST turbulence model is higher than SA turbulence model for
all the grids. A grid independent solution is not achieved with the supplied grids. Further grid refinement is
recommended in order to achieve grid independent solutions. In figure 13 force coefficient monitors indicates that
solutions converge very well within 100 iterations for both the turbulence models on all the grids.

(a) Lift Coefficient. (b) Drag Coefficient. (c) Moment Coefficient.

Figure 12. ANSYS CFX grid convergence.

10
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 13. ANSYS CFX, RANS, Force coefficients convergence with SA and SST turbulence models on supplied grids.

Figure 14 shows the comparison between computed and experiment pressure coefficient at 60% wing span. Both
SA and SST turbulence models show excellent match with the experiment pressure coefficient. Suction peak on
upper surface near wing leading edge is slightly under predicted. Fine grid results for both the turbulence models
predict suction peak values closer to experiments. Predicted pressure coefficients using ANSY CFX SA and SST
turbulence models show no sensitivity to grid refinement.

Figure 14. ANSYS CFX, Case-1, RANS, Pressure coefficient comparison between computation and experiments at 60%
wing span.

11
American Institute of Aeronautics and Astronautics
 URANS forced oscillation predictions of force coefficient show good temporal convergence with respect to
varying time-step-size of 32, 64 and 128time steps per period on a given grid size (see Figure 15). There is a slight
sensitivity seen with coarse time-step of 32tt, but the force coefficients show nice convergence on medium and fine
time-step-size. Spatial convergence with the smallest time step size of 128tt per period and varying grid size does
shows sensitivity to force coefficients. Lift and moment coefficients monotonically increase and drag coefficient
monotonically decreases with grid refinement.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 15. ANSYS CFX, URANS, Force coefficients, temporal and spatial convergence.

The experimental pressure magnitude and phase FRFs are compared with the computationally obtained FRFs at
60% wing span location. Computational magnitude and phase FRFs do not show temporal sensitivity at any given
grid size, and hence the results are compared at 64 time-steps per period. The spatial sensitivity is clearly seen from
Figure 16. The SA turbulence model pressure magnitude FRFs on the upper surface at 60% wing span show lower
values at the suction peak for coarse grid and higher values for medium and fine grids compared to experiments.
The magnitude increases continuously with grid refinement. Lower surface pressure magnitude FRFs at 60 % wing
span location shows excellent match with experiments, with no sensitivity to grid refinement. Pressure phase FRFs
on the upper surface exactly matches with experiments at the critical suction peak region, and are slightly off near
the trailing edge. The upper surface phase shows sensitivity to grid refinement, where coarse grid prediction falls
slightly away from medium and fine grids. Lower surface phase shows good overall match with the experiments,
except near trailing edge. The lower surface phase does not show sensitivity to grid refinement, predicting the same
phase values for all the grids.

12
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 16. ANSYS CFX, Case-1, URANS, SA turbulence model, Pressure magnitude and phase FRFs comparison with
experiment at 60% wing span.

SST turbulence model pressure magnitude FRFs on the upper surface at 60% wing span show similar trends to
the SA turbulence model. The coarse grid underpredict the pressure magnitude, and medium and fine grids
overpredict the magnitude compared to experiments (see Figure 17). Lower surface pressure magnitude FRFs at
wing span location 60% show excellent match with experiments, with no sensitivity to grid refinement. Pressure
phase FRFs on the upper surface exactly match with experiments at the critical suction peak region, whereas,
slightly off near the trailing edge. Lower surface phase shows good over all match with experiments, except near
trailing edge. The lower phase does not show sensitivity to grid refinement, predicting same phase values for all the
grids.

13
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 17. ANSYS CFX, Case-1, URANS, SST turbulence model, Pressure Magnitude and Phase FRFs comparison with
experiment at 60% wing span.

Figure 18 shows pressure magnitude and phase FRFs comparing ANSYS CFX SA and SST turbulence model
with experiments on the fine grid and 64 time-steps per period. At the critical suction peak region on the upper
surface at 60% wing span, pressure magnitude FRFs show no sensitivity to turbulence model, whereas ANSYS
Fluent showed some sensitivity. Both the ANSYS CFX SA and SST turbulence models predict exactly the same
pressure magnitude FRFs over the entire wing upper surface. Again pressure magnitude FRFs on the lower surface
show no sensitivity to turbulence model and matches closely with experiments. A slight sensitivity was observed
for the ANSYS Fluent turbulence models. Similarly, pressure phase FRFs on the upper and lower surfaces show no
sensitivity to turbulence model and match very well with experiments.

Figure 18. ANSYS CFX, Case-1, Fine grid and 64tt, URANS, SST and SA turbulence models, Pressure magnitude and
phase FRFs comparison with experiment at 60% wing span.

14
American Institute of Aeronautics and Astronautics
 B. Case-2
The analysis matrix in Table 6 provides insight into detailed simulations carried out for Case-2 steady RANS and
flutter URANS calculations. RANS calculations were done with ANSYS Fluent and CFX solvers with SST and SST
k-ω turbulence models, respectively. Flutter URANS calculations were performed with ANSYS CFX SST
turbulence model. A modified mesh was used for this study in order to have better control of the two-degrees-of-
freedom plunge and pitch dynamic motion.

Table 6. ANSYS CFD Case 2 analysis matrix.

ANSYS CFD Codes CFX Fluent

Turbulence Models SST SST k-ω
Unstructured Coarse (Grid 1) RANS RANS
(Modified Grid, URANS --
ANSYS T-Grid) Medium (Grid 2) RANS RANS
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

URANS --
Fine (Grid 3) RANS RANS
URANS --

Figure 19 shows the pressure transducer locations for measuring the pressure at 60% and 95% wing span. Both
ANSYS RANS solvers predict exactly same pressure coefficient as reported by experiments at both the measuring
locations. Also, pressure coefficients did not show any sensitivity to grid refinement and are consistent with both
solvers for 60% and 95% wing span locations, respectively (see Figures 20 and 21).

Figure 19. Pressure measurement locations at 60% and 95% wing span.

15
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 20. ANSYS Fluent and CFX, Case-2, RANS, Pressure coefficient comparisons with experiments at 60% wing span.

16
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 21. ANSYS Fluent and CFX, Case-2, RANS, Pressure coefficient comparisons with experiments at 95% wing span.

1. Experimental on-set flutter dynamic pressure

The on-set flutter dynamic pressure measured in the wind tunnel is 168.8 psf, and the PAPA system responded
with a flutter frequency of 4.3 Hz. ANSYS CFX using a coupled 6DOF solver is used to predict the flutter dynamics
at this experimental flutter dynamic pressure. Pitch and plunge experimental and computational model
representations are shown in Figure 22. Computational force coefficient monitors on the finer grid with finer time
step size shows good convergence (see Figure 23). The pitch and plunge motion traces representing wing flutter at
coarse 0.001 sec time step and fine 0.00025 sec time step are shown in Figure 24. Further decrease in time-step-size
shows only a very small difference on the flutter amplitude.

(a) Experimental model and PAPA system. (b) Computational model.

Figure 22. Representation of pitch and plunge wing motion.

(a) Lift Coefficient. (b) Drag Coefficient. (c) Moment Coefficient.

Figure 23. ANSYS CFX, URANS, Force coefficient monitors.

17
American Institute of Aeronautics and Astronautics
 (a) Pitch motion. (b) Plunge motion.
Figure 24. ANSYS CFX, wing flutter trace.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figures 25 and 26 show the comparison between experimental and computational pressure magnitude and phase
FRFs at the experimental on-set flutter dynamic pressure of 168.8 psf. Experimentally obtained results from the
PAPA tests are available at 60% and 95% wing span locations. Computational results on the finer grid with finer
time-step of 0.00024395 sec (960 time step per period) show excellent match with experimental values. There are
small differences in pressure magnitude and phase FRFs observed near the leading and trailing edges of the wing
surface at both wing span locations. These differences are smaller at 95% wing span compared to 60% wing span.
The magnitude and phase FRFs trends are captured very well over the entire wing at both the locations. The phase
wrap step near the leading edge is not captured properly at both the wing span locations.

Figure 25. ANSYS CFX, Case-2, Fine grid and 960tt, Experimental on-set flutter dynamic pressure of 168.8 psf, Pressure
Magnitude and Phase FRFs comparison with experiment at 60% wing span.

18
American Institute of Aeronautics and Astronautics
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 26. ANSYS CFX, Case-2, Fine grid and 960tt, Experimental on-set flutter dynamic pressure of 168.8 psf, Pressure
Magnitude and Phase FRFs comparison with experiment at 95% wing span.

2. Computational predictions of on-set flutter dynamic pressure

ANSYS CFX predictions of the on-set wing flutter dynamic pressure is summarized in Table 7. Computational
on-set wing flutter is predicted at 10% less than the experimental flutter dynamic pressure. Figures 27 (a) and (b)
show pitch and plunge wing oscillation on the finer grid with finer time step and varying flutter dynamic pressure.
Pitch and plunge oscillation amplitude increases over time indicating wing flutter at 168.8 psf and 151.92 psf. The
oscillation amplitude remains constant at 143.48 psf indicating limit cycle oscillations and decreases at 135.04 psf
stabilizing the wing motion.

Table 7. ANSYS CFX, wing flutter prediction.

Dynamic pressure (psf) Time step Pitch, mean Plunge, mean Pitch Frequency Remark
(sec) (Ɵ, deg) (m) (Hz)
Exp. flutter 168.8 0.00025 -0.01761507 -0.01696012 4.273504 flutter
10% < Exp. flutter 151.92 0.00025 -0.01552567 -0.01664139 4.3290043 On-set flutter
15% < Exp. flutter 143.48 0.00025 -0.01418724 -0.01677924 4.329004 No-flutter
20% < Exp. flutter 135.04 0.00025 -0.01305917 -0.01686543 4.3290043 No-flutter

The pitch oscillation frequency predicted by ANSYS CFX at the experimental on-set flutter dynamic pressure
was 4.27 Hz and closely matches with the experimental value, 4.30 Hz. Figure 28 shows the calculation of pitch
frequency using Fast Fourier Transform (FFT) post-processing technique available within ANSYS Fluent. Pitch
frequencies at the on-set flutter and no-flutter dynamic pressure are identical and slightly higher than the
experiments. The mean pitch and plunge oscillations are shown in Table 7. With reduction in flutter dynamic
pressure the pitch mean angle Ɵ consistently reduces, whereas, plunge vertical mean, m, almost remains constant.

19
American Institute of Aeronautics and Astronautics
 (a) Pitch motion prediction. (b) Plunge motion prediction.
Figure 27. ANSYS CFX, Prediction of on-set flutter dynamic pressure.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 28. Wing flutter pitch frequency calculation and comparison with experiments.

Figure 29. ANSYS CFX, Case-2, Fine grid and 960tt, Computational on-set flutter dynamic pressure predicted at 151.92
psf, Pressure magnitude and phase FRFs at 60% wing span.

20
American Institute of Aeronautics and Astronautics
 Pressure magnitude and phase FRF plots at computational on-set flutter dynamic pressure on the finer grid with
finer time-step 0.00024057 sec (960tt) show similar trends with experimental measurements as well as
computational results derived from the known experimental flutter dynamic pressure. Figures 29 and 30 show the
magnitude and phase plots on the upper and lower surface at 60% and 95% wing span locations, respectively.
Magnitude and phase values at both the wing span locations on the upper and lower surfaces closely match with
experiments as well as computations for the known experimental on-set flutter dynamic pressure.
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

Figure 30 ANSYS CFX, Case-2, Fine grid and 960tt, Computational on-set flutter dynamic pressure of 151.92 psf,
Pressure magnitude and phase FRFs at 95% wing span..

Acknowledgments
The authors would like to thank the AePW-2 Computational, Experimental and Organizing teams for providing rich
data for our CFD solver comparisons. Also, the authors would like to recognize CRAY for generously providing
computational resources for performing these simulations. Their support made this investigation possible.

References
1
Heeg, J., et al., “Plans and Example Results for the 2 nd AIAA Aeroelastic Prediction Workshop,” AIAA SciTech Forum and
Exposition, 5-9 Jan 2015
2
Schuster, D. M., Chwalowski, P., Heeg, J., and Wieseman, C., “Summary of Data and Findings from the First Aeroelastic
Prediction Workshop,” ICCFD7, July 2012.
3
Heeg, J., Chwalowski, P., Schuster, D. M., Dalenbring, M., Jirasek, A., Taylor, P., Mavriplis, D. J., Boucke, A., Ballmann,
J., and Smith, M.,“Overview and Lessons Learned from the Aeroelastic Prediction Workshop,” IFASD Paper 2013-1A, June
2009.
4
Heeg, J., Chwalowski, P., Schuster, D. M., and Dalenbring, M., “Overview and lessons learned from the Aeroelastic
Prediction Workshop,” AIAA Paper 2013-1798, April 2013.
5
David J. Piatak, and Craig S. Cleckner, “Oscillating Turntable for the Measurement of Unsteady Aerodynamic Phenomena,”
Journal OF AIRCRAFT, Vol. 40, No. 1, January-February 2003
6
David J. Piatak, and Craig S. Cleckner, “A New Forced Oscillation Capability for the Transonic Dynamic Tunnel” AIAA
Paper AIAA 2002-0171
7
Ricketts, R., et al. “Geometric and Structural Properties of a Rectangular Supercritical Wing Oscillated in Pitch for
Measurement of Unsteady Pressure Distributions,” NASA TM-85763, November 1983.

21
American Institute of Aeronautics and Astronautics
 8
Ricketts, R., et al., “Transonic Pressure Distributions on a Rectangular Supercritical Wing Oscillating in Pitch,” NASA TM-
84616, March 1983.
9
Ricketts, R., et al., “Subsonic and Transonic Unsteady and Steady-Pressure Measurements on a Rectangular Supercritical
Wing Oscillated in Pitch,” NASA TM-85765, August 1984.
10
Bennett, R. M. and Walker, C. E., “Computational Test Cases for a Rectangular Supercritical Wing Undergoing Pitching
Oscillations,” NASA TM-1999-209130, April 1999.
11
Dansberry, B.E., et al., “Physical Properties of the Benchmark Models Program Supercritical Wing,” NASA TM-4457,
September 1993.
12
Dansberry, B. E., “Dynamic Characteristics of a Benchmark Models Program Supercritical Wing,” AIAA Paper AIAA 92-
2368, April 1992.
13
Dansberry, B. E., et al., “EXPERIMENTAL UNSTEADY PRESSURES AT FLUTTER ON THE SUPERCRITICAL
WING BENCHMARK MODEL” AIAA Paper AIAA-93-1592-CP, Jan 1993
14
Balasubramanyam S., et al., “VIRTUAL MODEL BASIN FOR RESISTANCE, MANEUVERING, AND SEAKEEPING -
A RANS CFD BASED APPROACH,” Proceedings of the 26th International Conference on Offshore Mechanics and Arctic
Engineering, June 10-15, 2007, San Diego, California, USA
15
Robert M. Bennett, “7E. TEST CASES FOR FLUTTER OF THE BENCHMARK MODELS RECTANGULAR WINGS
Downloaded by Balasubramanyam Sasanapuri on January 16, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-0192

ON THE PITCH AND PLUNGE APPARATUS” UNCLASSIFIED Defense Technical Information Center Compilation Part
Notice ADP010713
16
Aeroelastic Prediction Workshop, NASA “http://nescacademy.nasa.gov/workshops/AePW2/public/”
17
Drag Prediction Workshop, NASA, ”http://aaac.larc.nasa.gov/” cited July 2013.
18
High Lift Prediction Workshop, NASA, ”http://hiliftpw.larc.nasa.gov/” cited July 2013.
19
Mavriplis, D. J. et al., “Grid Quality and Resolution Issues from the Drag Prediction Workshop Series,” AIAA Paper 2008-
930, Jan. 2008.

22
American Institute of Aeronautics and Astronautics

View publication stats