Professional Documents
Culture Documents
AIAA –2004–0889
Improvements in Low Dimensional
Tools for Flow-Structure Interaction
Problems: Using Global POD
Ryan F. Schmit
Clarkson University, Potsdam, NY 13699
Mark N. Glauser
Syracuse University, Syracuse, NY 13244
1
2
1.5
3
4
Klinger Rail with 45 Deg. Mirror 1.5
7
8
6
5
Back Scatter PIV Camera 9
y,V 10
1.0 11
12
x,U
Tunnel Section
Fig. 4 Placement of strain gauges on the frame
Forward Scatter PIV Camera
(n)
λ(n) (x0 )φi (x0 , y 0 , z 0 ) (2)
Equation 2 is solved numerically for the eigenfunc-
(n)
tions of the two-point correlation. The φi are used
to reconstruct the original fluctuation quantities with
Equations 3.
m
(n)
X
Ui (x0 , y, z, t) = an (x0 , t)φi (x0 , y, z) (3)
n=1
Equation 4 determines the Global POD expansion Fig. 6 Instantaneous Snapshot of the Wake Flow
(n)∗ Field
coefficients, an (x0 , t) of the flow, where the φi de-
(n)
notes the complex conjugate of φi .
Table 1 Results Setup for Instantaneous Snapshot
Z of the Wake Flow Field
(n)∗
an (x0 , t) = Ui (x0 , y, z, t)φi (x0 , y, z)dy, dz (4) Upper Left U Component
D
Upper Right V Component
2 Using the simultaneously sampled strain gauge Lower Left W Component
measurements and the Global POD expansion coef- Lower Right V and W Components
ficients, the LSE coefficient, Anq , can be determined
for each streamwise location by:
In the works by Schmit9,10 the analysis technique
used solved the POD and mLSE equations for one
ε21 ε1 ε2 · · · ε1 εq
An1
an ε1
data set at a time and the results will be used for
ε2 ε1 ε22 · · · ε2 εq An2 an ε2 comparison in this paper. From Schmit Figure 7
.. .. .. ..
. =
..
shows the reconstruction of the velocity components
. ..
. . . . using the POD expansion coefficients from Equation 3
εq ε 1 εq ε 2 · · · ε2q Anq an εq and compares remarkably well with the instantaneous
(5) snapshot, Figure 6, with only the first 20 POD modes
3 Finally, estimate the Global POD expansion coef- used. Also, from Schmit Figure 8 shows the estima-
ficients from the temporal strain gauge measurements, tion of the velocity field using mLSE, Equations 6 and
shown in Equation 6, and combine them with the 7, and shows the low dimensional description of the
eigenvectors, to produce the instantaneous estimations flow field, which is well resolved with 20 POD modes
of flow field as expressed in Equation 7. compared to the original snapshot, Figure 6.
4 of 8
z mm
0
110 -0.013 110
100 100
-0.014 -2
90 90
80 -0.015 80 -4
40 20 0 40 20 0
y mm y mm
Eigenvectors in Z direction x 10
-3 Plot of Eigenvectors in Y and Z direction
6
150 150
Fig. 13 Rebuild of Instantaneous Snapshot with
140 4 140 20 Global POD Modes at α = 15◦ AOA, airspeed
130 130
120
2
120
= 44f ps, plane = c/2
z mm
z mm
110 0 110
100 100
-2
90 90
80 -4 80
Results of Estimation Procedure
40 20 0 50 40 30 20 10 0
y mm y mm
Figure 14 presents an estimation of the wake flow
Fig. 11 Eigenvector Mode 1 at 44f ps, 15◦ AOA, field using the procedure described by Equations 5, 6,
c/2 and 7 with 20 Global POD modes. From the results
presented in Figure 14, the estimate of the wake flow
field shows that the technique captures much of the
low dimensional features of the flow, when compared
to the instantaneous snapshot, Figure 6. But the es-
timation using the Global POD modes shows similar
results when comparing to the estimation of the wake
flow field using just the POD modes for each data set
shown in Figure 8.
the wing tip vortex, which is spread out over the wind. ment Based on Proper Orthogonal Decomposition Modeling of
With 20 out of 2400 POD modes, a reasonable recon- a Cylinder Wake,” 33rd AIAA Fluid Dynamics Conference and
Exhibit - AIAA 2003-4259 , 2003, pp. 1–11.
struction of the instantaneous snapshot of the wake 14 Siegel, S., Cohen, K., and McLaughlin, T., “Feedback Con-
flow field can be obtained. trol of a Circular Cylinder Wake in Experiment and Simulation
Utilizing the simultaneously sampled strain gauges (Invited),” 33rd AIAA Fluid Dynamics Conference and Exhibit
and the wake flow field measurements in conjunction - AIAA 2003-3569 , 2003, pp. 1–13.
15 Gerhard, J., Pastoor, M., King, R., Noack, B. R., Dill-
with mLSE, and estimation of the wake flow field with
mann, A., Morzynski, M., and Tadmor, G., “Model-Based
the lower Global POD modes, provides a reasonable Control of Vortex Shedding Using Low-Dimensional Galerkin
estimation of the wake flow field. Models,” 33rd AIAA Fluid Dynamics Conference and Exhibit -
From the results shown, using a low dimensional tool AIAA 2003-04262 , 2003, pp. 1–10.
16 Samimy, M., Debiasi, M., Caraballo, E., Ozbay, H., Efe,
for flow-structure interaction provides a valid method-
M. O., Yuan, X., DeBonis, J., and Myatt, J. H., “Development
ology for estimation the wake flow field from thy dy- of Closed-loop Control for Cavity Flows,” 33rd AIAA Fluid
namic strain gauges alone. This provides a starting Dynamics Conference and Exhibit - AIAA 2003-4258 , 2003,
point for developing control strategies for all types of pp. 1–17.
17 Carlosn, H. and Miller, R., “An Eigensystem for Predicting
flow-structure interaction problems.
and Controlling Unsteady Aeroelasticity,” 39th AIAA Aerospace
Sciences Meeting & Exhibit - AIAA 2001-0854 , 2001, pp. 1–11.
References 18 Hussain, A. K. M. F. and Hayakawa, M., “Eduction of
1 Morris, S. J., “Miniature Spy Planes: The Next Gener- large-scale organized structures in a turbulent plane wake,” J.
ation of Flying Robots,” Frontiers of Engineering Reports on Fluid Mech., Vol. 180, 1987, pp. 193–229.
Leading-edge Engineering from the March 2002 NAE Sympo-
sium on Frontiers of Engineering, 2002, pp. 10–20.
2 Lumley, J., “The Structure of Inhomogeneous Turbulent
set, D., Cole, D., Fiedler, H., Garem, J., Hilberg, D., Jeong,
J., Kevlahan, N., Ukeiley, L., and Vincendau, E., “Collabora-
tive Testing of Eddy Structure Identification Methods in Free
Turbulent Shear Flows,” Experiments in Fluids, Vol. 25, 1998,
pp. 197–225.
7 Taylor, J. and Glauser, M., Dynamics of Large Scale Struc-
8 of 8