Professional Documents
Culture Documents
DOI: 10.2514/1.J053291
An approach has been developed to use the strain sensed in a membrane wing to estimate aerodynamic loads.
Through experimental and theoretical correlation of membrane strain state due to a normally distributed load, the
pressure over the surface of the wing was estimated. Elastic deformations and strains of the membrane wing were
measured in a low-speed wind tunnel using digital image correlation. Simultaneously, aerodynamic loads were
measured using a six-component load cell, inclinometer, and pitot tube. Digital image correlation displacement
measurements were used to form a reduced-order model of the membrane displacements via a snapshot proper
orthogonal decomposition method. Digital image correlation strain measurements were used to form virtual strain
sensors, which served as the sensory input to the estimation. The Poisson equation for a two-dimensional linear-elastic
membrane with out-of-plane deformation was used to calculate the normal pressure distribution from virtual strain
sensors using proper orthogonal decomposition basis functions and a recursive least-squares minimization.
Estimated pressure distributions were compared with a high-fidelity three-dimensional computational fluid dynamic
model and pressures were calculated from digital image correlation deformations. Coefficients of lift and pitching
moment for steady-state flow conditions were estimated and compared with measured wind-tunnel loads. Results
show promise toward the application of a low-fidelity estimation approach for real-time load-estimation applications.
Nomenclature
CL
Cm
E
f
h
P
T
Uinf
u
v
w
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
coefficient of lift
coefficient of pitching moment
modulus of elasticity, Pa
interpolant function
membrane thickness, m
pressure, N
membrane tension, Nm
freestream flow velocity, ms
membrane displacements in x direction, m
membrane displacements in y direction, m
membrane displacements in z direction, m
angle of attack, deg
viscous dampening parameter, N sm
strain measurement, %
strain estimate, %
viscoelastic dampening parameter, N sm2
Poisson ratio
membrane density, kgm3
standard deviation
membrane domain
I.
Introduction
2070
wing with a reduced-order model. The motivation for this work was
twofold, the first of which was developing an approach that used
strain sensors, rather than displacements, to estimate aerodynamic
loads. Second, we sought to explore the accuracy of a low-fidelity
estimation approach for future development toward real-time estimation applications. In this paper, we present the analytic formulations
of the approach, our experimental procedures are explained, estimated values are compared with experimental measurements and
numeric models, and results found are discussed.
II.
work, shear strains were neglected for the formulation of the sensor
measurement. The process for acquiring these measurements can be
found in Sec. III.
More specifically, if xxDIC x; y and yyDIC x; y represent the
experimentally measured full-field DIC strain over the domain in
the x and y directions, respectively, and i represents the rectangular
domain of the individual ith sensor, then the magnitude of a single
virtual strain measurement for a given location xi ; yi is defined as
xxi
1
i
yyi
1
i
Methods
Z
i
Z
i
xxDIC x; y dx dy
(1)
yyDIC x; y dx dy
(2)
Using Eqs. (1) and (2), a scalar value for strain is given for each
xi ; yi location and a strain sensor vector can be constructed for n
number of strain sensors:
sensor f xx1 : : : xxn
yy1 : : : yyn g
(3)
fx
m
X
cx i gx i x; y
(4)
cy i gy i x; y
(5)
i1
fy
m
X
i1
2071
interp
9
8 R
1
>
>
>
i i f x 1 dx dy >
>
>
>
>
>
>
..
>
>
>
>
>
> R
.
>
>
>
>
=
<1
f
dx
dy
i Ri x n
1
>
>
>
f y 1 dx dy >
>
>
>
> i i .
>
>
>
>
.
>
>
>
>
.
>
> R
>
>
;
:1
f
dx
dy
n
y
i i
(6)
Using Eqs. (3) and (6), a relationship between the strain sensors
and strain estimates is formed:
sensor interp
(7)
t
t
tx2 ty2
2 w
2 w T w T y w
Tx 2 Ty 2 x
Px; y; t
x x
y y
x
y
2 w
(8)
s
2 2 2
v
u
v
w
1
yy 1 2
y
y
y
y
2 2 2
v 1 u
v
w
y 2 y
y
y
(10)
(11)
(12)
With this definition for arc length, the in-plane deformation ux; y
can then be approximated as
Fig. 2 Internal reaction forces for an infinitesimal element of a thin
membrane acted on by a normal pressure distribution.
The relationships developed in Eqs. (15) and (17) are derived through
empirical inference.
2072
Fig. 3
First four POD modes of wx;y used to form set of basis to estimate out-of-plane deformation: ad) POD modes 14, respectively.
ux; y x x
for
x
Sx x; y
S x; y
x a x
Sx a; y
Sx a; y
a
ux; y x a
Sx x; y
Sx a; y
(13)
fw
(14)
(15)
vx; y y b
Sy x; y
Sy x; b
(16)
(17)
Given Eqs. (9), (15), and (17), the pressure distribution in Eq. (8)
can be found if wx; y and its derivatives are known. Moreover, it can
be observed that every variable within the formulation of Eq. (8) is
ultimately a function of wx; y. Therefore, once wx; y is found,
any number of differentiation and integration schemes can be used to
solve Eq. (8) for the pressure distribution Px; y. For this work, a
second-order-accurate explicit finite difference scheme was used for
both integration and differentiation.
E. Surface Deformation Estimation
p
X
(18)
i1
where cwi is the ith weighting coefficient for the ith basis function
gwi x; y. With Eq. (18), the estimation of the strain field from the
membrane deformation is now possible. By substituting Eqs. (15),
(17), and (18) into Eqs. (10) and (11), and numerically integrating
and differentiating as necessary, full-field strain interpolants for xx
and yy were developed simply as a function of cwi , which can
be represented as fwx cwi and fwy cwi . A relationship to find the
interpolants weighting coefficients can now be formed in the same
manner as that in Eq. (7):
fwx cwi
f x
(19)
f y
fwy cwi
where, in this case, represents the error between the estimated
strain distribution (derived from the strain sensors) and the derived
strain values from the estimated surface deformation. Applying the
same least-squares optimization approach as before, the weighting
coefficients cwi from Eq. (18) are found, and an estimate of the outof-plane deformation west x; y is generated. Now, west x; y and
its subsequent derivatives can be substituted into Eqs. (8) and (9)
to calculate the pressure distribution Px; y over the membrane
domain .
III.
Experimental Setup
In this approach, the source from which the loads on the membrane
originate is irrelevant from a structural analysis standpoint. What is
of consequence to solving the structural problem is simply the total
pressure difference between the upper and lower surface of the membrane. To evaluate the validity and accuracy of the proposed method,
the pressure estimation must be compared with a known pressure
state before comparison with a highly dynamic and unknown pressure state as that induced by a fluid flow over a wing. For this reason,
two experiments were conducted to provide different loading scenarios for the estimation routine to evaluate. The first was a hydrostatic pressure case, where a pretensioned membrane was subjected
to a steady, evenly distributed, known pressure. DIC measurements
of strain and deformation were taken of the deformed membrane.
A pressure transducer measured the magnitude of gauge pressure
applied to the membrane surface. In the second scenario, the membrane wing, defined in Fig. 1, was placed in a low-speed wind tunnel
and was subjected to various steady conditions of wind speed and
angle of attack. Aerodynamic loads generated by the wing were
measured, and DIC measurements of the membrane deformation
were taken. A time-invariant CFD model was created of the wing in
each condition to have a numerical approximation of the pressure
distribution as a basis for comparison.
2073
Fig. 4 Speckled latex membrane: a) random digital pattern printed on surface, used for DIC measurements; b) wing showing out-of-plane deformation
during wind-tunnel testing (airflow from left to right).
2074
Fig. 5 View of a) bench-top hydrostatic pressure experiment; b) low speed wind-tunnel experiment.
of the wing; and 1 c to the outboard side of the wing. The volume
was discretized into approximately 1.2 million hexahedral volume
cells with an average cell width of 0.1 mm in the near-wake region,
and growing cells up to 10 mm near the simulation boundaries.
Zero slip boundary conditions were used on the walls and wing
surfaces, a constant velocity boundary condition was used on the inlet
surface, and a constant pressure boundary condition was used on the
outlet surface. The simulation volume is illustrated in Fig. 6a.
To create a three-dimensional model of the deformed wing, DIC
measurements were time averaged for each simulation case of
velocity and AOA, and imported into a CAD model. This is the same
process used in [11] and assumes the steady-state, time-averaged
simulation results will compare adequately with time-averaged
dynamic results (as investigated in [19]). The results of this process
can be seen in Fig. 6b for a membrane wing with 2% prestrain and
6 deg AOA and freestream velocity of 13 ms equating to
Re 60; 000. The Reynolds-averaged NavierStokes (RANS)
equations were solved for an incompressible flow in conjunction with
the SpalartAllmaras turbulence model. As shown in [20,21], the
SpalartAllmaras turbulence model proves to perform well in nearwake regions, such as flow around finite wings, while remaining
relatively computationally inexpensive.
IV.
Results
A. Material Properties
Latex rubber sheets were used to construct the membrane wings for
all tests. Because of the relatively low levels of deformation, resulting
in ii 30%, a linear-elastic constitutive model was used to identify
the elastic modulus and Poisson ratio using similar techniques as in
[14,22]. To model the dynamics of the system, a constant material
density was assumed, as well as an assumption that the system
behaved according to linear viscous dampening and KelvinVoigt
viscoelastic damping effects. Further discussion and derivation of
dampening parameters can be found in [8].
For all results, the following material parameters were used to
model the rubber membrane: nominal prestrain in x and y directions
of 2%, linear-elastic modulus of 1.6 MPa, Poisson ratio of 0.4, constant thickness of 0.14 mm, membrane density of 960 kgm3 ,
viscous damping coefficient of 5 N sm, and KelvinVoight viscoelastic damping coefficient of 0.0001 N sm2 .
B. Strain Fields
For both test cases, DIC was used to measure full-field strain and
displacement of the deformed membrane. A 5 5 grid of virtual
strain sensors, as seen in Fig. 7b, was distributed over the membrane
to provide sensed strain as input into the estimation routine. Optimization of sensor placement and selection may provide improvements
upon the results and is a topic for further study in future work. The
estimation routine then produced estimated strain fields xx x; y and
yy x; y, out-of-plane displacement wx; y, and pressure distribution Px; y using the methods previously described. An example
of the measured strain fields, distributed virtual sensors, and the
estimated strain field can be as seen in Fig. 7 for a particular windtunnel test.
The estimated strain fields compared well in terms of overall strain
energy and overall strain magnitude. Errors in total strain energy were
Fig. 6 Representation of a) half-span CFD simulation domain; b) meshed membrane wing analyzed in CFD. Membrane deformation was derived from
time-averaged DIC measurement from wind-tunnel testing.
2075
Fig. 7 Strain a) DIC xx measurements from wind-tunnel test; b) virtual xx sensors; and c) estimated xx field.
found to be less than 4% for all cases, and local errors in strain
magnitude were less than 12% for all cases. Because of the nature of
the selected basis, the estimated strain fields exhibit a stiffness to
high-frequency responses, which reduced the routines ability to
model spacial high-frequency phenomena. This characteristic of the
basis function coupled with the least-squares approach, inherently
rejects a certain amount of sensor noise. A similar behavior is seen in
the estimation of out-of-plane displacement.
C. Surface Deformation
2076
Pressure [Pa]
200
a)
b)
c)
Fig. 9 Plots of hydrostatic pressure distributions: a) ideal; b) DIC calculated; and c) estimated.
Ideal
Est.
DIC
Pressure [Pa]
200
0
0
Chord [y/a]
a)
b)
c)
Fig. 10 Section cuts of hydrostatic pressure distributions: a) xa 0.25, b) xa 0.5, and c) xa 0.75.
tunnel test at 6 deg AOA and Re 60; 000 (12 ms) are shown in
Fig. 12. Figure 12a shows the total pressure distribution derived
from the steady CFD simulation, Fig. 12b shows the calculated
pressure distribution from full-field DIC measurements, and Fig. 12c
shows the estimated pressure distribution from virtual strain sensors.
Figure 13 shows the same data as Fig. 12, but presented as crosswise
sectional cuts of the pressure distributions as a function of normalized
chord. Each figure represents a different spanwise position at
xa 0.25; 0.5; 0.75, also represented by black lines in Fig. 12 for
reference.
Observing the pressure distributions shown in Fig. 12, the CFD,
calculated DIC and estimated pressure distributions correspond well
to one another in both shape and magnitude. Again, variance is
2077
Pressure [Pa]
250
a)
b)
c)
Fig. 12 Membrane wing pressure distributions: a) CFD; b) DIC calculated; and c) estimated.
Ideal
Est.
DIC
Pressure [Pa]
250
0
0
a)
Chord [y/a]
b)
c)
Fig. 13 Section cuts of membrane wing pressure distributions: a) xa 0.25, b) xa 0.5, and c) xa 0.75.
V.
Conclusions
2078
a)
b)
c)
Fig. 14 Wing load coefficients: a) lift vs AOA, Re 60,000; b) lift vs AOA, various speeds; and c) pitching moment vs lift.
distributed load, input from virtual strain sensors was used to estimate
the aerodynamic load on the membrane. Two tests were conducted to
evaluate the accuracy of the estimation routine, as well as comparison
with a steady numeric fluid model.
A hydrostatic pressure test was conducted to compare a known
input load, in this case, a static pressure, to the estimated load. With a
known input to the system, the output from the estimation routine
could be directly evaluated for its accuracy. Results were favorable;
the average hydrostatic pressure estimate was within 5% error to the
Acknowledgments
The authors would like to thank and acknowledge the support of
the Aerospace Engineering Department at the Politecnico di Milano
University, Italy, the European Office of Aerospace Research &
Development, Air Force Office of Scientific Research, Air Force
Material Command, USAF, under grant number FA8655-12-1-2114,
technical monitor Gregg Abate, and Air Force Office of Scientific
Research, under DURIP grant number 10691449. The authors would
also like to thank Belinda Batten and Cody Ray for their continued
support.
References
[1] Schmit, R., and Glauser, M., Improvements in Low Dimensional Tools
for Flow-Structure Interaction Problems: Using Global POD, 42nd
AIAA Aerospace Sciences Meeting and Exhibit, AIAA, Reston, VA,
2004, pp. 11,39011,398.
[2] Schmit, R., and Glauser, M., Use of Low-Dimensional Methods for
Wake Flowfield Estimation from Dynamic Strain, AIAA Journal,
Vol. 43, No. 5, 2005, pp. 11331136.
doi:10.2514/1.1722
[3] Stanford, B., Albertani, R., and Ifju, P., Inverse Methods to Determine
the Aerodynamic Forces on a Membrane Wing, Structures, Structural
Dynamics, and Materials and Co-Located Conferences, AIAA, Reston,
VA, April 2007.
doi:10.2514/6.2007-1984
[4] Waldman, R., Song, A., Riskin, D., Swartz, S., and Breuer, K.,
Aerodynamic Behavior of Compliant Membranes as Related to Bat
Flight, Fluid Dynamics and Co-Located Conferences, AIAA, Reston,
VA, 2008.
doi:10.2514/6.2008-3716
[5] Waldman, R. M., and Breuer, K. S., Shape, Lift, and Vibrations of
Highly Compliant Membrane Wings, Fluid Dynamics and Co-Located
Conferences, AIAA, Reston, VA, June 2013.
doi:10.2514/6.2013-3177
[6] Song, A., Tian, X., Israeli, E., Galvao, R., Bishop, K., Swartz, S., and
Breuer, K., Aeromechanics of Membrane Wings with Implications for
Animal Flight, AIAA Journal, Vol. 46, No. 8, 2008, pp. 20962106.
doi:10.2514/1.36694
[7] Dickinson, B. T., and Singler, J. R., Distributed Load Estimation from
Noisy Structural Measurements, Journal of Applied Mechanics,
Vol. 80, No. 4, 2013, Paper 041011.
doi:10.1115/1.4007794
[8] Ray, C., Modeling, Control, and Estimation of Flexible, Aerodynamic
Structures, Ph.D. Thesis, Oregon State Univ., Corvallis, OR, 2012.
[9] Stanford, B., Albertani, R., Lacore, D., and Parker, G., Proper
Orthogonal Decomposition of Flexible Clap and Fling Elastic Motions
via High-Speed Deformation Measurements, Experimental Mechanics, Vol. 53, No. 7, 2013, pp. 11271141.
doi:10.1007/s11340-013-9743-4
2079
[10] Albertani, R., Ray, C., and Batten, B., Tip Vorticity and MembraneStrain Analysis on a MAV Elliptical Pliant Membrane-Wing, Fluid
Dynamics and Co-Located Conferences, AIAA, Reston, VA, June 2011.
doi:10.2514/6.2011-3119
[11] Carpenter, T., Ray, C., and Albertani, R., Correlation of Structural
Strain to Tip Vorticity and Lift for a MAV Pliant Membrane Wing,
Experimental and Applied Mechanics, edited by Ventura, C. E., Crone,
W. C., and Furlong, C., Vol. 4, Conference Proceedings of the Society
for Experimental Mechanics Series, Springer, New York, 2013, pp. 1
13.
[12] Branch, M. A., Coleman, T. F., and Li, Y., Subspace, Interior, and
Conjugate Gradient Method for Large-Scale Bound-Constrained
Minimization Problems, Journal of Scientific Computation, Vol. 21,
No. 1, 1999, pp. 123.
doi:10.1137/S1064827595289108
[13] Byrd, R., Schnabel, R., and Shultz, G., Approximate Solution of the
Trust Region Problem by Minimization over Two-Dimensional
Subspaces, Mathematical Programming, Vol. 40, Nos. 13, 1988,
pp. 247263.
doi:10.1007/BF01580735
[14] Stanford, B., and Ifju, P., Validity Range of Low Fidelity Structural
Membrane Models, Experimental Mechanics, Vol. 48, No. 6, 2008,
pp. 697711.
doi:10.1007/s11340-008-9152-2
[15] Cohen, K., Siegel, S. J., Aradag, S., and McLaughlin, T., Nonlinear
Estimation of Transient Flow Field Low Dimensional States Using
Artificial Neural Nets, Expert Systems with Applications, Vol. 39,
No. 1, 2010, pp. 12641272.
doi:10.1016/j.eswa.2011.07.135
[16] Durgesh, V., and Naughton, J., Multi-Time-Delay LSE-POD Complementary Approach Applied to Unsteady High-Reynolds-Number near
Wake Flow, Experiments in Fluids, Vol. 49, No. 3, 2010, pp. 571583.
doi:10.1007/s00348-010-0821-4
[17] Sutton, M., Mingqi, C., Peters, W., Chao, Y., and McNeill, S.,
Application of an Optimized Digital Correlation Method to Planar
Deformation Analysis, Image and Vision Computing, Vol. 4, No. 3,
1986, pp. 143150.
doi:10.1016/0262-8856(86)90057-0
[18] Sutton, M., Turner, J., Bruck, H., and Chae, T., Full-Field Representation of Discretely Sampled Surface Deformation for Displacement
and Strain Analysis, Experimental Mechanics, Vol. 31, No. 2, 1991,
pp. 168177.
doi:10.1007/BF02327571
[19] LaCroix, B., Gardiner, C., and Ifju, P., Investigating the Aerodynamic Effects of Flexible Membrane Wings, Imaging Methods for
Novel Materials and Challenging Applications, edited by Jin, H.,
Sciammarella, C., Furlong, C., and Yoshida, S., Vol. 3, Conference Proceedings of the Society for Experimental Mechanics Series, Springer,
New York, 2013, pp. 345352.
[20] Posada, J. A., Numerical Study of Wingtip Shed Vorticity Reduction by
Wing Boundary Layer Control, Ph.D. Thesis, West Virginia Univ.,
Morgantown, WV, 2007.
[21] Imamura, T., Enomoto, S., and Yamamoto, K., Noise Simulation
Around NACA0012 Wingtip Using Large Eddy Simulation, Japan
Society of Aeronautical Space Sciences Transactions, Vol. 55, No. 4,
2012, pp. 214221.
doi:10.2322/tjsass.55.214
[22] Stanford, B., Viieru, D., Albertani, R., Shyy, W., and Ifju, P., Numerical
and Experimental Investigation of Flexible Micro Air Vehicle Wing
Deformation, Aerospace Sciences Meetings, AIAA, Reston, VA,
Jan. 2006.
doi:10.2514/6.2006-440
N. Wereley
Associate Editor