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

net/publication/306356544

Nonlinear Aerodynamics and Nonlinear Structures Interaction for F-16 Limit

Cycle Oscillation Prediction

Conference Paper · January 2016

DOI: 10.2514/6.2016-1796

CITATIONS READS

3 576

4 authors, including:

Zhichao Zhang Chen Chih Ping

zona technology Fu Jen Catholic University
30 PUBLICATIONS   235 CITATIONS    29 PUBLICATIONS   137 CITATIONS   

SEE PROFILE SEE PROFILE

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

Sensitivity Analysis for Complex, Multidisciplinary Systems Using an Unsteady Adjoint Method View project

All content following this page was uploaded by Zhichao Zhang on 19 March 2018.

The user has requested enhancement of the downloaded file.

 Nonlinear Aerodynamics and Nonlinear Structures
Interaction for F-16 Limit Cycle Oscillation Prediction
1 2
Z. Zhang and P. C. Chen
ZONA Technology, Scottsdale, AZ 85258
and
3 4
X.Q. Wang and Marc P. Mignolet ,
Arizona State University, Tempe, AZ 85287-6106

Abstract
In this paper, a physical-rooted Nonlinear Structural Damping (NSD) model based on a
generalized van der Pol model is integrated into ZEUS; a ZONA's Euler Unsteady Solver, to
establish a Nonlinear Aerodynamics and Nonlinear Structures Interaction (NANSI) simulation
tool for predicting limit cycle oscillation (LCO) amplitude. The parameter involved in the
generalized van der Pol model is estimated from the flight test data of an F-16 with external
stores configuration at a given flight condition. Once this parameter is identified, NANSI
simulation is carried out to predict the LCO amplitudes at all flight conditions. Results show that
the predicted LCO amplitudes for the F-16 Typical-LCO and Non-typical configurations
correlate very well with the flight test data; demonstrating that the NSD is one of the key
elements involved in the LCO of the F-16 with external stores configurations.

1. Background
Several current fighter aircraft with external store configurations persistently encounter Limit
Cycle Oscillation (LCO) problems. LCO is a self-excited, sustained vibration of limited
amplitude which can impact a pilot's control authority over the aircraft, ride quality, and weapon
aiming. It can also induce structural fatigue and, under certain circumstances, flutter. Denegri
[1] provided a detailed description of the aircraft/store LCO phenomenon. Norton [2] gave an
excellent overview of LCO for a fighter aircraft carrying external stores and its sensitivity to
store carriage configuration and mass properties. Because of this sensitivity, the LCO clearance
of a modern fighter aircraft should be addressed for all possible store/weapon configurations.
Given the drastic number of such configurations, this effort is a major engineering task in
aircraft/store weapon compatibility certification. It requires accurate aeroelastic predictions
within a short-time frame as demanded by rapid military responses when facing today’s ever-
changing international situation. Further, since there can be thousands of store/weapon
combinations for a typical fighter aircraft, the LCO predictions must also be computationally
efficient to rapidly identify the critical cases. A robust post-processing procedure is also needed
to identify a wide variety of aeroelastic response characteristics including flutter, divergence and
LCO.
It is generally believed that LCO of an aircraft with stores is a post flutter phenomenon that
belongs to the so-called supercritical LCO mechanism. When the flight condition of the aircraft
is beyond its flutter boundary, the aircraft's aeroelastic system is unstable and a divergent
response of the structure occurs if the aeroelastic system is linear. However, if the aeroelastic
system is nonlinear and includes a “LCO bounding mechanism” dependent on the amplitude of
1
Research and Development Engineering Specialist, Senior Member AIAA
2
President, Associate Fellow AIAA
3
Associate Research Scientist, SEMTE, Faculties of Mechanical and Aerospace Engineering, Senior Member AIAA,
4
Professor, SEMTE, Faculties of Mechanical and Aerospace Engineering, Fellow AIAA
the structural response, then the growth of the divergent response due to flutter can be limited
resulting in LCO at a particular amplitude. The source of the LCO bounding mechanism, which
could be from the aerodynamics, structure, or both, still remains to be fully understood and is a
long-standing research issue. Many researchers believe that the nonlinearity involved in the
LCO bounding mechanism is solely induced by oscillating transonic shocks and/or shock
induced flow separation. This type of approach for predicting LCO is defined herein as the sole
nonlinear aerodynamic approach. If this is the correct bounding mechanism, the LCO can be
predicted using high fidelity Computational Fluid Dynamics (CFD) tools coupled with a linear
structural model. Using a CFD tool called the AERO-F/S Suite developed by Farhat [3]; Pasiliao
[4] performed a LCO study on an F-16 with stores configuration that experienced LCO during
flight tests and first found good correlation of the onset LCO Mach number (the flutter
boundary) between the predicted and flight test measured results. This good matching results
from the CFD code accurately capturing the transonic shock effects that normally lower the
predicted flutter boundary in transonic flow regions as compared to that predicted by the linear
unsteady aerodynamic methods such as the Doublet Lattice Method (DLM) and ZAERO [5].

However, even with this good transonic flutter predictive capability, Pasiliao’s investigation
failed to predict LCO. On this basis, it appears that the nonlinear aerodynamics provided by the
CFD methodology alone is not sufficient as a predictive LCO bounding mechanism. Another
sole nonlinear aerodynamic approach was adopted by Prananta et al. [6] using the ENFLOW
CFD system developed by the National Research Laboratory NLR. It predicted LCO of an F-16
configuration with stores at Mach number (M) = 0.9 and Angle of Attack (AoA) = 7°. However,
it is known from the flight observation that the F-16 LCO could occur at cruise angle of attack
normally in the range of 2° to 3°. Therefore, it is highly possible that Prananta et al. simulated
the oscillating dynamic loads due to wing buffet, but not LCO, on the F-16 at that moderate
angle of attack. The strongest evidences to show that the nonlinear aerodynamics cannot be the
sole LCO bounding mechanism is the flight test data of two F-16 with store configurations
presented by Brignac [7]. The F-16 with the first set of stores are the tip launchers 16S200 at
weapon stations 1 and 9 and AIM-7F missiles at weapon stations 3 and 7. The flight test data of
the first F-16 with stores configuration shows that LCO occurs within M=0.9 and M=1.4.
However, at M=1.4 the transonic shock is absent from the F-16 wing and therefore the LCO
bounding mechanism at M=1.4 cannot be induced by the oscillating shock. The second F-16
with stores configuration is a Block 40 F-16 with AIM-9P missiles and LAU-129 launchers at
weapon stations 1 and 9, LAU-129 launchers at weapon stations 2 and 8, MK-84 bombs at
weapon stations 3 and 7, 370 gallon tanks at weapon stations 4 and 6 and a 300 gallon tank at
weapon station 5. The Mach number at which LCO begins for this F-16 configuration with
stores is 0.6, which is far below the transonic Mach numbers; showing once again that the LCO
bounding mechanism at M=0.6 cannot be the oscillating shock. Therefore, it can be stated that
the sole nonlinear aerodynamic approach supported by a computational methodology cannot
adequately address the LCO phenomenon thus far.

In 1998, ZONA Technology Inc. (ZONA) proposed nonlinear structural damping (NSD) as a
LCO bounding mechanism [8]. The original justification for the appearance of structural
nonlinearity was rooted in friction. In this perspective, note first an aircraft with stores consists
of many mechanical joints to connect structural components to each other; the stores with their
respective pylon/launcher and the pylons/launchers to the wing. Further, the dry friction in each

2
mechanical joint could provide a stabilizing nonlinear structural damping to the aeroelastic
system. Indeed, when flutter starts and the structural oscillating amplitude is small, the resulting
forces due to the low-amplitude oscillation of the joints also are small, smaller than the static
friction limit; thus no slip takes place and the oscillating amplitude continues to increase due to
flutter. When the amplitude of response becomes large enough, the forces in the mechanical
joins are sufficient to induce slip and thus dissipation takes place through friction. Note that the
various joints of the aircraft act in series and thus the occurrence of slip progresses as the
amplitude of response increases. Thus, the nonlinear structural damping of the aeroelastic
system increases gradually as the oscillating amplitude due to flutter increases. If the flutter
mechanism is not explosive, the friction damping in the aeroelastic system (the LCO bounding
mechanism) may equate the energy introduced into it through aerodynamics and a LCO may
result. The existence of Coulomb friction in the F-16 structure at large amplitude oscillation has
been confirmed by the ground vibration test performed by Dossogne et al [9].

The NSD mechanism is not limited to friction; it can include other forms of damping and
dissipation. In fact, Sharma and Denegri [10] performed a time integration of the nonlinear
aeroelastic equations (TINA) that reads:
 1   1 
M hh q   Bhh ( g )   bVQhhi (k )  q   K hh  V 2Qhhr (k )  q  0 (1)
 4k   2 
where Mhh, Khh, and Bhh(g) are the generalized stiffness, mass and damping matrices,
respectively. ρ, b, V and q are the air density, reference chord, free stream velocity and
generalized coordinates, respectively. Qhhi (k ) and Qhhr (k ) are the imaginary part and real part of
the generalized aerodynamic forces (GAFs) that are pre-computed by the linear Doublet Lattice
Method (DLM). k is the reduced frequency which is determined by an iterative process so that
the frequency and damping of the time domain response computed by Eq. (1) match those
computed by the frequency-domain flutter equation.
The NSD mechanism was modeled globally in Eq. (1) through the generalized damping matrix
Bhh(g) which is defined as a prescribed, monotonically increasing nonlinear function of the
structural acceleration g at a reference point. During the time-domain simulation, the
generalized damping matrix is updated at each time step according to g. The TINA computed
LCO amplitudes compared favorably with the flight test data of an F-16 with stores
configuration; proving that the NSD mechanism is a strong contender as a LCO bounding
mechanism. In this perspective, it is useful to recognize the parallel between Eq. (1) and the
classic van der Pol equation
m q   q 2  c  q  k q  0 (2)
which is well known to exhibit LCOs [11]. The similarity between Eqs. (1) and (2) is
particularly clear if the unstable aerodynamic forces are modeled by the negative damping term
c q present in Eq. (2).
However, because of the linear aerodynamics from DLM, the TINA computed onset LCO
Mach numbers do not correlate very well with the flight test data. This is because the linear
DLM cannot capture the transonic shock effects; thereby, its predicted flutter boundary which is
the onset of the LCO condition is not accurate. The linear aerodynamic and nonlinear structural
damping model in TINA is defined herein as the sole nonlinear structural approach.
The ideal computational methodology for predicting LCO is the combination of the sole
nonlinear aerodynamic approach and the sole nonlinear structural approach. This calls for a
Nonlinear Aerodynamics and Nonlinear Structures Interaction (NANSI) simulation tool that

3
must have the following features: (1) the unsteady aerodynamic method must be nonlinear and
able to capture transonic shocks, (2) flow viscosity should be included, (3) all components of the
configuration, particularly the external stores, must be modeled both structurally and
aerodynamically, and (4) structural nonlinearities can be modeled in the numerical simulation.
In addition, the simulation tool must be computationally efficient so that results can be generated
for a massive number of aircraft with stores configurations at various flight conditions with an
affordable computational cost. One such a NANSI simulation tool is the nonlinear flutter
module in ZONA’s Euler Unsteady Solver (ZEUS).

2. ZEUS (ZONA’S EULER UNSTEADY SOLVER)

ZEUS [12] is a ZONA Euler Unsteady Solver that integrates the essential disciplines required
for aeroelastic design/analysis. It uses the Cartesian Euler flow solver with a boundary-layer
option to include viscous effects as the underlying aerodynamic force generator coupled with
structural modal solution to solve various aeroelastic problems, such as maneuver loads, store
ejection loads, gust loads, static aeroelastic/trim and nonlinear flutter analysis. The structural
modal solution can be directly imported from commercial finite element software such as
Nastran, ANSYS, IDEAS and ABAQUS.

(a) ZAERO Surface Panel Model

(b) 3-D View

(c) X-Y View (d) Y-Z View

Figure 1. Overset Mesh to Model the F-16 Typical-LCO Configuartion

ZEUS is driven by the need of higher fidelity CFD methods for aeroelastic analysis, yet the
aerospace industry is still accustomed to using aerodynamic panel methods such as the doublet
lattice method and ZAERO. This is because these methods utilize a paneling scheme, which is
far simpler than the grid generation procedure required by the CFD methods. Towards this end,
ZEUS has been developed to use the input format that is very similar to that of Nastran and
ZAERO. In fact, the majority of the input data cards of ZEUS are nearly identical to those of
ZAERO. ZEUS also is equipped with an automated mesh-generation scheme that can generate a
flowfield grid/mesh by automatically extending from a surface mesh. This automated mesh-
generation greatly relieves users from the tedious CFD grid generation procedure, and is one of

4
the primary advantages of using ZEUS over other CFD codes. In addition, ZEUS has an overset
mesh capability that can handle very complex aircraft configurations such as a complete aircraft
with external stores, in which aircraft and stores are modeled by different blocks of mesh. The
communication of the flow solution among the different blocks of mesh is handled through the
interpolation of the flow solutions in the overlapped-mesh regions. Solution convergence is
achieved by Newton sub-iterations. An F-16 with external stores configuration, known as the
Typical-LCO configuration which encountered LCO during flight tests [1], is selected as the test
case to demonstrate the overset mesh capability of ZEUS. The surface mesh represented by a
panel model of this typical-LCO configuration is shown in Figure 1(a) where three underwing
stores/missiles mounted under each side of the wing can be seen. To model such a complex
configuration, we use 24 blocks of mesh whose 3-D view, X-Y view and Y-Z view are shown in
Figures 1(b), 1(c), and 1(d), respectively, rendering an overset mesh for the ZEUS computation.
Because the overset mesh scheme allows the ease of modeling complex configurations, ZEUS
is an ideal aerodynamic solver for NANSI simulation of massive number of aircraft with stores
configurations. If a wide range of store aerodynamic models can be pre-established and saved on
a weapons database, a ZEUS aerodynamic model of a given aircraft with stores can be easily
obtained by simply retrieving the store aerodynamic models from the weapons database and
including these models in the clean wing aerodynamic model, rendering an overset mesh that can
accurately account for the aerodynamic interference between aircraft and stores

2 THE NONLINEAR FLUTTER MODULE IN ZEUS

ZEUS has been validated with many wind tunnel data ranging from steady aerodynamic
pressure distributions to flutter boundaries [13]. ZEUS has a built-in nonlinear flutter module
that allows for the inclusion of nonlinear structural damping effects in the aeroelastic simulation.
It should be noted that prior to dynamic aeroelastic simulation, a static aeroelastic analysis is
always first performed by ZEUS to determine the static deformation of the structure from which
the dynamic aeroelastic simulation starts. The inclusion of the static aeroelastic effects in the
dynamic aeroelastic simulation is important for the transonic flows because the statically
deformed shape has major impact on the transonic shock position and strength that dominate the
nonlinear aerodynamic characteristics of the unsteady flow. Two sets of the aeroelastic
simulations were performed using the nonlinear flutter module in ZEUS for the F-16 Typical-
LCO configuration at M=0.75 and AoA=2.7°, M=0.8 and AoA=2.4°, M=0.85 and AoA=2.3°,
M=0.9 and AoA=2.0°, M=0.95 and AoA=2.0° and at sea level for which the time histories of the
accelerations (G's) near the leading edge of the tip launcher were determined. Note that the
angle of attack at each Mach number is obtained from the flight test data at the cruise condition.
Since the LCO response of the F-16 Typical-LCO configuration is known to be anti-symmetric,
we employed an one-cycle anti-symmetric flap oscillation with amplitude of ±0.1° to excite the
anti-symmetric modes
5
4 0.75 Mach
3
2
Frequency = 8.3835Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15
Time (seconds)

5
 5
4 0.80 Mach
3
2
Frequency = 8.3699Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15
Time (seconds)
5
4 0.85 Mach
3
2
Frequency = 8.3562Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15
Time (seconds)
5
4 0.90 Mach
3
2
Frequency = 8.3555Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15
Time (seconds)
100
0.95 Mach
50 Frequency = 8.3597Hz
G's

-50

-100 0 5 10 15
Time (seconds)

Figure 2. Tip Launcher Accelerations of the F-16 Typical-LCO Configuration Computed by

ZEUS using the Sole Nonlinear Aerodynamic Approach.

The first aeroelastic simulation used the sole nonlinear aerodynamic approach for which the
NSD effects are excluded from the nonlinear flutter
module and the results are presented in Figure 2. It 0.04
can be seen that the acceleration responses using the
sole nonlinear aerodynamic approach at all Mach
0.03
numbers are divergent motions; demonstrating once
Damping ratio

again that the nonlinear aerodynamics alone is not

sufficient to yield LCO. The frequencies of 0.02

oscillations observed at all Mach numbers are

approximately 8.3 Hz which correlates well with the 0.01
LCO frequencies observed during the flight tests. In
addition, the slow divergent response at M=0.75
0
indicates that the stability boundary (flutter boundary) -4 -3 -2 -1 0 1 2 3 4
Response amplitude (g)
of the F-16 Typical-LCO configuration is slightly
below M=0.75 which also correlates well with the
onset LCO Mach number observed during the flight Figure.3. The NSD Model for the F-16
tests. These results computed by the sole nonlinear Typical-LCO Configuration.
aerodynamic approach using ZEUS show that,

6
without the nonlinear structural damping effects,
ZEUS only is a good flutter predictive tool but not a
good LCO predictive tool.
5
4 0.75 Mach
3
2
LCO Frequency = 8.3856Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15 20
Time (seconds)
5
4 0.80 Mach
3
2
LCO Frequency = 8.3717Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15 20
Time (seconds)
5
4 0.85 Mach
3
2
LCO Frequency = 8.3566Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15 20 25 30 35 40
Time (seconds)
5
4 0.90 Mach
3
2
LCO Frequency = 8.3557Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15 20 25 30 35 40
Time (seconds)
5
4 0.95 Mach
3
2
LCO Frequency = 8.3523Hz
1
G's

0
-1
-2
-3
-4
-5 0 5 10 15 20
Time (seconds)
Figure 4. Tip Launcher Accelerations of the F-16 Typical-LCO Configuration Computed by
ZEUS using the NANSI Approach.

The second aeroelastic simulation used the NANSI approach in which a NSD model is
included in the aeroelastic equations of motion by assuming that the generalized damping matrix
is scaled by a function of the acceleration near the leading edge of the tip launcher. This
function was identified from the flight data and is shown in Figure 3.

7
Note that the maximum damping ratio of this
NSD model is not very large; only
approximately 3.5% at ±3.5g. Yet, this small
nonlinear damping is sufficient to induce LCO
at all Mach numbers as shown from the time 4
Flight test 10000 ft
histories of Figure 4 predicted by the ZEUS Flight test 5000 ft
3.5 Flight test 2000 ft

Response amplitude (+/-g)

nonlinear flutter module with the NSD in ZEUS fwd accel
ZEUS aft accel
Figure 3. Excellent agreement of the ZEUS 3
predicted LCO amplitudes using the NANSI 2.5
approach with those measured by flight testing
is obtained and is shown in Figure 5. In this 2

figure, the ZEUS results are represented by the 1.5

solid triangular symbols (solid line for the
1
acceleration near the leading edge of the tip
launcher and dashed line for that near the 0.5
trailing edge) and the flight test data at various
0
altitudes represented by the open symbols. 0.6 0.7 0.8 0.9 1
Mach number
This case clearly demonstrates that the
Figure 5. Comparison of the F-16 Typical-
nonlinear flutter module in ZEUS using the
LCO Amplitudes between the Results Computed
NANSI approach can be used as a good tool to
by ZEUS using the NANSI Approach and
determine the flutter and LCO characteristics
Measured by Flight Test.
of aircraft so that potentially dangerous
configurations can be identified with
confidence to reduce the need for flight testing.

3 DEVELOPMENT OF A PHYSICAL-ROOTED NSD MODEL-THE GENERALIZED VAN DER

POL MODEL
The approach currently available in the ZEUS nonlinear flutter module was intended as a first,
simple effort to include NSD in the ZEUS aeroelastic framework. The recent validation efforts
of [14] strongly support the inclusion of NSD for the accurate prediction of LCO but also
demonstrate that the approach currently implemented in ZEUS is too informal: the damping
model introduced to match the behavior of one aircraft configuration is not always appropriate
for a different aircraft configuration.
To resolve this issue, it is necessary to introduce a NSD model that is physically rooted. In this
regard, it is well recognized that the sources of dissipation in structures are numerous and often
complex, e.g. internal damping, friction, and thus, it is not expected to have a fully predictable
model from first principle. Rather, it is desired to have a NSD model form which includes a
minimum of “tunable” coefficients, say one to two, to be determined from one set of tests but
enabling the prediction of the behavior in other tests. This situation is completely similar to what
is done in linear structural dynamics: the damping matrix is postulated in a particular form, e.g.
Rayleigh damping or modal damping, the coefficients of which (e.g. damping ratios) are
obtained from a particular set of tests but generally apply to similar structures as well. One of
the key properties required of the damping matrix is that it be positive definite so that dissipation
always takes place. This is an essential property of any damping modeling, linear or nonlinear,
to ensure the absence of spurious, non-physical responses.

8
 The above discussion demonstrates that the proposed NSD model to be implemented should:
(i) be physically rooted, e.g. in continuum mechanics, but must be consistent with the use of
vibration modes to represent the structural response.
(ii) guarantee that dissipation always takes place to avoid spurious responses.
(iii) be able to drive the post flutter response to LCO.
(iv) involve a limited set (one to three) of coefficients to be identified from one set of tests
but are predictable of the response in other sets for similar aircraft (i.e. still F-16 but with
different stores).
Since the global NSD mechanism to be modeled arises when the amplitude of the motions
becomes “large enough”, an appropriate framework for its formulation is finite deformation
viscoelasticity in which the response is further expressed as a superposition of the linear
structural modes to satisfy condition (i) above. Then, this effort is effectively the extension to
viscoelasticity of the nonlinear reduced order modeling of elastic structures developed by the
Arizona State University (ASU) group. This extension can be carried out by replacing the elastic
constitutive relation (Hooke’s law) of Eq. (6) by a viscoelastic one. Here, we adopt a simple
Maxwell [15] viscoelastic model. That is, the stress tensor S of components Sij will be expressed
in terms of its strain counterparts Eij as (summation over repeated indices implied)
Sij  Cijkl Ekl  Dijkl Ekl (7)
where Cijkl denotes the fourth order elasticity tensor but Dijkl is a dissipation tensor relating the
stress to the strain rate. Under the model of Eq. (7), the power dissipated in the infinitesimal
element considered is
Pdissip  E ij S ij  E ij Dijkl E kl (8)
which will be positive at all times and for all elements if Dijkl is positive definite, which is thus
the necessary and sufficient condition to impose the constraint (ii) above. A simple choice of a
tensor with this property [15] is
Dijkl   d ij  kl   d ik  jl  il  jk  (9)
where δij is the Kronecker delta (δij = 1 if i = j, 0 otherwise) and λd and μd are parameters. The
positive definiteness of Dijkl is guaranteed when λd and μd are both positive. Continuing the
analogy with this tensor, one can also define a dissipation modulus E D and ratio  D (similar to
the Young’s modulus and Poisson’s ratio) through the transformation

ED ED  D
d  and d  (10)
2 1   D  1   1  2 
D D

Adopting the dissipation model of Eq. (9) leads to a 2-parameter model, i.e., with either  d and
 d or E D and  D characterizing it. In the following section, it will be shown that the ratio  D
has only a very small effect on the aeroelastic response of the F-16 leaving the dissipation
modulus E D as sole parameter of the nonlinear structural damping model to be identified from
flight test data. Note that the dissipation tensor of Eq. (9) can be different in different
sections/elements of the structural model to reflect different materials.

9
 Then, paralleling the derivations in references [16-19] it is found that the governing equations
for the generalized coordinates of the structural response with the constitutive model of Eq. (7)
are
M ij q j   Dij(1)  Dijl(2) ql  Dijlp
(3)
ql q p  q j  Kij(1) q j  Kijl(2) q j ql  Kijlp
(3)
q j ql q p  Fi (11)
where Fi denote the generalized aerodynamic forces. Note that Eq. (11) includes both nonlinear
damping and nonlinear stiffness terms. Since the amplitude of observed LCO during flight
testing is notably less than the threshold of 15% of span below which the nonlinear stiffness
terms are typically insignificant in wings, it is proposed here to ignore these nonlinear stiffness
terms in Eq. (11) leading to the equations
M ij q j   Dij(1)  Dijl(2) ql  Dijlp
(3)
ql q p  q j  Kij(1) q j  Fi (12)
where Kij(1) and Dij(1) are the standard linear stiffness and damping matrices.
Note that the model of Eq. (12) is a generalized van der Pol model, i.e. a natural extension of
the classic van der Pol equation of Eq. (2) to a multi degree of freedom system. Given the well
known LCO features of the classic van der Pol equation, it is surmised that Eq. (12) will indeed
exhibit the capability to curb the increasing structural response post flutter and lead to LCO.
Thus, condition (iii) is also expected to be satisfied by the model of Eq. (12).
It remains to address the estimation of the nonlinear damping parameters Dijl(2) and Dijlp
(3)
. For a
particular set of values of the parameters λd and μd of Eq. (9), the parameters Dijl and Dijlp
(2) (3)
can
be determined from a commercial finite element model of the wing/aircraft by following the
same procedure as the one for the estimation of the nonlinear stiffness terms Kijl(2) and Kijlp
(3)
in the
ASU nonlinear structural reduced order modeling approach [16-19]. Further, the damping
parameters Dijl(2) and Dijlp
(3)
can be seen to depend linearly on λd and μd. Thus, estimating the
parameters Dijl and Dijlp for two sets of values of λd and μd will permit the
(2) (3)

interpolation/extrapolation of the damping parameters to other values of λd and μd. This

discussion demonstrates that condition (iv) is also satisfied, i.e. the damping model can be
identified from a commercial finite element model of the aircraft and includes the two tunable
parameters λd and μd. These two parameters may then be determined by minimizing the
differences between one set of measured LCO amplitudes and the corresponding predictions
from the model of Eq. (12). Once λd and μd are determined, the coefficients Dijl(2) and Dijlp (3)
can be
evaluated and the generalized van der Pol model can be applied to other configurations/altitudes.

4 DEVELOPMENT OF THE NANSI MODULE IN ZEUS

The NANSI module in ZEUS essentially solves Eq. (12) using the Newmark-β scheme. After
the generalized coordinates q are obtained at each time step, the physical deformation on the
ZEUS surface mesh h is computed by first superimposing the natural mode shapes φ with q then
by mapping the physical deformation from the structural grids onto the ZEUS surface mesh via
the spline matrix Sp . Thus, the physical deformation on the ZEUS surface mesh is computed
using the following equation:

h  S p q (13)
The physical deformation on the ZEUS surface mesh is plugged into the ZEUS unsteady Euler
boundary condition to compute the aerodynamic forces f by the ZEUS Euler solver such as:

10
 f  f (h, h) (14)
Then, the generalized aerodynamic forces are simply computed by the following equation:

Fi   T f (h, h) (15)

Substituting Fi into the right hand side of Eq. (12) yields to the solution of q at the next time step;
leading to the complete time history of the structural response.

5 VALIDATION OF THE NANSI MODULE WITH F-16 TYPICAL-LCO FILGHT TEST DATA

Figure 6 presents the structural finite element model of the F-16 Typical LCO configuration.
To construct the structural reduced order model, its first 26 elastic modes were retained spanning
the frequency range of 0 to 25 Hz. Conforming to standard aeroelastic analyses, the linear
damping matrix was assumed negligible and only the quadratic and cubic nonlinear damping
terms were included. They were identified as described in section 3 first for  d 0 and  d =0,
then for  d =0 and  d 0. Then, proceeding with a linear combination, the nonlinear damping
coefficients corresponding to nonzero values of both  d and  d are obtained.

Figure 6. Finite Element of the F-16 Typical LCO Configuration.

In a first effort, the ratio  D in each element was selected as the Poisson’s ratio of the F-16
material of that element. Then, the elemental dissipation modulus E D was selected as a common
fraction  of the corresponding Young’s modulus. Then,  was tuned so that the ZEUS predicted
LCO amplitude closely matched the flight test data at Mach number = 0.9 and the altitude of
-4
5000 ft. This optimum value of  was found to be  = 0.8 ×10 seconds. It should be noted that
the ZEUS computational mesh for the F-16 Typical LCO configuration is already shown, in
Figure 1.

11
Then, shown in Figure 7 are the time histories of the acceleration at the tip launcher of the F-16
Typical-LCO configuration computed by the NANSI module of ZEUS at various Mach numbers
and at 5000 ft. Comparing the results shown in Figure 2, clearly, without the NSD model the
acceleration at the tip launcher is a divergent response. But with the NSD model, LCO at all
Mach numbers is obtained. Note that the LCO frequencies predicted by the NANSI module of
ZEUS with and without NSD model are around 8.3 Hz which are in good agreement with the
flight test frequencies of 8.0 Hz. The comparison of the LCO amplitudes predicted by the
NANSI module of ZEUS and the flight test data at Mach numbers ranging from 0.7 to 0.95 and
at the three altitudes of 2000 ft, 5000 ft, and 10000 ft is shown in Figure 8. Note that at M=0.7,
the flight test data is not available at altitudes of 10000 ft and 2000 ft.
8
Mach = 0.75; LCO Frequency = 8.36Hz

4
Accel. (g)

-4

-8
0 20 40 60 80 100 120
Time (s)
8
Mach = 0.80; LCO Frequency = 8.35Hz

4
Accel. (g)

-4

-8
0 5 10 15 20 25 30
Time (s)
8
Mach = 0.85; LCO Frequency = 8.34Hz

4
Accel. (g)

-4

-8
0 5 10 15 20
Time (s)
8
Mach = 0.90; LCO Frequency = 8.33Hz

4
Accel. (g)

-4

-8
0 5 10 15 20
Time (s)
8
Mach = 0.95; LCO Frequency = 8.32Hz

4
Accel. (g)

-4

-8
0 5 10 15 20
Time (s)

Figure 7. Tip Launcher Accelerations of the F-16 Typical-LCO Configuration Computed by

the NANSI Module of ZEUS.

12
 Figure 8. Comparison of the F-16 Typical-LCO Amplitudes between the Results Computed
by the NANSI Module of ZEUS and Measured by Flight Test

The above effort focused solely on tuning the value of  with the Poisson’s ratios fixed. To
assess the influence of the ratio  D on the ZEUS predicted LCO amplitude, the computations
were also repeated with  D = 0. Shown in Figure 9 are the comparisons of the predicted LCO
-4
amplitudes between  D = 0 and  D  0 at  = 0.8 ×10 seconds. Clearly, these two sets of
results are practically identical. It is thus concluded that the ratio  D has insignificant effect on
the LCO amplitude and thus only one parameter () of the dissipation model needs to be tuned to
match flight test data at only one flight condition.

ZEUS 10000 ft, vD=v

3.5 ZEUS 5000 ft, vD=v
Response amplitude (+/-G)

ZEUS 2000 ft, vD=v

ZEUS 10000 ft, vD=0
3 ZEUS 5000 ft, vD=0
ZEUS 2000 ft, vD=0
2.5

1.5

0.5

0
0.7 0.8 0.9
Mach number

13
 Figure 9. Comparison of the ZEUS Predicted LCO Amplitude
for  D = 0 and  D  0 at  =0.8×10 seconds
-4

6 VALIDATION OF THE NANSI MODULE WITH F-16 NON-TYPICAL LCO FLIGHT TEST
DATA
The structural finite element model of the F-16 Non-Typical LCO configuration is presented in
Figure 10. The ZEUS computational mesh is depicted in Figure 11 which consists of 42 blocks
of mesh to model the F-16 aircraft and all the external stores. The F-16 Non-Typical LCO flight
test data at Mach number = 0.95 and altitude = 5000 ft was selected to identify the parameter 
-3
for the NANSI computations which was found to be =1.0×10 s. Shown in Figure 12 are the
time histories of the acceleration at the tip launcher of the F-16 Non-Typical LCO configuration
computed by ZEUS Euler solver with and without the NANSI module at various Mach numbers
and at 5000 ft altitude. Similarly to the F-16 Typical LCO configuration, without the NSD
model the acceleration at the tip launcher is a divergent response but with the NSD model, LCO
at all Mach numbers is obtained; once again, showing that the NSD is one of the key elements
involved in the LCO bounding mechanism.

Figure 10. Finite Element of the F-16 Non-Typical LCO Configuration

14
 Figure 11. 42-Block of Mesh to Model the F-16 Non-Typical LCO Configuration

The comparison of the LCO amplitudes predicted by the ZEUS NANSI module and the flight
test data at Mach numbers ranging from 0.7 to 0.98 and at the three altitudes of 2000 ft, 5000 ft,
and 10000 ft is shown in Figure 13. Unlike the F-16 Typical LCO configuration the LCO
amplitude of which increased with Mach number, the LCO amplitude of the F-16 Non-Typical
LCO flight test data exhibits a peak in the range of Mach number between 0.8 and 0.98. This
trend is captured by the NANSI module except that its peak occurs at Mach number 0.95
whereas for the flight test data it occurs at 0.90.

Without NSD With NSD

10 3
Mach = 0.86; Flutter Frequency = 4.3555Hz Mach = 0.86; LCO Frequency = 4.3555Hz

2
5
1
Accel. (g)

Accel. (g)
0 0

-1
-5
-2

-10 -3
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20
Time (s) Time (s)
80 3
Mach = 0.90; Flutter Frequency = 4.3645Hz Mach = 0.90; LCO Frequency = 4.2996Hz

2
40
1
Accel. (g)
Accel. (g)

0 0

-1
-40
-2

-80 -3
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

80 3
Mach = 0.95; Flutter Frequency = 4.3875Hz Mach = 0.95; LCO Frequency = 4.2716Hz

2
40
1
Accel. (g)

Accel. (g)

0 0

-1
-40
-2

-80 -3
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)
80 3
Mach = 0.98; Flutter Frequency = 4.3609Hz Mach = 0.98; LCO Frequency = 4.3037Hz

2
40
1
Accel. (g)
Accel. (g)

0 0

-1
-40
-2

-80 -3
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

Figure 12. Structural Responses for F-16 Non-Typical LCO Configuration by the NANSI
Module of ZEUS ( = 1.0×10-3 s.) and Without NSD

15
 2
Flight test 10000 ft
Flight test 5000 ft
Flight test 2000 ft

Response amplitude (+/-g)

ZEUS 10000 ft
ZEUS 5000 ft
1.5 ZEUS 2000 ft

0.5

0
0.7 0.8 0.9 1 1.1
Mach number
Figure 13. Comparison of the F-16 Non-Typical LCO Amplitudes Computed by the NANSI
Module of ZEUS and Measured by Flight Test

5. CONCLUSIONS
A nonlinear structural damping model was first developed as a LCO bounding mechanism. It is
based on a linear viscoelastic material modeling undergoing large deformations. The model was
developed in a reduced order modeling format assuming a Kelvin-Voigt constitutive relation
expressed in the undeformed configuration, which led to governing equations for the generalized
coordinates in the form of van der Pol – Duffing equations. These equations which generalize
similar reduced order models obtained for elastic structures are parametric, i.e., involve a series
of coefficients which are functions of the material properties, geometry of the structure, and basis
functions used to represent the response. An existing identification strategy of the stiffness
parameters in such reduced order models was next modified to permit the determination of the
linear and nonlinear damping coefficients from a finite element model of the structure.

In this study, the above identification process was carried out on the F-16 Typical and Non-
Typical LCO finite element configurations under the assumption of an isotropic dissipation
model. Such a model is characterized by only two parameters, either Lamé-type parameters or
dissipation modulus and ratio, the values of which can be calibrated from experimentally
observed LCO amplitudes.

The nonlinear structural damping model was coupled with the time accurate Euler solver in
ZEUS to establish a Nonlinear Aerodynamics and Nonlinear Structure Interaction (NANSI)
capability. Through the use of the NANSI module in ZEUS, the following key findings were
observed:

(1) The dissipation ratio,  D , has an insignificant effect on the LCO characteristics
predicted by the NANSI module in ZEUS leaving the dissipation modulus, E D , or more
specifically the parameter   E D / E , where E is the Young’s modulus, as the only
parameter needing to be identified from flight test data.

16
 (2) Calibrating the parameter  using only one flight test data point leads to good matching
of the LCO amplitudes at all other flight conditions between the NANSI module
predictions and the flight test data.

(3) Without the nonlinear structural damping model, the time accurate Euler solver in ZEUS
predicts divergence of the response at flight conditions beyond the flutter boundary. This
finding confirms that the sole nonlinear aerodynamics provided by the time accurate
Euler solver is not a sufficient LCO bounding mechanism.

(4) A good matching of the flight test LCO amplitudes by the NANSI module of ZEUS
predictions was obtained for both the F-16 Typical and Non-Typical LCO
configurations.

Based on this work, the NANSI moduel of ZEUS can be applied to a wide range of F-16 with
stores configurations of which the flight test data is available so that a database of the tunable
parameter in the generalized van der Pol model can be established. Once established, the tunable
parameter in the generalized van der Pol model of a new store configuration can be determined
from the database prior to flight test; rendering ZEUS with the generalized van der Pol model a
truly LCO predictive tool.

8. ACKNOWLEDGEMENT
This work was sponsored by the Air Force Office of Scientific Research under the Small
Business Technology Transfer Research program (Contract Number: FA9550-15-C-0006). The
contract monitor was Dr. Michael Kendra. The authors would like to thank Dr. Charles Denegri
and Dr. Vin Sharma of SEEK Eagle Office for their valuable suggestions throughout the period
of this project.

7. REFERENCES
[1] Denegri Jr., C.M., “Limit Cycle Oscillation Flight Test Results of a Fighter with External
stores,” Journal of Aircraft, Vol. 37, No. 5, Sept.-Oct. 2000.
[2] Norton, W.J., “Limit Cycle Oscillation and Flight Flutter Testing,” Proceeds of the 21st
Annual Symposium, Society of Flight Test Engineers, Lancaster, CA, 1990, pp. 3.4-3.4-12.
[3] Farhat, C., Geuzaine, P., and Brown, G., "Application of a Three-Field Nonlinear Fluid-
Structure Formulation to the Prediction of the Aeroelastic Parameters of an F-16 Fighter,"
Computers and Fluids, Vol. 32, No. 1, 2003, pp. 3-29.
[4] Pasiliao, C. L., "Characterization of Aero-Structural Interaction Flow-Field Physics,"
Aerospace Flutter and Dynamics Council Fall Meeting. Shalimar, FL. 25-26 October 2012.
[5] ZAERO Theoretical Manual, Version 9.0, 2014, ZONA Technology Inc.
[6] Parananta, B.B., Kok, J.C., Spekreijse, S.P. Hounjet, M.H.L., and Meijer, J.J., "Simulation
of Limit Cycle Oscillation of Fighter Aircraft at Moderate Angle of Attack," NLR-TP-2003-
526, October 2003.
[7] Brignac W.J., "Limit Cycle Oscillation F-16 Experience,", Aerospace Flutter and Dynamics
Council Fall. 10-11 May, 1989.
[8] Chen, P.C., Sarhaddi, D., and Liu, D.D., "Limit Cycle Oscillation Studies of a Fighter with
External Stores," AIAA 98-1727, AIAA, 1998.

17
 [9] Dossogne, T, Noel, J.P., Grappasonni, C., Kerschen, G., Peeter, B., Debille, J. Vaes, M. and
Schoukems, J., "Nonlinear Ground Vibration Identification of an F-16 Aircraft - Part II:
Understanding Nonlinear Behaviour in Aerospace Structure Using Sine-Sweep Testing,"
Proceedings of the International Forum on Aeroelasticity and Structural Dynamics Society,
Saint Petersburg, 2015, IFASD-2015-035.
[10] Sharma, V.K., and Dengeri, C.M., "Time Domain Aeroelastic Solution Using Exact
Aerodynamic Influence Coefficients and Nonlinear Damping," Proceedings of the
International Forum on Aeroelasticity and Structural Dynamics Society, London, 2013,
IFASD-2013-29D. Royal Aeronautical Society, London, 2013, IFASD-2013-29D.
[11] Nayfeh, A.H., and Mook, D.T., Nonlinear Oscillations, Wiley 1979.
[12] Chen, P.C., Zhang, Z., Sengupta, A., and Liu, D., “Overset Euler/Boundary-Layer Solver
with Panel-Based Aerodynamics Modeling for Aeroelastic Applications”. Journal of
Aircraft, 2009. 46(6): p. 2054-2068.
[13] ZEUS Applications Manual, ZONA Technology Inc. 2013.
[14] Sharma, V.K., and Dengeri, C.M., ., "F-16 Limit Cycle Oscillation Simulation Using
ZONA Euler Unsteady Solver (ZEUS),"", Aerospace Flutter and Dynamics Council Fall.
10-11 May, 1989.
[15] Fung, Y. C., and Tong, P., Classical and Computational Solid Mechanics, World Scientific,
River Edge, New Jersey, 2001.
[16] Kim, K., Radu, A.G., Wang, X.Q., and Mignolet, M.P., “Nonlinear Reduced Order
Modeling of Isotropic and Functionally Graded Plates,” International Journal of Non-
Linear Mechanics, Vol. 49, pp. 100-110, 2013.
[17] Perez, R.A., Wang, X.Q., Matney, A.K., and Mignolet, M.P., “Reduced Order Modeling for
the Static and Dynamic Geometric Nonlinear Responses Of a Complex Multi-Bay
Structure,” Proceedings of the 11th International Conference on Recent Advances in
Structural Dynamics,” Pisa, Italy, Jul. 1-3, 2013.
[18] Perez, R.A., Wang, X.Q., and Mignolet, M.P., “Non-Intrusive Structural Dynamic Reduced
Order Modeling for Large Deformations: Enhancements for Complex Structures,” Journal
of Computational and Nonlinear Dynamics. In Press.
[19] Mignolet, M.P., Przekop, A., Rizzi, S.A, and Spottswood, S.M., “A Review of Indirect/Non-
Intrusive Reduced Order Modeling of Nonlinear Geometric Structures,” Invited Paper,
Journal of Sound and Vibration, Vol. 332, No. 10, pp. 2437-2460, 2013.

18

View publication stats