You are on page 1of 25

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

net/publication/346927868

Correlations Between UM/NAST Nonlinear Aeroelastic Simulations and the Pre-


Pazy Wing Experiment

Conference Paper · December 2020


DOI: 10.2514/6.2021-1712

CITATIONS READS
12 901

2 authors:

Cristina Riso Carlos E. S. Cesnik


Georgia Institute of Technology University of Michigan
28 PUBLICATIONS   222 CITATIONS    331 PUBLICATIONS   10,818 CITATIONS   

SEE PROFILE SEE PROFILE

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

AePW3 - Large Deflection Working Group View project

X-HALE View project

All content following this page was uploaded by Cristina Riso on 15 April 2021.

The user has requested enhancement of the downloaded file.


This is a preprint of the following article:
Cristina Riso and Carlos E. S. Cesnik. Correlations Between UM/NAST Nonlinear Aeroelastic Simulations and the Pre-Pazy Wing
Experiment, 2021 AIAA SciTech Forum, Virtual Event, January 11–15 and 19–21, 2021.
The published article may differ from this preprint and is available at:
https://doi.org/10.2514/6.2021-1712.

Correlations Between UM/NAST Nonlinear Aeroelastic


Simulations and the Pre-Pazy Wing Experiment

Cristina Riso ∗ and Carlos E. S. Cesnik †


University of Michigan, Ann Arbor, MI, 48109

The Third Aeroelastic Prediction Workshop Large Deflection Group is invested in validating
computational models to analyze geometrically nonlinear wings. The effort is based on the
Pazy wing, a wind-tunnel aeroelastic benchmark developed at the Technion—Israel Institute
of Technology. To consolidate the techniques for testing the Pazy wing, the Technion developed a
preliminary experimental model named the Pre-Pazy wing. This paper assesses the accuracy of
a low-order, geometrically exact beam model with potential-flow aerodynamics of the Pre-Pazy
wing against results from a built-up finite element model and experimental data. The Pre-
Pazy wing low-order model is developed in the University of Michigan’s Nonlinear Aeroelastic
Simulation Toolbox, a multidisciplinary framework for analyzing aircraft with very flexible
wings. The correlations show that the low-order model of the Pre-Pazy wing captures its
frequencies and static response with accuracy comparable to the built-up finite element model
while requiring only a fraction of the degrees of freedom.

I. Introduction
The pursuit of energy efficient flight is a key driver of novel aircraft designs [1]. Energy efficient flight enables the
sustainable growth of global aviation along with other capabilities that require long endurance, such as telecommuni-
cations, internet distribution, and remote sensing. Achieving energy efficient flight, however, requires aircraft with
lightweight and slender wings that are very flexible. These very flexible wings experience large structural deflections
that result in geometrically nonlinear aeroelastic behaviors, which may anticipate the flutter onset [2, 3] and cause
undesirable behaviors such as subcritical limit-cycle oscillations [4, 5]. This problem requires accurate and efficient
models able to predict the dynamics in very flexible wings, especially their flutter boundaries.
There has been a significant effort in developing geometrically nonlinear aeroelastic models of very flexible
wings [6]. These models remove linear assumptions used in state-of-the-art aeroelastic analysis software [7] to capture
phenomena such as the variation in the flutter boundary with the deformed shape and instabilities caused by flight
dynamics-aeroelastic interactions [8]. While many approaches exist for modeling very flexible wings, time-critical
applications in design and control have used low-order, geometrically exact beam structural formulations [9–12] coupled
to potential-flow aerodynamics [13–15] or reduced-order aerodynamic models [16–18]. These modeling approaches are
the basis for multidisciplinary frameworks for simulating very flexible aircraft in free flight [19–24].
Maturing aeroelastic models of aircraft with very flexible wings requires extensive experimental validation. Wind-
tunnel experiments in very flexible wings, however, are limited in the literature (e.g., [25]). To address this problem,
the Third Aeroelastic Prediction Workshop (AePW3) Large Deflection Group is studying a wind-tunnel testbed for
geometrically nonlinear aeroelastic experiments—the Pazy wing [26, 27]. This wing was developed at the Technion as a
benchmark to ease the comparison and validation of different geometrically nonlinear aeroelastic models and increase
the understanding of large-amplitude phenomena by exploring aeroelastic deflections up to 60% semispan.
∗ Research Fellow, Department of Aerospace Engineering, 1320 Beal Avenue, AIAA Member, criso@umich.edu
† Clarence “Kelly” Johnson Professor, Department of Aerospace Engineering, 1320 Beal Avenue, AIAA Fellow, cesnik@umich.edu

1
(a) Undeformed configuration (b) Sample deformed configuration

Fig. 1 Pre-Pazy wing experimental model (image source: Technion).

To consolidate the techniques for testing the Pazy wing, the Technion developed a preliminary experimental
model named the Pre-Pazy wing [26, 27]. The objective of this paper is to assess a low-order, geometrically exact
beam numerical model with potential-flow aerodynamics of the Pre-Pazy wing against results from a built-up finite
element model (FEM) and experimental data from static and ground vibration tests (GVT) [26, 27]. The low-order
model is developed in the University of Michigan’s Nonlinear Aeroelastic Simulation Toolbox (UM/NAST) [5], a
multidisciplinary framework for analyzing aircraft with very flexible wings. The overall objective of this work is to
validate the low-order, geometrically nonlinear aeroelastic formulation implemented in UM/NAST with data for the
Pazy wing—the benchmark that will be considered under the frame of the AePW3—to investigate its ability to capture
the dynamics in very flexible wings with a focus on their flutter boundaries.
The paper is organized as follows. Section II describes the Pre-Pazy wing experiment and Section III introduces the
numerical models used in this paper. Correlations between numerical and experimental are presented in Sec. IV. A
section of concluding remarks summarizes the paper and outlines future work.

II. The Pre-Pazy Wing Experimental Model


The Pre-Pazy wing (see Fig. 1) is a very flexible wind-tunnel experimental model developed by Avin et al. [26] for
geometrically nonlinear aeroelastic experiments in low-speed flow. This wing was developed in preparation for the Pazy
wing experimental campaign. Both the Pre-Pazy and the Pazy wings were designed to achieve tip vertical displacements
up to 60% semispan in wind-tunnel tests to collect experimental data beyond the 0–25% semispan deflection range
considered in previous work [25]. This larger range of deflections will enable to extensively validate numerical models
of very flexible wings and to further investigate geometrically nonlinear aeroelastic phenomena.
The Pre-Pazy wing is made of an Aluminum 7075 spar and a 3D-printed PA2200 frame covered with a Oralight
Polyester skin. The wing has a semispan 𝑙 = 0.55 m and no sweep, dihedral, or twist. The cross section is shaped as a
NACA 0018 airfoil with a uniform chord 𝑐 = 0.1 m. A tip rod is 3D-printed as part of the frame for attaching weights
during testing. Further details on the Pre-Pazy wing can be found in Refs. [26, 27].

III. Low-Order Geometrically Exact Aeroelastic Model of the Pre-Pazy Wing


The Pre-Pazy wing low-order numerical model used in this paper consists of a geometrically exact, strain-based
beam structural formulation [12] coupled to potential-flow thin airfoil theory [14] with tip loss factors. The model is
developed in UM/NAST [5], a multidisciplinary framework for analyzing very flexible aircraft based on a fully coupled
flight dynamics-nonlinear aeroelastic formulation.
This section summarizes the formulation implemented into UM/NAST that is relevant to this study where there are
no rigid-body degrees of freedom (DOFs). Next, the section discusses the process of modeling the Pre-Pazy wing in
UM/NAST starting from its built-up FEM provided by the Technion [26].

2
(a) Without skin (b) With skin

Fig. 2 Pazy wing built-up FEM.

A. Theoretical Formulation
UM/NAST describes an aircraft as a set of geometrically exact beam-type members representing the vehicle
components. The aircraft motion is modeled by introducing a global inertial frame 𝐺, a body frame 𝐵, and local frames
𝑤 along the beam reference axis of each member. The 𝐵 frame position in the 𝐺 frame is given by the vector 𝑝 𝐵 , which
is resolved in the 𝐵 frame. The 𝐵 frame is oriented such that 𝐵 𝑥 points downstream, 𝐵 𝑦 points toward the right wingtip,
and 𝐵 𝑧 points along the cross product of 𝐵 𝑥 and 𝐵 𝑦 . Each local frame 𝑤 has its origin at a point 𝑝 along the beam
reference axis and is oriented such that 𝑤 𝑥 is along that axis, 𝑤 𝑦 points toward the leading edge, and 𝑤 𝑧 points along
the cross product of 𝑤 𝑥 and 𝑤 𝑦 .
The position and orientation of the beam reference axis point are described by the 12-component column vector

ℎ(𝑠)𝑇 = 𝑝(𝑠)𝑇 𝑤 𝑥 (𝑠)𝑇 𝑤 𝑦 (𝑠)𝑇 𝑤 𝑧 (𝑠)𝑇 (1)

In Eq. (1), 𝑠 is the curvilinear coordinate along the beam reference axis and 𝑝, 𝑤 𝑥 , 𝑤 𝑦 , and 𝑤 𝑧 are column vectors
resolved in the 𝐵 frame. In UM/NAST, the model is discretized in three-node beam elements each one having four strain
degrees of freedom (DOFs) denoted by 𝜀 𝑥 (axial extension), 𝜅 𝑥 (torsion curvature), 𝜅 𝑦 (out-of-plane bending curvature),
and 𝜅 𝑧 (in-plane bending curvature). The structural dynamics is solved in terms of strain DOFs and the configuration
vector ℎ is recovered as a dependent quantity by integrating geometrically exact strain-based kinematic relations [5, 12].
The Pre-Pazy wing aerodynamics is modeled using the potential-flow thin airfoil theory of Peters et al. [14] with tip
loss factors at local wing stations. Each wing station is defined by its own local aerodynamic frame 𝑎, where the unit
vector 𝑎 𝑥 is along the beam reference axis, 𝑎 𝑦 is oriented along the chord and points toward the leading edge, and 𝑎 𝑧
completes the right-hand triad.
In the case of no rigid-body DOFs, the UM/NAST equations of motion take the form [12]

𝑀𝐹 𝐹 (𝜀) 𝜀¥ + 𝐶𝐹 𝐹 (𝜀, 𝜀)
¤ 𝜀¤ + 𝐾 𝐹 𝐹 𝜀 = 𝑅𝐹 (𝜀, 𝜀,
¤ 𝜆) (2)

𝜆¤ = 𝐹1 𝜀¥ + 𝐹2 𝜀¤ + 𝐹3 𝜆 (3)
In Eq. (2), 𝜀 is the model strain vector, 𝑀𝐹 𝐹 and 𝐶𝐹 𝐹 are generalized mass and damping matrices (dependent on the
strain DOFs), 𝐾 𝐹 𝐹 is the constant stiffness matrix of the strain-based formulation, 𝑅𝐹 is the vector of generalized forces,
and 𝜆 is a vector of aerodynamic inflow states. These states account for unsteady wake effects and are governed by
first-order equations with coefficients given in the matrices 𝐹1 , 𝐹2 , and 𝐹3 .

B. Built-Up Structural Model


The Pre-Pazy wing low-order model in UM/NAST is derived from the built-up FEMs in Fig. 2, which differ only for
the skin. In this paper, the original built-up FEMs provided by the Technion are modified by making the tip rod rigid to
eliminate flexibility effects along the tip chord. Table 1 shows the frequencies of the original and modified built-up
FEMs computed using the MSC Nastran modal analysis solver SOL 103 [28]. The results show that making the tip rod
rigid does not significantly impact the frequencies of the first two out-of-plane bending modes (OOP1 and OOP2) and
of the first in-plane bending mode (IP1). The frequencies of the first torsion mode (T1) and of the third out-of-plane
bending mode (OOP3) slightly increase by 3.0–3.2% and 1.1–1.3%, respectively. The Pre-Pazy wing first five mode
shapes for the case without skin are shown in Fig. 2. The model with skin has about 5% higher OOP1, OOP2, OOP3,
and T1 frequencies with practically the same IP1 frequency (see Table 1).

3
Table 1 Pre-Pazy wing built-up FEM natural frequencies.

Without skin With skin


Frequency (Hz) Frequency (Hz)
# Mode type Mode label Original Rigid rod Δ (%) Original Rigid rod Δ (%)
1 Out-of-plane bending OOP1 4.42 4.42 0.01 4.67 4.67 0.01
2 Out-of-plane bending OOP2 28.98 29.02 0.15 30.63 30.68 0.16
3 Torsion T1 40.33 41.53 2.96 42.22 43.57 3.19
4 Out-of-plane bending OOP3 82.41 83.34 1.13 86.87 87.97 1.27
5 In-plane bending IP1 112.27 112.56 0.26 112.60 112.88 0.25

(a) Mode #1 (OOP1), 4.42 Hz (b) Mode #2 (OOP2), 29.02 Hz

(c) Mode #3 (T1), 41.53 Hz (d) Mode #4 (OOP3), 83.34 Hz

(e) Mode #5 (IP1), 112.56 Hz

Fig. 3 Pre-Pazy wing built-up FEM mode shapes and corresponding natural frequencies (without skin).

4
Built-up FEM

MSC Nastran SOL 103

pi, mi, δi, Ii

EF2S inertia module


Masses Mass offsets Inertia tensors
PNj 1 PNj PNj
m̂j = i=1 mi
δ̂j = m̂j i=1 mi (pi + δi − pj ) Îj = i=1 (Ii + Iij )

Rigid-body elements

UM/NAST

Fig. 4 Schematic of EF2S inertia module.

C. Structural Model Order Reduction


The low-order structural model of the Pre-Pazy wing in UM/NAST is derived from the built-up FEMs in Fig. 2 using
the University of Michigan’s Enhanced FEM2Stick (EF2S) framework [29]. This framework computes the equivalent
inertia and stiffness distributions of a built-up FEM along a user-specified beam reference axis discretized in a number
of nodes. The inertia and stiffness distributions are computed based on data from MSC Nastran—or alternate finite
element solver—and define a beam-type representation of the built-up FEM consistent with the UM/NAST geometrically
exact strain-based formulation [12]. The equivalent inertia distributions consist of constitutive properties that relate the
beam generalized forces to the corresponding generalized velocities. The equivalent stiffness distributions consist of
constitutive properties that relate the beam generalized forces to the corresponding generalized strains.
Figure 4 shows a schematic of the EF2S process for obtaining the equivalent inertia distributions. These distributions
are provided as a set of rigid-body elements given at the beam reference axis nodes with masses, mass offsets, and
inertia tensors computed using data from the lumped mass matrix of the built-up FEM.
The lumped mass matrix of a built-up FEM represents its distributed inertia as a set of nodal masses 𝑚 𝑖 with
positions 𝑝 𝑖 + 𝛿𝑖 in the 𝐵 frame, where 𝑝 𝑖 are the positions of the built-up FEM nodes and 𝛿𝑖 the mass offsets, and
inertia tensors I𝑖 with respect to the positions 𝑝 𝑖 + 𝛿𝑖 (𝑖 = 1, . . . , 𝑁, where 𝑁 is the number of built-up FEM nodes).
The equivalent mass 𝑚ˆ 𝑗 , mass offset 𝛿ˆ 𝑗 , and inertia tensor Î𝑗 associated with the 𝑗-th beam reference axis node are
obtained by summing the contributions from the 𝑁 𝑗 built-up FEM nodes in its nearest neighbor:
𝑁𝑗 𝑁𝑗 𝑁𝑗
∑︁ 1 ∑︁ ∑︁
𝑚ˆ 𝑗 = 𝑚𝑖 𝛿ˆ 𝑗 = 𝑚 𝑖 ( 𝑝 𝑖 + 𝛿𝑖 − 𝑝 𝑗 ) Î𝑗 = (I𝑖 + I𝑖 𝑗 ) (4)
𝑖=1
𝑚ˆ 𝑗 𝑖=1 𝑖=1

In Eq. (4), 𝑝 𝑗 is the position of the 𝑗-th beam reference axis node in the 𝐵 frame and I𝑖 𝑗 is the transport moment of
inertia of 𝑚 𝑖 from the local center of mass 𝑝 𝑖 + 𝛿𝑖 to the beam reference axis node 𝑝 𝑗 . The equivalent inertia distributions
in Eq. (4) are implemented in UM/NAST as rigid-body elements akin to the MSC Nastran CONM2 elements.
Figure 5 shows a schematic of the EF2S process for obtaining the equivalent stiffness distributions. These
distributions are given in terms of the elements of the cross-sectional stiffness matrix 𝑘 of the UM/NAST strain-based
formulation [12]. A potentially different stiffness matrix is identified for each element along the beam reference axis

5
Six load cases Built-up FEM, RBE3s

MSC Nastran SOL 101

Ftip Uj pj

EF2S stiffness module


Convert Uj → uj

uj

Compute ∆uj+1

∆uj+1

Compute fj Solve fj+1 = Kj ∆uj+1

Kj

Enforce symmetry

Kj

Solve Kj−1 Q−1 −1 −1 −1


j = kj Hj Qj + Ej kj Compute Ej , Hj , Qj

kj−1

Enforce symmetry

kj−1

Reduce kj−1 to 4 × 4

kj−1

Invert kj−1

kj

Enforce symmetry

Stiffness distributions

UM/NAST

Fig. 5 Schematic of EF2S stiffness module.

6
defined by a pair of consecutive nodes 𝑝 𝑗 and 𝑝 𝑗+1 . This matrix is computed using a structural identification in the
linear elastic regime using data from the MSC Nastran linear static analysis solver SOL 101 [30].
The built-up FEM is subject to six independent static load cases—three tip forces and three tip moments—defining
the 6 × 1 tip load vector 𝐹tip . The three-dimensional displacement fields obtained for the six static load cases using MSC
Nastran SOL 101 are reduced to the beam reference axis using RBE3 interpolation elements. The one-dimensional
displacements of the beam reference axis nodes resolved in the 𝐵 frame, denoted by 𝑈 𝑗 , are the basis for identifying the
strain-based stiffness matrix associated to each beam element.
The one-dimensional displacements are first transformed from the 𝐵 frame to the local 𝑤 frames along the beam
reference axis to obtain the displacements 𝑢 𝑗 . For a given beam element, the local displacement and load are related by
𝑓 𝑗+1 = 𝐾 𝑗 Δ𝑢 𝑗+1 (5)
In Eq. (5), 𝑓 𝑗+1 is the load vector at the beam reference axis node 𝑝 𝑗+1 due to the applied tip load, Δ𝑢 𝑗+1 := 𝑢 𝑗+1 − 𝑢 𝑗
is the relative displacement vector from 𝑝 𝑗 to 𝑝 𝑗+1 , and 𝐾 𝑗 is the displacement-based stiffness matrix. The vector
equation (5) is resolved in the local 𝑤 frame associated with the current beam element and gives six scalar equations per
static load case, leading to 36 equations that allow one to solve for the elements of 𝐾 𝑗 .
Once 𝐾 𝑗 is known and symmetry is enforced to compensate for numerical noise, the strain-based stiffness matrix 𝑘 𝑗
is obtained using the method of Ref. [31] by solving the following equation for 𝑘 −1 𝑗 :

𝐾 −1 −1 −1 −1 −1
𝑗 𝑄 𝑗 = 𝑘 𝑗 𝐻𝑗 𝑄 𝑗 + 𝐸𝑗 𝑘 𝑗 (6)
The matrices 𝑄 𝑗 , 𝐸 𝑗 , and 𝐻 𝑗 in Eq. (6) are only functions of the beam element length defined by the positions 𝑝 𝑗 and
𝑝 𝑗+1 . The 6×6 flexibility matrix 𝑘 −1
𝑗 obtained by solving Eq. (6) is reduced by eliminating rows and columns associated
with the shear strains and is inverted to obtain the 4×4 stiffness matrix used in UM/NAST. The process is repeated for
each pair of beam reference axis nodes for obtaining the stiffness distributions of the model.

D. Pre-Pazy Low-Order Model


The EF2S process described in Sec. III.C is applied to the Pre-Pazy wing built-up FEMs shown in Fig. 2 for deriving
their equivalent UM/NAST beam models for the cases with and without skin. The UM/NAST models are verified by
comparing their frequencies with the frequencies of the built-up FEMs computed using MSC Nastran SOL 103.
The equivalent beam model accuracy depends on the choice of EF2S parameters such as the beam reference axis
location along the chord, the number of beam reference axis nodes, and the layout of the RBE3s used in the stiffness
identification. The values of these parameters for the Pre-Pazy wing are identified via a sensitivity study for the model
without skin, reported below, and are then used also for the model with skin.
The RBE3 layouts used in the sensitivity study are shown in Fig. 6 for the case of beam reference axis at 44% of the
chord. The layout in Fig. 6(a) uses 16 beam reference axis nodes connected to all the nodes of the corresponding rib.
The layouts in Fig. 6(b) and 6(c) use 16 and 6 beam reference axis nodes, respectively, connected to a subset of the rib
nodes. The beam reference axis location is varied between 40% and 60% of the chord. For each RBE3 layout and beam
reference axis location, the equivalent inertia distributions are computed from the lumped mass matrix data associated
with the built-up FEM nodes in the nearest neighbor of each beam reference axis node.
The beam model accuracy for varying beam reference axis location along the chord is shown in Table 2(a) for the
case of the 16-node RBE3s in Fig. 6(a). The OOP1 and OOP2 frequency errors are minimum, about 0.1–0.2%, when
the beam reference axis is at 44% of the chord. The OOP3 frequency error is minimum when the beam reference axis is
at 53% of the chord, but it remains below 0.6% when the reference axis is at 44% of the chord. The T1 frequency error
decreases from 3.2% to 2.4% as the beam reference axis moves toward the trailing edge. The IP1 frequency error is
insensitive to the beam reference axis location along the chord and remains fixed to 12.4%.
The beam model accuracy for different choices of the RBE3s is shown in Table 2(b). These results assume the
beam reference axis at 44% of the chord and consider 16 nodes. While the RBE3 layouts in Figs. 6(a) and 6(b) give
comparable frequency errors for the out-of-plane bending modes, the layout in Fig. 6(b) gives lower errors for the T1 and
IP1 modes. For the IP1 mode, the frequency error drops from 12.4% to 0.9% using the subset RBE3 layout in Fig. 6(b).
The beam model accuracy for varying number of reference axis nodes is shown in Table 2(c). These results assume
the beam reference axis at 44% of the chord and the RBE3 layouts in Figs. 6(b) and 6(c). The 6-node discretization
gives higher frequency errors for the bending modes, but it reduces the error for the frequency of the T1 mode to 1%.
Based on these results, the Pre-Pazy wing UM/NAST model for the case without skin is developed by choosing
the beam reference axis at 44% of the chord. The axial and out-of-plane/in-plane bending stiffness distributions and

7
(a) Full RBE3 layout, 16 nodes

(b) Subset RBE3 layout, 16 nodes (c) Subset RBE3 layout, 6 nodes

Fig. 6 RBE3 layouts used for identifying the Pre-Pazy wing equivalent stiffness distributions.

the associated coupling terms are computed using the 16-node subset RBE3 layout in Fig. 6(b). The torsion stiffness
distribution and the torsion–axial and torsion–bending coupling terms are computed using the 6-node subset RBE3
layout in Fig. 6(c) because this coarser discretization leads to lower errors in the T1 frequency as shown in Table 2(c).
The distributions obtained with the 6-node discretization are then interpolated to the 16-node discretization used for
obtaining the other distributions to assemble the complete stiffness model. The 16-node discretization is also used for
computing the inertia distributions based on the nearest neighbors associated to the beam reference axis nodes. The
Pre-Pazy wing UM/NAST model for the case with skin is developed using the same parameters. The equivalent inertia
and stiffness distributions of the UM/NAST models with and without skin used in this study are reported in the appendix.
For aeroelastic calculations, the beam models of the Pre-Pazy wing with and without skin are coupled to the
potential-flow thin airfoil theory of Peters et al. [14] augmented with tip loss factors. The tip loss factors are obtained
from static aeroelastic responses of the built-up FEM without skin as discussed in Sec. IV.D.

IV. Results
This section correlates results from modal and static analyses of the Pre-Pazy wing models in UM/NAST against
results from the built-up FEMs in Fig. 2 and experimental data from Avin et al. [26, 27]. These comparisons investigate
the accuracy of the Pre-Pazy wing UM/NAST models in capturing its frequencies and static response in preparation for
further correlations to be conducted for the Pazy wing with a focus on flutter boundaries.

A. Undeformed Modal Results


The frequencies of the Pre-Pazy wing with and without skin in undeformed configuration are reported in Table 3.
The UM/NAST frequencies are based on the equivalent beam models derived from the built-up FEMs as described in
Sec. III.D. These frequencies are correlated against the results from the built-up FEMs computed using MSC Nastran
SOL 103 and the GVT data by Avin et al. [26, 27]. In Table 3, the errors ΔTest associated with the MSC Nastran SOL
103 results are relative to the GVT data. The errors ΔFEM associated with the UM/NAST results are relative to MSC
Nastran SOL 103 to highlight the effect of the structural model reduction to a beam.
The frequency comparisons in Table 3 show that the built-up FEMs capture the Pre-Pazy wing frequencies of the
first three modes with errors from 3% to 5.2% for the case without skin and from 1.6% to 4% for the case with skin.
The UM/NAST models capture the frequencies of the built-up FEMs with errors below 1% for the first five modes. The
OOP1 frequency is captured with an error below 0.1%, the OOP2 and T1 frequencies with errors below 0.5%, and the
OOP3 and IP1 frequencies with errors below 1%.
These results show that the equivalent UM/NAST beam models of the Pre-Pazy wing with and without skin capture

8
Table 2 Sensitivity of the Pre-Pazy wing natural frequencies to the EF2S parameters (without skin).

(a) Varying beam reference axis location along the chord (full-rib RBE3 layout, 16 nodes)

SOL 103 UM/NAST, 0.44𝑐 UM/NAST, 0.53𝑐 UM/NAST, 0.62𝑐


# Mode type Freq. (Hz) Freq. (Hz) Δ (%) Freq. (Hz) Δ (%) Freq. (Hz) Δ (%)
1 OOP1 4.42 4.41 −0.25 4.40 −0.31 4.40 −0.45
2 OOP2 29.02 29.06 0.11 28.92 −0.37 28.60 −1.45
3 T1 41.53 40.17 −3.25 40.29 −2.97 40.51 −2.44
4 OOP3 83.34 83.80 0.56 83.20 −0.17 81.55 −2.15
5 IP1 112.56 98.63 −12.38 98.63 −12.38 98.63 −12.38

(b) Varying RBE3 layout (beam reference axis at 0.44𝑐, 16 nodes)

SOL 103 UM/NAST, full RBE3 UM/NAST, subset RBE3


# Mode type Freq. (Hz) Freq. (Hz) Δ (%) Freq. (Hz) Δ (%)
1 OOP1 4.42 4.41 −0.25 4.41 −0.10
2 OOP2 29.02 29.06 0.11 29.09 0.23
3 T1 41.53 40.17 −3.25 40.70 −1.99
4 OOP3 83.34 83.80 0.56 83.89 0.66
5 IP1 112.56 98.63 −12.38 113.53 0.86

(c) Varying number of beam reference axis nodes (beam reference axis at 0.44𝑐, subset RBE3 layout)

SOL 103 UM/NAST, 6 nodes UM/NAST, 16 nodes


# Mode type Freq. (Hz) Freq. (Hz) Δ (%) Freq. (Hz) Δ (%)
1 OOP1 4.42 4.48 1.34 4.41 −0.10
2 OOP2 29.02 30.26 4.27 29.09 0.23
3 T1 41.53 41.12 −0.97 40.70 −1.99
4 OOP3 83.34 85.01 2.01 83.89 0.66
5 IP1 112.56 115.35 2.48 113.53 0.86

9
Table 3 Pre-Pazy wing natural frequencies.

(a) Without skin

GVT SOL 103 UM/NAST


# Mode type Freq. (Hz) Freq. (Hz) ΔTest (%) Freq. (Hz) ΔFEM (%)
1 OOP1 4.20 4.42 5.17 4.42 −0.04
2 OOP2 28.10 29.02 3.29 29.11 0.28
3 T1 40.30 41.53 3.04 41.44 −0.20
4 OOP3 – 83.34 – 83.92 0.69
5 IP1 – 112.56 – 113.53 0.86

(b) With skin

GVT SOL 103 UM/NAST


# Mode type Freq. (Hz) Freq. (Hz) ΔTest (%) Freq. (Hz) ΔFEM (%)
1 OOP1 4.60 4.67 1.60 4.68 0.09
2 OOP2 29.50 30.68 4.00 30.82 0.46
3 T1 42.60 43.57 2.28 43.51 −0.13
4 OOP3 – 87.97 – 88.82 0.97
5 IP1 – 112.88 – 113.92 0.92

the first five natural frequencies in the undeformed configuration with practically the same accuracy as the built-up
FEMs while requiring a fraction of the DOFs. The Pre-Pazy wing models in UM/NAST consist of 15 elements (16
beam reference axis nodes) each having 4 strain DOFs resulting in 60 total DOFs. The built-up FEM without skin in
Fig. 2(a) consists of 3,050 nodes, resulting in 18,300 DOFs; and the built-up FEM with skin consists of 6,971 nodes,
resulting in 41,826 DOFs.

B. Static Structural Results


The Pre-Pazy wing was tested in horizontal configuration as shown in Fig. 1(a) by attaching masses at the tip half
chord (bending test) and at a point 0.08 m ahead of the tip leading edge (torsion test) [26]. The UM/NAST results for
the tip vertical displacement and twist are compared to the results from the MSC Nastran nonlinear structural solver
SOL 400 [32] and to the experimental data by Avin et al. [26]. The comparison to MSC Nastran SOL 400 is done
only for the case without skin because the built-up FEM with skin fails in nonlinear analyses due to convergence issues
caused by the skin elements.
The tip vertical displacements from the numerical models and the tests are shown in Figs. 7(a) and 7(b) (bending
test), and Figs. 7(c) and 7(d) (torsion test). The tip twist in the torsion test is shown in Figs. 7(e) and 7(f). The tip
vertical displacements are compared at the half chord and the twist is computed from the differential positions of the
tip leading and trailing edges. All the results in Fig. 7 are presented by eliminating the offsets due to gravity. The
correlation of numerical and experimental results for a 3-kg tip mass is shown in Table 4.
The tip vertical displacements from MSC Nastran SOL 400 capture the experimental data for the case without
skin with maximum errors below 3.4% (bending test) and 7.9% (torsion test). The displacement errors between the
UM/NAST results and MSC Nastran SOL 400 is below 0.2% in the 0–50% semispan deflection range, which verifies
the beam model accuracy for large deflections. The agreement between the UM/NAST and MSC Nastran SOL 400
results is also observed in the deformed configurations in Fig. 8, which include the gravity offsets.
Because MSC Nastran SOL 400 was not able to generate results for the case with skin, the UM/NAST tip vertical
displacements for this case are compared to the experimental data directly. The maximum errors are below 6.4%
(bending test) and 8.1% (torsion test). The UM/NAST models with and without skin underestimate the tip vertical
displacements from the tests by an error that increases with the tip mass. This can be the result of a constant beam

10
Table 4 Pre-Pazy wing tip vertical displacement and twist for a 3-kg tip mass.

(a) Without skin

Quantity Unit Case Test SOL 400 ΔTest (%) UM/NAST ΔFEM (%)
Vert. displ. % semispan Bending −49.10 −47.44 −3.39 −47.51 0.15
Vert. displ. % semispan Torsion −50.10 −46.16 −7.86 −46.20 0.10
Twist deg Torsion −10.66 −10.55 −1.02 −10.94 3.70

(b) With skin

Quantity Unit Case Test UM/NAST ΔTest (%)


Vert. displ. % semispan Bending −47.58 −44.55 −6.36
Vert. displ. % semispan Torsion −47.24 −43.39 −8.14
Twist deg Torsion −10.68 −10.52 −1.46

constitutive relation that does not account for internal stress. The errors are larger in the case with skin because the
beam model used in this paper does not account for the skin wrinkling that occurs in the tests [22].
The tip twist from UM/NAST captures well the experimental data with exception for the tip mass range 1–2.8 kg in
the case without skin. In that range, the experimental data shows irregularities in the measurements and a rigorous
comparison cannot be made. While the UM/NAST results for the tip twist show larger errors compared to MSC Nastran
SOL 400 than for the tip vertical displacements, the maximum error remains below 3.7%.
These results show that the equivalent UM/NAST model of the Pre-Pazy wing without skin captures the tip vertical
displacements and twist measured in the tests with practically the same accuracy as the relatively large, built-up FEM
over the 0–50% semispan deflection range. The UM/NAST model with skin shows a similar level of accuracy and does
not suffer from the convergence issues that prevent from using the corresponding built-up FEM in MSC Nastran SOL
400 nonlinear static analyses.

C. Deformed Modal Results


The Pre-Pazy wing beam model in UM/NAST is further assessed by comparing frequencies about statically deformed
configurations with MSC Nastran SOL 400 solutions for the case without skin. The deformed configurations are
obtained by applying a follower vertical tip force at the beam reference axis with no gravity.
Deformed configurations for selected follower tip force values are shown in Fig. 9(a) and the evolution of the first five
frequencies with the tip vertical displacement is shown in Fig. 9(b). The frequencies about the deformed configurations
in Fig. 9(a) are compared in Table 5. The frequency labels used in Fig. 9 and Table 5 are based on the mode shape
classification in the undeformed configuration. The OOP2 and T1 frequencies approach each other as the tip vertical
displacement increases until a mode switch occurs. A similar behavior is observed for the OOP3 and IP1 frequencies.
The OOP1 frequency predicted by MSC Nastran SOL 400 decreases with the tip vertical displacement while UM/NAST
predicts a practically constant value.
Table 5 shows that the errors in the OOP1, OOP2, T1, and IP1 frequencies predicted by UM/NAST increase with
the tip vertical displacement while the error in the OOP3 frequency remains about 0.5%. While the maximum errors in
the OOP2, T1, and IP1 frequencies remain below 3.5% in the 0–50% semispan deflection range, the error in the OOP1
frequency increases up to 28.9% because the UM/NAST model does not capture the frequency decrease predicted by
MSC Nastran SOL 400. The results in Fig. 9 and Table 5 show that the beam model of the Pre-Pazy wing captures the
frequencies of the statically deformed built-up FEM with errors below 4.6% up to tip vertical displacements of the order
20–25% semispan, and with larger errors in the OOP1 frequency for larger tip deflections up to 50% semispan.

D. Static Aeroelastic Results


The Pre-Pazy wing was tested in a wind tunnel at a root angle of attack 𝛼𝑟 = 5 and 7 deg in the flow speed range
𝑈 = 0 to 60 m/s. The tests were conducted in the vertical configuration as shown in Fig. 1(b). The UM/NAST results

11
0 0
UM/NAST
Vertical displ. (% semispan)

Vertical displ. (% semispan)


SOL 400
-10 Experimental data -10

-20 -20

-30 -30

-40 -40

-50 -50

-60 -60
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Tip mass (kg) Tip mass (kg)
(a) Bending test, without skin (b) Bending test, with skin

0 0
Vertical displ. (% semispan)

Vertical displ. (% semispan)


-10 -10

-20 -20

-30 -30

-40 -40

-50 -50

-60 -60
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Tip mass (kg) Tip mass (kg)
(c) Torsion test, without skin (d) Torsion test, with skin

0 0

-2 -2

-4 -4
Twist (deg)

Twist (deg)

-6 -6

-8 -8

-10 -10

-12 -12
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Tip mass (kg) Tip mass (kg)
(e) Torsion test, without skin (f) Torsion test, with skin

Fig. 7 Pre-Pazy wing tip vertical displacement and twist for varying tip mass (without gravity offset).

12
0 0
Vertical displ. (% semispan)

Vertical displ. (% semispan)


0.5 kg 0.5 kg
-10 -10

-20 -20
1 kg 1 kg
-30 -30

-40 2 kg -40 2 kg

-50 -50
UM/NAST 3 kg 3 kg
SOL 400
-60 -60
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise position (m) Spanwise position (m)
(a) Bending test, without skin (b) Torsion test, without skin

Fig. 8 Pre-Pazy wing deformed configurations for selected tip masses (with gravity offset).

60 125
UM/NAST IP1
25 N
Vertical displ. (% semispan)

SOL 400
50
100
Frequency (Hz)

40 OOP3
75
30
10 N 50 T1
20
5N
25 OOP2
10
2.5 N OOP1
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 10 20 30 40 50
Spanwise position (m) Vertical displ. (% semispan)
(a) Deformed configurations (b) Frequency evolutions with the tip vertical displacement

Fig. 9 Pre-Pazy wing natural frequencies for varying follower vertical tip force (without skin).

0.8
Weighting factor (-)

0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise position (m)

Fig. 10 Pre-Pazy wing aerodynamic load weighting factor at 𝛼 = 5 deg and 𝑈 = 0 m/s.

13
Table 5 Pre-Pazy wing natural frequencies for selected follower vertical tip forces (without skin).

(a) Tip force = 2.5 N , tip vertical displacement = 5.7% semispan (b) Tip force = 5 N, tip vertical displacement = 11.4% semispan

SOL 400 UM/NAST SOL 400 UM/NAST


# Mode type Freq. (Hz) Freq. (Hz) Δ (%) # Mode type Freq. (Hz) Freq. (Hz) Δ (%)
1 OOP1 4.41 4.42 0.23 1 OOP1 4.37 4.42 1.10
2 OOP2 28.99 29.08 0.31 2 OOP2 28.88 28.99 0.39
3 T1 40.77 40.65 −0.30 3 T1 38.52 38.26 −0.66
4 OOP3 83.25 83.83 0.70 4 IP1 82.12 81.07 −1.27
5 IP1 99.00 98.42 −0.59 5 OOP3 83.68 83.63 −0.05

(c) Tip force = 10 N, tip vertical displacement = 22.4% semispan (d) Tip force = 25 N, tip vertical displacement = 50.7% semispan

SOL 400 UM/NAST SOL 400 UM/NAST


# Mode type Freq. (Hz) Freq. (Hz) Δ (%) # Mode type Freq. (Hz) Freq. (Hz) Δ (%)
1 OOP1 4.23 4.43 4.60 1 OOP1 3.48 4.49 28.88
2 OOP2 28.44 28.65 0.75 2 T1 16.75 16.16 −3.51
3 T1 31.25 30.62 −2.04 3 OOP2 26.10 26.63 2.02
4 IP1 66.99 65.14 −2.76 4 IP1 59.29 57.25 −3.43
5 OOP3 82.27 82.72 0.55 5 OOP3 77.78 78.08 0.38

Table 6 Pre-Pazy wing tip loss factor coefficients.

Quantity Unit 𝛼𝑟 = 5 deg 𝛼𝑟 = 7 deg


𝜏0 – 6.48 6.39
𝜏1 s/m 6.86 × 10−3 1.05 × 10−2
𝜏2 s2 /m2 −3.57 × 10−4 −4.99 × 10−4

Table 7 Pre-Pazy wing tip vertical displacement in % semispan at 𝑈 = 50 m/s.

(a) Without skin

𝛼𝑟 (deg) SOL 400-VLM SOL 144 ΔFEM (%) UM/NAST ΔFEM (%)
5 31.09 36.53 17.51 31.07 0.66
7 39.32 51.15 30.07 39.51 0.48

(b) With skin

𝛼𝑟 (deg) Test SOL 144 ΔTest (%) UM/NAST ΔTest (%)


5 28.65 30.88 7.79 27.42 −4.30
7 39.97 43.23 8.17 35.19 −11.96

14
60 60
UM/NAST
Vertical displ. (% semispan)

Vertical displ. (% semispan)


SOL 144
50 SOL 400-VLM 50
Experimental data
40 40

30 30

20 20

10 10

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Speed (m/s) Speed (m/s)
(a) 𝛼𝑟 = 5 deg, without skin (b) 𝛼𝑟 = 7 deg, without skin

60 60
Vertical displ. (% semispan)

Vertical displ. (% semispan)

50 50

40 40

30 30

20 20

10 10

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Speed (m/s) Speed (m/s)
(c) 𝛼𝑟 = 5 deg, with skin (d) 𝛼𝑟 = 7 deg, with skin

Fig. 11 Pre-Pazy wing tip vertical displacement for varying flow speed.

15
60 60
UM/NAST 60 m/s
Vertical displ. (% semispan)

Vertical displ. (% semispan)


SOL 144
50 SOL 400-VLM 50
60 m/s

40 40 50 m/s

30 50 m/s 30
40 m/s
20 20
40 m/s
30 m/s
10 10
30 m/s

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise position (m) Spanwise position (m)
(a) 𝛼𝑟 = 5 deg, without skin (b) 𝛼𝑟 = 7 deg, without skin

60 60
Vertical displ. (% semispan)

Vertical displ. (% semispan)

50 50 60 m/s

60 m/s
40 40 50 m/s

30 30
50 m/s
20 20 40 m/s
40 m/s
10 10 30 m/s
30 m/s

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise position (m) Spanwise position (m)
(c) 𝛼𝑟 = 5 deg, with skin (d) 𝛼𝑟 = 7 deg, with skin

Fig. 12 Pre-Pazy wing deformed configurations for selected flow speeds.

16
for the case without skin are compared to solutions from a fully nonlinear static aeroelastic analysis of the built-up FEM
conducted using an in-house solver, which couples MSC Nastran SOL 400 to the vortex lattice method (VLM) [33].
The UM/NAST results for the case with skin are compared to the experimental data from the wind-tunnel tests by Avin
et al. [26, 27]. Linear static aeroelastic results for the built-up FEMs are also obtained using the MSC Nastran linear
static aeroelastic solver SOL 144 [7] to highlight the limit of validity of the small-displacement assumption. The linear
static aeroelastic analysis can be conducted for both the built-up FEMs with and without skin.
Because the potential-flow thin airfoil theory by Peters et al. [14] used in UM/NAST is a strip-theory-type
aerodynamic formulation, an aerodynamic load weighting factor 𝑊 (𝑠) is introduced to approximate three-dimensional
(tip) effects. The weighting factor is defined as

𝑊 (𝑠, 𝑈) = 1 − exp[1 − 𝑠𝜏(𝑈)/𝑙] (7)

where the tip loss factor 𝜏(𝑈) is assumed as a quadratic function of the flow speed, i.e.,

𝜏(𝑈) = 𝜏0 + 𝜏1𝑈 + 𝜏2𝑈 2 (8)

The coefficients 𝜏0 , 𝜏1 , and 𝜏2 in Eq. (8) are evaluated by minimizing the error between the tip vertical displacements
predicted by UM/NAST and the MSC Nastran SOL 400-VLM solver at the flow speeds 𝑈 = 10, 20, 40 m/s for each angle
of attack. The values obtained for the model without skin are shown in Table 6 and are also used for the model with skin
where MSC Nastran SOL 400-VLM reference solutions are unavailable due to convergence issues. The weighting factor
distribution in Eq. (7) obtained at 𝛼𝑟 = 5 deg and 𝑈 = 0 m/s with the coefficients in Table 6 is shown in Fig. 10.
The tip vertical displacements predicted by different numerical models and measured in the experimental tests are
summarized in Fig. 11. Displacements are compared at the half chord. Selected deformed configurations are shown in
Fig. 12 and the tip vertical displacement errors for the flow speed 𝑈 = 50 m/s are given in Table 7. The errors associated
with the results from SOL 144 and UM/NAST are relative to SOL 400-VLM in the case without skin and to the tests in
the case with skin.
The tip vertical displacements predicted by different numerical models overlap up to approximately 10% semispan
deflections. The MSC Nastran SOL 144 results overestimate the tip vertical displacements for larger deflections and do
not capture the wing shortening because they do not account for geometrically nonlinear effects. These limitations in
the linear model are also seen in the deformed configurations in Fig. 12 where the linear results are only shown for
𝑈 = 30, 40, 50 m/s. The UM/NAST results capture the SOL 400-VLM tip vertical displacements with errors below
0.7% over the flow speed range considered, as summarized in Table 7. These results show that the aerodynamic load
weighting factor in Eq. (7) with the quadratic law for the tip loss factor from Eq. (7) approximates three-dimensional
static aerodynamic effects while requiring only three MSC Nastran SOL 400-VLM solutions for tuning.
The UM/NAST tip vertical displacements for the case with skin capture the trend of the experimental results
with errors below 5% up to 30% semispan deflections. The maximum error compared to the experimental results
is approximately 12% in the flow speed range considered. The UM/NAST results underestimate the tip vertical
displacements from the tests, as observed in Sec. IV.B. One reason for this discrepancy is that the UM/NAST model
does not capture the skin wrinkling observed in the tests, which softens the wing. Another reason is that the tip loss
factor used in the UM/NAST model with skin is based SOL 400-VLM static aeroelastic solutions of the model without
skin, which experiences larger deflections for a given flow speed. Besides this, overall the low-order UM/NAST was
able to accurately capture the static response of the Pre-Pazy wing experiment.

V. Concluding Remarks
As part of the Third Aeroelastic Prediction Workshop (AePW3) activities, researchers at Technion developed the
Pre-Pazy wing, a preliminary experimental model for nonlinear aeroelastic tests in low-speed flow. This model was
conceived for maturing the techniques for testing the Pazy wing—the benchmark under study by the AePW3 Large
Deflection Group. This paper correlated a low-order, geometrically exact beam model with potential-flow aerodynamics
of the Pre-Pazy wing with results from a built-up FEM and experimental data. The Pre-Pazy wing low-order model
was developed in UM/NAST, a multidisciplinary framework for analyzing aircraft with very flexible wings. The
Pre-Pazy wing structure was modeled by deriving an equivalent beam representation of the available built-up FEM. The
aerodynamics was modeled using potential-flow thin airfoil theory with tip loss factors. The complete low-order static
aeroelastic models with and without skin consist of 60 DOFs, whereas the parent built-up FEMs consist of approximately
42,000 and 18,000 DOFs, respectively.

17
The low-order aeroelastic model of the Pre-Pazy wing was assessed in terms of natural frequencies and static
structural and aeroelastic response. The studies spanned tip vertical displacements in the 0–50% semispan deflections
range. The low-order model of the Pre-Pazy wing captured its static response and frequencies with accuracy comparable
to the built-up FEM while requiring much fewer DOFs. Additionally, the low-order model with skin could be run in
nonlinear analyses, while the built-up FEM with skin was found to suffer from convergence issues caused by the skin
shell elements. The errors in the structural and aeroelastic tip vertical displacements predicted by the low-order model
with respect to the experimental results were found to be below 12% in the 0–50% semispan deflection range, with
lower errors in the 0–30% semispan range expected in practical very flexible aircraft wing applications. These errors
were caused by the use of a constant constitutive relation, the lack of skin wrinkling effects, and the approximation of
three-dimensional aerodynamics using strip-theory with tip loss factors.
Future work under the frame of the AePW3 will assess the low-order aeroelastic modeling approach used in this
study with experimental data for the Pazy wing with a focus on flutter prediction. The low-order model will be refined
using a three-dimensional aerodynamic formulation to investigate the impact of structural and aerodynamic model
fidelity on the accuracy of aeroelastic predictions.

Acknowledgments
The authors thank Prof. Daniella Raveh and Arik Drachinsky (Technion—Israel Institute of Technology) for
proving the built-up FEMs of the Pre-Pazy wing used for developing the UM/NAST models used in this paper and the
experimental data.

Appendix: Pazy Wing Equivalent Inertia and Stiffness Distributions


This appendix reports the equivalent beam distributions of the Pre-Pazy wing built-up FEMs, shown in Fig. 2,
obtained using EF2S as discussed in Sec. III.D and implemented into the Pre-Pazy wing UM/NAST models used in this
paper. These distributions and the corresponding results reported in this paper are available for download from the website
of the AePW3 Large Deflection Group at https://nescacademy.nasa.gov/workshops/AePW3/public/wg/largedeflection.

Equivalent Inertia Distributions


The equivalent inertia distributions of the Pre-Pazy wing are given in the form of rigid-body elements at the beam
reference axis nodes (beam element ends). Each rigid-body element is defined by its mass, mass offsets with respect to
the corresponding beam reference axis node, and inertia tensor components about the local center of mass (given by the
beam reference axis node plus the mass offset).
The rigid-body mass and mass offsets distributions are shown in Fig. 13. The distributions of the rigid-body inertia
tensor components are shown in Fig. 14. All components are given in the UM/NAST body frame (origin at the beam
reference axis root, 𝑥 axis downstream, 𝑦 axis toward the right tip, and 𝑧 axis upward).

Equivalent Stiffness Distributions


The equivalent stiffness distributions of the Pre-Pazy wing are given in terms of the elements of the 4 × 4 stiffness
matrix of the geometrically exact, strain-based beam formulation reported in Ref. [5, 12].
The distributions of the diagonal stiffness matrix terms are shown in Fig. 15 and the distributions of the off-diagonal
stiffness matrix terms are shown in Fig. 16. All the stiffness matrix elements are given in the local beam reference axis
frame. While the distributions of the axial–out-of-plane-bending and out-of-plane–in-plane-bending coupling terms of
the model with skin present some numerical anomalies in the first and last elements, these anomalies do not influence
the UM/NAST results.

18
10-3
0.03 5
Without skin
With skin
0.025 2.5

Rigid-body x offset (m)


Rigid-body mass (kg)

0.02 0

0.015 -2.5

0.01 -5

0.005 -7.5

0 -10
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(a) Rigid-body mass 𝑚
ˆ𝑗 (b) Rigid-body mass offset 𝛿ˆ 𝑗 𝑥

10-3 10-4
10 0

8 -0.25
Rigid-body y offset (m)

Rigid-body z offset (m)

6 -0.5

4 -0.75

2 -1

0 -1.25

-2 -1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(c) Rigid-body mass offset 𝛿ˆ 𝑗 𝑦 (d) Rigid-body mass offset 𝛿ˆ 𝑗𝑧

Fig. 13 Pre-Pazy wing rigid-body mass and mass offset distributions.

19
10-6 10-7
3 3
Without skin
Rigid-body xx inertia (kg m2)

Rigid-body xy inertia (kg m2)


With skin
2.5 2

2 1

1.5 0

1 -1

0.5 -2

0 -3
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(a) Rigid-body inertia tensor component Î𝑗 𝑥 𝑥 (b) Rigid-body inertia tensor component Î𝑗 𝑥 𝑦

10-8 10-4
15 1.5
Rigid-body xz inertia (kg m2)

Rigid-body yy inertia (kg m2)

10 1

5 0.5

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(c) Rigid-body inertia tensor component Î𝑗 𝑥𝑧 (d) Rigid-body inertia tensor component Î𝑗 𝑦 𝑦

10-9 10-4
0 1.6
Rigid-body yz inertia (kg m2)

Rigid-body zz inertia (kg m2)

-2 1.2

-4 0.8

-6 0.4

-8 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(e) Rigid-body inertia tensor component Î𝑗 𝑦𝑧 (f) Rigid-body inertia tensor component Î𝑗𝑧𝑧

Fig. 14 Pre-Pazy wing rigid-body inertia tensor distributions about the local centers of mass.

20
106
12 24
Without skin
With skin
10 20

Torsion stiffness (N m2)


Axial stiffness (N)

8 16

6 12

4 8

2 4

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(a) Axial stiffness 𝑘11 (b) Torsion stiffness 𝑘22

6 3600
Out-of-plane bend. stiffness (N m2)

In-plane bend. stiffness (N m2)

5 3000

4 2400

3 1800

2 1200

1 600

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(c) Out-of-plane bending stiffness 𝑘33 (d) In-plane bending stiffness 𝑘44

Fig. 15 Pre-Pazy wing axial, torsion, and bending stiffness distributions.

21
Axial-out-of-plane-bend. coup. (N m)
10
Without skin 300
With skin
Axial-torsion coup. (N m)

8
250

6 200

4 150

2 100

0 50

-2 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(a) Axial–torsion coupling 𝑘12 (b) Axial–out-of-plane-bending coupling 𝑘13

Torsion-out-of-plane bend. coup. (N m2)


104
6 1.5
Axial-in-plane-bend. coup. (N m)

5 1

4 0.5

3 0

2 -0.5

1 -1

0 -1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(c) Axial–in-plane-bending coupling 𝑘14 (d) Torsion–out-of-plane bending coupling 𝑘23
Out-of-plane-in-plane bend. coup. (N m2)
Torsion-in-plane bend. coup. (N m2)

1 10

0.8 8

0.6 6

0.4 4

0.2 2

0 0

-0.2 -2
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Spanwise location (m) Spanwise location (m)
(e) Torsion–in-plane-bending coupling 𝑘24 (f) Out-of-plane–in-plane-bending coupling 𝑘34

Fig. 16 Pre-Pazy wing coupling stiffness distributions.

22
References
[1] “NASA Aeronautics Strategic Implementation Plan: 2019 Update,” , retrieved June 2020. URL https://www.nasa.gov/
aeroresearch/strategy.

[2] Patil, M. J., Hodges, D. H., and Cesnik, C. E. S., “Nonlinear Aeroelasticity and Flight Dynamics of High-Altitude Long-Endurance
Aircraft,” Journal of Aircraft, Vol. 38, No. 1, 2001, pp. 88–94. doi:10.2514/2.2738.

[3] Cesnik, C. E. S., and Su, W., “Nonlinear Aeroelastic Modeling and Analysis of Fully Flexible Aircraft,” 46th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, Austin, TX, 2005, pp. 1–27.
doi:10.2514/6.2005-2169, AIAA-2005-2169.

[4] Patil, M. J., Hodges, D. H., and Cesnik, C. E. S., “Limit-Cycle Oscillations in High-Aspect-Ratio Wings,” Journal of Fluids
and Structures, Vol. 15, No. 1, 2001, pp. 107–132. doi:10.1006/jfls.2000.0329.

[5] Su, W., and Cesnik, C. E. S., “Nonlinear Aeroelasticity of a Very Flexible Blended-Wing-Body Aircraft,” Journal of Aircraft,
Vol. 47, No. 5, 2010, pp. 1539–1553. doi:10.2514/1.47317.

[6] Afonso, F., Vale, J., Oliveira, E., Lau, F., and Suleman, A., “A Review on Non-Linear Aeroelasticity of High Aspect-Ratio
Wings,” Progress in Aerospace Sciences, Vol. 89, 2017, pp. 40–57. doi:10.1016/j.paerosci.2016.12.004.

[7] Anon., MSC Nastran Aeroelastic Analysis User’s Guide, 2017.

[8] Cesnik, C. E. S., Palacios, R., and Reichenbach, E. Y., “Reexamined Structural Design Procedures for Very Flexible Aircraft,”
Journal of Aircraft, Vol. 51, No. 5, 2014, pp. 1580–1591. doi:10.2514/1.C032464.

[9] Hodges, D. H., “A Mixed Variational Formulation based on Exact Intrinsic Equations for Dynamics of Moving Beams,”
International Journal of Solids and Structures, Vol. 26, No. 11, 1990, pp. 1253–1273. doi:10.1016/0020-7683(90)90060-9.

[10] Géradin, M., and Cardona, A., Flexible Multibody Dynamics: A Finite Element Approach, Wiley, 2001.

[11] Hodges, D. H., “Geometrically Exact, Intrinsic Theory for Dynamics of Curved and Twisted Anisotropic Beams,” AIAA Journal,
Vol. 41, No. 6, 2003, pp. 1131–1137. doi:10.2514/2.2054.

[12] Su, W., and Cesnik, C. E. S., “Strain-Based Geometrically Nonlinear Beam Formulation for Modeling Very Flexible Aircraft,”
International Journal of Solids and Structures, Vol. 48, No. 16-17, 2011, pp. 2349–2360. doi:10.1016/j.ijsolstr.2011.04.012.

[13] Theodorsen, T., “General Theory of Aerodynamic Instability and the Mechanism of Flutter,” Tech. rep., NACA Report 496,
1935.

[14] Peters, D. A., Hsieh, M. C. A., and Torrero, A., “A State-Space Airloads Theory for Flexible Airfoils,” Journal of the American
Helicopter Society, Vol. 52, No. 4, 2007, pp. 329–342. doi:10.4050/JAHS.52.329.

[15] Katz, J., and Plotkin, A., Low-Speed Aerodynamics, Mac Graw-Hill, 1991.

[16] Skujins, T., and Cesnik, C. E. S., “Reduced-Order Modeling of Unsteady Aerodynamics Across Multiple Mach Regimes,”
Journal of Aircraft, Vol. 51, No. 6, 2014, pp. 1681–1704. doi:10.2514/1.c032222.

[17] Mallik, W., Schetz, J. A., and Kapania, R. K., “Rapid Transonic Flutter Analysis for Aircraft Conceptual Design Applications,”
AIAA Journal, Vol. 56, No. 6, 2018, pp. 2389–2402. doi:10.2514/1.J056218.

[18] Opgenoord, M. M. J., Drela, M., and Willcox, K. E., “Physics-Based Low-Order Model for Transonic Flutter Prediction,” AIAA
Journal, Vol. 56, No. 4, 2018, pp. 1519–1531. doi:10.2514/1.J056710.

[19] Drela, M., “Integrated Simulation Model for Preliminary Aerodynamic, Structural, and Control-Law Design of Aircraft,” 40th
Structures, Structural Dynamics, and Materials Conference and Exhibit, St. Louis, MO, 1999, pp. 1–13. doi:10.2514/6.1999-1394,
AIAA-1999-1394.

[20] Cesnik, C. E. S., and Brown, E. L., “Modeling of High Aspect Ratio Active Flexible Wings for Roll Control,” 43rd
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Denver, CO, 2002, pp. 1–15.
doi:10.2514/6.2002-1719, AIAA-2002-1719.

[21] Patil, M. J., and Hodges, D. H., “Flight Dynamics of Highly Flexible Flying Wings,” Journal of Aircraft, Vol. 43, No. 6, 2006,
pp. 1790–1799. doi:10.2514/1.17640.

23
[22] Su, W., and Cesnik, C. E. S., “Dynamic Response of Highly Flexible Flying Wings,” AIAA Journal, Vol. 49, No. 2, 2011, pp.
324–339. doi:10.2514/1.J050496.

[23] Wang, Z., Chen, P. C., Liu, D. D., and Mook, D. T., “Nonlinear-Aerodynamics/Nonlinear-Structure Interaction Methodology for
a High-Altitude Long-Endurance Wing,” Journal of Aircraft, Vol. 47, No. 2, 2010, pp. 556–566. doi:10.2514/1.45694.

[24] del Carre, A., noz Simón, A. M., Goizueta, N., and Palacios, R., “SHARPy: A Dynamic Aeroelastic Simulation Toolbox for Very
Flexible Aircraft and Wind Turbines,” Journal of Open Source Software, Vol. 4, No. 44, 2019, p. 1885. doi:10.21105/joss.01885.

[25] Tang, D., and Dowell, E. H., “Experimental and Theoretical Study on Aeroelastic Response of High-Aspect-Ratio Wings,”
AIAA Journal, Vol. 39, 2001, pp. 1430–1441. doi:10.2514/2.1484.

[26] Avin, O., Drachinskyy, A., Ben-Shmuelz, Y., and Raveh, D. E., “Design of an Experimental Benchmark of a Highly Flexible
Wing,” 60th Israel Annual Conference on Aerospace Sciences (IACAS), Haifa, Israel, 2020, pp. 1–24.

[27] Avin, O., Raveh, D. E., Drachinskyy, A., Ben-Shmuelz, Y., and Tur, M., “An Experimental Benchmark of a Very Flexible
Wing,” 2021 AIAA SciTech Forum, Virtual Event, 2021, pp. 1–32.

[28] Anon., MSC Nastran Dynamic Analysis User’s Guide, 2017.

[29] Riso, C., Sanghi, D., Cesnik, C. E. S., Vetrano, F., and Teufel, P., “Parametric Nonlinear Aeroelastic Analysis of a High-Aspect-
Ratio-Wing Civil Transport Aircraft,” 2020 AIAA SciTech Forum, Orlando, FL, 2020, pp. 1–23. doi:10.2514/6.2020-1191,
AIAA-2020-1191.

[30] Anon., MSC Nastran Linear Static Analysis User’s Guide, 2017.

[31] Malcolm, D. J., and Laird, D. L., “Extraction of Equivalent Beam Properties from Blade Models,” Wind Energy, Vol. 10, No. 2,
2007, pp. 135–157. doi:10.1002/we.213.

[32] Anon., MSC Nastran Non Linear User’s Guide SOL400, 2016.

[33] Riso, C., Di Vincenzo, F. G., Ritter, M., Cesnik, C. E. S., and Mastroddi, F., “Nonlinear Aeroelastic Trim of Very Flexible
Aircraft Described by Detailed Models,” Journal of Aircraft, Vol. 55, No. 6, 2018, pp. 2338–2346. doi:10.2514/1.C034787.

24

View publication stats

You might also like