Proceedings of the IASTED International Conference

Modelling, Identification and Control (MIC 2017)

February 20 - 21, 2017 Innsbruck, Austria

EXPERIMENTAL MODAL ANALYSIS OF A SMALL UNMANNED AERIAL

VEHICLE WITH VARIABLE ASPECT RATIO WING

Razvan Rusovici, Ryan Haughey

Florida Institute of Technology
150 W. University Blvd, Melbourne, FL 32901, U.S.A.
[email protected], [email protected]

Jordan Provost
Ecole de l’Air
Salon de Provence, France
[email protected]

ABSTRACT conditions, usually during landing or takeoff, while a high

The main purpose of this paper is to present the natural aspect ratio is desirable during cruise. One such
frequencies and mode shapes of a small unmanned aerial configuration, the telescoping wing, has been adopted by
vehicle (sUAV), designed and test-flown at the Florida Raytheon Missile Systems on cruise missiles.
Institute of Technology obtained via experimental modal
analysis (EMA). The EMA was performed in order to 𝑏2
𝐴𝑅 = (1)
understand the UAV's modal characteristics as a first step 𝑆
aimed towards understanding the aeroelastic characteristics
of the aircraft. The Variable Aspect Ratio (VAR) Morphing was defined as the ability of a structure
unmanned aerial vehicle (UAV) studied employs a to change shape and wing morphing was identified as a
telescoping wing which, by adaptively-changing its span method to allow optimal performance in various flight
can evolve optimally in all flight regimes. Data acquisition conditions without the need of employing complicated
and analysis were performed with ModalVIEW. The modal secondary control surfaces such as slats or flaps [2]. The
tests employed hammer excitation and accelerometer Wright Brothers employed morphing wings as early as
response measurements. In particular, the study focused on 1903 with their Wright Flyer. While most 20th century
the difference between modal characteristics of two aircraft were built with fixed wing configurations, the
different VAR wing configurations: fully retracted vs. fully current decade has brought new research into the concept
extended wing. of an active, morphing, aeroelastic wing. The Boeing X-53,
for example, employed an Active Aeroelastic Wing
KEY WORDS (AAW) [3] whose aeroelastic flexibility allowed the
Experimental Modal Analysis, UAV optimization of aircraft performance in all flight regimes.
The principle behind the AAW was to build a wing that
could be used itself as a control surface. Several morphing
wing designs [4] are currently proposed and implemented;
1. Introduction perhaps the simplest one is the telescopic wing concept.
The telescoping wing is capable of modifying its
Unmanned aerial vehicles (UAV) or systems (UAS) are span during flight and thus increase its aspect ratio. The
used in many civilian and military applications [1]. Often extension direction is perpendicular to the chord and the
time, these UAS are required to perform missions that telescoping wing is composed of a fixed center section and
require short-field take-off and landing, rapid cruise two extendable outer sections, one for the port and one for
airspeed and maneuvering at fast airspeeds. These the starboard wings. An overlapping spar system is
performance characteristics often are best performed with responsible for extending the wing.
wing configurations that have different aspect ratios [1]. A morphing wing aircraft was designed,
The aspect ratio (AR) [1], one of the most manufactured and successfully flight tested at the Florida
important parameters in aircraft design and is a function of Institute of Technology (Figure 1). The mission of the
the wing span b and wing area S (Equation 1). A change in aircraft was to perform low level photography missions, as
the aspect ratio leads to a change in the performance of the well as to be a technology demonstrator for the telescoping
aircraft (i.e. lift). A high aspect ratio aspect ratio would
allow an aircraft to perform better in low airspeed

DOI: 10.2316/P.2017.848-057 94
wing concept and an educational tool for the senior design 2.1 Brief Overview of Experimental Modal Analysis
class. The plane is shown in Flight in Figure 2.
A mode shape or a normal mode of a vibrating system is a
pattern of motion in which all parts of the system move
sinusoidally with the same frequency (natural frequency
associated with the mode) and with a fixed phase relation.
An object has, in theory, an infinite number of mode shapes
and natural frequencies. If a structure is excited by a
periodic force whose frequency matches any of the natural
frequencies of the structure to which it is applied, a
resonance condition occurs. At resonance, the structure is
able to admit an unlimited amount of mechanical energy,
which is why (in the absence of damping), the structure
undergoes increasingly larger deformations and ultimately
failure.

2.2 Experimental Modal Analysis Software

ModalVIEWTM software was developed with NI

LabVIEWTM, an open environment designed to make
Figure 1VAR UAV and Data Acquisition System interfacing with any measurement hardware fast and
simple. Developed by ABSignal, ModalVIEW was
designed to acquire multichannel sound or vibration signals
from a running machine or to obtain the static or dynamic
loading of a structure by using data acquisition hardware.
ModalVIEW allows the user to perform an experimental
modal analysis. After acquiring a set of acceleration,
velocity or displacement time histories of certain points on
a structure generated by a time-varying excitation signal
(force, acceleration or velocity, etc.), and the software can
animate the response of a structure and show the structure's
vibration at the measurement points. It also helps extract
and visualize useful modal parameter information from
acquired time-and frequency-domain experimental data.
This software uses the Fast Fourier Transform (FFT)
algorithm [7] which computes a sampled version of the
frequency spectrum from sampled time signal. The
Figure 2 VAR UAV in Flight discretized and finite-length spectrum is called Discrete
Fourier Transform (DFT). The software produces the
One research goal was to investigate the
frequency response functions from the spectra of the
aeroelastic [5] characteristics of the wing (i.e. flutter
response and excitation produced at each point on the
speed), which could not be obtained without a preliminary
structure under test. The software then performs a curve-
modal analysis needed to identify the torsional and bending
fitting algorithm to extract the natural frequencies,
modes of the wing and horizontal stabilizer. Modal
damping ratios and mode shapes from the FRF.
characteristics of wings are obtained using experimental
Theoretically, the FRF describes the input/output
modal analysis (EMA) [6]. It was thus needed to find the
relationship between two points on a structure as in the
modal frequencies for the fully extended wing
frequency domain. In other words, the FRF is a measure of
configuration first, as it would be expected that this
how much displacement, velocity, or acceleration response
configuration would be more prone to flutter. Two studies
a structure has at an output point, per unit of excitation
were conducted: aircraft on ground and aircraft suspended
force at an input point. The Frequency Response Function
(in flight). The study presented in the current document
investigated and compared the modal characteristics of the (FRF) H(j ) is generally computed by
VAR UAV with extended wing in the two situations. 𝑋(𝑗𝜔)
𝐻(𝑗𝜔) = (2)
𝐹(𝑗𝜔)
where X is the autopower spectrum of the response (in this
2. Experimental Modal Analysis case displacement) and F is the autopower spectrum of the
excitation, with  being the circular frequency and j the
The methods and procedures used to conduct the series of imaginary number. The spectra X and F are a result of the
modal analyses are presented in this section [6]. Fourier transform applied to the dynamic equations of

95
motion of the structure, which performs a transition of the had to have the correct spatial resolution. The model
equation from the time domain (t is time) to the frequency consisted initially of 36 points (suspended configuration)
domain. The equations of motion are: and then was modified to include 72 measurement points
(Figure 4) for the on-ground configuration due to spatial
𝑀{𝑥̈ (𝑡)} + [𝐾]{𝑥(𝑡)} + [𝐶]{𝑥̇ (𝑡)} = {𝐹(𝑡)} (3) resolution needs dictated by the “harder” boundary
condition and ensuing coupling.
where [M], [K], [C] are the mass, stiffness and damping
matrices, respectively; {𝑥̈ }, {𝑥̇ }, {x} are the time-
dependent, acceleration, velocity and displacement,
respectively; {F} is the force vector.

2.3 Methods

The methods used to perform the EMA was the roving

response-fixed impact method. The response was measured
by triaxial accelerometers, PCB Piezotronics Model
Y356A16 with 100 mV/g nominal sensitivity for each of
the three axes (1g=9.8 m/s2). The excitation was provided
by an instrumented (force gauge) PCB Piezotronics
080C01 impact hammer with a nominal 12.3 mV/N
sensitivity (Figure 3).
Impact testing is relatively fast, convenient and low
cost; in this experiment one single driving impact point was
chosen. This particular point was situated be situated at a
location that would allow excitation of the main modes of
interest, which must be away from a vibrational node. One
wing, in its most simple approximation, could be modeled
as a cantilevered beam. The highest vibration displacement Figure 3 Data Acquisition System with Accelerometers and
on a single degree of freedom, cantilevered beam is located Impact Hammer on Wing
at the end of the beam: the wingtip would have the highest
vibrational displacement as the wing is similar to a
cantilevered beam in a first approximation. In the modal 2.3.2 Measurements
analysis experiment the middle of the wingtip had been
chosen as the impact point for each measurement. The measurement sessions were performed for the two
Accelerometers situated at various points on the structure configurations: aircraft resting on the ground and aircraft
recorded the accelerations caused by the hammer impact. suspended by the wings [8]; the same acquisition
FRFs are computed one at a time after each impact. For parameters were used for both. After the hammer impact
each measured point, seven impacts (averages) are [9] and response measurement, ModalVIEW automatically
performed for accuracy and good coherence. Modal generated a FRF for each channel from the response that
parameters are defined by curve fitting the resulting set of was measured. The acquisition were performed using the
FRFs. The impact hammer was connected to the first following parameters:
channel in the NI DAQ 3174 data acquisition card. The
accelerometers (two were used simultaneously) were  Sampling rate: 1652 Hz
connected to the rest of the available channels. In total, 7  Resolution: 1 Hz
data acquisition channels were used.  Block size / number of points: 1652

2.3.1 Base Model The hammer impact technique needed to impart energy
into exciting the modes of interest. The response at each
From a testing point of view, a real structure can be DOF was acquired 7 times (corresponding to 7 impacts)
sampled spatially at as many DOFs as needed. For practical and linearly averaged. Each impact had to be carefully
reasons, the UAV has been represented with a finite made in order to avoid double-hammering and
number points; at each point the accelerations along three overloading. The coherence, a function that indicated
DOFs (X, Y and Z) were made. The accelerometer whether the response of the structure was the result of the
measurements were made at each one of the DOF. The hammer impact, was for all FRF’s close to a value of 1
finite set of DOFs is called the base model. The locations (except at the resonance peaks), which indicated that the
of the DOF were measured and input into the software. The measurements were performed correctly and repeatedly
retracted and the extended wing configuration were and that no extraneous noise influenced the measurements.
represented via a separate base model. The base model was
used to visualize the mode shapes of interest and a result

96
 the Least-Square Complex Function (LSCF) [5] and a
model order using 40 terms. The mode was positively
identified when the natural frequency, mode shape and the
corresponding damping were determined; on the
stabilization chart a stable mode was denoted by a green
“s” symbol. The natural frequency and mode shape list is
shown on the right side of Figure 7.

Figure 4 Extended Configuration Base Model on Ground

The choice of the sampling rates came from the Nyquist-

Shannon sampling, which states that the sampling
frequency needs to be at least twice the highest frequency
of interest. The lower modes usually have the highest
participation factors and are usually the ones of interest.
Without prior knowledge of the modes of interest, a
suitably high frequency sampling rate was chosen. Figure Figure 6 Sample FRF obtained (Magnitude and Phase vs.
5 shows the aircraft right before the modal analysis. Frequency)

Figure 7 FRF Curve-Fit and Stabilization Chart

The mode shapes extracted were checked using

the Modal Assurance Criterion (MAC). The Modal
Assurance Criterion is a widespread tools for the
quantitative comparison of two modal vectors. This
Figure 5 Preparation of the Aircraft for Modal Analysis statistical indicator is used to measure the differences
between the modes shape, in conjunction with a separate
3. Results frequency analysis. The MAC gives a solid statistical
indicator and a degree of consistency between mode
The results obtained are presented. Figure 6 presents a shapes. Finally, a frequency comparison enables to
sample of the FRF obtained. The FRF phase changed from determine the correlated mode pairs. The MAC is bounded
+1800 to -1800 at each natural frequency. The magnitude between 0 and 1: values close to 0 indicate a non-consistent
FRF plot displayed local peaks at each natural frequency. mode and a value around 1 indicates a fully consistent
From the plots, it was evident that careful curve-fitting of mode shape. A value of 1 generally signifies that the two
the FRF and mode shape extraction had to be made over mode shapes compared are the same, especially if these are
portions of the entire frequency range. Figure 7 shows a found very close together in frequency. While in an
sample FRF plot curve-fitted using the mode shapes found analytic modal analysis (i.e. finite element method) is easy
using a stabilization chart. The stabilization chart to make the distinction between the characteristics of each
employed by the code to determine the mode shapes used separate mode, experimentally this task becomes much

97
harder. Due to imperfect spatial resolution, frequency
discretization or imperfect choice of mode extraction
algorithm, it is hard to distinguish between the same mode
estimated twice and different modes found at close to one
another in the frequency domain. That is why it is useful to
develop a quantitative method needed to assess the
orthogonality of estimated modes. The MAC:
 Assures uniqueness and orthogonality of the
experimental modes.
 May be used to compare an experimentally-
found mode to an analytically-found one. Figure 10 First Bending mode of Wings at 10.7 Hz
The MAC plot seen in Figure 8 was generated by the
software. From the plot in Figure 8 it was observed that in
general the correlation between two adjacent modes was
much less than1: thus the modes were independent of one The suspended modes of the aircraft should be the most
another. A summary of the list of modes found is presented relevant ones to an aeroelastic analysis; unfortunately any
in Table 1. Further Figures will depict the actual mode suspension mechanism introduces a change in the
shapes. boundary conditions, no matter how compliant the
suspension method may be (see ropes used to suspend the
VAR in Figure 1). However, it was interesting to see how
the same modes would be affected by the different
boundary condition (landing gear on ground). The more
shapes corresponding to an aircraft placed on ground
would be more difficult to analyze due to the coupling
introduced by the landing gear. The advantage of having
the aircraft placed on the ground would be that the tail and
wings would be free. Figure 12 shows the first bending
mode of the wing on the ground. Some tail empennage
bending was also involved in the mode shape.

Figure 8 List of Modes and the Modal Assurance Criterion Plot

Table 1 Mode Shape Comparison for Aircraft On-Ground vs.

Suspended with Extended Configuration

Mode Suspended ground

Figure 11 First Torsion around fuselage of wing and tail at 8.5
1st bending 10.7 Hz 7.8 Hz Hz

Antisymmetric/ 8.5 Hz 13.8 Hz

torsion

Figure 10 shows a mode shape at 10.7 Hz which

corresponds to a symmetric flapping of wing extensions
combined with tail bending (arrows indicate deflection
from initial position). An antisymmetric bending coupled
with stabilizer torsion was observed at 8.5 Hz. Figure 12 First Bending mode of Wings on Ground at 7.8 Hz

98
Figure 13 shows the torsion/ antisymmetric wing bending
at 13.8 Hz

Figure 15 First Torsion around fuselage of empennage at 5 Hz

on Ground

4. Conclusion
Figure 13 First Torsion around Fuselage of Wing and
Empennage at 13.8 Hz on Ground The results have shown that the natural mode frequencies
and mode shapes differ upon the boundary condition:
While the first bending (in phase bending of the wing aircraft grounded vs. suspended (mimicking the free flight
extensions) combined with empennage bending in the condition). The modal analysis provided a reference set of
vertical plane was actually higher in the suspended modes and aeroelastic investigations would follow to
condition than on the ground condition, one probable determine flutter speed. More investigations need to be
reason was that the wing was suspended at the edge of the conducted in order to compare similar modes in different
wing extensions. The antisymmetric tail and wing torsion boundary condition: for example the aircraft could be
mode was higher in frequency when the aircraft was on the suspended by tail and nose. A comparison must be
ground, which would be expected since the landing gear conducted in order to determine as well as possible the
provides additional stiffness. modes that may be first excited. Flutter could occur first on
the empennage or on the wings extensions due to their less
At 27 Hz, Figure 14, there was an additional mode stiff attachments to the wings.
shape corresponding to wing extension torsion around the
lateral (pitch) axis. This mode would be very important for
aeroelastic investigations along with the bending modes.
Acknowledgement(s)
This work was performed in part through funding provided
by Northrop Grumman Corporation.

References
[1] J.D. Anderson Introduction to Flight (McGraw-Hill
Education 2015).
[2] S. Barbarino, O. Bilgen, A.M. Rajaj, Friswell, M., & D.
Inman, A review of morphing aircraft, Journal of
Figure 14 First Wing Torsion around Lateral Pitch Axis at 27 Hz Intelligent Material Systems and Structures, 22(9), 2011,
on Ground 823-855.
[3] E. Pendleton, P. Flick, D. Voracek, E. Reichenbach, K.
This mode was found in the on-ground configuration,
Griffin, D. Paul, “The X-53 A Summary of the Active
which allowed the wing to be free. Aeroelastic Wing Flight Research Program,” Paper 07-
Yet another mode of interest was the torsion of the 1855, Proceedings of the 48th AIAA Structures, Structural
empennage around the long thin tail. This occurred at 5 Hz Dynamics, and Materials Conference, Honolulu, Hawaii,
and although some motion of the wing extensions was April 23–26, 2007.
[4] A. Guenther, F. Rossi, L.B. Machado, Rusovici, R.,
found, the predominant motion in the mode was that of the
“Roll Control of Aircraft through Active Winglets”, ICAST
empennage (Figure 15).
2016 Proceedings, Lake George, NY, October 2016.
[5] R.L. Bisplinghoff, H. Ashley, R.L. Halfman
Aeroelasticity (Dover, 1955).
[6] J.M.M. Silva and N.M.M. Maia, editors, Modal
analysis and testing (NATO Series, Series E, Applied
Sciences Volume 383: Springer Science+Business
Dordrecht, 1999).

99
[7] J.W. Tedesco, W.G. McDougal, C.A. Ross, Structural
dynamics (Pearson 1999).
[8] T. G. Woehrle. W. Costerus, C.L. Lee, Modal Analysis
of Pathfinder Unmanned Aerial Vehicle, Proceedings of
the 13th International Modal Analysis Conference, Paper
13-45-1, 2015, 1687-1693.
[9] S. Pedreira, D. Zuñiga, J. de Lima, W. de Alencar, et
al., "Impact Modal Analysis of UAV Vector-P's Wing
Using LabVIEW," SAE Technical Paper 2013-36-0460,
2013.

100