AIAA SciTech Forum 10.2514/6.

2019-2132
7-11 January 2019, San Diego, California
AIAA Scitech 2019 Forum

Transient CFD/CSD Tiltrotor Stability Analysis

Ethan Corle∗ and Matthew Floros†

US Army Research Laboratory, Aberdeen Proving Ground, MD, 21005, USA

Sven Schmitz‡
The Pennsylvania State University, University Park, PA, 16802, USA

This work is part of a research program to improve the efficiency, stability, and productiv-
ity of tiltrotor aircraft in order to meet the performance and sustainment goals of the Future
Vertical Lift program. The XV-15 is used as a testbed to investigate the impact of rotor aerody-
namics on whirl-flutter stability. Improvements to the state-of-the-art in prediction capability
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

of tiltrotor aeroelastic stability through the use of time-domain coupled computational fluid
dynamics (CFD) and computational structural dynamics simulations are presented. The Ro-
torcraft Comprehensive Analysis System (RCAS) is coupled to the rotorcraft CFD suite, Helios.
Comparisons are made between linearized stability analysis, transient stability analysis using
RCAS and transient stability analysis using CFD/CSD. For transient RCAS analysis, an algo-
rithm based on the log decrement method is then used to determine the system stability. For
transient CFD/CSD analysis, Prony’s method is used to determine modal frequency and damp-
ing. Good agreement is shown between the three methods of predicting tiltrotor whirl-flutter
stability at speeds less than 250 knots. In a high speed case at 377 knots, large differences
between the predicted values are displayed between CFD/CSD analysis and comprehensive
analysis.

I. Nomenclature
fˆ Curve fit response
f¯ Mean response
fi Peak response values
v Airspeed, kts
∆z Z-displacement of wing tip, ft
ωn Natural frequency
ωd Damped natural frequency
φ Phase angle
ζ Damping ratio

II. Introduction
For tiltrotor aircraft, aeroelasticity considerations have a much more significant effect on size, weight, and power
relative to runway-based aircraft. The elastic motion of the large proprotors can couple with wing bending motions and
becomes unstable at a critical airspeed. To achieve the desired airspeed, a thick wing is required to increase torsional
stiffness and the 3-blade gimbal hub design is largely unchanged in forty years from the XV-15 demonstrator aircraft
through the V-280 "Valor" aircraft being built as part of the Joint Multirole Technology Demonstrator program. These
design compromises, while necessary for stability, result in additional aerodynamic drag and weight which could be used
to increase payload, range, or speed. Furthermore, increased system stiffness adds to fatigue and creates maintenance
and reliability concerns. Technologies such as active winglets [1], wing extensions [2]–[3], and structural tailoring [4]
have been proposed which allow a thinner, more efficient wing, but the risk associated with wind-tunnel or flight testing
these technologies has been too high to make them viable to manufacturers. The ability to research technologies in a
∗ PhD Candidate, Penn State Department of Aerospace Engineering, Student AIAA Member.
† Aerospace Engineer, Vehicle Technology Directorate, Senior AIAA Member.
‡ Associate Professor, Department of Aerospace Engineering, Senior AIAA Member.

Copyright © 2019 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
 computational environment instead reduces the risk of promising technologies by allowing them to be evaluated safely
and at much lower cost.
Unfortunately, current computational predictions of tiltrotor stability have mixed success when compared to
experimental data. Kang et al. [5] performed comparison studies between predictions from two comprehensive analyses,
RCAS and Dymore, and experimental results for a number of configurations of the WRATS rotor. While the physics in
some configurations, such as changes in control stiffness, are well captured, other configurations, such as changes in
blade pitch-flap coupling, δ3 , show a very poor correlation of critical whirl-flutter speed.
Different comprehensive analysis tools handle the calculation of stability analysis in different ways. Previously, many
computations for tiltrotor aeroelasticity have used simplified aerodynamics models and modal representations of the
aircraft wing structure which has no aerodynamics at all. Only the proprotor blades and hub were represented directly
by physics-based structural models with coupled aerodynamics. Possible deficiencies in current prediction capabilities
include low-fidelity lifting-line aerodynamic models that are a staple in comprehensive analysis, model simplification
through linearization, uncertainty of wind-tunnel model structural properties, and a modal representation of the wing.
The current effort seeks to address three of these deficiencies. First, aerodynamic loads are calculated using
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

computational fluid dynamics, CFD, and both the wing and rotor are represented using multi-body dynamics and
finite-element models. While still taking advantage of RCAS as a rotorcraft dynamics analysis tool, now the airloads
include physics-based, nonlinear, and interactional aerodynamic effects. Linearized stability analysis could be performed
using the periodic steady-state solution with CFD airloads, however in this work; tight coupling is implemented to view
the nonlinear, transient response of the system following a perturbation to achieve a full dynamic response. In tight
coupling, data are passed between CFD and CSD solvers at every time step without a relaxation factor on airloads.
This is in contrast to the loose coupling methodology in which data between CFD and CSD solvers are passed after a
large azimuthal sweep and airloads are not directly applied to the structural model. The first demonstration of using a
CFD/CSD coupled model of a tiltrotor was presented in [6]. This work begins to identify the impact of both higher
fidelity airloads and structural nonlinearities on whirl flutter predictions.
The effort leverages DoD investments in high performance computing in the Computational Research and Engineering
Acquisition Tools and Environments (CREATE) program. In the Air Vehicles branch of CREATE, a software program,
Helios, has been developed specifically toward multi-physics analysis for rotary-wing aircraft. CREATE developers
have added the capability of modeling general elastic structures within the Helios framework. Leveraging this new
capability in Helios, two high-fidelity CFD packages, FUN3D and SAMCART, are coupled to the well-validated
rotorcraft comprehensive analysis tool, RCAS, to make time-domain tiltrotor stability predictions.

III. Numerical Methods

A. Tiltrotor Model (XV-15)

The model used in this work is a semi-span wing and rotor system based on the XV-15. Table 1 shows many of the
relevant geometric quantities used. One challenging aspect of using the XV-15 is the variation of the vehicle design
through its life cycle. Two XV-15 airframes exist, one operated by Bell Helicopter and one operated by NASA. On
these, different combinations of rotor hubs and blades have been tested [7]. In addition, early wind tunnel work uses a
different wing model than was used in the flight test vehicle. Stability data do exist for the XV-15, but are primarily for
airspeeds less than 250 knots, and there is significant variation in calculated damping across the tests [8]. With this
much uncertainty, validating a model is not possible. However, the focus of this work is to use a realistic tiltrotor model
to demonstrate the transient analysis method, so comparisons are shown to work with a previously used CAMRAD II
XV-15 model [9] to establish that the model used is a realistic design.

B. Computational Structural Dynamics - RCAS

RCAS is a program formulated using finite-element based, flexible, multi-body dynamics. It allows for the
construction of arbitrarily complex models of rotor systems from a library of primitive elements, including nonlinear
beams, rigid-body masses, rigid bars, springs, dampers, hinges, slides, and mechanically applied loads. RCAS
equilibrium trim and periodic solution processes are based on the solution of the nonlinear system of equations at a
series of time steps. The Newmark-Beta time integration approach, extended by the Hilber-Hughes-Taylor method,
is used for integration. Within each time step, the iterative Newton-Raphson method is used to achieve a converged
solution of the system equations.

2
 Table 1 XV-15 Model Parameters

Property Value
Rotor Radius 150 in
Precone 1.5 deg
δ3 -15 deg
Wing Semispan 193 in
Wing Chord 63 in
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

Fig. 1 XV-15 control connections for a Blade-1 in RCAS.

The aerodynamic model in RCAS is made up of 10 aerodynamic segments for the wing and 14 for the rotor with
linear coefficients for airfoil properties. When RCAS is used without CFD, uniform inflow is used to calculate induced
velocity for the rotor. The wing does not use an inflow model. The wing structure is constructed using 10 nonlinear
beam elements. From the tip of the wing, a distributed-mass, rigid beam, which is representative of the pylon, is used to
connect to the rotor gimbal and hub. Each blade is modeled using 20 nonlinear beam elements with structural properties
that are adapted from a legacy CAMRAD II model [9].
The control system is derived from the CAMRAD II model described in [9]. The actual XV-15 aircraft has an
unconventional control system compared to most helicopters with a number of linkages which are not modeled explicitly.
In the XV-15 hub, the pitch links are inside the spinner above the rotor plane rather than below it as they would be
in most rotary-wing aircraft. This is represented in the CAMRAD II model from [9] with the swashplate above the
rotor and is modeled similarly in the RCAS model in this effort. The control system consists of a controlled slider for
collective pitch and controlled hinges for cyclic pitch. These components are collocated above the rotor disk along
the shaft axis. Rigid bars are used to represent a swash plate which connects the controls to the pitch links which are
modeled as high stiffness springs to replicate control system frequencies of the actual aircraft. Finally, a rigid bar is
used as a pitch horn connecting the pitch link to the blade, outboard of a pitch bearing. This configuration is shown in
Fig. 1 for a single blade.
After a baseline model was established, discrepancies between predicted modal frequencies and previous results
were seen. Integrated mass for the wing, blades, gimbal, and pylon were tuned to match documented values. Finally,
natural frequencies of rotor modes and coupled system modes were achieved through scaling of existing distributions

3
 of stiffness values. Notably, the blade lag stiffness needed to achieve the desired frequency is roughly half of the
documented value. Finally, flap-lag stiffness is used to tune the progressive and regressive modes. A fan plot showing
the net result is presented in the results.
At high speeds, converging on a trim solution is challenging, particularly as the model approaches an aeroelastic
instability. Previous work [6] leveraged the ‘trim spring’ feature in RCAS to stabilize the model in trim while not
influencing the dynamic response. The trim spring is attached at the wing tip, as shown in Fig. 1, and is set to an
arbitrarily high stiffness. After trim is complete, the stiffness of this spring is set to zero for eigenanalysis or transient
analysis. This model predicts a rotor mode which goes unstable prior to the chord bending mode. To stabilize this
during trim, a ‘trim-damper’ on the gimbal joint is also used; again, with no impact on the stability prediction and an
insignificant impact on the trim state.

C. Computational Fluid Dynamics – Helios

Helios is a framework that incorporates a suite of individual software packages to create an analysis tool capable
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

of overset CFD simulations using a number of different structured and unstructured solvers. The philosophy behind
this overset method is that a near-body mesh system is developed for volumes in close proximity to solid walls, and a
Cartesian off-body mesh is used away from these regions. One benefit of overset CFD which is particularly useful
for rotorcraft simulations is that solid bodies can have arbitrary motion. Domain connectivity and flow interpolation
between the different mesh systems is carried out using Parallel Unsteady Domain Information Transfer (PUNDIT).
Finally, near-body mesh movement, mesh deformation, and fluid structure interactions are carried out through the
Mesh Motion, Loading, and Deformation Interface (MELODI) module [10]. MELODI is capable of handling arbitrary
configurations with prescribed rigid-body motion, prescribed elastic deformations, and CSD-coupled motions and
deformations.
FUN3D is used as the unstructured near-body solver and is easily used within the Helios framework. This solver was
developed and is maintained by the NASA Langley Research Center. In this study, it is used primarily with default Helios
settings, including the Spalart-Allmaras turbulence model. Subiterations are used to converge the solution at each time
step and 20 are found to give at least two orders of magnitude reduction in near-body density residual. The structured
off-body solver is SAMCART. The automatic mesh refinement capability of SAMCART builds a Cartesian domain
based on the size of the outer boundary of the near-body cells and the far-field cell size. Adaptive mesh refinement
(AMR) provides the ability to maintain flow features with little dissipation far into the wake, without the prohibitive
computational cost of having a fine off-body cell size in the entire domain. AMR is used to adapt to near-body mesh
motion throughout the simulation.
Near-body mesh systems for the blade and wing are shown in Figures 2 and 3, respectively. The blade mesh is
trimmed to roughly 1 chord length from all solid boundaries and after trimming, each blade has approximately 3 million
cells. The wing is trimmed to a similar distance to provide a quality overset interpolation boundary with the off-body
grid. After trimming, the wing has roughly a half of a million cells. The initial off-body mesh, with geometry-based
adaption, has 7.5 million cells. Convergence difficulties necessitated solution-based refinements from the beginning of
the simulation but were limited to 3 rotor radii aft of the wing to avoid an excessive number of cells. The off-body mesh
then adapts to the starting vortex and reaches a maximum cell count of 30 million, before leveling off at approximately
20 million. As airspeed increases, this final cell count decreases by 1–2 million because the wake is propagated out of
the domain more quickly.

D. CFD-CSD Coupling Method

While loose coupling is beneficial for finding the trim state of a vehicle or rotor, tight coupling must be used to
simulate transient behavior such as whirl flutter. In this method, the airloads from CFD are applied directly to the CSD
body at each time step. As mentioned previously, this method is prohibitively computationally expensive for obtaining a
trimmed solution. The method for using tight coupling for whirl flutter predictions is to obtain a trimmed state using
loose coupling first and then use tight coupling to observe a transient response. Initial attempts at tight coupling with
the configuration described in [11] demonstrated that a computational instability forms immediately after switching
from loose to tight coupling where an oscillation of the wing motion at every time step grows until either CFD or RCAS
fails. While not ideal, one solution to work past this instability is to restrict CFD/CSD coupling to only one rotor. In this
method, rigid-body rotor motion and rotor blade elastic deformation are used to move and deform the CFD meshes. In
RCAS, CFD airloads for the rotor are used, while lifting-line aerodynamics are used for the wing. Here, many of the
important physical aspects of the problem are still captured: rotor dynamic motion, high-fidelity rotor airloads, and

4
 aerodynamic interference from the wing on the rotor. This coupling method will not capture high-fidelity airloads on
the wing and will not capture aerodynamic interference from the rotor on the wing.

E. Signal Processing
Frequency-domain methods of predicting system stability naturally find modal frequency and damping through
eigenanalysis of the system equations of motion. Time-domain methods need to be analyzed with a signal processing
method to extract modal frequency and damping. In experimental work, the moving-block method [12] is often used
to process signal data as it is received from the data acquisition system, and early indicators of instabilities can be
identified. The moving block method relies on tracking frequency amplitudes with time as determined using a Fast
Fourier Transform (FFT). Damping is determined by the decay or growth of dominant frequencies over time. Simulation
time needs to be minimized in CFD/CSD simulations because of the increase in computational cost associated with
wind-tunnel time periods. This short simulation time, less than 10-rotor revolutions, creates FFT frequency resolution
issues.
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

In this study, a time history of the wing-tip displacement is used for frequency and damping identification. For
RCAS results, the signal is overwhelmingly dominated by a single frequency which is the wing beam-bending mode.
Due to the low signal complexity, the log-decrement method is used for frequency and damping identification. This
method is sensitive to the sampling frequency because it relies on accurate amplitude values and identifying when these
maxima or minima occur. Here, several peaks are used, and the values of frequency and damping that best re-create the
original signal are recorded to mitigate sample frequency errors. The log decrement method assumes an under-damped,
single degree-of-freedom system and fits the system response to the curve shown in Equation 1. In Ref [13], it is shown
how the method can be extended to multiple degree-of-freedom systems using bandpass filters.
" #
−ζω n t
ˆf (t) = f0 + ( f¯ − f0 ) 1 − pe sin (ωd t + φ) (1)
1 − ζ2
The log decrement algorithm used in this work first identifies both maximum and minimum points in the signal. The
maximum values are processed first, and then the same method is used for the minimum values. Equation 2 is used to
find the δ for a given peak. Here, f1 is the first peak, f1+n is the nth peak, f¯ is the mean value of the signal and n is the
peak number being used.

f1 − f¯
 
1
δ = ln (2)
n f1+n − f¯
Damping ratio (ζ), phase angle (φ), damped natural frequency (ωd ), and natural frequency (ωn ) are then found in
equations 3–6.

Fig. 2 Near-body mesh at multiple stations on the rotor blade.

5
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

Fig. 3 Near-body mesh at two cuts on the semi-span wing.

δ
ζ = q 2π (3)
δ 2
1 + 2π

p !
1 − ζ 2
φ = tan−1 (4)
ζ

2π
ωd = (5)
Tn
ωd
ωn = p (6)
1 − ζ2
Finally, signals are reconstructed with values of frequency and damping ratio for each of the peaks used. Each curve
fit quality is measured using R2 and the best set of values is selected.
When CFD/CSD time histories are compared to RCAS, higher frequency content is clearly visible, particularly at
high speeds. For multi-modal frequencies, time-domain stability analysis can be performed via Prony analysis to gather
modal frequency and damping information from a transient response of the system. The Prony method is curve fit
technique which uses a linear combination of n damped sinusoidal signals, where n is the model rank. In [14], damping
predictions via Prony analysis are shown to be sensitive to model rank number and sampling frequency and a similar
result is shown here. To mitigate risk of selecting a singular model rank to evaluate the signal, an average is taken for
both frequency and damping over a range of model ranks.

IV. Results
The results from this work are separated into three different sections. First, the baseline RCAS structural model is
compared to previous work using the XV-15. Next, transient stability predictions are made with RCAS and compared to
frequency domain predictions. Finally, CFD/CSD predictions are presented, and comparisons are made with the other
analysis methods.

A. CSD - Baseline Results

Prior to transient stability analysis, the RCAS model developed in this work is compared to a XV-15 CAMRAD
II model developed in [9]. First, the full semispan model is constructed in RCAS and linearized stability analysis is
performed. Figure 4 shows the frequency domain stability analysis predicted by RCAS compared with the 2.5 degree
precone CAMRAD II model. While modes are labeled for their primary motion, there is significant coupling between
different responses in a single mode. For example, the wing torsion modes often have both beam bending and chord
bending deformations. This coupling is exaggerated in rotor modes because of the high collective pitch required to trim

6
 in high speed flight. The frequency results show good agreement between the two analysis codes for the wing modes.
The low-frequency rotor modes have not been published for the CAMRAD II model. This result is expected as the wing
mode frequencies are primarily governed by wing root stiffness and system mass, principally in the pylon and rotor;
both of which do not change with blade precone.
More significant differences are observed between the two analysis codes for the predicted damping ratio. The
critical value which identifies an instability is where a given mode crosses the 0-damping line and becomes negatively
damped. Here, the mode would begin to grow in amplitude with any perturbation. While trends of the wing beam
bending mode damping with airspeed do not agree well, the stability boundary (ζ = 0) shows similar predictions between
RCAS and CAMRAD. The chord bending mode damping curves follow a similar trend of a relatively constant damping
until a critical speed where a sharp drop-off occurs. The RCAS model predicts a much later onset of this instability. The
torsion mode does not show good agreement between RCAS and CAMRAD II analyses. While this could indicate a
sensitivity of the mode to precone angle, further investigation would be needed to find the source of this discrepancy.
Finally, the current RCAS model predicts an unstable rotor progressive lag mode at 425 knots. The rest of the results
will focus on the wing beam bending mode for two reasons. First, it has the lowest instability speed. Second, it is the
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

most straight forward mode to excite in the current analysis procedure.

15
8
Frequency (Hz)

10
ζ (% Critical)
6

4 5

2
0
0
150 200 250 300 350 400 450 150 200 250 300 350 400 450
v (kts) v (kts)
CAMRAD II - Beam Mode CAMRAD II - Chord Mode CAMRAD II - Torsion Mode
RCAS - Regressive Gimbal Mode RCAS - Progressive Lag Mode RCAS - Beam Mode
RCAS - Chord Mode RCAS - Torsion Mode

Fig. 4 Semispan XV-15 frequencies and damping ratio as a function of airspeed. CAMRAD II results from [9]
for a 2.5 degree precone model. RCAS model has 1.5 degrees of precone.

In the current transient analysis procedure, the rotor is first trimmed to zero torque using collective pitch. The zero
torque trim state represents an unpowered wind-tunnel test scenario, also known as wind-milling. While cyclic pitch
would be necessary to control the rotor thrust direction when the aerodynamic influence from the wing on the rotor is
included, this effect is not captured in the RCAS analysis and is expected to have minimal impact on overall system
stability. As airspeeds approach stability boundaries, difficulty in achieving periodic solutions between trim updates is
experienced.
Two modeling mechanisms are used to overcome this problem. First, a six degree-of-freedom spring system is
placed at the wing tip. This stabilizes the wing system but does not affect rotor or gimbal degrees-of-freedom. As
shown in Fig 4, the progressive lag mode becomes unstable at high speeds. For stabilization of this mode, a damper is
added to the gimbal. Both the spring system and the damper are only active during trim and are prescribed to go to zero
stiffness and damping prior to system linearization or transient analysis. The trim spring at the wing tip has a secondary
effect on the model as supplying an initial system perturbation at the transition from trim to transient analysis. While the
wing is lifting during trim, the z-displacement at the wing tip is held close to zero. When released, the wing’s initial
response is to displace as a result of the wing lift which is an excitation of the wing beam bending mode.
This preliminary system perturbation is used to conduct a baseline comparison between the RCAS linearized stability
predictions and RCAS transient stability analysis for the wing beam bending mode. During the transient analysis of the
system, the motion of a node located at the wing tip is recorded. For RCAS results, the log decrement method described

7
 3.6 6

3.4 4
Frequency (Hz)

ζ (% Critical)
3.2 2

3 0

2.8 −2
150 200 250 300 350 400 450 150 200 250 300 350 400 450
Speed (knots) Speed (knots)
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

RCAS - Transient RCAS - Linearized

Fig. 5 Linearized stability analysis and transient stability analysis results for XV-15 wing beam bending mode.

previously is used to make a modal estimation of the signal. The predicted frequencies and damping ratios are shown
for a sweep of airspeed in Fig. 5. Excellent agreement is shown between the two analysis techniques. For airspeeds less
than 250 knots, both frequency and damping values are nearly identical. At speeds above that, some variations are
displayed but the predicted whirl-flutter speed of roughly 370 knots is consistent. Small differences between the two
methods are expected because any nonlinear structural dynamics are removed during the linearization process. Overall,
this result gives confidence in the transient analysis method.

B. CSD - Perturbation Sensitivity

While a small excitation is natively provided, a study of the effect of applying an external force as a perturbation is
shown here. Two forces, in the form of sinusoidal inputs with amplitude values of 1000 and 5000 pounds are applied for
half of a rotor revolution in the positive z-direction. Figure 6 shows time histories of z-displacement at the wing tip.
Note that the signals have been clipped to remove initial transient responses which corrupt the signal processing. The
clipped amount is determined by finding the first peak response following the perturbation. The x-axis indicates the
physical time after this peak value occurs. As expected, the larger amplitude perturbation yield a larger system response.
Also shown in Fig. 6 are the curve fits which are created using predicted values from the log decrement method. Here,
excellent agreement is shown between the recreated signals and the original responses which gives more reassurance in
the analysis method. Finally, general trends in both frequency and damping can be observed from these time histories.
With increasing airspeed, the signals show a longer wavelength, indicating a reduction in frequency. From 89 knots to
266 knots, an increase in damping is seen as the peak amplitudes decrease to a greater extent over time. This is followed
by a sharp reduction in damping from 266 knots to 355 knots.
Figure 7 shows the predicted frequencies and damping ratios for a sweep of airspeeds for each forcing function and
compares the results for linearized stability predictions. It is shown that for each of the three perturbation methods, the
predicted beam bending mode stability and damping ratios are nearly identical. With this in mind, it is clearly seen that
adding a tip force perturbation does not influence stability predictions for the beam bending mode. As a result, a force
perturbation is not needed which saves computational time. It is noted that in order to study the other wing modes, a
force would be needed, since they are not activated by the initial response of releasing the wing.

C. CFD/CSD - Transient Analysis

With a transient analysis method established and tested using computationally inexpensive RCAS simulations, it is
then applied to CFD/CSD simulations. A trimmed solution is achieved using loose coupling for seven coupling iterations
over three revolutions. The trim solution is achieved quickly in comparison with other rotorcraft simulations for several
reasons. First, a single control, collective pitch, is used to trim the rotor to zero torque. Second, the rotor wake is
weak because of the windmilling trim state and is quickly washed away from the rotor due to the system geometry and
high-speed axial flow. Finally, whirl-flutter calculations are not anticipated to be sensitive to small deviations from

8
 0.6 0.6
v = 89 Kts v = 177 Kts
0.55 0.55
∆z (ft)

∆z (ft)
0.5 0.5

0.45 0.45

0.4 0.4
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Time (s) Time (s)
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

0.6 0.6
v = 266 Kts v = 355 Kts
0.55 0.55
∆z (ft)

∆z (ft)
0.5 0.5

0.45 0.45

0.4 0.4
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Time (s) Time (s)
F=0 F = 1000 lb F = 5000 lb
F = 0, Prony F = 1000 lb, Prony F = 5000 lb, Prony

Fig. 6 Transient motion in the z-direction for a node located at the wing tip after three different system
perturbations are exerted. Also shown are the log decrement curve fit signals with predicted frequency and
damping values.

an exact trim state, so the convergence tolerance can be less than what is needed for simulations investigating rotor
performance.
First, the wing response can be evaluated qualitatively against RCAS results. Figure 8 shows this comparison for
three airspeeds. While a consistent trend of CFD/CSD results predicting a smaller initial response than RCAS, the
signals at both 177 and 266 knots show very good agreement between RCAS and CFD/CSD. The 355 knot simulation
demonstrates more exaggerated differences between the analysis methods. Both an increase in frequency and an increase
in damping is observed in the CFD/CSD simulation. Additionally, the highest speed case shows slight contribution from
frequencies other than the dominant beam bending mode.
Due to the slight mixing of frequency content in the CFD/CSD solutions, the log decrement method is seen to not
be robust, primarily because defining singular peak values is challenging. Signal processing of the CFD/CSD cases
is therefore performed using Prony Analysis. As previously mentioned, this method is known to be very sensitive
to the model rank number used and to the signal sample which is used. For modal content predictions made for
these simulations, signal sensitivity is not investigated. However, the sensitivity of model rank is considered and
mitigated by averaging values over a broad range of mode number values. Figure 9 illustrates this sensitivity by plotting
predicted frequencies and damping for the three CFD/CSD simulation airspeeds as a function of Prony modal rank.
While frequency predictions show relatively consistent values, damping predictions range over roughly 2 percent.
Unfortunately, the predicted values do not converge with increased mode number, so an average overall mode numbers
is used as the result.

9
 3.6 6

3.4 4
Frequency (Hz)

ζ (% Critical)
3.2 2

3 0

2.8 −2
150 200 250 300 350 400 150 200 250 300 350 400
v (kts) v (kts)
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

F = 0 lb F = 1000 lb F = 5000 lb RCAS - Linearized

Fig. 7 Rotor natural frequencies as a function of rotor rotational speed. Collective pitch is set to 0 degrees and
aerodynamic and gravity effects neglected.

Finally, frequency and damping values for the three CFD/CSD simulations are plotted with the previously shown
RCAS stability predictions in Fig. 10. As expected from the time history signal, at 177 knots very good agreement is
seen between the analysis methods. With increasing airspeed, the separation between RCAS and CFD/CSD increases
with the high-speed CFD/CSD case, 355 knots, demonstrating a different trend than previous results. This result agrees
well with the predictions for a notional tiltrotor shown in [6] where good agreement was shown for low-speed cases but
large differences where displayed between RCAS and CFD/CSD for high speed cases. This indicates that aerodynamic
causes of these differences need to be explored. Additionally, for this model, higher speed cases would need to be tested
to investigate two things: first to see if the predicted peak in damping using CFD/CSD simply happens later than RCAS
predictions and then turns sharply negative, and second, to investigate how the other modes in the system are predicted
using CFD/CSD.

V. Conclusion
This work begins to explore the effect of incorporating high-fidelity aerodynamics into transient whirl-flutter
predictions and opens several questions for future research. Baseline results are established and compared to previous
XV-15 comprehensive analysis results to demonstrate a realistic configuration. The wing beam bending mode is the
primary focus because it is natively activated through the analysis procedure, and it is the mode with the lowest unstable
airspeed. RCAS transient stability analysis is performed and compared to the traditional linearized eigenanalysis method
of predicting system stability. Small differences between the two methods are attributed to the removal of nonlinear
structural dynamics in eigenanalysis. Finally, CFD/CSD simulations are performed for three airspeeds, and stability
predictions are made via averaging Prony Analysis predictions. At 177 knots and 266 knots, relatively good agreement
is seen between the low- and high-fidelity transient analysis methods. At 355 knots, large differences are shown in
frequency and damping of the wing beam bending mode.
The deviation at high speeds is a consistent finding with previous CFD/CSD predictions and needs further
investigations. Rotor airloads from RCAS should be compared with CFD/CSD results and if possible, the RCAS model
should be updated to more accurately capture the aerodynamic physics occurring at high speed. One possibility could
be compressibility effects as airspeed increases. Finally, the effect of high-fidelity wing airloads and the influence of
rotor wake on the wing should be included for a more complete model.

Acknowledgments
This work was sponsored by the High Performance Computing Modernization Program (HPCMP) through the
HPC Internship Program (HIP). The computational time for this work was provided as part of a Frontier Project.

10
 0.55 v = 177 kts
∆z (ft) 0.5

0.45

0.55 v = 266 kts

∆z (ft)

0.5

0.45

0.55 v = 355 kts

∆z (ft)
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

0.5

0.45
RCAS CFD
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time (s)

Fig. 8 Wing response as calculated by RCAS and CFD/CSD transient analysis for 3 airspeeds.

Research was sponsored by the Army Research Laboratory and was accomplished under Cooperative Agreement
Number W911NF-18-2-0206. The views and conclusions contained in this document are those of the authors and should
not be interpreted as representing the official policies, either expressed or implied, of the Army Research Laboratory or
the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes
notwithstanding any copyright notation herein.

References
[1] Floros, M., and Kang, H., “Tiltrotor Whirl Flutter Stability Augmentation Using Active Wing Tips,” AHS International 73rd
Annual Forum and Technology Display, Fort Worth, TX, 2017.

[2] Kambampati, S., and Smith., E. C., “Aeroelastic Optimization of High-Speed Tiltrotor Wings with Wing Extensions and
Winglets,” Journal of Aircraft, Vol. 54, No. 5, 2017, pp. 1718–1727.

[3] Cole, J. A., Maughmer, M. D., and Bramesfeld, G., “Aerodynamic Design Considerations for Tiltrotor Wing Extensions and
Winglets,” 51st AIAA Aerospace Sciences Meeting, Grapevine, TX, 2013.

[4] Nixon, M. W., Piatak, D. J., Corso, L. M., and Popelka, D. A., “Aeroelastic Tailoring for Stability Augmentation and Performance
Enhancements of Tiltrotor Aircraft,” Journal of the American Helicopter Society, Vol. 45, No. 4, 2000, pp. 270–279.

[5] Kang, H., Shen, J., and Kreshock, A. R., “Validation of Comprehensive Analysis for Tiltrotor Whirl Flutter Predictions,”
Journal of Aircraft, Vol. 54, No. 4, 2017, pp. 1566–1570. doi:10.2514/1.C034272.

[6] Corle, E., Floros, M., and Schmitz, S., “CFD/CSD Aeroelastic Predictions of a Semi-Span Tiltrotor,” AHS 74th Annual Forum
& Technology Display, Phoenix, AZ, 2018.

[7] C. W. Acree, J., and Tischler, M. B., “Determining XV-15 Aeroelasatic Modes from Flight Data with Frequency-Domain
Methods,” Tech. Rep. TP 3330, NASA, 1993.

[8] Schroers, L. G., “Dynamic Structural Aeroelastic Stability Testing of the XV-15 Tilt Rotor Research Aircraft,” Tech. Rep. TM
84293, NASA, 1982.

[9] C. W. Acree, J., Peyran, R. J., and Johnson, W., “Rotor Design Options for Improving XV-15 Whirl-Flutter Stability Margins,”
Tech. Rep. TP 2291, NASA, 1984.

11
 8
v = 177 kts
v = 266 kts
3.4
v = 355 kts
Frequency (Hz)

ζ (% Critical)
3.35
4

3.3
2

0 100 200 300 400 500 600 0 100 200 300 400 500 600
Model Rank Model Rank
Downloaded by IOWA STATE UNIVERSITY on January 19, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2019-2132

Fig. 9 Frequency and damping ratio values as predicted via Prony Analysis for three airspeeds using a sweep
of mode numbers. Also shown are mean values for each airspeed.

3.6 6

3.4 4
Frequency (Hz)

ζ (% Critical)

3.2 2

3 0

2.8 −2
150 200 250 300 350 400 150 200 250 300 350 400
v (kts) v (kts)
RCAS - Linearized F=0 CFD

Fig. 10 Stability predictions using linearized RCAS, transient RCAS, and CFD/CSD.

[10] Roget, B., Sitaraman, J., and Wissink, A., “Maneuvering Rotorcraft Simulations Using CREATE A/VTM Helios,” 54th AIAA
Aerospace Sciences Meeting, San Diego, CA, 2016.

[11] Corle, E., Schmitz, S., and Floros, M., “Coupled CFD/CSD Predictions of Semi-Span XV-15,” 2018 AHS Technical Meeting on
Aeromechanics Design for Verical Flight, San Francisco, CA, 2018.

[12] Hammond, C., and Jr., R. D., “Determination of Supercritical Damping by Moving-Block/Randomdec Applications,” Tech.
Rep. NASA-SP-415, NASA, 1975.

[13] Liao, Y., and Wells, V., “Modal Parameter Identification Using the Log Decrement Method and Band-Pass Filters,” Journal of
Sound and Vibration, Vol. 330, No. 12, 2011, pp. 5014–5023.

[14] Zaki, A. A., “Using Tightly-Coupled CFD/CSD Simulation for Rotorcraft Stability Analysis,” Ph.D. thesis, Georgia Institute of
Technology, May 2012.

12