AIAA AVIATION Forum 10.2514/6.

2018-2992
June 25-29, 2018, Atlanta, Georgia
2018 Applied Aerodynamics Conference

Stability Derivative Estimation: Methods and Practical Considerations for Conventional

Transonic Aircraft
Steve Klausmeyer*, Caleb Fisher†, Kelly Laflin‡
Textron Aviation, Wichita, Kansas

Abstract

An assessment of various stability derivative prediction methods is presented and comparisons are made to
flight derived static and dynamic stability derivatives for a transonic business jet and the Douglas D558-II
research aircraft. Prediction methods include the forced oscillation method using FUN3D, the vortex lattice
code AVL, and handbook methods. Flight-based stability derivatives were extracted through analysis of
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

aircraft response to control surface doublets. A second component of this paper discusses the accuracy
requirements for stability and control derivatives, their effect on the basic aircraft lateral/directional and
longitudinal modes, and subsequent impact on aircraft design.

I. NOMENCLATURE

b = wing span 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝜕𝜕𝐶𝐶𝑌𝑌

𝐶𝐶𝐿𝐿 = 𝐶𝐶𝑌𝑌𝑟𝑟 =
𝑞𝑞𝑞𝑞 𝑟𝑟𝑏𝑏
𝑐𝑐̅ = wing mean aerodynamic chord 𝜕𝜕 � �
2𝑉𝑉∞
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀
kp = reduced frequency for pitch oscillations 𝐶𝐶ℓ = 𝜕𝜕𝐶𝐶𝑌𝑌
𝑞𝑞𝑞𝑞𝑞𝑞 𝐶𝐶𝑌𝑌𝛽𝛽̇ =
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃ℎ𝑖𝑖𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝛽𝛽̇ 𝑏𝑏
kβ = reduced frequency for yaw oscillations 𝜕𝜕 � �
𝐶𝐶𝑚𝑚 = 2𝑉𝑉∞
l = characteristic length 𝑞𝑞𝑞𝑞𝑐𝑐̅
𝜕𝜕𝐶𝐶𝑛𝑛
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃ℎ𝑖𝑖𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝐶𝐶𝑛𝑛𝑝𝑝 =
p = roll rate 𝐶𝐶𝑛𝑛 = 𝑝𝑝𝑏𝑏
𝑞𝑞𝑞𝑞𝑐𝑐̅ 𝜕𝜕 � �
2𝑉𝑉∞
q = pitch rate 𝜕𝜕𝐶𝐶𝐿𝐿
𝐶𝐶𝐿𝐿𝛼𝛼 = 𝜕𝜕𝐶𝐶𝑛𝑛
r = yaw rate 𝜕𝜕𝜕𝜕 𝐶𝐶𝑛𝑛𝑟𝑟 =
𝑟𝑟𝑏𝑏
𝜕𝜕𝐶𝐶𝑚𝑚 𝜕𝜕 � �
2𝑉𝑉∞
𝑉𝑉∞ = freestream velocity 𝐶𝐶𝑚𝑚𝛼𝛼 =
𝜕𝜕𝜕𝜕 𝜕𝜕𝐶𝐶ℓ
α = angle of attack 𝜕𝜕𝐶𝐶𝑌𝑌 𝐶𝐶ℓ𝛽𝛽̇ =
𝐶𝐶𝑌𝑌𝛽𝛽 = 𝛽𝛽̇ 𝑏𝑏
𝜕𝜕𝜕𝜕 𝜕𝜕 � �
𝛼𝛼̇ = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑜𝑜𝑓𝑓 𝑐𝑐ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑜𝑜𝑜𝑜 𝛼𝛼 2𝑉𝑉∞
𝜕𝜕𝐶𝐶ℓ 𝜕𝜕𝐶𝐶ℓ
β = angle of sideslip 𝐶𝐶ℓ𝛽𝛽 = 𝐶𝐶ℓ𝑝𝑝 =
𝜕𝜕𝛽𝛽 𝑝𝑝𝑏𝑏
𝛽𝛽̇ = rate of change of β 𝜕𝜕 � �
𝜕𝜕𝐶𝐶𝑛𝑛 2𝑉𝑉∞
𝐶𝐶𝑛𝑛𝛽𝛽 =
ωp = pitch axes forced oscillation frequency 𝜕𝜕𝜕𝜕 𝜕𝜕𝐶𝐶ℓ
𝐶𝐶ℓ𝑟𝑟 =
𝜕𝜕𝐶𝐶𝐿𝐿 𝑟𝑟𝑏𝑏
𝜕𝜕 � �
ωβ = yaw axes forced oscillation frequency 𝐶𝐶𝐿𝐿𝑞𝑞 =
𝑞𝑞𝑐𝑐̅
2𝑉𝑉∞
𝜕𝜕 � � 𝜕𝜕𝐶𝐶𝑛𝑛
ωdr = Dutch roll frequency 2𝑉𝑉∞
𝐶𝐶𝑛𝑛𝛽𝛽̇ =
𝜕𝜕𝐶𝐶𝐿𝐿 𝛽𝛽̇ 𝑏𝑏
ζdr = Dutch roll damping ratio 𝐶𝐶𝐿𝐿𝛼𝛼̇ = 𝜕𝜕 � �
𝛼𝛼̇ 𝑐𝑐̅ 2𝑉𝑉∞
𝜕𝜕 � �
ωsp = Short period frequency 2𝑉𝑉∞
i = subscript indicating drag, lift, or
𝜕𝜕𝐶𝐶𝑚𝑚 pitching moment
ζsp = Short period damping ratio 𝐶𝐶𝑚𝑚𝑞𝑞 =
𝑞𝑞𝑐𝑐̅
𝜕𝜕 � �
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 2𝑉𝑉∞ j = subscript indicating sideforce,
𝐶𝐶𝐷𝐷 = rolling moment, or yawing
𝑞𝑞𝑞𝑞 𝜕𝜕𝐶𝐶𝑚𝑚
𝐶𝐶𝑚𝑚𝛼𝛼̇ = moment
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝛼𝛼̇ 𝑐𝑐̅
𝐶𝐶𝑌𝑌 = 𝜕𝜕 � �
2𝑉𝑉∞
𝑞𝑞𝑞𝑞

*
Senior Engineering Specialist, Ph.D., Aerodynamics
†
Engineer, Leadership Development Program, Aerodynamics
‡
Senior Engineering Specialist, Ph.D., Aerodynamics
Copyright © 2018 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
 II. INTRODUCTION
The dynamic response of an aircraft to atmospheric disturbances and control inputs is traditionally modeled using
non-linear equations of motion. These equations are often simplified using linear small-perturbation approximations
where the external aerodynamic forces and moments are assumed to vary linearly about a steady-state flight condition.
These linear variations are the well-known static and dynamic stability derivatives and are used for assessing vehicle
dynamics, sizing tail surfaces, establishing control surface layout, and designing the control system architecture.

Prior to first flight of a new aircraft, stability derivative estimates are obtained from a variety of sources. Wind tunnel
tests provide static derivatives, but this data may not be available until well into the design cycle and does not typically
provide dynamic derivative estimates. Analytical methods are used for early estimates of static derivatives until wind
tunnel data becomes available. Dynamic derivative estimates rely heavily on analytical methods throughout the design
cycle until flight data becomes available, at which point flight-matched derivatives can be produced.

A variety of analytical methods exist for estimation of static and dynamic stability derivatives. The simplest of these
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

are the “handbook” methods consisting of ESDU, DATCOM, and various text books. These methods offer elaborate
curvefits to large sets of wind tunnel data, occasionally supplemented with theoretical methods and are valuable for
rapid assessment of configuration changes during conceptual design. Simple compressibility corrections are
implemented but don’t capture the changes in the aerodynamic derivatives due to shock formation and thus must be
used with caution for aircraft designed to operate at transonic speeds (e.g. business jets and commercial transports).
These Mach-induced changes are most prominent as the aircraft approaches its maximum operating Mach number
(MMO). Over many decades of aircraft development, an aircraft manufacturer typically accumulates a large wind
tunnel database which is used to tune the handbook methods to better fit company-specific products. One limitation
is that inaccurate estimates may be produced when a new configuration is analyzed that differs substantially from the
existing product line to which the methods were calibrated.

Vortex-lattice and panel codes can also be used to estimate stability derivatives but are limited to lower subsonic
speeds, although many have some form of compressibility correction that extend their applicability to higher speeds
if shock waves are not present. These methods require more user effort than handbook methods but the cost is not
prohibitive. Advantages of these methods include the ability to better model the effects of subtle configuration
changes that are beyond handbook methods and the ability to compute finer aerodynamic detail such as wing span
loading, induced drag, etc. Most of these methods are limited to steady-state solutions and therefore cannot predict
some of the time-dependent dynamic derivative components such as the 𝛼𝛼̇ contribution to pitch damping.

Euler and RANS solvers can be used after details of the configuration outer mold line firm up. At this development
stage, preliminary wing and body contours are complete including airfoil definition for the wing and tail surfaces,
wing twist/dihedral, engine/nacelle and aft fuselage integration. Static derivatives are estimated using sweeps in angle-
of-attack, sideslip, and Mach number. Prediction of dynamic derivatives typically involves running forced oscillation
time dependent solutions and extracting stability derivative information from the aircraft’s force and moment
responses3-8.

Calibration of analytical methods takes different forms. The static derivatives may initially be calibrated to wind
tunnel data and later tuned to match the actual aircraft. Dynamic derivatives are rarely measured in a wind tunnel due
to the cost and complexity of such a test and as such, flight data is used to calibrate these values.

This paper begins with a brief review of stability derivatives and the fundamental modes of aircraft motion about the
longitudinal and lateral/directional axes. Next a discussion of required precision is presented, followed by comparison
of predicted and measured stability derivatives and subsequent modal behavior. Comparisons are made to wind tunnel
and flight derived values for both a transonic business jet and the Douglas D558-II research aircraft1 2 with focus on
the static and dynamic stability derivatives rather than control derivatives. Flight estimated derivatives were obtained
by analyzing aircraft response to control surface doublets.

Note that the equations of motion for transport category aircraft configurations are typically separated into longitudinal
and lateral/directional. Cross interaction between these two motions is relatively weak and the separation results in a
simplification of the equations of motion without appreciable loss of accuracy.

2
 A. Longitudinal Modes and Stability Derivatives

The longitudinal modes for a conventional aircraft consist of two oscillatory motions, a highly damped, high frequency
oscillation called the short period and a low frequency oscillation with very low damping called the phugoid. The
short period oscillation occurs at virtually constant speed while the phugoid takes place at nearly constant angle-of-
attack.

The most important longitudinal stability derivatives for conventional aircraft are 𝐶𝐶𝐿𝐿𝛼𝛼 , 𝐶𝐶𝑚𝑚𝛼𝛼 , 𝐶𝐶𝑚𝑚𝑞𝑞 , 𝐶𝐶𝑚𝑚𝑉𝑉 , 𝐶𝐶𝑚𝑚𝛼𝛼̇ , 𝑎𝑎𝑎𝑎𝑎𝑎 𝐶𝐶𝐷𝐷𝑉𝑉 .
The derivative 𝐶𝐶𝐿𝐿𝛼𝛼 is the lift curve slope and is typically in the range 1 to 8 rad-1 for the linear region of the lift curve
below stall. It increases with Mach number as the transonic range is approached, peaks and then falls off rapidly at
high subsonic speeds. It has a strong influence on the turbulence response and the maneuverability of the aircraft.
𝐶𝐶𝑚𝑚𝛼𝛼 is the basic longitudinal stability and is a key determinant of short-period mode natural frequency. This derivative
varies with Mach number and may change erratically in the transonic range. The derivatives 𝐶𝐶𝑚𝑚𝑞𝑞 and 𝐶𝐶𝑚𝑚𝛼𝛼̇ are pitch
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

damping derivatives and are the primary sources of short period damping. 𝐶𝐶𝑚𝑚𝑞𝑞 is usually of much greater magnitude
than 𝐶𝐶𝑚𝑚𝛼𝛼̇ and typically has a value in the range from -20 to -40 rad-1.

The effect of speed variations on pitching moment is contained in the derivative 𝐶𝐶𝑚𝑚𝑉𝑉 . If this derivative is positive,
the aircraft will tend to pitch-up with an increase in speed. The resulting increase in drag and the greater component
of gravity along the x-axis will then tend to slow the aircraft down. This is stable behavior. In practice, the effective
center of pressure of an aircraft may move aft with increasing subsonic Mach number, thus creating a tendency to dive
(negative 𝐶𝐶𝑚𝑚𝑉𝑉 ) with associated unstable speed behavior. Depending on the aircraft, this tuck-under effect is commonly
referred to as “Mach tuck” and may require an additional system (Mach trim) that counters this instability.

The derivative 𝐶𝐶𝐷𝐷𝑉𝑉 forms the bulk of the damping for the phugoid mode.

B. Lateral/Directional Modes and Stability Derivatives

The fundamental lateral/directional modes of motion consist of the Dutch roll mode (oscillatory) and two non-
oscillatory modes: roll subsidence and spiral stability. Dutch roll is usually very lightly damped or unstable for swept
wing aircraft and an artificial means of enhancing damping must be implemented which usually takes the form of a
rudder-based yaw damper. Loss of the primary yaw damper must be considered during aircraft design and requires a
backup yaw damping method. Vertical fin loads (and subsequently airframe weight) may be reduced by subtle
configuration changes early in the design cycle. Thus an accurate knowledge of aerodynamic derivatives is important.

Spiral stability is the tendency of an aircraft to return to wings-level attitude after a roll upset.

Roll subsidence is a non-oscillatory mode that represents the damping of rolling motion and is an indicator of
resistance to roll rate (not angle).

The most important lateral-directional stability derivatives are 𝐶𝐶𝑦𝑦𝛽𝛽 , 𝐶𝐶ℓ𝛽𝛽 , 𝐶𝐶𝑛𝑛𝛽𝛽 , 𝐶𝐶ℓ𝑝𝑝 , 𝐶𝐶𝑛𝑛𝑝𝑝 , 𝐶𝐶ℓ𝑟𝑟 𝑎𝑎𝑎𝑎𝑎𝑎 𝐶𝐶𝑛𝑛𝑟𝑟 . The derivative 𝐶𝐶𝑦𝑦𝛽𝛽
is known as the side-force derivative and typically has values between -0.1 and -2 rad-1. The derivative 𝐶𝐶ℓ𝛽𝛽 is the
rolling moment due to sideslip and is known as the dihedral stability derivative. If this term is negative, the aircraft
will generally have a stable spiral mode, e.g., a roll upset will produce a restoring moment that will roll the aircraft
back to a wings-level attitude. However, more negative 𝐶𝐶ℓ𝛽𝛽 usually degrades Dutch roll damping which has
implications on selection and design of artificial yaw damping systems.

Directional stability, 𝐶𝐶𝑛𝑛𝛽𝛽 , describes the tendency of the aircraft to return to a zero-sideslip condition after a yawing
motion disturbance. Dutch roll frequency is proportional to 𝐶𝐶𝑛𝑛𝛽𝛽 . This derivative is analogous to the spring component
of a spring-mass-damper. The yaw damping derivative 𝐶𝐶𝑛𝑛𝑟𝑟 provides the bulk of Dutch roll damping. This derivative
is often artificially increased through implementation of a yaw damper which is required for aircraft certification and
acceptable ride comfort. 𝐶𝐶𝑛𝑛𝛽𝛽̇ is another contributor to yaw damping but is usually quite small and neglected for

3
 conventional aircraft evaluations. Roll damping, 𝐶𝐶ℓ𝑝𝑝 , is a factor in selection of roll control devices and deflection
scheduling. Yawing moment due to roll rate, 𝐶𝐶𝑛𝑛𝑝𝑝 , affects Dutch roll damping and is typically small but can still be
significant, especially when damping is low.

III. METHODS
The following methods were used to estimate static and dynamic stability and control derivatives for a transonic
business jet configuration with aft-mounted engines, T-tail, and moderate wing sweep angle, and the 1950’s era
Douglas D558-II research aircraft. Publicly available dynamic wind tunnel and flight data are available for the
Douglas configuration, hence its selection for analysis.

A. FUN3D/Forced Oscillation

Time-dependent, forced-oscillation simulations were performed using the Reynolds-Averaged Navier-Stokes

(RANS), finite-volume flow solver FUN3D. FUN3D is actively being developed by NASA Langley Research
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Center and has been used extensively to assess aircraft aerodynamics for a variety of applications. The reader is
referred to the FUN3D website (Ref: https://fun3d.larc.nasa.gov) for an overview of the flow solver, user’s manual,
list of publications, and a sampling of applications.

All simulations of the current study were conducted assuming compressible, viscous and fully-turbulent flow. The
Spalart-Allmaras turbulence model was used with the rotation/curvature correction (denoted SA-RC) activated.
Inviscid flux residual construction was performed using Roe flux difference splitting with the MUSCL upwinding
reconstruction scheme. Flux limiting was not employed during any of the simulations.

FUN3D can perform both forced, sinusoidal rotational (pitch) and translational (plunge) oscillations on a rigid grid
about an arbitrary axis. Sinusoidal motion frequency and magnitude are specified by the user, as are normal vector
components of either the rotational axis or translational direction, as appropriate. This capability was used in the
current study to perform separate, time-dependent aircraft roll, pitch, and yaw rotation oscillations about the x, y,
and z aircraft body axes, respectively. Additionally, time-dependent plunge oscillations were performed through the
sinusoidal translation of the aircraft along its z axis.

Time-accurate temporal time integration was performed using 30 sub-iterations of the FUN3D optimized second-
order backward differencing scheme, denoted 2ndorderOPT, at each time step. Sub-iteration convergence was
deemed acceptable, with each of the root-mean-square (RMS) residuals typically dropping 4 to 5 orders of
magnitude, as shown in Figure 1 through Figure 4.

4
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 1. Pitching Motion Sub-iteration Convergence, M=0.81, FUN3D Forced Oscillation

Figure 2. Plunging Motion Sub-iteration Convergence, M=0.81, FUN3D Forced Oscillation

5
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 3. Rolling Motion Sub-iteration Convergence, M=0.81, FUN3D Forced Oscillation

Figure 4. Yawing Motion Sub-iteration Convergence, M=0.81, FUN3D Forced Oscillation

An exception to this satisfactory convergence behavior occurred for the forced yaw-oscillation simulations at Mach
numbers of 0.84 and 0.85, for which the turbulence RMS residual decreased by only 2.5 orders of magnitude, while
all other RMS residuals decreased nearly 5 orders of magnitude (Figure 5). The degraded convergence appears to
correspond to the development of a boundary-layer separation bubble at the trailing edge of the wing-fuselage
junction.

6
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 5. Yawing Motion Sub-iteration Convergence, M=0.85, FUN3D Forced Oscillation

All simulations considered in this study were performed using 400 time steps per oscillatory cycle. It was
determined that the simulation was adequately ‘mature’ by the end of the second cycle, and so simulations were
concluded at the end of the third cycle. Data from the first two cycles was discarded and the third cycle data was
used in the subsequent stability analysis.

All CFD simulations were performed using a single, common computational grid. Grid viscous surface spacing is
appropriate for flight Reynolds numbers, with a non-dimensional wall distance, y-plus, value no greater than 1 for the
Reynolds numbers considered. The mixed-element unstructured viscous grid was constructed using Pointwise [Ref]
and is comprised of approximately 34 million prism cells, 200 thousand pyramidal cells, and 16 million tetrahedron
cells. The aircraft engines were modeled as flow-through nacelles so engine power effects are not included in the
analysis.

B. AVL

AVL is a program for the aerodynamic and flight-dynamic analysis of rigid-body aircraft. It implements an extended
vortex lattice model for the lifting surfaces, together with a slender-body model for fuselages and nacelles. The flight
dynamic analysis combines a full linearization of the aerodynamic model about any flight state, together with specified
mass properties.

To more accurately capture the lift generated by the fuselage and nacelles, the slender-body model capability of AVL
was not utilized. The slender-body model represents the fuselage as a doublet that does not contribute any lift and is
more likely to create erroneous solutions. Instead the fuselage is modeled by creating a lifting body projection of the
planform area of the fuselage and nacelle onto the vertical x-z and horizontal x-y geometric planes. These lifting
bodies are represented as low aspect ratio lifting surfaces within AVL. Figure 6 depicts how an aircraft would be
modeled in this manner within AVL.

7
 Figure 6. Vortex Lattice (AVL) Representation of a Business Jet
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

It was determined in a fuselage study done within AVL that it is better to make separate models for the longitudinal
and lateral-directional analysis. Therefore, the above model with intersecting low aspect ratio lifting surfaces was then
separated into a longitudinal model and a lateral-directional model. These models are shown in Figure 7 and
Longitudinal Modes Figure 8.

Figure 7. AVL Model for Longitudinal Modes Figure 8. AVL Model for Lateral-Directional Modes

C. ESDU/DATCOM

Engineering Sciences Data Units are “handbook” methods for predicting derivatives using basic aircraft geometry.
These methods are essentially intricate multi-dimensional curvefits to aerodynamic databases and have the advantage
of speed and computational efficiency during conceptual design. They can also be adjusted to match an existing
aerodynamic database for an aircraft manufacturer’s product line.

D. Forced Oscillation Method

A method for developing stability and control rate derivatives uses forced oscillation data from wind tunnel
measurements or CFD methods. In this method the model undergoes forced (not free) sinusoidal motion during which
variation of aerodynamic coefficients are measured. Static and dynamic stability and control derivatives can then be
extracted from aircraft response to the forced oscillation.

A basic assumption of the method is that the aerodynamic coefficient variation with time can be described by the first
order terms of a Fourier series. For example a sinusoidal oscillation imposed on an aircraft model will produce a
sinusoidal variation in the aerodynamic coefficients. This assumption is valid if the forced oscillation magnitude and
frequency is such that non-linear aerodynamic effects are avoided.

8
 A thorough description of the forced oscillation method is first described by Da Ronch, et. al.3 and Mialon, et. al.

Assume that the aircraft model is driven by a sinusoidal pitch oscillation about the moment reference center and that
lift, drag, and pitching moment are measured as a function of time. These oscillatory responses provide sufficient
information to determine key static and dynamic stability derivatives such as lift curve slope 𝐶𝐶𝐿𝐿𝛼𝛼 , static longitudinal
stability 𝐶𝐶𝑀𝑀𝛼𝛼 and pitch damping coefficient (𝐶𝐶𝑀𝑀𝑞𝑞 + 𝐶𝐶𝑀𝑀𝛼𝛼̇ ). Separate values for the q and 𝛼𝛼̇ pitch damping
contributions can be determined by performing a 2nd test that subjects the aircraft to a pure plunging motion (q=0)
from which 𝐶𝐶𝑀𝑀𝛼𝛼̇ can be computed. The derivative 𝐶𝐶𝑀𝑀𝑞𝑞 can then be estimated as the difference between the
combined derivative (𝐶𝐶𝑀𝑀𝑞𝑞 + 𝐶𝐶𝑀𝑀𝛼𝛼̇ ) and 𝐶𝐶𝑀𝑀𝛼𝛼̇ .

A detailed buildup of the equations required to estimate longitudinal stability derivatives follows.

The traditional model for an increment in force or moment coefficient from a given initial condition can be
expressed using a first order Taylor series approximation:
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

𝑙𝑙 𝑙𝑙 𝑙𝑙 2 1
∆𝐶𝐶𝑖𝑖 = 𝐶𝐶𝑖𝑖𝛼𝛼 Δ𝛼𝛼 +
𝐶𝐶𝑖𝑖𝑞𝑞 𝑞𝑞 + 𝐶𝐶𝑖𝑖𝛼𝛼̇ 𝛼𝛼̇ + � � 𝐶𝐶𝑖𝑖𝑞𝑞̇ 𝑞𝑞̇
𝑉𝑉∞ 𝑉𝑉∞ 𝑉𝑉∞
Where the subscript i represents either lift, drag, or pitching moment. The pitch acceleration (𝑞𝑞̇ ) terms are typically
not used in the analysis of commercial transport aircraft although they could be important for highly maneuverable
combat fighters or aerobatic airplanes.

Defining the characteristic length as 𝑐𝑐̅⁄2 produces the following equation which is consistent with the traditional
definition of rate derivatives

𝑐𝑐̅ 𝑐𝑐̅ 𝑐𝑐̅ 2 2

∆𝐶𝐶𝑖𝑖 = 𝐶𝐶𝑖𝑖𝛼𝛼 Δ𝛼𝛼 +
𝐶𝐶𝑖𝑖𝑞𝑞 𝑞𝑞 + 𝐶𝐶𝑖𝑖𝛼𝛼̇ 𝛼𝛼̇ + � � 𝐶𝐶𝑖𝑖𝑞𝑞̇ 𝑞𝑞̇
2𝑉𝑉∞ 2𝑉𝑉∞ 2𝑉𝑉∞
The next step is to develop equations describing the forced oscillation. If we assume a harmonic pitch oscillation
with amplitude 𝛼𝛼𝐴𝐴 and frequency 𝜔𝜔𝑝𝑝 (rad/sec) we can write the following kinematic equations:

Δ𝛼𝛼 = 𝛼𝛼𝐴𝐴 sin�𝜔𝜔𝑝𝑝 𝑡𝑡� 3

𝛼𝛼̇ = ω𝛼𝛼𝐴𝐴 cos�𝜔𝜔𝑝𝑝 𝑡𝑡� 4

𝛼𝛼̈ = −(ω)2 𝛼𝛼𝐴𝐴 sin�𝜔𝜔𝑝𝑝 𝑡𝑡� 5

For pure pitching motion in the absence of plunging, we can write

𝑞𝑞 = 𝛼𝛼̇ 6

𝑞𝑞̇ = 𝛼𝛼̈ 7

A formula for the change in coefficient ∆𝐶𝐶𝑖𝑖 from an initial state as a function of time can be derived by substituting
equations 3 to 5 (kinematic equations) into equation 2 (response ∆𝐶𝐶𝑖𝑖 ):
𝑐𝑐̅ 𝑐𝑐̅ 8
∆𝐶𝐶𝑖𝑖 = 𝐶𝐶𝑖𝑖𝛼𝛼 𝛼𝛼𝐴𝐴 sin�𝜔𝜔𝑝𝑝 𝑡𝑡� + 𝐶𝐶𝑖𝑖𝑞𝑞 𝜔𝜔𝑝𝑝 𝛼𝛼𝐴𝐴 cos�𝜔𝜔𝑝𝑝 𝑡𝑡� + 𝐶𝐶 𝜔𝜔 𝛼𝛼 cos�𝜔𝜔𝑝𝑝 𝑡𝑡�
2𝑉𝑉∞ 2𝑉𝑉∞ 𝑖𝑖𝛼𝛼̇ 𝑝𝑝 𝐴𝐴
𝑐𝑐̅ 2 2
+ � � 𝐶𝐶𝑖𝑖𝑞𝑞̇ �−�𝜔𝜔𝑝𝑝 � 𝛼𝛼𝐴𝐴 sin�𝜔𝜔𝑝𝑝 𝑡𝑡��
2𝑉𝑉∞

Defining a non-dimensional reduced frequency as

𝑐𝑐̅𝜔𝜔𝑝𝑝 9
𝑘𝑘𝑝𝑝 =
2𝑉𝑉∞

equation 8 can be written as

9
 ∆𝐶𝐶𝑖𝑖 = 𝛼𝛼𝐴𝐴 �𝐶𝐶𝑖𝑖𝛼𝛼 − 𝑘𝑘𝑝𝑝2 𝐶𝐶𝑖𝑖𝑞𝑞̇ � sin�𝜔𝜔𝑝𝑝 𝑡𝑡� + 𝛼𝛼𝐴𝐴 𝑘𝑘𝑝𝑝 �𝐶𝐶𝑖𝑖𝑞𝑞 + 𝐶𝐶𝑖𝑖𝛼𝛼̇ � cos�𝜔𝜔𝑝𝑝 𝑡𝑡� 10

Equation 10 describes the variation of the force or moment coefficient relative to a steady-state initial condition, and
is thus an incremental value. The first term on the right-hand side of the equation (sine term) is often referred to as
the “in-phase” response since it is in phase with Δ𝛼𝛼 while the 2nd term (cosine) is referred to as the “out-of-phase”
component since it leads the forcing function by 90 degrees (e.g., it is 90 degrees out of phase with Δ𝛼𝛼).

The final longitudinal response equation is

∆𝐶𝐶𝑖𝑖 = 𝛼𝛼𝐴𝐴 𝐶𝐶𝑖𝑖𝛼𝛼 sin�𝜔𝜔𝑝𝑝 𝑡𝑡� + 𝛼𝛼𝐴𝐴 𝑘𝑘𝑝𝑝 �𝐶𝐶𝑖𝑖𝑞𝑞 + 𝐶𝐶𝑖𝑖𝛼𝛼̇ � cos�𝜔𝜔𝑝𝑝 𝑡𝑡� 11

The response described in equation 11 can be determined in a wind tunnel suitably equipped for forced oscillations,
or in a time-accurate CFD code such as FUN3D. Steady-state CFD solvers and lower order methods such as vortex-
lattice codes do not capture out-of-phase dynamics and thus cannot be used with the forced oscillation method.
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Static and dynamic derivatives are determined by first fitting aerodynamic coefficient response data with a 1st order
Fourier series

∆𝐶𝐶𝑖𝑖 = 𝐴𝐴0𝑖𝑖 + 𝐴𝐴1𝑖𝑖 sin�𝜔𝜔𝑝𝑝 𝑡𝑡� + 𝐵𝐵1𝑖𝑖 cos�𝜔𝜔𝑝𝑝 𝑡𝑡� 12

Equating equations 12 and 11 yields:

𝐴𝐴0𝑖𝑖 = 0 𝐴𝐴1𝑖𝑖 = 𝛼𝛼𝐴𝐴 𝐶𝐶𝑖𝑖𝛼𝛼 𝐵𝐵1𝑖𝑖 = 𝛼𝛼𝐴𝐴 𝑘𝑘𝑝𝑝 �𝐶𝐶𝑖𝑖𝑞𝑞 + 𝐶𝐶𝑖𝑖𝛼𝛼̇ �

And solving for the aerodynamic derivatives:

𝐴𝐴1𝑖𝑖
𝐶𝐶𝑖𝑖𝛼𝛼 = 13
𝛼𝛼𝐴𝐴
𝐵𝐵1𝑖𝑖
�𝐶𝐶𝑖𝑖𝑞𝑞 + 𝐶𝐶𝑖𝑖𝛼𝛼̇ � = 14
𝑘𝑘𝑝𝑝 𝛼𝛼𝐴𝐴

The coefficients 𝐴𝐴1𝑖𝑖 , and 𝐵𝐵1𝑖𝑖 can be determined using a least-squares curvefit to the lift and pitching moment
responses (e.g. lsqcurvefit in MATLAB).

A similar set of equations based on oscillations around the z-axis (yawing motion) can be used to derive equations
for directional stability (𝐶𝐶𝑛𝑛𝛽𝛽 ), side force derivative (𝐶𝐶𝑦𝑦𝛽𝛽 ), dihedral stability (𝐶𝐶ℓ𝛽𝛽 ), and yaw damping (𝐶𝐶𝑛𝑛𝑟𝑟 + 𝐶𝐶𝑛𝑛𝛽𝛽̇ ).
Separate values for the r and 𝛽𝛽̇ yaw damping contributions can be determined by performing a 2nd test that subjects
the aircraft to a pure side acceleration motion (r=0) from which 𝐶𝐶𝑁𝑁𝛽𝛽̇ can be computed. However, for transonic
transports yaw damping is dominated by 𝐶𝐶𝑛𝑛𝑟𝑟 and the additional steps of computing 𝐶𝐶𝑁𝑁𝛽𝛽̇ are excluded.

The characteristic length is defined as b/2 which is consistent with the traditional definition of the directional
stability derivatives.

A harmonic yaw oscillation with amplitude 𝛽𝛽𝐴𝐴 and frequency 𝜔𝜔𝛽𝛽 (rad/sec) is represented in the following kinematic
equations:

Δ𝛽𝛽 = 𝛽𝛽𝐴𝐴 sin�𝜔𝜔𝛽𝛽 𝑡𝑡� 15

𝛽𝛽̇ = 𝜔𝜔𝛽𝛽 𝛽𝛽𝐴𝐴 cos�𝜔𝜔𝛽𝛽 𝑡𝑡� 16

2 17
𝛽𝛽̈ = −�𝜔𝜔𝛽𝛽 � 𝛽𝛽𝐴𝐴 sin�𝜔𝜔𝛽𝛽 𝑡𝑡�

For pure yawing motion in the absence of side motion, we can write

10
 𝑟𝑟 = 𝛽𝛽̇ 18

𝑟𝑟̇ = 𝛽𝛽̈ 19

A formula for the change in force or moment coefficient (∆𝐶𝐶𝑖𝑖 ) from an initial state as a function of time can be given
by
𝑏𝑏 𝑏𝑏
∆𝐶𝐶𝑗𝑗 = 𝐶𝐶𝑗𝑗𝛽𝛽 𝛽𝛽𝐴𝐴 sin�ω𝛽𝛽 𝑡𝑡� + 𝐶𝐶 ω 𝛽𝛽 cos�ω𝛽𝛽 𝑡𝑡� + 𝐶𝐶 ω 𝛽𝛽 cos�ω𝛽𝛽 𝑡𝑡�
2𝑉𝑉∞ 𝑗𝑗𝑟𝑟 𝛽𝛽 𝐴𝐴 2𝑉𝑉∞ 𝑗𝑗𝛽𝛽̇ 𝛽𝛽 𝐴𝐴
20
𝑏𝑏 2 2
+ � � 𝐶𝐶𝑗𝑗𝑟𝑟̇ �−�ω𝛽𝛽 � 𝛽𝛽𝐴𝐴 sin�ω𝛽𝛽 𝑡𝑡��
2𝑉𝑉∞

Where the subscript j represents either sideforce, rolling moment, or yawing moment.

Defining a non-dimensional reduced frequency as

Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

𝑏𝑏ω𝛽𝛽 21
𝑘𝑘𝛽𝛽 =
2𝑉𝑉∞

equation 20 can be written as

∆𝐶𝐶𝑗𝑗 = 𝛽𝛽𝐴𝐴 �𝐶𝐶𝑗𝑗𝛽𝛽 − 𝑘𝑘𝛽𝛽2 𝐶𝐶𝑗𝑗𝑟𝑟̇ � sin�ω𝛽𝛽 𝑡𝑡� + 𝛽𝛽𝐴𝐴 𝑘𝑘𝛽𝛽 �𝐶𝐶𝑗𝑗𝑟𝑟 + 𝐶𝐶𝑗𝑗𝛽𝛽̇ � cos�ω𝛽𝛽 𝑡𝑡� 22

Equation 22 describes the variation of the side force relative to a steady-state initial condition, and is thus an
incremental value. The first term on the right-hand side of the equations (sine term) is often referred to as the “in-
phase” response since it is in phase with Δ𝛽𝛽 while the 2nd term (cosine) is referred to as the “out-of-phase”
component since it leads the forcing function by 90 degrees (e.g., it is 90 degrees out of phase with Δ𝛽𝛽).

Static and dynamic derivatives are determined by first fitting the sideforce and moment responses with a 1st order
Fourier series (consisting of a single sine and cosine term)

∆𝐶𝐶𝑗𝑗 = 𝐴𝐴0𝑗𝑗 + 𝐴𝐴1𝑗𝑗 sin�ω𝛽𝛽 𝑡𝑡� + 𝐵𝐵1𝑗𝑗 cos�ω𝛽𝛽 𝑡𝑡� 23

Equating equation 23 and 22 yields the following:

𝐴𝐴0𝑗𝑗 = 0 𝐴𝐴1𝑗𝑗 = 𝛽𝛽𝐴𝐴 𝐶𝐶𝑌𝑌𝛼𝛼 𝐵𝐵1𝑗𝑗 = 𝛽𝛽𝐴𝐴 𝑘𝑘𝛽𝛽 �𝐶𝐶𝑗𝑗𝑟𝑟 + 𝐶𝐶𝑗𝑗𝛽𝛽̇ �

And solving for the aerodynamic derivatives:

𝐴𝐴1𝑗𝑗
𝐶𝐶𝑗𝑗𝛽𝛽 = 24
𝛽𝛽𝐴𝐴

𝐵𝐵1𝑗𝑗
�𝐶𝐶𝑗𝑗𝑟𝑟 + 𝐶𝐶𝑗𝑗𝛽𝛽̇ � = 25
𝑘𝑘𝛽𝛽 𝛽𝛽𝐴𝐴

The coefficients 𝐴𝐴0𝑗𝑗 , 𝐴𝐴1𝑗𝑗 , and 𝐵𝐵1𝑗𝑗 can be determined using a least-squares curvefit to the lift and pitching moment
responses (e.g. lsqcurvefit in MATLAB).

Roll axis derivatives can be estimated using the forced oscillation approach or steady state methods. A forced
oscillation approach used for roll-rate derivatives presented later in this paper. The equations are identical to those
for the directional axes after substituting roll angle and roll rate for sideslip angle and yaw rate. For the roll axes,
the only derivatives of interest are those with roll rate.

IV. ACCURACY REQUIREMENTS

The effect of stability derivative uncertainty depends on flight conditions and the dynamic modes that are of most
concern to the designer. For example, many business jets have a lightly damped or even unstable Dutch Roll mode at

11
 high altitudes and are equipped with a rudder-based yaw damper that is used to control the mode. The designer must
consider failures of this damping system and subsequent impact to safe aircraft operation. The design solution to loss
of the yaw damper depends on the aircraft’s bare-airframe (un-augmented) Dutch Roll characteristics. If natural Dutch
Roll damping is sufficiently positive, an “escape envelope” may be defined where the aircraft retreats after loss of the
yaw damper for continued safe flight and landing. Another option may be to add a backup system that engages upon
failure of the yaw damper. Possible failure scenarios per regulation must include a rudder mechanical jam so simply
adding another actuator and controller to the rudder is insufficient and backup system solutions must utilize other
control surfaces. Example solutions include splitting the rudder or adding a roll-based Dutch Roll control system.
Alternatives to adding a 2nd system include increasing the airframe’s natural Dutch roll damping via configuration
changes such as reduced wing dihedral, tail changes, and/or strakes on the aft end of the aircraft. In any case, a
relatively precise evaluation of lightly damped Dutch roll modes is of value. As an example, assume a notional aircraft
with the aerodynamic characteristics defined in Table 1.
Table 1. Notional Aircraft Characteristics for Dutch Roll Sensitivity Evaluation
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Dihedral
𝐶𝐶ℓ𝛽𝛽 -0.15 1/rad
Stability

Directional
𝐶𝐶𝑛𝑛𝛽𝛽 0.125 1/rad
Stability

Yaw Damping 𝐶𝐶𝑛𝑛𝑟𝑟 -0.25 1/rad

Roll Damping 𝐶𝐶ℓ𝑝𝑝 -0.5 1/rad

𝐶𝐶𝑛𝑛𝑝𝑝 0.05 1/rad

𝐶𝐶ℓ𝑟𝑟 0.075 1/rad

Given mass properties, one can develop linearized aerodynamic models and perform modal analysis in a program
such as MATLAB. Using this approach, the yaw rate response of the notional aircraft to a rudder doublet was
calculated and is shown in Figure 9. This response is typical of a lightly damped Dutch Roll mode with a damping
ratio of 0.06.

12
 Figure 9. Yaw Rate Response to a Rudder Doublet for a Notional Aircraft
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 10 shows how uncertainty in the various stability derivatives affects Dutch roll damping. Derivative
variations of 20%, 40%, and 60% are indicated by the small solid symbols while the nominal value is the large solid
symbol.

Dihedral stability, 𝐶𝐶ℓ𝛽𝛽 , shown in the first plot of Figure 10 has a large influence on Dutch roll damping. Suppose
that predicted dihedral stability was the nominal value of -0.15 but in flight the actual value turned out to be -0.20.
This would degrade Dutch roll damping from the predicted value of 0.06 to an actual value of 0.048. If the
minimum required damping ratio was 0.05, this error could be problematic if not discovered until a flight test,
leading to possible program delays and lost revenue.

Dutch roll damping varies slightly with directional stability (𝐶𝐶𝑛𝑛𝛽𝛽 ) as shown in the 2nd plot of Figure 10. Reducing
directional stability increases damping ratio although the effect is small. 𝐶𝐶𝑛𝑛𝛽𝛽 , however, has a large effect on the
Dutch roll frequency.

Yaw damping and roll damping are shown in the 3rd and 4th plots respectively. As one would expect, the Dutch roll
mode is heavily influenced by yaw damping. If actual 𝐶𝐶𝑛𝑛𝑟𝑟 were 20% lower than predicted, damping ratio would
drop to 0.048, again potentially problematic.

The cross derivatives 𝐶𝐶𝑛𝑛𝑝𝑝 and 𝐶𝐶ℓ𝑟𝑟 are shown in the last two plots of Figure 10. Rolling moment due to yaw rate has
little effect on damping while yawing moment due to roll rate has a pronounced effect. A 20% error in the 𝐶𝐶𝑛𝑛𝑝𝑝
estimate can swing the damping ratio estimate by 0.01.

Clearly, the accumulated effect of errors in each derivative could make matters even worse.

Although not presented here, similar sensitivity studies can be performed for other aircraft modes. An example
would be the roll mode time constant which influences roll performance metrics that must be satisfied for acceptable
handling qualities sand certification requirements.

13
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 10. Effect of Stability Derivative Uncertainty on Damping Ratio for a Notional Aircraft with a Lightly Damped
Dutch Roll Mode

V. RESULTS

Stability derivative estimates using various methods are presented in this section. Pitch axis dynamic motion is
typically composed of two longitudinal oscillatory modes, the short period and the phugoid. Results in the longitudinal
prediction sections are focused on the short period mode since the Phugoid mode is generally a less critical mode for
consideration in the design of a business jet. Lateral/directional modes include the oscillatory Dutch roll mode and
two non-oscillatory modes consisting of the spiral mode and roll subsidence.

One of the authors’ goals was to evaluate the forced oscillation method and evaluate its efficacy for estimating static
and dynamic stability derivatives. A series of trade studies were conducted to determine proper FUN3D input settings
that control the solver characteristics. The effects of various FUN3D numerical input parameters such as the number
of time steps per cycle, time scheme (2nd and 3rd order), mesh size, and viscous/inviscid solution mode were

14
 investigated using the Douglas D558-II aircraft. These settings were then carried over to the business jet model where
the effects of oscillation magnitude and frequency were evaluated.

A. Douglas D558-II

The Douglas D558-II aircraft shown in Figure 11 was chosen for preliminary evaluations given publicly available
wind tunnel and flight data for static and dynamic derivatives. Wind tunnel data for dynamic derivatives is somewhat
rare, so this was a consideration in selection of the D558-II. The Douglas D558-II test vehicle was developed to
expand knowledge of transonic and supersonic flight. The propulsion system could be configured in either turbojet
or rocket mode and in the turbojet configuration, the single engine exhausted from the lower portion of the aft fuselage.
Stability derivatives for this aircraft were estimated using AVL and the FUN3D/Forced oscillation approach.
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 11. Douglas D558-II Research Aircraft Used for FUN3D/Forced Oscillation and AVL Predictions

A trade study was conducted to determine the influence of various computational parameters for the FUN3D/forced
oscillation method:

• number time steps per cycle

• time scheme
• mesh size
• inviscid/viscous fluid dynamic equations
• jet eflux

This list is by no means exhaustive, but formed a manageable level of effort in terms of computational resources and
personnel availability.

These studies were limited to the pitch axis and a flight Mach number of 0.64 and a pressure altitude of 25,000 feet.
Flight data was extracted from time history and frequency domain responses to elevator doublets8 using in-house flight
data processing tools.

AVL results were computed using a modeling technique where the fuselage is represented by a low-aspect ratio lifting
surface oriented horizontally similar to Figure 7.

FUN3D/forced oscillation results were computed using Euler grids with 5.2 million cells (nominal) and 7.2 million
cells (dense) and a viscous grid with 106 million cells. Both 2nd and 3rd order time stepping schemes were evaluated
along with various numbers of time steps per oscillation cycle. Finally, a power-on condition was evaluated with a
cold jet emanating from the exhaust port on the lower aft fuselage.

15
 Results for lift curve slope are presented in Figure 12 which shows the flight, AVL, and wind tunnel values along with
a series of forced oscillation results. The forced oscillation results were quite uniform, regardless of changes in the
numerical parameter and agree closely with the AVL estimate. Working from left to right through the forced
oscillation results in Figure 12: the Navier-Stokes solution produced a slightly lower value of 𝐶𝐶𝐿𝐿𝛼𝛼 which is an expected
result given the typical reduction in lift slope due to viscosity. Varying the time stepping scheme from 2nd to 3rd order
had no effect and increasing steps per cycle from 50 up to 800 had very little effect on the results. Increasing mesh
size had no noticeable effect.

A side-study was conducted to determine the effect of engine exhaust since it exits from the bottom of the fuselage.
However, modeling the jet exhaust in FUN3D produced very little change in the forced oscillation lift curve slope
(last column in Figure 12).
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 12. Lift Curve Slope from Flight, AVL, and Forced Oscillation CFD Parameter Study, Douglas D558-II, M=0.64, 25,000
feet

Pitching moment slope calculations are shown in Figure 13. Very little difference was noted for the various forced
oscillation permutations. The AVL result compares well to forced oscillation, but the flight value is slightly lower.
The flight-derived 𝐶𝐶𝑀𝑀𝛼𝛼 reproduces the flight short period mode frequency quite well, which lends confidence to the
flight values.

16
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 13. Pitching Moment Slope from Flight, AVL, and Forced Oscillation CFD Parameter Study, Douglas D558-II, M=0.64,
25,000 feet

Pitch damping is shown in Figure 14. Slightly more variation with forced-oscillation parameters is apparent. The
AVL result is lower than the forced oscillation but this is expected since AVL only provides the 𝐶𝐶𝑀𝑀𝑞𝑞 component of
pitch damping. The forced oscillation method compared well to the flight-derived damping.

Figure 14. Pitch Damping from Flight, AVL, and Forced Oscillation CFD Parameter Study, Douglas D558-II, M=0.64, 25,000
feet

17
 Lift damping is shown in Figure 14 and was more sensitive to the forced oscillation parameter variation. Inspecting
forced oscillation results from left to right reveals that lift damping increased when changing the basic fluid
equations from Euler to Navier-Stokes. Order of the time stepping scheme (2nd or 3rd) had virtually no effect. Lift
damping increased as time steps per cycle varied from 50 to 800, but the time scheme, seemed to reach a plateau
after reaching 400 time steps per cycle. Increasing mesh density had no effect and adding power reduced lift
damping as indicated by the last bar.
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 15. Lift Damping from Flight, AVL, and Forced Oscillation CFD Parameter Study, Douglas D558-II, M=0.64, 25,000
feet

Based on the results of this section, attention was turned to the business jet example using

• 400 time steps per cycle.

• 2nd order time stepping scheme
• the Navier-Stokes equations

B. Transonic Business Jet

A transonic business jet configuration with aft-mounted engines, T-tail, and winglets (Figure 16) was used for this
study. Prediction methods included FUN3D/Forced oscillation, AVL, and handbook methods for both longitudinal
and lateral/directional axes. Longitudinal and lateral/directional static and dynamic derivatives were computed using
FUN3D/Forced Oscillation, AVL, and handbook methods at a variety of flight conditions. These derivatives were
then used to produce linear models for the aircraft about each flight condition, followed by modal analysis in
MATLAB. Results were then compared to flight-based estimates.

18
 Figure 16. Business Jet Aircraft Used for FUN3D/Forced Oscillation, AVL, Handbook Method Predictions

i. Business Jet Pitch Axes Derivatives

Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

The effect of the forced oscillation frequency and amplitude on static stability (𝑑𝑑𝐶𝐶𝑀𝑀 ⁄𝑑𝑑𝐶𝐶𝐿𝐿 ) and pitch damping (𝐶𝐶𝑀𝑀𝑞𝑞 +
𝐶𝐶𝑀𝑀𝛼𝛼̇ ) were evaluated for a low-speed flight condition. Lower freestream velocity was selected with the intent of
avoiding potential non-linear aerodynamics associated with high transonic speeds due to local shock formation and
movement with AOA. Also, a moderate angle-of-attack was selected which was well within the expected linear region
of lift and pitching moment variation with angle-of-attack, thus avoiding separated flow. A sinusoidal pitch oscillation
(no vertical translation) was imposed about the moment reference center on the CFD model. Body axes forces and
moments from FUN3D were converted to stability axes by a rotation through AOA. Typical lift and pitching moment
responses are presented in Figure 17 over the specified four cycles and show that CFD solutions converged to a
repeatable oscillation pattern after the 1st cycle, so only the last 3 cycles were used for data analysis.

0.10

CL CM

0.10

Curvefit applied to this data Curvefit applied to this data

Time Time

Figure 17. Lift and Pitching Moment Response to a Forced Pitch Oscillation. Amplitude = 3 deg, frequency = ωsp

The resulting lift and pitching moment responses were processed to extract pitch damping and aircraft neutral point.
A heavy-weight/forward-CG loading condition was selected for this study. Three frequencies were evaluated which
were 1, 2, and 3 times the short period natural frequency (ωsp) and yielded reduced frequencies of 0.0268, 0.0536, and
0.0804. Lift curve slope and pitching moment slopes are shown in Figure 18 which shows a decline in lift slope with
increasing oscillation frequency. Neutral point calculated from the lift and pitching moment slopes was observed to
move about 2% MAC aft with increasing frequency. This variation is on the border of acceptability, particularly for
an aircraft with minimal static margin (distance between the aft center-of-gravity limit and neutral point). A typical
business jet tends to have a large static margin, since the aft CG position is often constrained by landing gear placement
and on-ground tip-back angles rather than aerodynamic limits. Pitch damping exhibited very minor changes with
frequency.

The effect of pitch amplitude at a constant reduced frequency kp=0.0268 is also shown in Figure 18. Three amplitudes
were evaluated: the baseline of 3 degrees and two other amplitudes of 1 and 4 degrees which bracketed the baseline.
Results show relatively small changes in the derivatives, even for the largest amplitude of 4 degrees. This is not
surprising given that the steady state angle-of-attack was chosen such that flow separation would not be encountered
even with the higher amplitude oscillation. At Mach numbers greater than 0.5, one would expect the 3 and 4 degree

19
 amplitudes to be susceptible to non-linear aerodynamic effects due to shock formation and movement, hence the lower
magnitude of 1 degree was chosen for further studies.
Table 2. Forced Oscillation Frequency and Amplitude Variation Study

Mean
Reduced
Amplitude Frequency Angle of
Frequency
attack
3 ωsp .0268 5.6
3 2ωsp .0536 5.6
3 3ωsp .0804 5.6
1 ωsp .0268 5.6
4 ωsp .0268 5.6
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 18. Effect of Pitch Oscillation Frequency and Amplitude on Pitch Derivatives. M<0. 5, Fwd/Heavy Aircraft Loading

Static and dynamic pitch axes derivatives as estimated using FUN3D forced oscillation, AVL, and handbook
methods are shown in Figure 19 for four flight conditions along with a comparison to flight derived derivatives. The
forced oscillation method provided acceptable estimates (relative to flight) at all flight conditions. The AVL
estimates were close to flight values at lower speeds began to deviate at the highest Mach number, likely due to
AVL’s limited ability to handle compressibility effects. The handbook methods included in Figure 19 were

20
 produced using unmodified ESDU and DATCOM methods without in-house calibrations which would improve the
predictions.
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 19. Pitch Axes Stability Derivatives for a Transonic Business Jet

ii. Business Jet Pitch Axes Modes

Frequency and damping ratio for the short period mode for each of the analytical methods were evaluated with a linear
model constructed using the estimated stability derivatives and flight mass properties. Short period characteristics are
presented in Figure 20 along with corresponding flight values. All three methods estimated short period damping
ratio within about 0.05 of the flight value (typical business jet aircraft have a heavily damped short period mode with
average damping ratios varying between 0.2 and 0.4). This level of agreement is acceptable for the heavily damped
conditions in Figure 20 but would likely be unacceptable under lightly damped flight conditions or in a military
applications where rapid pitch changes combined with very precise pitch attitude control are of high importance. In
these examples, the handbook methods provide sufficiently accurate predictions such that the added expense of the
forced oscillation approach would not necessarily be required. However, the handbook methods do not handle
configurations that deviate from the calibration database and often do not have the granularity necessary to capture
the effects of small configuration changes. In these cases, higher order methods are necessary.

The AVL method is also quite efficient and can sometimes be used to compute changes due to configuration
modifications but it is limited to simpler geometries and does not account for viscous effects or the formation of shock
waves. A simple Prandtl-Gluaert type compressibility correction is implemented which perform acceptably into the
low transonic speed range.

21
 The first two flight conditions in Figure 20 show the effect of changing aircraft weight and center-of-gravity
location. All three analytical methods captured the change in damping ratio associated with this aircraft loading
change. The last two conditions represent the effect of changing Mach number. Flight data suggests an increase in
damping and natural frequency with speed and all three analytic methods correctly estimate these behaviors. Again,
the FUN3D forced oscillation method provided the best overall agreement with flight 𝐶𝐶𝑀𝑀𝛼𝛼 is the primary contributor
to short period frequency, so good agreement with this parameter also suggests accurate natural frequency estimates.
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 20. Short Period Mode for a Business Jet

iii. Business Jet Lateral/Directional Axes Derivatives

Lateral/directional stability derivatives were computed using FUN3D/Forced Oscillation, AVL, and handbook
methods. Predicted quantities are compared to wind tunnel and flight estimates for the static derivatives, and flight
estimates only for the dynamic derivatives.

For the FUN3D/Forced Oscillation method the sensitivity to oscillation frequency and magnitude were evaluated for
a low speed flight condition (Mach Number < 0.50). Lower freestream velocity was selected with the intent of
avoiding potential non-linear aerodynamics associated with high transonic speeds due to local shock formation and
movement with sideslip angle. Also, moderate sideslip magnitudes were selected which were within the expected
linear region of sideforce, rolling moment, and pitching moment variation with sideslip angle, thus avoiding separated
flow. A sinusoidal yaw oscillation (no lateral translation) was imposed about the moment reference center on the
CFD model. Body axes forces and moments from FUN3D were converted to stability axes by a rotation through
angle-of-attack. The number of cycles were reduced to three for expediency instead of the four cycles uses for the
pitch axes. The CFD solutions converged to a repeatable oscillation pattern after the 1st cycle, so only the last 2 cycles
were used for data analysis.

Force and moment responses were then processed to extract directional and dihedral stability along with yaw damping.
Three frequencies were evaluated which were 1, 2, and 3 times the Dutch roll natural frequency (ωdr) and yielded
reduced frequencies of 0.0104, 0.0208, and 0.0417. Results are shown in Figure 21 which indicates a decline in
magnitude of all quantities with increasing oscillation frequency.

The effect of yaw amplitude at a constant reduced frequency (kβ) of 0.0104 is also shown in Figure 21. Three
amplitudes were evaluated: the baseline of 3 degrees and two other amplitudes of 1 and 4 degrees which bracketed
the baseline. Results show relatively small changes in the derivatives for the largest amplitude of 4 degrees.

22
 Table 3. Forced Oscillation Frequency and Amplitude Variation Study

Reduced Mean Angle

Amplitude Frequency
Frequency of attack
3 ωsp .0268 5.6
3 2ωsp .0536 5.6
3 3ωsp .0804 5.6
1 ωsp .0268 5.6
4 ωsp .0268 5.6
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 21. Lateral/Directional Stability Derivatives from FUN3D with a Forced Yaw Oscillation

Static and dynamic stability derivatives computed using FUN3D/Forced oscillation, AVL, and handbook methods
and compared to flight data in Figure 22Error! Reference source not found.. The FUN3D/Forced oscillation
method provided the best agreement overall. AVL provided reasonable agreement with flight measurements but
overestimates directional stability and yaw damping, likely due to the purely inviscid equation set. Modeling the
fuselage as a lifting surface rather than a simple body was required for improved accuracy. The yawing moment due
to roll rate derivative was off for the M0.5 condition. The cause of this disagreement among the various methods is
unknown, however, the magnitude of difference is noticeable in Dutch roll predictions.

23
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 22. Lateral/Directional Stability Derivatives for a Business Jet.

iv. Business Jet Lateral/Directional Axes Derivatives

Modal characteristics (frequency and damping ratio for oscillatory modes, time constant and time to ½ amplitude or
double for non-oscillatory modes) for each of the analytical methods were evaluated with linear models constructed
using estimated stability derivatives and flight mass properties. Dutch roll damping and natural frequency are shown
in Figure 23. The FUN3D/Forced oscillation results are in relatively close agreement with flight except for the lowest

24
 Mach number where much of the disagreement is due to 𝐶𝐶𝑛𝑛𝑝𝑝 mismatch. There is good agreement for natural frequency
which is consistent with accurate prediction of directional stability, 𝐶𝐶𝑛𝑛𝛽𝛽
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

Figure 23. Dutch Roll Mode for a Business Jet

Figure 24. Spiral Mode Time to Double - Business Jet Figure 25. Roll Mode Time Constant – Business Jet

VI. CONCLUSION
Three methods for determining static and dynamic stability derivatives were assessed using the Douglass D558-II
research aircraft and a transonic business jet:

1) FUN3D coupled with the forced oscillation method

2) AVL – an open source vortex lattice code
3) Handbook methods consisting of ESDU and DATCOM

Comparisons to flight derived modes and stability derivatives were discussed. The forced oscillation method
coupled with the FUN3D flow solver provided the best agreement to flight for both the longitudinal and
lateral/directional axes. In some cases, the lower order AVL method provided sufficient accuracy, however,
predictions deteriorated at higher freestream Mach numbers. Handbook methods performed less satisfactorily,
although these estimates were made in “raw” form and did not include in-house calibrations.

25
 The overall conclusion is that the FUN3D/forced oscillation method is the desired method for predicting stability
and control derivatives prior to flight. Although the method is compute resource intensive, the workload is
manageable by a small aerodynamics team. The method provides reasonable stability derivative magnitude
estimates, handles compressibility effects including local shock waves, and has the potential to resolve subtle
configuration changes which are often “overlooked” by lower order methods

References
1Queijo,M.J.,Goodman, A., “Calculations of the Dynamic Lateral Stability Characteristics of the Douglas D-558-II Airplane in
High-Speed Flight for Various Wing Loadings and Altitudes,” NACA L50H16a, 1950.
2Fletcher,
H.., “Comparison of Several Methods for Estimating Low-Speed Stability Derivatives for Two Airplane
Configurations,” NASA TN D-6531, 1971
Downloaded by IOWA STATE UNIVERSITY on January 2, 2019 | http://arc.aiaa.org | DOI: 10.2514/6.2018-2992

3Da Ronch, A, Vallespin, D., Ghoreyshi, M., Badcock, K.J., “Computation of Dynamic Derivatives Using CFD,” AIAA Paper
2010-4817, 2010.
4Mialon, B., Khelil, S.B., Huebner, A., Jouhaud, J., Roge, G., Hitzel, S., Badcock, K., Eliasson, P., Khrabov, A., Lahuta, M..,

“European Benchmark on Numerical Prediction of Stability and Control Derivatives,” AIAA Paper 2009-4116, 2009.
5Green, L., Spence, A., Murphy, P., “Computational Methods for Dynamic Stability and Control Derivatives,” AIAA Paper 2005-
15, 2005.
6Lee,H., Kim, B., Lee, S., “Computational Study of the Stability Derivatives for the Standard Dynamic Model,” AIAA Paper
2013-2658, 2013.
7Thompson, J., Frink, N., Murphy, P., “Guidelines for Computing Longitudinal Dynamic Stability Characteristics of a Subsonic
Transport” AIAA Paper 2010-4819, 2010.
8Mialon, B., Khrabrov, A., Da Ronch, A., Cavagna, L., Zhang, M., Ricci, S. “Benchmarking the Prediction of Dynamic

Derivatives: Wind Tunnel Tests, Validation, Acceleration Methods” AIAA Paper 2010-8244, 2010.
9Holleman,Euclid C., “Longitudinal Frequency-Response and Stability Characteristics of the Douglas D-558-II airplane as
Determined from Transient Response to a Mach Number of 0.96,” NACA RM L52E02, 1952.

26