Goman, Khramtsovsky, Kolesnikov (2008 draft) - Evaluation of Aircraft Performance and Maneuverability by Computation of Attainable Equilibrium Sets

1
Evaluation of Aircraft Performance and Maneuverability by
Computation of Attainable Equilibrium Sets
M. G. Goman1
, A. V. Khramtsovsky2
and E.N. Kolesnikov3
De Montfort University, Leicester, LE1 9BH, UK
An aircraft performance and maneuvering capabilities in steady flight conditions are usually analyzed
considering the steady-states of the rigid body equations of motion. A systematic way of computation of the
set of all attainable steady states for a general class of helical trajectories is presented. The proposed
reconstruction of attainable equilibrium states and their local stability maps provides a comprehensive and
consistent representation of the aircraft flight and maneuvering envelopes. The numerical procedure is
outlined and computational examples of attainable equilibrium sets in the form of two-dimensional cross-
sections of steady-state maneuver parameters are presented for three different aircraft models.
Nomenclature
A, B = state and control matrices of the linearized system
g = kinematical constraints vector function
f = right-hand sides of aircraft rigid-body equations
xf = Jacobian matrix
H, M = altitude and Mach number
ph = angular momentum of the engine rotors
p, q, r = body-axis angular rates
R = radius of helical trajectory
T = thrust force
V = flight velocity
x = state vector
µy = maneuver parameter vector
α = angle of attack
β = angle of sideslip
γ = flight path angle
θ,φ = pitch and roll angles
δe, δa, δr = elevator, aileron and rudder deflections
δT = thrust throttle position
δ = control vector
= total angular velocity
V = velocity-vector roll rate
Subscripts
ε = equilibrium
µ = maneuver
1
Professor, Faculty of Engineering and Computing Sciences. Senior Member AIAA.
2
Visiting Research Fellow, Faculty of Engineering and Computing Sciences.
3
Now Engineer, Flight Sciences, Bombardier Aéronautique, 400 Côte-Vertu, Montréal, H4S 1Y9, Canada

2
I. Introduction
The evaluation of aircraft flight performance and maneuvering capabilities, from conceptual design to the final stage
of flight clearance, is an important task of aircraft development. Flight performance reflects an aircraft’s ability to
perform steady coordinated maneuvers at different speed and altitude, i.e. its capability of maintaining a straight-
and-level flight, steady level turn, climbing or gliding turns, etc. The maneuverability characterizes an aircraft’s
ability to alter its steady flight trajectory via rotation with respect to the flight velocity vector. The level of
maneuverability is directly linked with attainable values for the angle of attack or the normal load factor and the
angular rate in the velocity-vector roll maneuver.
The steady performance and maneuvering capabilities are usually evaluated by solving the steady-state problem
for the rigid body equations of motion, which depend on the aircraft propulsion, aerodynamic, mass and inertia
characteristics [1-3]. The state and control parameters are interchangeable for the purpose of finding solutions to the
steady-state problem. For example, the steady states can be determined for any particular combination of control
inputs without setting any requirements for maneuver parameters. Alternatively, some steady-state and control
variables can be chosen as independent and the remaining variables can be determined according to the imposed
kinematical constraints [2,14,15,17].
The continuation and bifurcation analysis methods first introduced in flight dynamics in the late seventies [5,6]
have evolved into a powerful tool for aircraft trim and stability analysis [20]. Within this computational framework
the equilibrium states are computed in the extended space of state variables and one selected control parameter,
while all other control effectors are kept constant. The steady states are represented as one-parameter continuation
diagram or as two-dimensional equilibrium surface. At some ranges of control parameters “folded” dependencies
may appear in the steady-state diagrams and surfaces, which indicate potential non-local departures in motion
variables at limit points of the folds. Other types of bifurcations or changes in steady state local stability are also
identified during continuation procedure thus allowing effective prediction of aircraft instability and loss of control.
A number of important flight dynamics phenomena such as departures at high incidence flight, onset of wing rock
motion, spin dynamics and roll-coupled manoeuvres have been successfully investigated using continuation method
[5-13,18].
Imposing additional maneuver-specific kinematical constraints, as it was demonstrated in [14-16], allowed the
continuation of steady states representing a specified maneuver. In this constrained problem several control surfaces

3
are deflected simultaneously so that this method can be used for designing aircraft steady-state maneuvers. The one-
parameter continuation method, however, may be difficult and computationally intensive in reconstruction of all
attainable steady states and the maneuver flight envelope due to non-uniqueness of solutions in the extended state
and control parameter space.
The goal of this paper is to present a method for the reconstruction of aircraft flight envelope considering a
general class of steady maneuvers via computation of the set of all attainable steady states. The results are
represented in the state and maneuver parameters space while the associated control parameters are stored in a
special data structure for processing and analysis at a later stage. Steady states are computed for the equations of
motion augmented by a number of auxiliary equations specifying maneuver kinematics. The set of all attainable
equilibrium states is computed on a grid of state and maneuver parameter points and visualized in the form of two-
dimensional cross-sections. The local stability maps, reflecting the structure of the eigenvalues of the linearized
equations of motion, complement this representation.
The next section of this paper discusses equations of spatial motion and kinematics of steady-state solutions. The
proposed computational method is presented in Section III. Section IV contains a number of examples of two-
dimensional attainable equilibrium sets and local stability maps for the open-loop mathematical models for the F-18
HARV, F-16 and the F-14 aircraft, based on data from [25-27]. Section V summarizes the main results and presents
our conclusions.
II. Equations of Motion
The rigid aircraft equations of motion can be presented in the form of a set of nonlinear, autonomous, ordinary
differential equations describing rotational/translational accelerations and rotational kinematics using Euler angles.
These equations have the following form [6]
( )
( )
( )
1
0
0 0
0
y z lx xz
y z x m p
xz z x y n p
I I qr qSbCp I I
q I I I pr qScC h r
r I I I I pq qSbC h q
−  − +−   
    = − + +    
    − − + −      
(1)

4
( ) ( )
( )
sin sec
cos sin tan sin cos sec
cos cos cos sin sin sec
x z
qS T
q p r C C
mV mV
g
V
α β
α α α β α α β
θ φ α θ α β
= − + − − −
+ +
(2)
( )
( )
cos sin
sin cos cos sin cos sin sin
cos sin cos sin sin cos cos cos sin sin
x y z
qS T
p r C C C
mV mV
g
V
α β
β α α α β β α β
θ φ β θ β α θ φ β α
= − − − + −
+ + −
(3)
( )
( )
cos cos sin sin cos cos cos
sin cos cos cos sin sin cos cos sin cos
x y z
V qS T
C C C
V mV mV
g
V
α β β α β α β
θ α β θ φ β θ φ α β
= + + +
+ + +
(4)
cos sinq rθ φ φ= − (5)
( )sin cos tanp q rφ φ φ θ= + + (6)
( )sin cos secq rψ φ φ θ= + (7)
where , , ,x y z xzI I I I are the moments of inertia,
2
2
V
q
ρ
= is the dynamic pressure, V is the airspeed, ρ is the air
density, S is the wing reference area, b is the wing span, c is the mean aerodynamic chord of the wing, g is the
gravitational acceleration, and m is the aircraft mass.
In general, the non-dimensional aerodynamic force , ,x y zC and moment , ,l m nC coefficients depend nonlinearly on
angle of attack ( )α and sideslip angle ( )β , angular velocity vector components ( , , )p q r , elevator, aileron and
rudder deflections ( , , )e a rδ δ δ and Mach number M . The static thrust T depends nonlinearly on altitude H ,
Mach number M and throttle position Tδ .
The aerodynamic coefficients and the thrust force are usually defined as tabular functions of the motion
parameters and control inputs. The look-up data tables used for interpolating the aerodynamic coefficients are
obtained from static, forced oscillations and rotary-balance wind tunnel tests. The tabulated data are interpolated
using linear or cubic spline function approximations, depending on requirements for smoothness of the aerodynamic
coefficient functions.
The steady states of the first eight equations (1- 6) describe a general class of helical trajectories with rotation
around a vertical axis [1]. In this motion the aerodynamic pressure distribution, the integrated forces and moments,

5
and the gravity projections are all constant with time in the rigid body frame of reference. The steady-state
conditions are characterized by the state vector, composed of motion variables in the selected reference frames, and
the control vector, composed of aerodynamic surface deflections and throttle position. This description of a steady
state motion may be expanded by a number of parameters, characterizing maneuver kinematics. These parameters
include the flight path angle
sin cos cos sin sin sin cos sin cos cos cosγ α β θ β φ θ α β φ θ= − − , (8)
the steady turning rate Ω or the radius R of the helical trajectory, which are related as
R
V γcos
=Ω (see Fig. 1a).
The kinematical triad ,V γ and Ω (or ,V γ and R ) completely specifies the helical trajectory [11]. The yaw
angle in steady helical motion varies with time and its time derivative defines the steady turning rate ψ = Ω , which
can be expressed (Eq. 7) along with the radius of steady helical trajectory R through the state variables as follows
( )sin cos secq rφ φ θΩ = + (9)
cos cos
sin cos
V
R
q r
γ θ
φ φ
=
+
(10)
The steady helical motion may be a straight-and-level flight with 0Ω = and 0γ = , or a level turn with
0γ = and 0Ω ≠ . Note that aircraft steady spin motion is also a helical solution of system (1 - 6). It is
characterized by high angle of attack, very small trajectory radius comparable with the wing span ( / 1R b ≈ ) and
near vertical flight path angle / 2γ π≈ − .
The first eight equations (1- 6) can be expressed in the general form of a nonlinear first-order vector differential
equation
8 4
( , ), X , U ,= ∈ ⊂ ℜ ∈ ⊂ ℜx f x δ x δ (11)
where the aircraft state vector ( , , , , , , , )T
p q r Vα β θ φ=x is composed of the angular velocity components in the
body-fixed reference frame ( , , )p q r , the linear velocity components in the wind-body reference frame
( , , )Vα β and the two Euler angles ( , )θ φ . The aircraft state vector x belongs to the bounded region X ,
specified by the validity of aerodynamic model and the imposed constraints on the aerodynamic loads. The control

6
vector ( , , , )T
e a r Tδ δ δ δ=δ is composed of control surface deflections ( , , )e a rδ δ δ and throttle position Tδ . It
belongs to the bounded region U, which is specified by control deflection limits: min maxi i iδ δ δ≤ ≤ , where
i = e,a,r,T .
The attainable turning rate in steady helical motion is normally much lower than the maximum roll rate
attainable in intensive manoeuvring [22]. For example, during intensive rotation in roll, shown in Fig. 1b, the state
variables of Eq. (11) can be approximated by a sum of pseudosteady mean values and superimposed small amplitude
oscillations, produced by harmonic gravity projections in the body-fixed axes. The pseudosteady components can be
analyzed as steady states of the reduced 5th
order system for the first five state variables ( , , , , )T
p q r α β=x and
the reduced control vector ( , , )T
e a rδ δ δ=δ [24]. The flight velocity is assumed constant and the gravitational
force is neglected. The steady states of this 5th
order dynamic system are analyzed in Section IV for evaluation of
attainable values for angle of attack, sideslip and the velocity-vector roll rate
cos cos cos sin sin cosV p q rα β α β α βΩ = + + (the projection of the total angular velocity vector Ω on
the velocity vector V ). Note that the steady-state solution of the reduced 5th
order system also corresponds to a
helical trajectory, but with an axis arbitrarily oriented in space.
Trim and linearization
The investigation of the steady states of system (11) and the reduced 5th
order sub-system constitutes a fundamental
problem in flight dynamics. The steady states are important for determining sustainable performance and
maneuvering capabilities, they can be used as initial conditions for flight simulation and for linearization of the
aircraft nonlinear dynamics. Since the steady states are characterized by zero rate of change of the linear and angular
velocity components in the selected reference frames, the steady-state problem is formulated by setting all time
derivatives in the equations of motion equal to zero and solving the resulting set of nonlinear algebraic equations
( , ) 0, X , U ,n m
ε ε ε ε= ∈ ⊂ ℜ ∈ ⊂ ℜf x δ x δ (12)
where εx and εδ are the steady-state vector and trim control vector, respectively.
For small state and control vector perturbations, ε= −x x x and ε= −δ δ δ , the equations of motion (11)
can be approximated by the linear time invariant system

7
= +x A x B δ , (13)
where
( , ), ( , )ε ε ε ε= =x δA f x δ B f x δ
are the state and control matrices, respectively. A steady state εx is locally stable if the real parts of all the
eigenvalues of the A matrix are negative. If the real part of any eigenvalue of the A matrix is positive, the steady
state is locally unstable. The system is attracted to a stable steady state in its neighborhood and repelled if the steady
state is unstable. The eigenvectors of the A matrix characterize subspaces for different modes of disturbed motion.
The linearized response to control input δ is characterized by the control B matrix. The A and B matrices
calculated on a grid of points for different flight regimes and steady maneuvers provide a linear parameter varying
approximation for the original nonlinear system (11), which is equally important for the open-loop dynamics
analysis and for control law design.
Variation of flight regime and/or maneuver parameters can lead to various bifurcational changes in dynamics of
system (11). Local bifurcations can be identified via changes in eigenvalues of the linearized system (13). More
information on dynamical systems theory, qualitative and bifurcation analysis can be found in [4].
III. Computation of attainable equilibrium sets
A systematic reconstruction of all the attainable equilibrium states is described in this section as a method for
evaluation of aircraft performance and maneuvering flight envelopes. The computational problem for steady states is
formulated for equations of motions augmented by auxiliary equations specifying maneuver kinematics and
maneuver parameters. The solutions of this constrained trim problem are presented in the form of attainable
equilibrium sets computed on a grid of points in the plane of two selected parameters characterizing flight regime
and steady maneuver. All steady states are classified according to the eigenvalue spectrum of the linearized system
(13). This classification generates local stability maps giving a qualitative insight into the aircraft nonlinear
dynamics. For every steady state the A and B matrices of the linearized system (13), the eigenvalues and
eigenvectors of the Jacobian matrix A are stored in a special data structure along with the trim control parameters,
so that they can be easily retrieved for later detailed analysis.

8
a)
b)
Fig. 1 Kinematics of steady state helical motion (a) and velocity-vector roll maneuver (b).

9
Deflection constraints on aircraft control inputs limit the available control power so that the set of equilibrium
states for the system described by Eq. (11) is bounded. The set of attainable equilibrium states Sε incorporates all
possible trim states generated by the available bounded control U∈δ in the allowable region of state variables
X∈x and is defined as:
{ }S : ( , ) 0, X , Un m
ε ε ε ε ε ε= = ∈ ⊂ ℜ ∈ ⊂ ℜx f x δ x δ (14)
The attainable equilibrium states in (14) belong to a wide class of aircraft steady helical trajectories with vertical
axis of rotation, when 8n = , and arbitrary oriented axis, when 5n = . Some additional kinematical constraints are
required to specify a particular steady maneuver for performance or maneuverability evaluation.
A. Constrained trim formulation
. Every steady maneuver from the complete set of equilibrium solutions (14) may be additionally characterized
by a vector of kinematical parameters µ
k
∈ℜy (see Eqs. 8-10), which is functionally linked with the state vector
( )µ =y g x . To determine the required trim control vector εδ and the equilibrium state vector εx for a particular
maneuver, specified by parameter vector µy , the following augmented system of trim equations should be solved:
µ
( , ) 0, , X , U
( ), ,
n n m
k
ε ε ε ε
ε
= ∈ℜ ∈ ⊂ ℜ ∈ ⊂ ℜ
= ∈ℜ
f x δ f x δ
y g x g
(15)
where k -dimensional )( mk ≤ vector function g defines the required maneuver. When the number of constraints
in (15) equals the number of control parameters, i.e. mk = , the system (15) represents a set of )( mn + equations
with )( mn + unknowns. This means that for every particular maneuver specified by vector µy there exists an
isolated solution of the system (15) consisting of steady-state vector εx and control vector εδ . When mk < the
solution of augmented trim system (15) will not be unique and isolated and the )( km − free control parameters can
be used for continuation along the )( km − -dimensional manifold of equilibrium solutions.

10
In trimming procedure for the system (12) the control input εδ is known a priori, while in the system (15) the
trim control vector εδ is defined implicitly by the augmented system of )( kn + -equations. In Eq. (12), εδ is the
input vector and εx is to be determined. In Eqs. (15), maneuver specification µy is the input vector, while steady
state εx and trim control εδ are to be determined. The equilibrium state vector εx and corresponding trim control
vector εδ can be determined on a grid of points in the maneuver parameter space µy allowing the reconstruction of
the steady flight envelopes, where boundaries are specified by bounded regions X and U .
The vector function g and the parameter vector µy in (15) specify the maneuver type and the maneuver
kinematical parameters, respectively. To have a unique isolated solution in (15), when 8n = and 4m = , it
requires four different constraints. A helical trajectory with vertical axis can be described by three kinematical
parameters such as flight velocity V , flight path angle γ and either the turn radius R or the angular rate Ω (see
Eqs. 8 - 10). One additional constraint can be imposed on aircraft attitude to complete the trim problem, such as a
prescribed value for the sideslip angle β or for the roll angleφ . So, the full vector of constraint functions can be
selected as one of the following combinations: 1) ( , , , )T
V R γ β=g , 2) ( , , , )T
V γ β= Ωg , 3) ( , , , )T
V R γ φ=g
or 4) ( , , , )T
V γ φ= Ωg . The steady straight-and-level flight, for example, can be defined by vector
µ µ µ µ µ( , 0, 0, )T
V γ β= Ω = =y , the steady level turn by vector µ µ µ µ µ( , , 0, )T
V R γ β= =y . The functional
representations for , ,Rγ Ω are given in Eqs. (8 – 10). The velocity, sideslip and roll angle ( , , )V β φ are the
components of the state vector x , hence the kinematical constraints can be expressed in a straightforward form
µ µ µ, ,V V β β φ φ= = = .
For analyzing steady states in the velocity-vector roll maneuver system (11) is considered for
( , , , , )T
p q rα β=x and ( , , )T
e a rδ δ δ=δ , the reduced state and control vectors, respectively. To have a unique
isolated solution the constrained trim formulation (15) should include three kinematical constraints, for example,
T
V( , , )α β= Ωg or alternatively V( , , )T
z yn n= Ωg , where zn and yn are the normal and side load factors,
respectively. In the latter case the constraint vector function will depend also on control input ( , )g x δ . The

11
constraint equations for the investigation of the velocity-vector roll maneuver will have the following form:
µ µ Vµ, , cos cos cos sin sin cosp q rα α β β α β α β α β= = + + = Ω . The two other projections of angular
velocity vector in the wind-body axes are defined as function of ,α β and VΩ . Solution of system (15) in this case
again will be unique and isolated since the dimension of control vector ( , , )T
e a rδ δ δ=δ is equal to the number of
constraint equations.
B. Computational procedure
The constrained trim problem (15) is solved numerically using the Newton-Raphson method
11
ε εε ε
1
ε ε µε ε
( , )
( , )
k kk k
k kk k
−+
+
 
= −  
 
x δ
x δ
f f f x δx x
g g g x δ - yδ δ
, (16)
where matrices xf , δf , xg and δg are evaluated numerically. The convergence process is terminated when all
components of the increment vectors
1k k k
ε ε ε
+
= −x x x and
1k k k
ε ε ε
+
= −δ δ δ become less than some prescribed
accuracy levels
k
i max minx /( )i i ix x ε− and max min/( )k
j j j jδ δ δ ε− , respectively. The convergence
process is also terminated if the accuracy criteria have not been met after a preset maximum number of iterations.
Note that an alternative way of solving this problem is described in [2] using a scalar cost function and a function
minimization algorithm.
For every new maneuver parameter vector µy the initial guess for solution
0 0
ε ε,x δ is taken equal to a successful
solution obtained for previous value of parameter vector µy . In computation of the starting solution of system (15)
the modules for increments of the state and control vectors during iterative process are confined by some preset
values, but the maximum number of iterations is increased.
Maneuver parameter vector µy is classified as attainable one if the steady state vector εx and trim control
vector εδ have successfully converged to the internal points of the bounded regions X and U , respectively. If
during the Newton-Raphson iterations the state vector εx is pushed out of the region X or the control vector εδ is
saturated on the boundary of the region U further reduction of the error norms becomes impossible. In this case the

12
selected maneuver parameter µy is classified as unattainable. Note that after successful convergence the iterative
algorithm (16) automatically provides the state and control matrices for the linearized system (13).
To represent multi-dimensional set of all equilibrium states Sε (14), two-dimensional cross-sections are
computed for different, fixed values of the other flight regime and maneuver parameters. These cross-sections, or
attainable equilibrium sets, are computed on a square grid of points with spacing allowing accurate representation of
external boundaries and boundaries for internal regions of steady states with different local dynamics. The grid
points with attainable steady-state parameters are displayed using different markers reflecting their local stability
properties, and the remaining grid points are left unmarked. The marked area on a selected grid approximates the
two-dimensional cross-section of the attainable equilibrium set and provides a local stability map for attainable
equilibria [23]. The steady state local stability properties are defined by the eigenvalues of the system A matrix in
(13).
Fig. 2 Qualitative classification of eigenvalues spectrum

13
A set of markers, shown in Fig. 2, was selected to visualize the local stability of steady-state solutions. Stable
steady states are marked by solid circles, while for unstable steady states several different markers are used to reflect
topology of the local dynamics: 1) saddle points, or aperiodically unstable equilibria with one positive real
eigenvalue, are marked by crosses; 2) unstable focuses, or oscillatory unstable equilbria with unstable complex
conjugate pair of eigenvalues, are marked by empty circles; 3) unstable nodes, or equilibria with two positive real
eigenvalues, are marked by star markers; 4) saddle-focus points, or equilibria with three unstable eigenvalues (one
real and one complex conjugate pair), are marked by diamonds; and 6) saddle-node points, or equilibria with three
unstable real eigenvalues, are marked by pentagrams.
For each marked grid point all computed data for a given steady state are stored in a special data structure for
later use and visualization. The stored data include the state vector εx and associated trim control vector εδ , the
state and control matrices of the linearized system (13), and also the eigenvalues and the eigenvectors of the
A matrix.
IV. Computational results
The mathematical models for the F-18 HARV, F-16 and the F-14 aircraft (corresponding geometric, mass, inertia
and aerodynamic data can be found in [25-27]) are used here as the case studies for testing the proposed
computational method. The presented analysis is limited by consideration of the bare-airframe models without
inclusion of stability and control augmentation system. The computation of attainable equilibrium sets of the closed-
loop system may be a useful method for post-design validation of control laws, but this is beyond the scope of this
paper. The example of such closed-loop system analysis is presented in [21] for the ADMIRE model with the
nonlinear dynamic inversion control law.
The computation of attainable regions for a class of steady helical trajectories using the 8th
order system (11) has
been performed only for the first aircraft model, while the investigation of attainable equilibrium sets using the
reduced 5th
order system (11) has been performed for all three aircraft models.
Aircraft A
The source of aerodynamic data used for the F-18 HARV aircraft is a simplified model available in [25] along
with aircraft mass and inertia parameters. The model is nonlinear with respect to angle of attack and is linear with
respect to all other state variables and control surface deflections. The model is valid for small velocities

14
( 0.6M ≤ ) and covers a wide range of angles of attack (-14 to 90 degrees). The reference center-of-gravity
position is 0.25cgx c= .
Figures 3a,b show the altitude-Mach number flight envelope for steady level turn ( 0)γ = with 800 mR =
and 0β = . The boundary of this flight envelope at high Mach numbers is specified by maximum throttle position
max( )T Tδ δ= ; the angle of attack at this boundary is small (
0
10α ≈ ), the roll angle is very high (
0
80φ ≈ ).
a) b)
c) d)
Fig. 3 The F-18 HARV altitude-Mach number flight envelope for level turn ( 0)γ = with 800 mR = and
0β = - a) right rotation and b) left rotation; c) the eigenvalues for two steady turns with left rotation at
M=0.6, H=1.5 km and M=0.15, H=1.5 km ; d) set of attainable steady helices for H=5 km, M=0.4 .

15
The low Mach number boundary corresponds to saturated deflection of elevator min( )e eδ δ= ; the angle of attack at
this boundary is very high (
0
50α ≈ ) and the roll angle is moderate (
0
30φ ≈ ). The regions for attainable steady
maneuvers with right rotation (Fig. 3a) and with left rotation (Fig. 3b) are practically identical, but the local stability
maps are different. The left level turns (Fig. 3b) are more stable than the right level turns (Fig. 3a) due to effect of
the engine rotors. The angular momentum of the engine rotors (
2
ph =15000 kg m /s⋅ ) destabilizes the slow
phugoid and spiral motion modes in maneuvers with right rotation.
The eigenvalues of steady level turns with left rotation are shown in Fig. 3c for two different flight regime
points. At M=0.6, H=1.5 km the eigenvalues (solid circles) are all stable including the spiral and phugoid
modes. At the left boundary of flight envelope (M=0.15, H=1.5 km) the eigenvalues (empty circles) indicate
slower dynamics in all modes and oscillatory instability of the phugoid mode.
The attainable set of all steady turns is presented in Fig. 3d for flight envelope point H=5 km, M=0.4 . The
attainable maneuvers are reconstructed in the plane of flight path angle γ and turning angular rate Ω . The
attainable set includes ascending and descending straight trajectories on a segment 0Ω = and
0 0
8 30γ− < < ,
level turns on a segment 0γ = and 9 deg/sΩ ≤ , gliding ( 0, 0)γΩ ≠ > and diving ( 0, 0)γΩ ≠ < helical
trajectories. Helical trajectories with negative rotation are relatively more stable, than helical trajectories with
opposite rotation due to effect of the engine rotors. The right and left boundaries of the attainable region correspond
to the maximum ( )maxT Tδ δ= and zero throttle position ( )0Tδ = , respectively. Two “sleeves” of the attainable
region at negative path angles indicate attainable “steep dive” maneuvers, which can be stable or unstable depending
on path angle and turning angular rate.

16
a) b)
c) d)
Fig. 4 The F-18 HARV attainable equilibrium sets for the velocity-vector roll maneuver (M=0.4, H=5 km)
a) ( , )α β −cross-section at 0VΩ = ; b) −Ω ),( Vα cross-section at 0β = ; c) ( ),V βΩ − cross-section
at 10α = deg; d) ( ),V βΩ − cross-section at 20α = deg.
Four different cross-sections of the attainable equilibrium states of the reduced 5th
order system (11) are
presented in Figs. 4a,b,c,d for flight regime H=5 km, M=0.4 . The first cross-section (Fig. 4a) shows the
attainable region for steady states in the plane of angle of attack and sideslip angle ( ),α β , when the velocity-vector
roll rate is kept at a zero value ( 0)VΩ = . The rotation vector in this case is normal to the linear velocity vector

17
and the aircraft follows a complete loop, with the radius
2
/R mV L= and angular rate /L mVΩ = , where L is
the lift force (the side force D is assumed equal zero). At non-zero value 0VΩ ≠ , the loop is transformed into a
helical trajectory. The radius of this helical trajectory will decrease with increase of VΩ (see Fig. 1b).
The second cross-section (Fig. 4b) shows the attainable region for angle of attack α and the velocity-vector roll
rate VΩ at zero value of sideslip angle ( 0)β = . The last two cross-sections (Figs. 4c,d) show the attainable
regions for the velocity-vector roll rate VΩ and sideslip angle β at two values of angle of attack 10α = deg and
20α = deg, respectively.
Each boundary of attainable steady-state region, presented in Figs. 4 a,b,c,d, is associated with saturation of one
of the control surfaces, i.e. rudder, aileron or elevator. The trim control inputs at the boundary clearly indicate the
control surface which limits the expansion of the attainable region. Characterization of local stability of steady
states using different markers (see Fig. 2) provides a basis for prediction of aircraft dynamics at different
combinations of control inputs. For example, the aircraft A is stable at low angles of attack ( )0
27α < and
a) b)
Fig. 5 a) Trim deflections of control surface as function of velocity-vector roll rate VΩ in maneuver (see
arrow in Figs. 4 b,c) with constant angle of attack 10α = deg and 0β = ; b) The locus of eigenvalues for
selected sequence of equilibria.

18
moderate values of rotation rate ( )0 0
60 deg/s, 8 20V αΩ ≤ < < (Figs. 4a,b). Maneuvers with intensive rate of
rotation ( )0 0
60 deg/s, 8 20V αΩ > < < become oscillatory unstable due to destabilizing effect of nonlinear
inertia coupling terms in equations (1).
Note that the accumulated data for trim control values and eigenvalues of the steady states can be used for
dynamics analysis in a way similar to continuation and bifurcation analysis method. Figs. 5a,b present trim
deflections of the control surfaces and the locus of eigenvalues stored in a special data structure for the maneuver
specified in Figs. 4 d,c by the horizontal and vertical arrows, respectively. The maneuver holds angle of attack and
sideslip constant ( 10α = deg, 0β = ) and velocity-vector roll rate VΩ varies from zero to 145deg/s. The trim
control inputs and the locus of eigenvalues are presented as function of the velocity-vector roll rate VΩ . The
transition of the Dutch roll mode eigenvalues to the right hand side of the complex plane at 60VΩ = deg/s
indicates the Hopf bifurcation of the steady state. At 60VΩ > deg/s the steady states become oscillatory unstable
and dynamics converge to stable limit cycle oscillations (LCO). Qualitative analysis of the aircraft dynamics using
attainable equilibrium sets and local stability maps can be effectively complemented by direct numerical simulation.
However, this is out of the scope of the present paper.
Aircraft B
The F-16 aircraft mathematical model used here as a second case study is based on the wind tunnel low-speed
aerodynamic data from [26], where they were used in piloted simulation for investigation of the maneuvering
characteristics and stall/poststall behaviour of a relaxed static stability airplane. The aerodynamic data cover a wide
range of angle of attack (-20 to 90 degrees), and of sideslip angle (-30 to 30 degrees). Most of the aerodynamic
coefficients are interpolated using the two-dimensional data tables with angle of attack and sideslip as independent
variables. The reduced aerodynamic data set derived from [26] for the F-16 model and the mass and inertia
characteristics are presented in [2].
Figs. 6 a,b,c,d show four cross-sections of the attainable equilibrium states similar to those shown in Figs. 4
a,b,c,d for aircraft A. The same flight regime and graphical layout of computed cross-sections allow consistent
comparison of flight mechanics parameters for different airframes. For example, at low angles of attack aircraft A
can be trimmed at higher sideslip angles, but its maximum attainable roll rate is approximately three times less than

19
maximum attainable role rate of aircraft B (Figs. 4a,b and Figs. 6 a,b). The maximum velocity vector roll rate of
aircraft A is limited due to saturation of rudder and ailerons (Fig. 4b), while for aircraft B the limiting factor is
a) b)
c) d)
Fig. 6 The F-16 attainable equilibrium sets for the velocity-vector roll maneuver (M=0.4, H=5 km) a)
( , )α β −cross-section at 0VΩ = ; b) −Ω ),( Vα cross-section at 0β = ; c) ( ),V βΩ − cross-section at
10α = deg; d) ( ),V βΩ − cross-section at 35α = deg.
saturation of elevator deflection (Fig. 6b). These two airframes have also different attainable steady-state regions in
( ),V βΩ - cross-sections, they are presented in Figs. 4c,d and Figs. 6c,d.
Aircraft A is aerodynamically stable at low angles of attack, while the aircraft B is unstable in pitch at the same
flight conditions at a reference location of center of gravity 0.35cgx c= . Positive real eigenvalue 0.5λ = rad/s

20
(Fig. 7b) indicates the level of instability in pitch. Stability augmentation feedback control law can change the
aircraft closed-loop eigenvalues and ensure stable dynamics. The control law gains will depend on the open-loop
system eigenstructure, which can vary with a change of maneuver [19]. To illustrate this the locus of eigenvalues
and associated sequence of trim control inputs are presented in Fig. 7b as a function of velocity-vector roll rate VΩ
(the maneuver is specified by two arrows in Figs. 6b,c). The eigenvalues of an open-loop system during intensive
roll maneuver due to inertia coupling are changed dramatically. In this example there are two qualitative changes in
the eigenvalues spectrum. The first one is when the unstable eigenvalue of the short-period mode moves to the stable
semi-plane with potential onset of the pitch-fork bifurcation of steady states. Following this event the system
becomes stable for a small range of velocity-vector roll rate (Fig. 7a). But transition of the Dutch roll mode
eigenvalues to the right-hand side of the semi-plane makes the system unstable again. Note that after the Hopf
a) b)
Fig. 7 a) Trim deflections of control surface as function of velocity-vector roll rate VΩ in maneuver (see
arrow in Figs. 6 b,c) with constant angle of attack 10α = deg and 0β = ; b) The locus of eigenvalues for
selected sequence of equilibria.
bifurcation, the system will be attracted to the stable limit cycle oscillations. The amplitude of these oscillations
increases with the level of equilibrium instability.
At high angles of attack the attainable steady state regions shown in Fig. 6 are asymmetrical vs sideslip angle
due to asymmetry in aerodynamic data. This asymmetry is quite typical for fighter aircraft configuration due to
onset of asymmetrical vortical flow structures, which are observed both in wind tunnel tests and in flight. The cross-
section of attainable equilibrium set presented in Fig. 6b indicates existence of two groups of equilibrium states

21
with positive and negative rotation at very high angles of attack. These equilibria represent aircraft flat spin motion
at zero sideslip with appropriate deflection of all control surfaces. The cross-sections of attainable equilibrium set
additionally computed for different values of sideslip angle reveal that these two steady-state “islands” are
continuously connected with the “mainland” of steady states.
Aircraft C
The third case study deals with the F-14 aircraft model, which is based on the wind tunnel aerodynamic data
from [27]. These data were used in piloted simulation for investigation of high-angle-of-attack maneuver-limiting
factors. The aerodynamic model of the F-14 aircraft is presented also in [9], where the aircraft nonlinear dynamics
have been investigated using continuation and bifurcation analysis methods. The aerodynamic data cover a wide
range of angle of attack (0 to 90 degrees), and of sideslip angle (-20 to 20 degrees), and correspond to the wings
fully swept forward. They are valid for low speeds ( )0.6M < and describe nonlinear dependencies of
aerodynamic coefficients on angle of attack, sideslip angle and elevator deflection.
Examples of the attainable equilibrium sets for aircraft C are shown in Figs. 8 a,b,c,d. They are computed for the
same flight regime with similar cross-sections as in two previous cases shown in Figs. 4 a,b,c,d for aircraft A and in
Figs. 6a,b,c,d for aircraft B. The attainable equilibrium regions in all cross-sections have well-defined boundaries,
which correspond to saturation of rudder, ailerons and elevator. At low angles of attack ( )10 degα < the
attainable range of sideslip angle ( )15 degβ < is bigger than in the case of aircraft B and less than in the case of
aircraft A. However, the attainable range for the roll rate is much less ( )40 deg/sVΩ < than in the first two
cases. There are two detached regions with stable steady states at low angles of attack ( )14 degα < and high
angles of attack ( )21 degα > . These two regions are separated by region with oscillatory unstable steady states.
Note that at high angles of attack the maximum attainable roll rate significantly increases, but the steady states are
oscillatory unstable (Fig. 8b). Performing longitudinal and velocity-vector roll maneuvers the aircraft can enter the
unstable region of steady states, which will produce a destabilizing effect on aircraft dynamics. For example, Fig. 9b
shows the locus of eigenvalues for a longitudinal pull-up maneuver, denoted in Figs. 8a,b by vertical arrows, and

22
Fig. 9a shows the associated control surface deflections. In the angle of attack range ( )0 0
14 21α≤ ≤ , where the
Dutch roll mode eigenvalues are relocated to the right-hand side semiplane, the aircraft will experience oscillatory
a) b)
c) d)
Fig. 8 The F-14 attainable equilibrium sets for the velocity-vector roll maneuver (M=0.4, H=5 km) a)
( , )α β −cross-section at 0VΩ = ; b) −Ω ),( Vα cross-section at 0β = ; c) ( ),V βΩ − cross-section at
10α = deg; d) ( ),V βΩ − cross-section at 20α = deg.
instability in the lateral directional motion (Fig. 9 b). There are two Hopf bifurcation points at 14 degα = and at
21 degα = (Fig. 9a) similar to the results presented in [9], where they were obtained using the continuation
method. The bifurcation parameters in our case are slightly different, because the speed is fixed accordingly to
selected flight regime H=5 km, M=0.4. The simulation results showing the buildup of the wing rock amplitude are

23
shown in [9]. The Hopf bifurcation will also take place when the aircraft is performing the velocity-vector roll
maneuver as it was shown for aircraft A (Fig. 4b).
a) b)
Fig. 9 a) Trim deflections of control surface as function of angle of attack α in maneuver (see arrows in Figs.
8 a,b) with zero sideslip 0β = and zero roll rate 0VΩ = ; b) The locus of eigenvalues for selected sequence of
equilibria.
The computational time for one attainable equilibrium set in the form of two-dimensional cross-section depends
on a size of the grid and also on a number of unattainable points. This is because each unattainable point will use the
maximum number of iterations before moving to the next point. Attainable points can normally be converged in a
few iterations. An average computation time for presented examples of attainable equilibrium sets with a grid 40x40
points is about 15 minutes on a PC with Intel©
Centrino Duo processor.
V. Conclusions
Consideration of the nonlinear models of the F-18 HARV, F-16 and the F-14 aircraft allows us to conclude that:
1) The computation of the set of all attainable steady states of the rigid aircraft equations of motion with imposed
constraints assigning maneuver type and its parameters provides a feasible and consistent way of investigating the
nonlinear aircraft dynamics which allow evaluation of the aircraft performance and maneuvering capabilities more
efficiently.
2) A number of representative attainable equilibrium sets computed on a two-dimensional grid of points in the
maneuver parameter space can reveal a global structure of the multidimensional set of all steady states which

24
includes the aircraft maneuver limitations produced by saturation of control inputs and local stability properties in
the form of stability maps.
3) The proposed method:
- possesses similarities and complementary features when compared with the bifurcation analysis method based
on one-parameter continuation of equilibrium states and it is expected that the two approaches can be used in a
complementary way;
- can significantly facilitate the mathematical and piloted simulation methods by determining the critical regions
of operation where simulation is necessary to verify the model’s dynamic behavior and it can be used for
comparative analysis of different airframes;
- has a high potential for designing control laws (particularly for high-angle-of-attack flight and intensive roll
maneuvers) via provision of interconnect schedules between different control surfaces, generation of linearized
equations of motion consistent with performing maneuver and also for post-design evaluation of the designed
control laws;
- in combination with high fidelity aircraft models validated in flight tests will provide reliable quantitative
results for actual performance and maneuvering capabilities.
Acknowledgments
The work partly stems from the research project funded at De Montfort University by DERA/QinetiQ, Bedford
UK in 1998-2004. The authors gratefully acknowledge this support with special thanks to Yoge Patel for
encouraging cooperation. The contribution of Hobina Rajakaruna along with the insightful comments and
recommendations form the reviewers, which helped to improve the presentation, are gratefully acknowledged.
References
[1] Etkin, B., Dynamics of Atmospheric Flight, J.Wiley & Sons, New York, 1972.
[2] Stevens, B.L., and Lewis, F.L., Aircraft Control and Simulation, 2nd
edition, J.Wiley & Sons, New York, 2003.
[3] Pamadi, B.N., Performance, Stability, Dynamics, and Control of Airplanes, AIAA Education Series, 1998.
[4] Guckenheimer, J. , and Holmes, P., Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector
Fields, Springer-Verlag, New York, 1983.
[5] Mehra, R.K., Kessel, W.C., and Carrol, J.V., “Global Stability and Control Analysis of Aircraft at High Angles
of Attack,“ Annual Technical Reports 1/2/3, ONR-CR215-1/2/3 Scientific Systems Inc., USA, 1977/1978/1979.
[6] Carroll, J. V., and Mehra, R. K., “Bifurcation Analysis of Nonlinear Aircraft Dynamics”, Journal of Guidance,
Control, and Dynamics, Vol. 5, No. 5, 1982, pp. 529-536.
[7] Guicheteau, P., “Bifurcation theory applied to the study of control losses on combat aircraft”, La Recherche
Aѐrospatiale, 2, 61-73, 1982.

25
[8] Zagaynov, G. I., and Goman, M. G., “Bifurcation Analysis of Critical Aircraft Flight Regimes”, Proceedings
ICAS-84-4.2.1, 1984.
[9] Jahnke, C. C., and Culick, F. E. C., “Application of Bifurcation Theory to the High-Angle-of-Attack Dynamics
of the F-14”, Journal of Aircraft, Vol. 31, No. 1, 1994, pp. 26-34.
[10] Lowenberg, M.H., “Stability and controllability evaluation of sustained flight manoeuvres”. AIAA
Atmospheric Flight Mechanics Conference, San Diego, CA, July 29-31, 1996, AIAA-1996-3422, Technical
Papers (A96-35084 09-08).
[11] Avanzini, G., de Matteis, G., and de Socio, L.M., “Aircraft Equilibrium and Response to Control in Intrinsic
Coordinates”, AIAA Paper 97-0323, 35th
Aerospace Sciences Meeting and Exhibit, Erno (NV), January 1997.
[12] Avanzini, G., and de Matteis, G., “Bifurcation Analysis of a Highly Augmented Aircraft Model”, Journal of
Guidance, Control and Dynamics, Vol. 20, No. 4, 1997, pp. 754-759.
[13] Goman, M. G., Zagaynov, G. I., and Khramtsovsky, A. V., “ Application of Bifurcation Theory to Nonlinear
Flight Dynamics Problems, ” Progress in Aerospace Sciences, Vol. 33, No. 9, 1997, pp. 539-586.
[14] Pashilkar, A.A., and Pradeep, S., “Computation of Flight Mechanics Parameters Using Continuation
Techniques”, Journal of Guidance, Control, and Dynamics, Vol. 24, No. 2, March-April 2001, pp.324-329.
[15] Ananthkrishnan, N., and Sinha, N. K., “ Level Flight Trim and Stability Analysis using an Extended
Bifurcation and Continuation Procedure, “ Journal of Guidance, Control, and Dynamics, Vol. 24, No. 6, 2001,
pp. 1225-1228.
[16] Charles G, Lowenberg M, Stoten D, Wang X, di Bernardo M. “Aircraft Flight Dynamics Analysis and
Controller Design Using Bifurcation Tailoring,” AIAA 2002-4751, AIAA Guidance, Navigation, and Control
Conference and Exhibit, Monterey, CA , 2002.
[17] Chudoba, B., and Cook, M.V., “Trim Equations of Motion For Airrcaft Design: Turning Flight, Pull-Up and
Push-Over”, AIAA Atmospheric Flight Mechanics Conference and Exhibit, Austin, Texas, 2003, Paper AIAA
2003-5693.
[18] Goman, M.G., Patel, Y., and Khramtsovsky, A.V., “Flight Clearance Tools Using a Non-linear Bifurcation
Analysis Framework”, AIAA Guidance, Navigation, and Control Conf. and Exhibit. AIAA Paper 2003-5557,
Austin, TX, 2003.
[19] Richardson, T.S., and Lowenberg, M.H. “A Continuation Design Framework for Nonlinear Flight Control
Problems”, The Aeronautical Journal, Vol. 110, No.1104, pp. 85-96, 2006.
[20] Paranjape A, Sinha NK, Ananthkrishnan N. “Use of Bifurcation and Continuation Methods for Aircraft Trim
and Stability Analysis - A State-of-the-Art,” AIAA 2007-1051, 45th AIAA Aerospace Scineces Meeting and
Exhibit, Reno, Nevada, 2007.
[21] Goman, M.G., Khramtsovsky, A.V., and Kolesnikov, E.N. “Investigation of the ADMIRE Manoeuvring
Capabilities Using Qualitative Methods”, in Nonlinear Analysis and Synthesis Techniques for Aircraft Control
Series: Lecture Notes in Control and Information Sciences,Vol. 365 Bates, Declan; Hagström, Martin (Eds.)
2007, VIII, 360 p. Also available online., Softcover ISBN: 978-3-540-73718-6
[22] Durham, W.C., Lutze, F.H., and Mason, W. “Kinematics and Aerodynamics of Velocity-Vector Roll”, Journal
of Guidance, Control, and Dynamics, Vol. 17, No. 6, 1994, pp. 1228-1233.
[23] “Maneuver Limitations of Combat Aircraft”, AGARD Advisory Report No.155A, AGARD-AR-155A, 1979.
[24] Young, J.W., Schy, A.A., and Johnson, K.G., “Pseudosteady-state Analysis of Nonlinear Aircraft Maneuvers”,
NASA Technical Paper 1758, NASA Langley Research Centre, Hampton (VA), 1980.
[25] F-18 High Alpha Research Vehicle (HARV) Working Documents
(http://www.nasa.gov/centers/dryden/history/pastprojects/HARV/Work/NASA2/nasa2.html), February 2007.
[26] Nguyen, L.T., Ogburn, M.E., Gilbert, W.P., Kibler, K.S., Brown, P.W., and Deal, P.L.”Simulator Study of
Stall/Post-Stall Characteristics of a Fighter Airplane With Relaxed Longitudinal Static Stability”, NASA
Technical Paper 1538, LaRC, Hampton, Virginia, 1979.
[27] Mitchell, D.G., Myers, T.T., Teper, G.L., Jonston, D.E., “Investigation of High-Angle-of-Attack Maneuver-
Limiting Factors”, Part I: Analysis and Simulation, Part II: Piloted Simulation Assessment of Bihrle Departure
Criteria, Part III: Appendices – Aerodynamic Models, Flight Dynamics Laboratory, Wright-Patterson AFB,
Technical report AFWAL-TR-80-3141, 1980

Goman, Khramtsovsky, Kolesnikov (2008 draft) - Evaluation of Aircraft Performance and Maneuverability by Computation of Attainable Equilibrium Sets

More Related Content

What's hot (19)

Similar to Goman, Khramtsovsky, Kolesnikov (2008 draft) - Evaluation of Aircraft Performance and Maneuverability by Computation of Attainable Equilibrium Sets (20)

More from Project KRIT (18)

Recently uploaded (20)

Goman, Khramtsovsky, Kolesnikov (2008 draft) - Evaluation of Aircraft Performance and Maneuverability by Computation of Attainable Equilibrium Sets