Chapter 5

Wing Rock

Yalew Mekonnen Fenta

12/10/2023 1
5.1. Wing Rock Onset And Its Prediction

The wing rock is an oscillation, predominantly in the lateral-directional variables,

induced when the Dutch roll mode loses stability.

The onset of wing rock corresponds to the occurrence of a Hopf bifurcation.

Wing rock is a fairly commonly observed phenomenon on military airplanes, hence
identifying the occurrence of wing rock is a matter of considerable interest.

 However, of equal or greater interest is the prediction of the possibility of wing

rock during the aircraft design process and ways to avoid the occurrence of wing
rock.

 Therefore, the development of a relatively simple criterion for wing rock onset is
a significant issue.

12/10/2023 2
 The onset of wing rock is related with a nonlinear variation of roll damping
derivative with α, sideslip angle β, oscillation frequency and amplitude [1].
 Wing rock is primarily observed by means of wind tunnel free-to-roll experiments
for very slender delta wings.

Fig. 5.2: Wing rock build-up of an 800 delta wing at a = 270

12/10/2023 3
12/10/2023
Fig. 5.1: Separated vortex fows and some undesirable airplane rotary motions 4
 Fig. 5.2: Emergence of stable limit cycle family at a supercritical Hopf bifurcation.

12/10/2023 5
 The onset of wing rock is related with a nonlinear variation of roll damping
derivative with α, sideslip angle β, oscillation frequency and amplitude [1].
 Wing rock is primarily observed by means of wind tunnel free-to-roll experiments
for very slender delta wings.

Fig. 5.3: Wing rock build-up of an 800 delta wing at a = 270

12/10/2023 6
 Let us consider the F-18 low-AOA model for this exercise.

 We shall carry out an Extended Bifurcation Analysis (EBA) computation with a

level flight constraint (zero sideslip, wings level, zero flight path angle).

12/10/2023 7
 Fig. 5.4: F-18/HARV longitudinal aerodynamic and control derivatives from Table 2.3
12/10/2023 8
 The resulting bifurcation diagram is shown in Fig. 5.5. The top plot shows the
required variation of the throttle to maintain level flight as a function of the trim
Mach number. The various points of instability are marked therein.

 For instance, the point marked “A” refers to the onset of spiral instability at a low
AOA. Points “B” and “C” mark the phugoid instability onset and recovery of
stability, respectively, at moderate values of AOA.

 Finally, the Hopf bifurcation at the point labeled “D” indicates the onset of Dutch
roll instability—that is, wing rock.

 This can be verified by referring to the bottom plot in Fig. 5.1, which shows the
movement of the various modal eigenvalues in the complex plane corresponding
to the bifurcation diagram.
12/10/2023 9
 Fig. 5.5: Bifurcation diagram for the low-angle-of-attack F-18 airplane model and onset of
12/10/2023 10
various instabilities.
 Moving from lower (−0.05 rad) to higher (0.42 rad) AOA,
 The spiral mode starts off marginally unstable, then moves into the left-half
plane where it encounters the roll mode eigenvalue; they collide and separate
into a complex conjugate pair signifying a lateral phugoid mode.
 The short period mode tends toward the imaginary axis, but remains stable all
through. The phugoid mode is initially marginally stable. The eigenvalue pair
briefly flit across the imaginary axis but then regain stability.
 The Dutch roll eigenvalue crosses the imaginary axis at an AOA of about 22
deg, following which the Dutch roll mode becomes unstable.

Thus, the onset of wing rock oscillations may be expected around 22 deg AOA for
this airplane model.

12/10/2023 11
5.2. Analytical Criterion for Wing Rock Onset
 Under the usual assumptions of small perturbations in the lateral-directional variables
alone, about a straight and level flight trim state, the lateral- directional equations of
motion may be written as
(5.1)

(5.2)

(5.3)

 where pb, rb are the body-axis roll and yaw rates, respectively, μ is the wind axis (velocity
vector) roll angle, χ is the wind-axis yaw angle, q is the dynamic pressure (= 1/2ρV2), S is a
reference area, usually the airplane wing planform area b is the wing span, g is the
acceleration due to gravity, W is the airplane weight, Ixx, Izz are the moments of inertia, and
V* is the trim velocity.
12/10/2023 12
 The first step is to derive the aircraft equations of motion under the following considerations:

 The aircraft is a rigid body.

 The aircraft has constant mass and mass properties.

 The x-z plane is a plane of symmetry.

 No small incidence angles assumption, all the angles are kept by their values.

 The three velocity components in the body axes frame of reference (u, v, and w) are
written in terms of the aircraft true velocity (V), and the incidence angles (α and ß).

u = V cosα cosß

v = Vsinß

w= V sinɑ cosß

 The first eight equations are independent on the angle of azimuth (Y), so the Y-equation is
discarded from the set of equations, leaving an autonomous system of eight first-order
ordinary differential equations
12/10/2023 13
 Insert the aerodynamic model of Equations in the lateral-directional Equations 5.1
to 5.3 to give the complete set of equations as below:

(5.4)

(5.5)

(5.6)

12/10/2023 14
 The rate derivatives of the side force coef ficient, CYp1, CYp2, CYr1, CYr2, are usually of
lesser importance and may be ignored.

 Then, Equation 5.4 reduces to

(5.7)

 The perturbed wind-axis Euler angle rates can be related to the body-axis angular
rates as follows:
(5.8)

 Differentiating the first of Equation 5.8 once with respect to time:

(5.9)

12/10/2023 15
 Using ∆� from Equation 5.7, we can complete Equation 5.9 as below

(5.10)

 which on collecting like terms, appears as

(5.11)

 Similarly, differentiating the second of Equation 5.8 yields

(5.12)

 The primed short symbols in Equation 5.13 are defined as below:

(5.13)
12/10/2023 16
 Again replacing for ∆� and its derivative from Equation 5.7, and collecting like
terms in Equation 5.12 gives the following equation for yaw:

(5.14)

 Equations 5.11 and 5.14 form the set of two second-order lateral-directional
small-perturbation equations.

 It can be verified that these two equations of second order in the variables Δμ
and Δβ are coupled because of the ∆�, Δβ terms in the Δμ (rolling moment)
equation, and the ∆�, Δμ terms in the Δβ (yawing moment) equation.

12/10/2023 17
 5.2

12/10/2023 18
Matrix Form of the Perturbed Lateral-Directional Equations
 For further analysis, Equations 5.11 and 5.14 need to be cast in matrix form, as
below:

(5.15)

 which can be briefly represented by the following matrix equation:

(5.16)
 where

12/10/2023 19
 and Clat, Klat are the damping and stiffness matrix, respectively.

 The criterion for static instability is given by det(Klat) = 0, which is equiv alent to
the Routh criterion requiring the constant term in the charac teristic polynomial
to be zero.

 This condition, labeled Slat, can be easily evaluated from Equation 5.15 to be

(5.17)

 The static instability criterion usually correlates with the onset of spi ral mode
instability; that is the spiral eigenvalue being at the origin on the complex plane.

12/10/2023 20
 5.3. Approximation by Hamiltonian Dynamical System

 Typically, at moderate values of AOA (α*), the eigenvalues of the lateral directional
model in Equation 5.13 turn out to be a complex pair of lightly damped poles
representing the Dutch roll mode and a second complex pair of low-damped poles,
called the lateral phugoid, formed by the merger of the roll (rate) and spiral
eigenvalues.

 This is as seen in Figure 5.5. Thus, the lateral-directional dynamics as a whole is

poorly damped around the point of onset of the dynamic instability.

 Therefore, one may justifiably approximate the lateral-directional dynamics as an

undamped system.

12/10/2023 21
 In the traditional way of presenting the lateral directional dynamic equations, it is
quite impossible to identify which terms represent the “damping” and to isolate
them in order to arrive at an undamped approximation.

 However, by transforming the lateral directional dynamics into second-order

equations and presenting them in a matrix form as in Equation 5.15, the damping
terms are clearly sepa rated. By setting the supposedly negligible Clat matrix to
zero, one can obtain an undamped approximation to the lateral-directional
dynamics as follows:

12/10/2023 22
 (5.18)

 which is equivalently represented by

(5.19)

 Equation 5.19 is in the form of a Hamiltonian system.

 The onset of dynamic instability in the damped lateral-directional dynamics Equation

5.12 can then be approximated by the dynamic instability of the undamped
(Hamiltonian) system in Equation 5.18.

12/10/2023 23
 The eigenvalues of a Hamiltonian system are constrained to be sym metric about
the origin in the complex plane.

 This means that only four possible arrangements of eigenvalues can occur:

1. Two pairs of complex conjugate eigenvalues on the imaginary axis; ±iω1, ±iω2

2. Two pairs of real eigenvalues; ±σ1, ±σ2

3. One pair of complex conjugate eigenvalues on the imaginary axis, another

pair of real eigenvalues; ±iω, ±σ

4. Two pairs of complex conjugate eigenvalues; ±(σ ± iω), σ ≠0

 Only the first option indicates stable behavior; the other three represent various
forms of instability.

12/10/2023 24