Professional Documents
Culture Documents
Lateral-Directional Dynamics
Narayan Ananthkrishnan1
Independent Consultant, Mumbai 400076, INDIA
Nandan K. Sinha2
Indian Institute of Technology Madras, Chennai 600036, INDIA
This paper offers a clean and correct pedagogical presentation of the theory of airplane
lateral-directional dynamics. In this work, the definition of the dynamic (rate) derivatives
has been corrected, distinct timescales corresponding to the standard lateral-directional
modes have been identified, and a multiple timescale procedure including static residuals has
been followed to derive new literal approximations to the modes. The lateral-directional
small-perturbation
equations
have
themselves been written without having
to first derive the complete 6-degree of
freedom equations of airplane motion.
New, physically meaningful results for
the Dutch roll and spiral mode
parameters are obtained and discussed.
This work complements the revised
presentation of airplane longitudinal
dynamics in a companion paper.
I. Introduction
Independent Consultant, B-257, IIT Campus, Powai, Mumbai 400076, INDIA; Associate Fellow AIAA.
Associate Professor, Department of Aerospace Engineering, IIT Madras, Chennai 600036, INDIA; Member AIAA.
1
American Institute of Aeronautics and Astronautics
Working along similar lines, we first define the various angles in the directional and lateral motions. Then we
write down the lateral-directional small-perturbation equations, identify the timescales, and present the aerodynamic
model. Subsequently, the small-perturbation equations are transformed into second-order form and literal
approximations to the modes are derived. Some comments are made on the results.
paper. In case of large perturbations, the angles about XB and about XW are notably different, in particular the
coning motion by allows the trim angle of attack to remain constant which is not the case for a body-axis roll by
the angle . However, in case of small disturbances in roll, denoted by the perturbation angles and , and in
case * is quite small, then to the first approximation, there is no difference between a disturbance by the angle
about the body axis XB and a disturbance by the angle about the wind axis XW. Hence, under these conditions, we
can take .
The angular velocity about the body XB axis is called the body-axis roll rate pb, and that about the XW axis is the
wind-axis roll rate pw. rb is the body-axis yaw rate about ZB, and rw is the wind-axis yaw rate about ZW. The body
and wind axis rates are related through the orientation angles , and their rates. For instance (Raghavan and
Ananthkrishnan3),
pb pw sin
rb rw cos
and
(1)
* 0,
* *,
* 0,
* 0,
* * 0
(2)
T D; Y 0; L W ;
L 0; M 0; N 0
(3)
3
American Institute of Aeronautics and Astronautics
Figure 3 shows the axes and the forces/moments and the various angles in case of small perturbations the yaw
angles and , the sideslip angle , and the roll angle . Under the assumption of and small *, the
angles limited to the directional plane may be related by
(4)
We consider a perturbation in yaw rate rb and rw about the body and wind axis, respectively. The wind axis
rate rw corresponds to the curvature in the flight path approximately in the horizontal plane. There is also a
perturbation in roll rate pb about the body axis and pw about the wind axis. The lift L is unchanged from its trim
value but could be tilted by the bank angle . Taking components of W along and normal to the lift direction,
there is an unbalanced component of W along the YB axis which is W sin mg mg . This is in
addition to the aerodynamic side force Y along YB marked in Fig. 3(a).
The lateral-directional equations under small perturbations can now be written by inspection. The centripetal
acceleration is equal to the net unbalanced force along the YB direction:
mV * Y mg
(5)
The moment equations about the XB and ZB axes are simply given by:
I xx L
I zz N
(6)
(7)
where Y is the sideforce, L is the rolling moment, and N is the yawing moment. Thus, Eqs. (5), (6) and (7) form the
small-perturbation lateral-directional dynamics equations.
A. Timescales
The perturbed force/moments are defined in terms of the corresponding aerodynamic coefficients as:
Y qS CY ; L qSbCl ; N qSbCn
(8)
where q is the dynamic pressure (= V2), S is a reference area, usually the airplane wing planform area, and b is
the wing span. With the definitions in Eq. (8), the small-perturbation equations (5), (6), and (7), appear as:
1
g
g qS
qS CY * * CY
*
mV
V
V W
(9)
qSb
I zz qSb
Cl
Cl
I xx
I xx I zz
(10)
qSb
Cn
I zz
(11)
where we have replaced with in Eq. (10). The underlined braces represent the timescales in Eqs. (9)
through (11). The various timescales are defined as follows:
g ~ 10sec;T
*
Ts V
I zz
b
~ 1sec; Tr * ~ 0.1sec
qSb
2V
4
American Institute of Aeronautics and Astronautics
(12)
The slow timescale Ts of the order of 10 sec and the faster timescale Tf of the order of 1 sec are obvious from Eqs.
(9) through (11). However, a third, even faster timescale Tr, which we shall encounter later, of the order of 0.1 sec,
also needs to be introduced in case of the lateral-directional dynamics.
For a conventional airplane, there are three lateral-directional modes corresponding to the three timescales, as
follows:
A fast roll (rate) mode, which goes with the timescale Tr;
An intermediate dutch roll mode, which corresponds to the timescale Tf;
And a slow spiral mode, which occurs at the slow timescale, Ts.
Before we work on these modes, we need to write out the perturbed aerodynamic force and moment coefficients in
terms of the aerodynamic derivatives.
B. Small-perturbation Aerodynamic Modeling
The perturbed lateral-directional force/moment coefficients CY, Cl, Cn are expressed in terms of the perturbed
aerodynamic variables. As in the case of longitudinal dynamics (Ananthkrishnan and Sinha1), there are four different
aerodynamic effects to model:
Static due to Mach number and the relative orientation of the aircraft (body-axis) to the wind (wind-axis)
given by the aerodynamic angles, ,.
Dynamic due to the angular velocity of the airplane (body-axis) with respect to the angular velocity of the
relative wind (wind-axis) given by the difference between the two vectors, b-w.
Flow curvature effect due to the angular velocity of the wind axis arising from the airplane flying along a
curved flight path, w.
Downwash lag effect due to the wing-tip trailing vortices impacting the aft lifting surfaces with a time
delay of approximately lt/V*.
Now remembering that we have held the velocity and all the longitudinal variables fixed for our present analysis, the
lateral-directional perturbed force/moment coefficients can only be functions of the lateral variables. Hence, we can
ignore effects due to the Mach number, angle of attack , and the body and wind axis pitch rates qb and qw. Also the
downwash lag effect in case of the lateral-directional dynamics, which happens because of the interaction between
the wing-tip trailing vortices and the vertical tail, is not usually such a dominant effect and can presently be set
aside. Thus, we are left with the following variables whose effect is to be modeled:
Static
Dynamic pb-pw, rb-rw
Flow curvature pw, rw
Besides these, there is the effect due to control deflection usually there are two lateral-directional controls of
interest, the aileron and the rudder, which we shall include later as desired. The perturbed aerodynamic
force/moment coefficients may now be modeled as:
(13)
(14)
(15)
The fastest timescale Tr=b/2V* from Eq. (12) now appears in Eqs. (13) through (15) when modeling the rate
derivatives. Each of the aerodynamic derivatives in Eqs. (13) through (15) is defined as follows, where the * refers
to the trim state.
CY
CY
;
*
Cl
Cl
;
*
Cn
Cn
*;
5
American Institute of Aeronautics and Astronautics
CYp1
CY
;
( pb pw )(b / 2V ) *
CYp2
CY
;
pw (b / 2V ) *
CYr1
CY
;
(rb rw )(b / 2V ) *
CYr 2
CY
;
rw (b / 2V ) *
Clp1
Clp 2
Cl
;
pw (b / 2V ) *
Clr1
Clr 2
Cl
;
( pb pw )(b / 2V ) *
Cnp2
Cl
;
(rb rw )(b / 2V ) *
Cl
;
rw (b / 2V ) *
Cn
pw (b / 2V ) *
Cnr1
Cnr 2
Cn
( pb pw )(b / 2V ) *
Cnp1
Cn
(rb rw )(b / 2V ) *
Cn
rw (b / 2V ) *
(16)
pb pw sin *
rb rw cos *
and
(17)
When * is small, the right hand side of the pb-pw equation can be taken to be the product of two small terms,
namely,
and sin * , and hence dropped. We may also assume cos * 1. Thus,
pb pw 0
and
rb rw
(18)
That is, we do not need to distinguish between a body-axis roll rate and a wind-axis roll rate when dealing with
small perturbations.
The wind axis angular rates can themselves be written in terms of the rate of change of the wind axis Euler
angles as follows (Raghavan and Ananthkrishnan3):
pw sin and
(19)
pw sin *
and
rw cos *cos
(20)
assuming cos 1.
With the relations in Eqs. (18) and (20), the perturbed aerodynamic force/moment coefficients in Eqs. (13)
through (15) may be updated as below:
(21)
Now we can insert the aerodynamic model of Eq. (21) in the lateral-directional equations (9) through (11) to give
the complete set of equations as below:
g q S
*
*
*
*
CY CYp2 (b / 2V ) CYr1 ( )(b / 2V ) CYr 2 (b / 2V )
V W
(22)
6
American Institute of Aeronautics and Astronautics
q Sb
Cl Clp 2 (b / 2V * ) Clr1 ( )(b / 2V * ) Clr 2 (b / 2V * )
I
xx
q Sb
Cn Cnp2 b 2V * Cnr1 b 2V * Cnr 2 b 2V *
I
zz
(23)
(24)
g q S
*
CY
V W
(25)
Before proceeding further, for ease of algebraic manipulation, let us define some short symbols:
qSb
qSb
qS
Cl L ;
Cn N ;
CY Y ;
W
I zz
I xx
qSb
qSb
qSb
b
b
b
Cnp 2 * N p 2 ;
Cnr1 * N r1;
Cnr 2 * N r 2 ;
2V
2V
2V
I zz
I zz
I zz
(26)
qSb
qSb
qSb
b
b
b
Clp 2 * Lp 2 ;
Clr1 * Lr1;
Clr 2 * Lr 2
2V
2V
2V
I xx
I xx
I xx
So, Eq. (25) can be compactly written as:
g
* Y
V
(27)
N N p 2 N r1 ( ) N r 2
(28)
(29)
g
* Y
V
(30)
g
g
N r1 * Y N N p 2 * N r 2 0
V
V
7
American Institute of Aeronautics and Astronautics
(31)
g
g
g
N r1 * Y N * Y N r 2 N p 2 *
V
V
V
g
* N r 2 0
V
(32)
Usually the derivative Np2 is not significant enough and may be dropped. Then, the yawing moment equation
appears as:
g
g
g
N r1 * Y N * Y N r 2 *
V
V
V
g
* N r 2 0
V
(33)
Turning our attention next to Eq. (23), in terms of the short symbols defined in Eq. (26),
L Lp 2 Lr1 ( ) Lr 2
Yet again we have
(34)
from Eq. (27), and using that we can rewrite Eq. (34) as:
g
g
L * Y Lr 2 L p 2 Lr1 ( ) * Lr 2
V
V
(35)
Equations (33) and (35) form the set of two second-order lateral-directional small-perturbation equations. These are
summarized in Table 1 which shows that the two equations of second order in the variables and are coupled
because of the , terms in the (rolling moment) equation, and the
moment) equation.
g
g
Rolling moment
Roll
(Tr) and
(35)
Lr1 L
Y Lr 2 L p 2 * Lr 2
and side force
spiral (Ts)
V*
V
Source
Yawing moment
and side force
g
g
g
g
N r1
Y N * Y N r 2 * * N r 2 0
*
V
V
V
(33)
Note that that (side force) equation has been absorbed into Eqs. (33) and (35); hence there is no separate equation
for . In fact, the variable has itself been eliminated. To that extent, the lateral-directional equations in Table 1
appear similar to the two equations in V and for the longitudinal case (Ananthkrishnan and Sinha1). However,
the two longitudinal equations are easily separated based on their timescales the V equation (for the phugoid
mode) operates at the slow timescale T2 whereas the dynamics (short period mode) occurs at the faster T1
timescale. Unfortunately, the division of the lateral-directional equations based on timescales is not so clear for most
conventional airplanes. As noted in the last column of Table 1, the (rolling moment) equation contains dynamics
both at the slowest (Ts) and the fastest (Tr) timescales, whereas the (yawing moment) equation operates at the
intermediate Tf timescale. This complicates matters significantly. In terms of the dynamics modes, (yawing
moment) equation represents a second-order dynamics that is called the dutch roll mode. The (rolling moment)
equation must split into two first-order dynamics a fast mode at the timescale Tr called the roll (rate) mode, and a
slow mode at the timescale Ts called the spiral mode.
8
American Institute of Aeronautics and Astronautics
L L
p2
ROLL MODE
where
g
g
L * Y Lr 2 L p 2 f * Lr 2
V
V
EQUALS ZERO
r1
(36)
r is the component at the faster time scale Tr and f is the component at the timescale Tf. The first set
of braces contains the fastest dynamics at the time scale Tr whereas at this timescale the terms in the second set of
braces are collectively equal to zero. Thus, the roll (rate) mode dynamics is given by the first-order equation:
Lp 2 r
(37)
The stability of the roll mode is given by the sign of the derivative Lp2, also called the roll eigenvalue r, being
negative. From the relations in Eq. (26), this can be written in terms of the aerodynamic derivatives as follows:
Clp 2 0
(38)
Clp 2 is usually always negative, so the roll mode is almost guaranteed to be stable for most conventional airplanes.
The second set of braces in Eq. (36) can be solved out to give,
1
L [ L ( g / V * )Y Lr 2 ] ( g / V * ) Lr 2
f
L r1
p2
(39)
Technically, this is called the residual the component of that remains after the faster dynamics at the time
scale Tr is complete and which now varies as per the next, intermediate time scale Tf.
B. Dutch-roll Mode
We move on to the next fastest mode at the time scale Tf, called the Dutch roll. The dominant variable for this mode
is and we examine the yaw equation (33) in Table 1 for .
g
g
g
g
N r1 * Y N * Y N r 2 * f * N r 2 0 (40)
V
V
V
V
where, as indicated, the in Eq. (40) is the residual f from the roll dynamics. So, we insert the expression for
f from Eq. (39) in the yaw equation (40) to obtain:
g
g
N r1 * Y N * Y N r 2
V
V
g
g 1
g
g
*
Lr1 L * Y Lr 2 * Lr 2 * N r 2 0
V
V
V Lp 2
V
9
American Institute of Aeronautics and Astronautics
(41)
g
g L
g
g L
N r1 * Y * r1 N * Y N r 2 *
V
V Lp 2
V
V Lp 2
(42)
g
* N r 2 0
V
Now we split into two components a faster one f at the intermediate timescale T f and a slower one
s at the slower timescale Ts . Then, rearranging the terms in Eq. (42), we can write,
g
g Lr1
g
g L
Y
N
r1
r2
f
*
*
*
*
V
V L p 2
V
V L p 2
g
g L
g
N * Y N r 2 * s * N r 2 0
V
V L p 2
V
EQUALS ZERO
The first set of braces in Eq. (43) contains the second-order dynamics at timescale
g
g L
g
g L
N r1 * Y * r1 N * Y N r 2 * 0
V
V L p 2
V
V L p 2
(44)
from which the damping and frequency of the Dutch roll mode may be easily read off as follows:
L
g
Y N r 2
*
L
V
p 2
2
nDR
N
(45)
L
g
(46)
2 DRnDR N r1 * Y r1
L
V
p 2
2
Assuming positive Dutch roll stiffness, that is, nDR 0 , then the condition for stability of the Dutch roll mode is
given by the requirement of the damping being positive. For simplicity, we ignore the less significant Y term in Eqs.
(46) and (47), and using the relations in Eq. (26), the conditions for positive stiffness and damping of the Dutch roll
mode may be obtained as:
Cn (Cl / Clp 2 ) 0
(47)
(48)
T f2 / TsTr
(49)
10
American Institute of Aeronautics and Astronautics
Since Tf~1 sec, Ts~10 sec, andTr~0.1 sec, as defined in Eq. (49), ~1.
From the second set of braces in Eq. (43), we have
g
* Nr2
V
g
g L
N * Y N r 2 *
V
V L p 2
(50)
This is the residual after the Dutch roll dynamics has subsided. This will vary as per the slowest time scale Ts.
C. Spiral Mode
To derive an approximation to the spiral mode, we go back to Eq. (36), and examine the terms in the second set of
braces:
g
g
Lr1 L * Y Lr 2 Lp 2 * Lr 2
V
V
EQUALS ZERO
(51)
which is what was left of the roll moment equation after the part representing the roll mode was detached. This may
be written as below:
g
g
(52)
L L * Y Lr 2 s * Lr 2
L r1
V
V
p2
where all the terms, including the s , vary at the slowest time scale Ts. We ignore the term in this equation,
and use the expression for
s obtained from the residual in Eq. (50) after the Dutch roll mode has been solved
L
p 2
N
Y
L
N
V * r2 r2
g
Lr 2 *
V
g
g L
* Y N r 2 *
V
V L p 2
Identifying Lp2 with the roll eigenvalue r, and the term in the square bracket in the denominator as
(45), and dropping terms in higher powers of
(53)
2
from Eq.
nDR
g
* , yields the following first-order equation for the spiral mode:
V
g L N r 2 N Lr 2
*
rd2
V
(54)
[ L N r 2 N Lr 2 ] 0
(55)
Using the relations in Eq. (26), we may rewrite the spiral stability condition in terms of the aerodynamic derivatives
as:
11
American Institute of Aeronautics and Astronautics
( )
()
(56)
()
where the usual signs of the derivatives are marked under each of them. Each product is positive, and the expression
in Eq. (56) turns out to be the difference between two positive terms. In general, the difference can turn out to be
either positive or negative in either case, it is usually quite small and close to zero, so that the spiral motion either
converges or diverges very slowly.
A summary of the stability requirements for the lateral-directional modes is presented in Table 2. Clearly, neither
Cn nor Cl alone provide any kind of stability criterion for the lateral-directional modes, though acting together in
the form of Eqs. (47) and (56), they impact the dutch roll stiffness and spiral mode eigenvalue, respectively. Thus,
the traditional notion of static stability based on Cn and Cl has no direct bearing on either directional or lateral
stability.
Table 2. Summary of lateral-directional stability requirements
Mode
Roll
Dutch Roll
Stability requirement
Clp 2 0
Cn (Cl / Clp 2 ) 0
(47),
(38)
Spiral
(48)
(56)
(57)
has been in use which roughly brings together Cn, Cl in a physically correct sense, but does not always correlate
well with the Dutch roll frequency. The correct combination is now obtained as in Eq. (47).
Likewise, the correct combination of Cnr1, Clr1 that affects the Dutch roll damping is as given in Eq. (48).
That Cn, Cl cooperate in the Dutch roll frequency but oppose each other when it comes to the spiral eigenvalue
is also a known fact. However, traditionally the derivatives Cnr1, Clr1 have been used in the expression for the spiral
mode eigenvalue. Instead, the derivatives Cnr2, Clr2 are the ones that matter for the spiral mode, as seen in Eq. (56).
The derivatives Cnr1, Clr1 matter when the airplane nose yaws relative to a reasonably steady velocity vector, as seen
in the Dutch roll mode. On the other hand, the derivatives Cnr2, Clr2 enter when both the yawing motion and the
curvature of the flight path (deviation of the velocity vector) are roughly coincident, as in the spiral mode. Hence,
the distinction between these two sets of derivatives is crucial a point that has been missed for the past century!
For a detailed discussion of each of these derivatives listed in this section, please refer the textbook (Sinha and
Ananthkrishnan4).
as the practicing engineer since the literal approximations help correlate various airplane configuration parameters to
its flight dynamic response for example, the dihedral angle to the Dutch roll frequency. However, it must be noted
that the dynamic (rate) derivatives such as Cnr1, Clr1 are also used to build the aerodynamic model for a complete,
nonlinear, 6-degree of freedom simulation, such as the ones used for flight testing, pilot training simulators, and
control law design. The corrections to the aerodynamic model have major implications here, as for instance below.
Traditionally, the yaw damping has been modeled by a single term, Cnrrb. Instead, it turns out that the correct way
to model this effect is by the sum of the two terms underlined below:
(58)
There is a fallout on control design as well. Earlier, the body-axis yaw rate rb would be fed back in a yaw damper
design to augment the Dutch roll damping. According to the corrected model in Eq. (58),
variable for feedback.
is a more appropriate
VIII. Conclusion
The various improvements and corrections suggested by Ananthkrishnan and Sinha1 to the existing presentation
of airplane flight dynamics have been carried over and applied to the lateral-directional dynamics in this paper. The
faulty modeling of the dynamic (rate) derivatives, a carry over from the original work by Bryan2, has been corrected.
Three timescales, corresponding to the three standard lateral-directional modes, have been clearly defined, and a
multi-timescale approach has been used to obtain literal approximations to the three modes. New, corrected
expressions for the Dutch roll and spiral mode parameters are presented and explained. The implications of this
development for pedagogy, aircraft design and flight simulation are obvious.
IX. References
1
Ananthkrishnan, N., and Sinha, N.K., A Simple, Correct Pedagogical Presentation of Airplane Longitudinal Dynamics,
AIAA Atmospheric Flight Mechanics Conference, Boston MA, Aug 19-22, 2013, submitted.
2
Bryan, G. H., Stability in Aviation, MacMillan, London, 1911.
3
Raghavan, B., and Ananthkrishnan, N., Small-Perturbation Analysis of Airplane Dynamics with Dynamic Stability
Derivatives Redefined, Journal of Aerospace Sciences and Technologies, Vol. 61, No. 3, 2009, pp. 365-380.
4
Sinha, N. K., and Ananthkrishnan, N., Elementary Flight Dynamics with an Introduction to Bifurcation and Continuation
Methods, CRC Press, Boca Raton FL, 2014 (to be published).
13
American Institute of Aeronautics and Astronautics