Vous êtes sur la page 1sur 16

LONGITUDINAL STABILITY CRITERIA FOR A PROPELLER-DRIVEN

AIRCRAFT

by

Gil Iosilevskii*

Faculty of Aerospace Engineering, Technion, Haifa 32000, Israel

Abstract

In this short exposition we use the Routh criterion to assess longitudinal dynamic stability

of a propeller-driven aircraft. Under a few plausible assumptions on possible ranges of the

pertinent stability derivatives we manage to reduce this criterion to a pair of simple

conditions. One of them imposes a traditional aft limit (the forward of the maneuver and the

neutral-speed-stability points) on the center-of-gravity position, whereas the other imposes an

upper limit on the longitudinal moment of inertia. It is demonstrated that most aircraft have

sufficiently small inertia to remain stable as long as their center-of-gravity is properly placed.

At the same time, sailplane-like aircraft (as, for example, long endurance UAVs), with an

engine installed at the rear extremity of the aircraft, may have sufficiently high inertia to

become unstable regardless of their center-of-gravity placement.

Keywords: longitudinal dynamic stability, propeller-driven aircraft

*
Email: igil@aerodyne.technion.ac.il

1
Nomenclature

at lift slope coefficient of the horizontal tail

a ,…, e coefficients in the characteristic equation

Aw aspect ratio of the wing

cw mean aerodynamic chord of the wing

CD , CL , CT drag, lift and thrust coefficients at trim; the unit is 1 2  V 2 S w

CD0 parasite (zero lift) drag coefficient

Cm , CM pitching moment coefficients, the former is referred to the center of

gravity, the latter is referred to the neutral point; the unit is 1 2  V 2 S w cw

f,g the third and the fourth terms in the Routh sequence

i0 ,…, i3 combinations of stability derivatives

i y , iy reduced moment of inertia; the units are 1 8   S w cw3 and 1 8   S w cw3 

J X  critical value of iy for a given X

J 0 , J1 min J  X  and min J  X  , respectively


X 1 X 1, CL  0

kq combination of the pitch damping derivatives defined in (25)

kT scaling exponent of the thrust coefficient variation with airspeed at

constant throttle

K induced drag coefficient

P parameter defined in (38)

q reduced pitch rate, 

ry reduced radius of gyration, 1 2  iy ; the unit is cw

R1 an upper stability limit for ry

2
Sw , St wing and tail areas

t reduced time; the unit is cw  2V 

u reduced airspeed perturbation from trim; the unit is V

V airspeed at trim

xcg reduced longitudinal center-of-gravity position; the unit is cw

xm , xn , xs maneuver, neutral and neutral-speed-stability points; the unit is cw

X reduced maneuver margin; the unit is xs  xm

X   iy  critical values of X for a given iy

 angle of attack, measured from trim

 flight path angle at trim

 pitch angle, measured from trim

 reduced mass; the unit is 1 2   S w cw

 air density

overheads


 (overdot) a derivative with respect to the reduced time


 (overbar) a nonstandard reduced quantity

subscripts

, partial derivative with respect to  , computed at trim

3
Introduction

Small perturbation (linear) analysis of longitudinal dynamics of a rigid aircraft can be

found practically in any textbook on flight mechanics. Under the assumptions of constant-

density atmosphere, flat non-rotating Earth, zero thrust angle, and small perturbations from a

steady straight-and-almost-level flight, equations of (free) motion can be recast as

B  x  A  x , (1)

where, with x  u,  , q,   and q   ,


T

 Cm,u Cm, Cm,q 0 


 CL,u  2CL CD  CL, CL,q  2  CL 
A , (2)
CT,u  CD,u  2 CL CL  CD, 0 CL 
 
 0 0 1 0 

0 Cm, iy 0
0   2  CL,  0 0 
B . (3)
 2 0 0 0
 
 0 0 0 1 

With a few small adjustments, the notation used here conforms that of Etkin [1] † – it can be

found in the Nomenclature section above; equations (1) - (3) are, essentially, his equations

(5.13, 16).

Setting x  x0 e st in (1), where t is the dimensionless time and x0 and s are obvious time-

independent quantities, leads to the generalized eigenvalue problem  sB  A   x0  0 , and, in

turn, to the fourth order characteristic equation,

as 4  bs 3  cs 2  ds  e  0 . (4)

The coefficients a, …, e in this equation can be identified with products of the elements of A


The deviations from the original notation are: (i) an over-dot is used to designate a derivative with respect to
reduced time instead of ‘D’; (ii) the subscript ‘e’ is omitted from unperturbed (equilibrium) values of the flight-
path angle and the lift and drag coefficients; (iii) ‘ i y ’ and ‘u’ are used instead of ‘ Iˆy ’ and ‘ Vˆ ’ for the reduced

moment of inertia and the reduced velocity.

4
and B, but except for

a  2i y   2  CL,  , (5)

all other coefficients are rather unwieldy (Appendix A).

Equation (4) has no analytical solution for s – it has to be solved either approximately [2]

or numerically. Each approach has its virtues, but in many cases the roots of (4) are not really

needed – what is needed is the knowledge whether the aircraft is stable or not. In principle,

stability of a system (the aircraft, in this case) can be assessed without actually solving its

characteristic equation – by the Routh criterion, the number of roots in the right-half-plane (of

equation (4)) is the number of sign alternations in the five-member sequence

S  a, b, f , g , e , where f  c  ad b and g  d  be f [3]. In practice, however, this

approach is hardly ever used with equation (4) – the unwieldiness of the members of S

renders it rather unattractive as compared with the instant solution of the numerical approach.

Notwithstanding the majority approach, we will demonstrate that for a propeller-driven

aircraft, an application of the Routh criterion to equation (4) yields a pair of very simple

conditions with immediate design implications. One of these conditions sets a traditional aft

limit (the forward of the maneuver and the neutral-speed-stability points) on the center-of-

gravity position. The other sets an upper limit on the longitudinal moment of inertia. The

details follow.

Approximate stability criterion

As a first step, we replace the moment derivatives Cm, in (2) and (3) (or, to the same end,

in (A1) – (A5) in Appendix A), all referred to the center of gravity, with combinations

Cm,  CM,  CL,  xcg  xn  (6)

of their neutral-point-referred counterparts, CM, , and lift coefficient derivatives, CL, . Here,

xcg and xn are the center-of-gravity and neutral point positions, and ‘  ’ stands for ‘  ’, ‘  ’,

5
‘q’, and ‘u’. Moreover,

CM,  0 , (7)

CM,q  2   xm  xn  , (8)

CM,u  2CL  xs  xn  , (9)

by definitions of the neutral ( xn ), maneuver ( xm ) and neutral-speed-stability ( xs ) points; the

last pair follows equations (6.10, 8) and (6.4, 26) of [1] by (6).

These substitutions do simplify the expressions for b, …, e, but essentially do not reduce

the number of terms comprising them. Hence, as a next step, we sort these terms by their

magnitudes, and retain a few dominant ones only. For a typical subsonic aircraft in an almost

level flight, invariably

CD , CD,u , CD, , CT,u , CL,u , CM,u ,  are of the order 10-1 , (10a)

CL is of the order 100, (10b)

CL, , CL, , CL,q , CM, and CM,q are of the order 101, (10c)

 is of the order 102, (10d)

iy is of the order 102 or 103. (10e)

The word ‘order’ can be understood here (and below) in its intuitive sense; nevertheless, we

shall remove possible ambiguity and define the order O of a quantity a as

round  log10 a 
O  a   10 . Thus, 4 and 30 are both of the order 101.

Sorting out the terms in (A2) – (A5) by (10), and discarding all those that are of the order

10-2 (or smaller) as compared with the leading ones, leaves

   
b  2  i y CL,  O 101    2   CL,q  CM,q  CM,   2   CL,q  CL,  xcg  xm  , (11)

c  4  2CL,  xcg  xm  1  O 103   , (12)

6

d  CL  CM,q  CM,   2CL  CL,u    CL, CM,q  O 101  

2CL  xcg  xs   CL,q  CL,  CL  2  CD,  O 1 
 
2  xcg  xm  CL,  CT,u  CD,u  3 CL   2CD, CL  O 102   O 102  , (13)

e  2CL2CL,  xcg  xs  1  O 102   , (14)

where

xm  xm   xm  xn  O 102 i y   , (15)

xs  xs   xs  xn  O 102  . (16)

Noting that xm  xn and xs  xn are of the order 10-1 by (8)-(10), xs  xs  O 103  and

xm  xm  O 103 i y   . Full expression for xs can be found in (A7).

It can be verified by direct substitution that the ratio b a is of the order 10-1, whereas the

ratios c a , d a and e a are of the orders 10  xcg  xm  i y1 , 10-3 i y1 and 103 i y1  xcg  xs  ,

respectively. Hence, as long as

xcg  xm is of the order 10-1 , (17)

the third and the fourth terms in the Routh sequence, f  c  ad b and g  d  be f , can be

approximated with c and d  be c ; the error of these approximations should be of the order

of 102 relative to the respective leading terms. Explicit expression for c is found in (12);

explicit (approximate) expression for g follows from (8) and (11) - (16). It can be regrouped

in various forms; the following two,


g  CL, CL2 i2 X   i1  i3  iy    i0  i3  iy  X 1  (18a)

X  1  i2  X  1   2i2  i1  i3   X  1   i0  i1  i2  
2

 CL, C 2
L
  iy  , (18b)
X  X 1 
will serve the needs of the following discussion. Here,

7
xcg  xm
X , (19)
xs  xm

4 kq 2kq CM,q
i0   xn  xm    , (20)
CL, CL,

2
i1 
CL, CL
 k Cq L,u 
  CL,   xn  xm   2CD,  xs  xm  , (21)

2
i2 
CL2
 CD,u  CT,u  3 CL   xs  xm  , (22)

CL,q kq CM,q CL,q


i3  i0  , (23)
2  CL,

iy  i y  , (24)

in which

kq  1  CM, CM,q . (25)

By the Routh criterion, the number of roots in the right-half-plane equals the number of

the sign alternations in S. Since, invariably, CM,q  0 , CM,  0 and CL,  0 , the first two

members of S, a and b, are positive by (5), (11), (10) and (17). Hence, the aircraft will be

stable only if the remaining members, f, g and e, are positive as well. The first of these

conditions yields xcg  xm (see equation (12)), or, approximately,

xcg  xm . (26a)

In view of (17) this condition is almost redundant. We shall assume that it is satisfied

unconditionally. The second of these conditions,

g  0, (26b)

will be discussed in the following sections. The last of these conditions yields xcg  xs (see

equation (14)), or, approximately,

xcg  xs . (26c)

8
It suggests that the center of gravity should be placed forward of the neutral speed stability

point. There is nothing new in this result (see [1, p. 335]), but it is satisfying to obtain it here.

In the analysis of (26b) we shall assume that this condition is satisfied as well. In conjunction

with (26a) and (19) it implies that either X  1 or X  0 .

Note that failing (26b), but not (26a) and (26c), implies two sign alternations in the Routh

sequence and hence two divergent (phugoid) roots. Failing (26c) only, or both (26c) and

(26b), implies one divergent root.

Condition (26b)

The consequences of (26b) can be inferred from either (18a) or (18b), depending on the

variable that is considered given. For example, the conjunction of (26b) and (18b) sets an

upper limit on iy ,

iy  J  X  , (27)

where

J  X   i2  X  1   2i2  i1  i3    i0  i2  i1   X  1 .
1
(28)

Conversely, given iy , the conjunction of (26b) and (18a) sets limits on X :

if i2  0 and iy  J 0 , then X   iy   X  0 or 0  X  X   iy  or 0  X   iy   X ; (29a)

if i2  0 and iy  J 0 , then 0  X ; (29b)

if i2  0 , then X  X   iy   0 or 0  X  X   iy  . (29c)

Here,

X   iy  
1
2i2

  i1  i3  iy   i  i
1 3  iy   4i2  i0  i3  iy 
2
 (30)

are the solutions of g  0 with respect to X , whereas

J 0  i1  i3  2i2  2 i2  i0  i1  i2  (31)

9
is the value of iy for which the square root in (30) vanishes with i2  0 . Note that

J 0  min J  X  by (28).
X 1

An example

The use of (27) and (29) can be elucidated on a practical example. Consider a fictitious

long-endurance UAV. For this aircraft we adopt a standard drag-lift relation

CD  CD0  KCL2 ; (32)

neglect CD,u , CL,u and CM,u ; and set

CT,u   kT  CD   CL  , (33)

where, kT is the scaling exponent of the propulsion system. It equals 2 for an ideal (constant

thrust) jet and 3 for an ideal (constant power) propeller [1, p. 263]. Values of CD0 , K ,  and

the relevant stability derivatives are listed in table 1.

quantity value quantity value


CL variable CM,u 0
CL, 6.3 CD0 0.03
CL, 1 K 0.03
CL,q 4.5 CD, 0.38 CL
CL,u 0 CD,u 0
CM, variable CT,u variable
CM, -4.0  150
CM,q -20.0 iy variable

Table 1: Dimensionless stability derivatives of the model aircraft.

Using this data, equation (4) was solved numerically with many combinations of lift

coefficient, stability margin ( CM, ) and reduced moment of inertia. Differentiating the stable

and unstable combinations yielded the stability boundaries drawn by the thick lines in figure

1. Approximate boundaries have been constructed as contours J  X   iy (see (27)) using

10
(28), (8) and (19)-(25); they appear as thin lines on the same figure. The fit between the two

is fair.

Figure 1: Stability boundaries for the model aircraft. Combinations of lift coefficient and
stability margin are unstable if situated to the right (b) or above (a, c, d) the respective line.
Thick lines mark the exact boundaries obtained from multiple solutions of (4); thin lines are
solutions of J  X   iy .

With no ‘u’ derivatives, xs  xs  xn , by (9) and (16); and since CM,q  0 , xs  xm  0 , by

(8). Thus, X  1 , by (19) and (26a). Moreover,

CM,q  CD, 
i1     2 , (34)
CL  CL, 
CM,q  CD 
i2    kT   kT  3    , (35)
C L  CL 

by (21) and (22).

11
With kT  2 (an ideal-jet), i2 turns negative once  exceeds 2CD CL . In this case (29c)

applies, inferring that in a (sufficiently) steep climb, the aircraft turns unstable with

X  X   iy  . This behavior is explicit in figure 1b. An exhaustive discussion on the stability

of a climbing jet aircraft can be found in Etkin [1, pp. 329-346] – it will not be repeated here.

With kT  3 (an ideal-propeller), i2 is positive unconditionally. In this case, (27) applies,

inferring that an ideal-propeller-driven aircraft turns unstable for sufficiently large moment of

inertia. This behavior is explicit in figures 1c and 1d. Since i0 , i2 and i3 are independent of 

by (20), (23) and (35), and i1 increases with  by (34), it follows from (28) that J  X 

increases with  as well. Consequently, the lowest (and hence the most constricting) limit on

the longitudinal moment of inertia of a propeller-driven aircraft is obtained in a power-off

glide, where    CD CL . Indeed, the curves of constant iy in figure 1d are lower than those

in 1c.

In a glide,

 3C   5C 2kq CL,q 
J  X    KCM,q   3  D02   X  1  1  D02  

 KCL   KCL KCL, 2  

 2k q 2C  1 
  2  D02  , (36)
 KCL, KCL  X  1 

by (28), (32)-(35), (20) and (23). Since X >1 (see above), J  X  decreases with increasing

lift coefficient. Indeed, the lines of constant i y in figure 1d are higher for lower values of i y .

Accordingly, the most constricting limit on the moment of inertia in a glide is obtained when

the aircraft flies at the highest possible lift coefficient.

Maximal design moment of inertia

Let J1  min J  X  be the absolute minimum of J  X  in a glide with respect to both


X 1, CL  0

12
X and CL . Under the same assumptions as those of the preceding section (the parabolic

drag polar and the neglect of the u-derivatives) it is given by

J1  2CM,q P , (37)

where
1 k CL,q 6k q 
PK  q  6. (38)
 2 KCL, 2 KCL, 
 

This minimum corresponds to CL   (which is, of course, beyond the reach) and

 2k q 2C   3C 
X  1   2  D02   3  D02  . (39)
 KCL, KCL   KCL 

Clearly, an aircraft which is designed with sufficiently small moment of inertia to satisfy

iy  J1 and with sufficiently forward center-of-gravity to satisfy (26c) (and (26a)), will be

stable at all flight conditions.

It is shown in Appendix B that P depends mainly on the aspect ratio of the wing. Since

the aspect ratio is dictated by performance (rather than stability) requirements, CM,q remains

the main design parameter affecting the maximal moment of inertia for which the aircraft is

stable.

For a conventional subsonic aircraft, CM,q is associated mainly with the contribution of its

horizontal tail; the contribution of the other parts is small and has the same (negative) sign as

the tail. Hence, the contribution of the tail,

CM,q  2at  xn  xt   S t S w  ,
2
(40)

provides a lower bound on CM,q . Here, S t S w is the tail to wing areas ratio, xn  xt is the

distance between the aircraft neutral point and the aerodynamic center of the tail, and at is

the lift slope coefficient of the tail [1, p. 269].

13
iy can be related with longitudinal (reduced) radius of gyration, ry . In fact, iy  4ry2 (see

the Nomenclature). Hence, one may set J1  4 R12 and use R1 as a conservative upper bound

of ry . Thus,

R1   xn  xt  at P S t S w , (41)

by (37), (38) and (40). With at  5 , S t S w  6 , and the data of our model aircraft, R1 equals,

approximately, 0.4  xn  xt  – almost half the distance between the tail and the center-of-

gravity (it is implied here that  xcg  xn    xn  xt  ). This result marks out a single

configuration which may be at risk of becoming unstable because of the excessive inertia.

This is the configuration with an aft engine, counterbalanced by the equipment in the nose.

All other configurations of propeller-driven aircraft should be stable as long as their center-

of-gravity is properly placed.

Concluding remarks

The key element underlying the preceding derivations is that under the assumptions (10)

and (17), the coefficients a, …, e of the characteristic polynomial can be approximated by a

few dominant terms only. These dominant terms were used in (11)-(14) and (18) to

approximate the respective members in the Routh sequence, leading to stability conditions

(26b) and (26c). The former sets an upper limit on the longitudinal moment of inertia; the

latter sets an aft limit on the center-of-gravity position.

Perhaps the most conspicuous result of this study is that most configurations of propeller-

driven aircraft are practically unlimited by inertia, and hence are stable as long as their center-

of-gravity is placed forward of the neutral-speed-stability (and maneuver) points – condition

(26c). At the same time, aircraft with high-aspect ratio wing and an aft engine are at risk of

becoming unstable because of an excessive inertia – condition (26b). The critical moment of

14
inertia (28) depends on the center-of-gravity position and the flight conditions, but J1 from

equation (37) may provide an early warning when longitudinal inertia becomes an issue.

References

[1] B. Etkin, Dynamics of Atmospheric Flight, John Wiley and Sons, N.Y., 1972, pp. 69, 70,

185, 221, 240, 263, 269, 282, 290, 328-346.

[2] R.D. Milne and G.D. Padfield, “The Strongly Controlled Aircraft,” The Aeronautical

Quarterly, May, 1971, pp. 146-168

[3] K. Ogata, Modern Control Engineering, Prentice-Hall, N.J., 1970, pp. 253-255.

[4] B. Thwaites, Incompressible Aerodynamics, Oxford University Press, Oxford, 1960,p. 310

Appendix A

The coefficients in equation (4) are

a  2i y   2  CL,  , (A1)


b  2  2   Cm,q  Cm,   i y  CL,  CD    Cm, CL,q  CL, Cm,q  
i y  2   CL,  CT,u  CD,u  2 CL  , (A2)


c  2    2   CL, q  Cm,  Cm,q  CL,  CD   CL    CT,u  CD,u  3 CL   Cm,q  Cm,  
i y   2CL  CL,u  CL  CD,    CL,  CD  CT,u  CD,u  2 CL  

  CT,u  CD,u  2 CL   Cm, CL, q  Cm, q CL,  , (A3)

d    2CL  CL,u   CL  Cm,q  Cm,   Cm,q CD, 


  CT,u  CD,u  2 CL   CL Cm,   CD  CL,  Cm,q   2  CL,q  Cm, 
Cm,u  CL  CL,q  CL,    2   CL,q  CD,   2  CL Cm, , (A4)

e  Cm, CL  2CL  CL,u   CT,u  CD,u  2CL     Cm,u CL  CL,  CD   CL  CD,    . (A5)

15
Using (6), (7) and (9), the last expression can be recast as

 C C  CL,u  CL  CD,  CT,u  CD,u 


e  2CL2CL, 1  D L,u     
 2CLCL, 
    xcg  xs  , (A6)
  2 C L C L, 2CL 

where

CD  CL  CL,u    CL  CD,  CL  CL,u   CT,u  CD,u 


1  2    
CM,u CLCL,  2CLCL, 2CL .
xs  xn  (A7)
2CL C C 
1  D L,u     
 CL  CD,  CL,u  CT,u  CD,u 

2CLCL,  2CLCL, 2CL 

Appendix B

In this appendix we seek the leading-order approximation for P, which was defined in

equation (38). It depends on four parameters: kq , CL, , K, and the ratio CL,q  . The former

is given by equation (25). Neglecting the respective contributions of the horizontal tail, and

assuming elliptic wing loading, the simplest approximations for the next two are

CL,  2πAw  Aw  2  , (B1)

K  1  πAw  , (B2)

where Aw is the aspect ratio of the wing [4]. With these,

CL,q  CM,  1 3 3 CM, 6 C  1  CL,q CL,q CM, 


P 1     2 M,  1    . (B3)
4π  CM,q  π Aw Aw CM,q Aw CM,q 2πAw    CM,q 

The ratio CM, CM,q is, approximately, the downwash derivative on the horizontal tail [1,

pp. 269, 284], about 0.3. The ratio CL,q  for a conventional subsonic aircraft (ultra-lights

excluded) is of the order of 10-1 – see (10c) and (10d); the aspect ratio Aw is of the order of

101. Hence, the leading-order (10-1) term in (B5) is π 1 3Aw1 .

16