Professional Documents
Culture Documents
Flightmechanics and Variational Problemsof A Linear Type: Douglas Aircraft Company, Inc
Flightmechanics and Variational Problemsof A Linear Type: Douglas Aircraft Company, Inc
a Linear Type f
ANGELO MIELE*
Douglas Aircraft Company, Inc.
Xl
[$(x, y) + yfSfr(x, y)]dx (1)
difficulties can be overcome. Rigorous sufficiency proofs are where $ a n d SP are given functions of t h e arguments x
developed for linear problems of both the simplest type and the and y a n d where yf = dy/dx. Assume t h a t , among all
isoperimetric type. Several physical problems, associated with arcs y{x) consistent with certain prescribed end condi-
the mechanics of flight of either a constant mass or a variable mass
vehicle, are treated within the general frame of problems of
tions, t h e special a r c which extremizes—-i.e., maximizes
linear type. or minimizes—the integral (1) is t o be found. As is
known from t h e theory of t h e first variation, t h e ex-
(1) INTRODUCTION tremal arc associated with t h e problem under consider-
ation is t o be a solution of t h e Euler's differential equa-
i
p,
i
pz
: p
^ F = $ + * / + \(y> + /) (10)
and \(x) a variable Lagrange multiplier. Eqs. (8) to
(10) lead to the following results:
FIG. 1. Parametric representation of a variable subject to
inequality. (d/dx) ( * + X) - (d$/dy) -
(d*/dy)y> - \(df/i>y) = 0 (11)
Downloaded by KINGSTON UNIVERSITY on March 14, 2015 | http://arc.aiaa.org | DOI: 10.2514/8.7785
fore, inferred t h a t b o t h the Legendre condition and the (df/dz) \{dz/dp) = 0 (12)
Weierstrass condition yield no information on the T h e Euler equation (12) associated with the p variable
minimal or maximal n a t u r e of the extremal path. is interesting, because it shows t h a t the extremal arc is
discontinuous, being generally composed, for df/dz ^ 0,
(3) L I N E A R P R O B L E M S INVOLVING A NONHOLONOMIC of subarcs
CONSTRAINT
X= 0 (13)
W i t h regard to engineering applications, a more im-
p o r t a n t t y p e of linear problem occurs in the case where and subarcs dz/dp = 0 (14)
the arc y(x) extremizing the integral (1) is to be con- For t h e subarc X = 0, t h e relationship between x and
sistent with a constraint of the form y is obtained from Eq. (11)
y' +f{x,y,z) = 0 (5) (d¥/ds) - (d$/dy) = 0 (15)
where z is a new variable such t h a t * For the subarcs dz/dp = 0 the z variable has a con-
s t a n t value^—-i.e., either z = %\ or z = z2 according to
Zi < z < z2 (6) the parametric representation devised in Fig. 1. T h e
relationship between x and y is obtained b y integrating
To account for t h e relation (6), the idea of para-
Eq. (5) subject to either the constraint z = Zi or z = z2.
metric representation of the variable subject to in-
Clearly, t h e present problem differs from t h e one of
equality is introduced. T h e variable z is represented
Section (2), insofar as subarcs z = const, m u s t be con-
(see Fig. 1) as a function z(p) of a parameter p having
sidered in the composition of the extremal arc in addi-
the following properties: (a) for — oo < p < ph
tion to the special subarc represented b y Eq. (15). T h e
z = zh dz/dp = 0, d2z/dp2 = 0; (b) for p2 < p <
role of these subarcs z = const, is fundamental in the
+ oo; z = z2, dz/dp = 0, d2z/dp2 = 0; and (c) for pi <
solution of the b o u n d a r y value problem: in fact, it is
p < p2, a one-to-one correspondence is assumed between
now entirely possible to satisfy a set of prescribed end
z and p—i.e., dz/dp ^ 0 everywhere.
conditions in the (x, y) plane—a, circumstance which,
With the above artifice, t h e independent parameter
on the contrary, is generally impossible for the problem
p is allowed to v a r y between — <» and + °°. T h e
formulated in Section 2.
variable z becomes a dependent quantity, variable be-
T h e analysis of t h e Legendre-Clebsch condition and
tween Zy and z2 according to t h e scheme of Fig. 1.
of the Weierstrass condition (omitted for the sake of
Notice t h a t the condition dz/dp = 0 represents either
brevity) leads to the following results: 1 1 (a) the
a line z = zv or a line z = z2\ on the other hand, the con-
Legendre- Clebsch condition yields no information on the
dition dz/dp ^ 0 is representative of any other value
minimal or maximal character of the solution for the
of z intermediate between Z\ and z2. Notice also t h a t
three subarcs X = 0, z = zh and z = z2; and (b) the
p is only a parameter and t h a t there is no necessity of
Weierstrassian function is E = 0 a t points of the sub-
attributing to it any geometrical or physical meaning.
arc X = 0. T h e analogy between the present conclusions
The main effect of the parametric representation is
and those of Section (2) should be emphasized.
to reduce a variational problem involving inequalities
to the same mathematical model useful in solving t h e
(4) L I N E A R P R O B L E M S INVOLVING B O T H A
case where all constraints involve equalities. Thus,
NONHOLONOMIC CONSTRAINT AND AN ISOPERIMETRIC
for the present problem, the two unknown functions
CONSTRAINT
y(x) and p(x) must be determined so as to extremize
* The subscripts 1 and 2 are employed here to denote, respec- As a further step, the problem of extremizing the
tively, lower bound and upper bound for the z variable. functional (1) is considered in connection with a
F L I G H T M E C H A N I C S AND V A R I A T I O N A L P R O B L E M S 583
nonholonomic constraint obeying E q . (7) plus a n iso- T h e horizontal distance (X) flown b y the aircraft is
perimetric constraint of linear t y p e : given b y 2 ' 1 2
mV VeV
[3>(x, y) + ty(x, y)y'\dx = given (16)
-a-
X dV (23)
Xl D(m, V) D(m, V) m
For t h e problem represented b y E q s . (1), (7), and (16) where mr = dm/dV. T h e m a x i m u m range problem
the two unknown functions y(x) a n d p(x) m u s t be con- consists of extremizing t h e integral (23), subject to t h e
sistent with t h e Euler equations (8) and (9) where t h e constraint
F function is now defined as follows:
*' + {mft/[Veft ~ D(m, V)]} 0 (24)
F= $ + Vy' + \(y' +f) + X($ + * / ) (17)
where 0 < 8 < 3max (25)
X(x) being a variable Lagrange multiplier and X a con-
s t a n t Lagrange multiplier. T h e development of E q s . I t is to be noted t h a t t h e problem represented b y t h e
(8), (9), and (17) leads to t h e following result: relations (23), (24), a n d (25) is mathematically identical
with the problem represented b y Eqs. (1), (5), and (6).
(d/dx) [tf + X* + X] - ( d $ / d y ) - ( d ^ / d y ) / - This fact is evident if t h e following substitutions are
X(5//by) - X[(5$/dy) + ( d t f / d y ) / ] = 0 (18) made:
Downloaded by KINGSTON UNIVERSITY on March 14, 2015 | http://arc.aiaa.org | DOI: 10.2514/8.7785
n-
dV (30)
(ft^ftmax)- D(m, V) D(m, V) m J
I n flight a t constant altitude, t h e drag D can be con-
ceived as a function D(V, L) of velocity (V) and lift subject to t h e constraints (24) and (25). T h e solution
(L) only. After accounting for t h e equation of motion involves subarcs /3 = 0, subarcs ft = ftmax, and variable-
on t h e normal to t h e flight p a t h (L — m g = 0, where t h r u s t subarcs satisfying t h e relationship
m is t h e mass and g t h e acceleration of gravity), t h e
(d/dV) (Ve/D) - (d/dm) (m/D) = 0 (31)
drag D is reduced to a function D(V, m) of velocity
and instantaneous mass of the vehicle. For t h e particular case of a drag polar obeying E q . (27),
584 JOURNAL OF T H E A E R O / S P A C E S C I E N C E S —SEPTEMBER, 1958
(d/dV) [VeV/mg] - (d/dm) [V/g] = h' + (V/g) [Ws'm 6/(Wsm 6 + D - T)] = 0 (43)
Ve/mg ^ 0 (36)
where 6 is t h e p a t h inclination with respect to a hori-
(5.4) Burning Program for the Maximum Increase in zontal plane and
Altitude of a Rocket in Vertical Flight (Quadratic Drag
Law) - 1 < sin-0 < + 1 (44)
T h e problem of m a x i m u m altitude increase is here This problem can be reduced t o the mathematical mode
investigated once more. All the hypotheses of Section treated in Section (3), b y using the substitutions
(5.3) are maintained, t h e only exception being t h a t a
quadratic drag law (D = KV2, w i t h X = const.) is now x = V, y = h, z = sin 6, zx = — 1, z2 = + 1
assumed.
$ = W/g(T - D), V = W/V(T - D)
The analytical question consists in extremizing t h e
integral 4 f = (V/g) [W sin 6/(W sin 6 + D - T)]
(5.6) A Degenerate Case: Burning Program Extremizing obeying Eq. (27), Eq. (50) reduces to
the Time of Flight for a Rocket in Vertical Flight (Zero
Drag)
With the same hypotheses of Section (5.3) the prob- m= Jl W JW±H±M±^l (5i)
lem of extremizing IB 1V(V+ 3 7.) + X(F + 2 TO
(6.2) Brachistocronic Burning Program of a Rocket in
[-(Vg) - (Ve/mg)m']dV (47) Vertical Flight for Given Altitude Increase (Quadratic
Vi Drag Law)
is considered, subject to the constraints (34) and (35). Under the same hypotheses of Section (5.4), the prob-
The solution formally involves subarcs (3 = 0, sub- lem of minimizing the integral 4
arcs 13 = /3max, and subarcs such t h a t
F
(d/dV) [VJmg] - (d/dm) [1/g] = 0 (48) t= p r - - ^ L — - - -m^[dV (52)
JVl L mg + KV* rng + KV* J
Notice t h a t Eq. (48) degenerates into an identity* this
is considered. T h e nonholonomic constraint repre-
is a clear indication of the fact t h a t an infinite number
sented b y the relations (38) and (39) is accounted for,
of extremals exist, 4 all leading from the initial point
together with the isoperimetric constraint represented
(Vi, Wi) to the final point (V2, m2) in equal times.
b y Eq. (37), where h is now a prescribed q u a n t i t y .
The latter circumstance is logical, insofar as Eq. (47)
Downloaded by KINGSTON UNIVERSITY on March 14, 2015 | http://arc.aiaa.org | DOI: 10.2514/8.7785
and the Weierstrass condition fails along a major por- step, the two limiting lines (3 = 0 and (3 = fimax starting
tion of the extremal arc. from 1 are calculated. Their respective equations are
In the present section, Green's theorem is employed (see Fig. 3)
to overcome the difficulties associated with the indirect
m = const. (61)
methods of the calculus of variations in the analysis of
linear problems. A method of attack is outlined b y V + Ve log m - (gm//3max) = const. (62)
means of particular examples having physical interest.
The technique here introduced is general. In fact, with As a second step the two other limiting lines /3 = 0
minor variants, it applies to all questions formulated and (3 = f3max arriving at 2 are computed (see Fig. 3).
in Sections (3) and (5), and to other problems, also. 8, 10 T h e broken line 13241 of Fig. 3 is drawn; it surrounds
a closed region to which all admissible p a t h s belong.
(7.1) Maximum Increase in Altitude for a Rocket in Now focus attention on the Eulerian arc 132 which
Vertical Flight (Zero Drag)
is formed b y one subarc 13 flown with maximum en-
T h e problem of Section (5.3) is now considered. gine o u t p u t followed b y one subarc 32 flown b y coast-
After introducing the definitions ing. T h e arbitrary, though admissible, comparison
p a t h 152 is considered and the difference in final alti-
* = ~(V/g), ¥ = -(VeV/mg) (57)
tude between the paths 132 and 152 calculated:
Eq. (33) is rewritten as
Downloaded by KINGSTON UNIVERSITY on March 14, 2015 | http://arc.aiaa.org | DOI: 10.2514/8.7785
h {$ + Vm')dV (58)
Ah = hU2 — hlb2 = = f (* + * m ' ) dV
J 132
Vi
I ( $ + Vm')dV = 0 ($dV + Vdm) (63)
(7.1.1) The co Function—The following function of J 152 ./13251
the variables V and m is defined:
B y means of Green's theorem, the above cyclic integral
co = (d^/dV) - (d<f>/dm) (59) can be transformed into a surface integral associated
with the area a (see Fig. 3) bounded b y the clockwise
After accounting for Eqs. (57) and (59), one obtains
circuit 13251:
co = ~{VJmg) (60)
state of flight (V, m) at time instant / and let P be its Since the co function, defined b y Eq. (59), is negative
representative point in the mass-velocity plane (see Fig. everywhere, it follows t h a t
2). After a time interval dt = t' — t, the new mass
Ah > 0 (65)
of the rocket is m' = m + dm, the new velocity is V =
V + dV, and the new vector position of the rocket in In view of the arbitrariness of the comparison p a t h 152
the (V, m) plane is OP 7 = OP + ¥F. it is concluded t h a t the Eulerian arc 132 maximizes
As is evident from Eq. (34) and from the inequality the altitude rise, under the limitation 0 < (3 < f3max.
dm < 0, there is one elementary displacement PP' for B y the use of the same technique, it can be proved
each value of the mass flow (3. Since 0 < (3 < f3max, one t h a t the other Eulerian p a t h 142 (see Fig. 3), formed
infers t h a t the class of admissible displacements b y a coasting subarc 14 followed by a subarc 42 flown
PP' is bound b y two limiting conditions—a displace- with maximum thrust, has the property of minimizing
m e n t PR(f3 — 0), corresponding to coasting flight; and the altitude rise h.
a displacement PT ((3 = /3max), corresponding to flight
with maximum engine output. (7.2) Maximum Range of a Rocket-Powered Aircraft in
Level Flight
While a displacement like PS is physically possible, a
displacement like PQ is physically impossible, because T h e problem of Section (5.1) is now considered.
the mass of the rocket cannot increase; analogously, a After introducing the definitions
displacement like PZ is also physically impossible, be-
$ = -(rnV/D), * = -(VeV/D) (66)
cause the constancy of the mass implies zero thrust and,
therefore, a negative acceleration. In conclusion, the Eq. (23) is rewritten as
two lines fi = 0 and f3 = /3max split the (V, m) plane into *V2
two regions associated with point P: a region (H), ($ + Vm')dV (67)
/ Vx
whose points are accessible to the rocket in vertical
flight; and a region (K), which is forbidden to the (7.2.1) The co Function—-The following function of
rocket, because of the physics of the motion. the variables V and m is defined:
(7.1.3) Analytical Proof—Consider a rocket which
= (d^/dV) - (d$/dm) (68)
m u s t be transferred from the initial state of flight
(Vi, mi) to the final state of flight (F 2 , m 2 ). As a first i.e.,
F L I G H T M E C H A N I C S AND V A R I A T I O N A L P R O B L E M S 587
Downloaded by KINGSTON UNIVERSITY on March 14, 2015 | http://arc.aiaa.org | DOI: 10.2514/8.7785
FIG. 2. Class of admissible displacements of a rocket in the mass- FIG. 5.W Comparison between the Eulerian arc 1562 and thp
velocity plane (vertical flight, zero drag). control path 172 (maximum range, level flight).
1i
m
p-o / v \ a ; # = C)
®
/ KV [ V+2Vel
>*•&» "*>°
®
(3-0
H <
| A< o
where I n view of the fact t h a t co* < 0 in the region SB and co*
> 0 in t h e region SA, it is concluded t h a t
<£* = $ + X<£ (80)
Ar = At + \Ah < 0 (86)
^* = ^ + X^ (81)
T h e isoperimetric condition concerning the altitude rise
(8.1.1) The co* Function—The following function of
requires t h a t the control p a t h 172 be chosen in such a
the variables V and m is introduced:
way t h a t
a,* = ( d ¥ * / d F ) - (d$*/dm) (82)
Ah = AiB62 - hm = 0 (87)
Simple manipulations yield the following results:
As a consequence, the inequality (86) leads to
At < 0 (88)
KV(V + 2Ve) + \KV\V + Ve) ~ XmgVe
(83)
\mg + KV2)2 T h e following theorem is, therefore, proved: "All
Eulerian arcs 1562 such t h a t the variable-thrust subarc
Notice that, for a given negative value of X, the co* 56 is associated with a negative value for X minimize the
function embodies the following properties (see Fig. time of flight for given altitude rise."
6): (a) co* = 0, at points of the variable-thrust sub- By means of similar techniques it can be proved
Downloaded by KINGSTON UNIVERSITY on March 14, 2015 | http://arc.aiaa.org | DOI: 10.2514/8.7785
arc, defined b y Eq. (54); (b) co* > 0, in the region A t h a t : "All Eulerian arcs 1562 such t h a t t h e variable-
bounded between the m axis and the subarc defined t h r u s t subarc 56 is associated with a positive value for
b y Eq. (54); and (c) co* < 0, in the region B bounded X maximize the time of flight for given altitude rise."
between the F axis and the subarc defined b y Eq. (54).
On the other hand, for a given positive value of X, CONCLUSIONS
the signs which the co* function takes into the two
regions A and B are reversed with respect to those T h e indirect methods of the calculus of variations
indicated in Fig. 6. are employed in the study of linear problems of b o t h
(8.1.2) Analytical Proof—-The first step for carrying the simplest type and the isoperimetric type, subject
out the sufficiency proof is to analyze t h e class of ad- to additional constraints involving inequalities. Dis-
missible displacements [Eq. (38) and inequalities (29) continuous solutions are found. T h e Legendre-Clebsch
and (39)] of the rocket in the mass-velocity plane. condition fails to supply any information for all sub-
Considerations analogous to those of Section (7.1.2) arcs forming one extremal arc. Moreover, the Weier-
must be carried out. By drawing t h e four limiting strassian function is zero along a major portion of the
lines /5 = 0 and (3 = f3max starting at 1 and arriving a t Eulerian arc.
2, the broken line 1362451 is determined: it surrounds By the use of a new and hitherto unexploited tech-
a closed region to which all admissible paths belong (see nique, based on Green's theorem, the afore-mentioned
Fig. 7). difficulties are overcome. Rigorous sufficiency proofs
T h e second step is to analyze all possible combina- are developed.
tions of subarcs. In this connection, the analysis shows I t is to be observed t h a t this new technique is of
t h a t the optimum p a t h 1562 includes a subarc 15, a limited interest as far as the calculus of variations is
subarc 56, and subarc 62. T h e equations of these concerned, because it only applies to linear problems.
subarcs are the same as in Table 1, the only difference Nevertheless, as far as the mechanics of flight is con-
being t h a t co is replaced b y co*. cerned, t h e use of Green's theorem is rather important,
Assume now t h a t the five quantities mh Vh m2, F 2 , h in view of the frequent occurrence of linear problems
are prescribed and t h a t the particular extremal path in t h e p a t h analysis of turbojet and rocket-powered
satisfying the above conditions has been found. As- vehicles.
sume also t h a t the value of X associated with this par-
ticular extremal p a t h is negative and t h a t the boundary REMARKS
ario degli Aeroplani, Atti della Accademia delle Scienze di Torino, Rocket Flight, Jet Propulsion, Vol. 28, No. 1, pp. 55, 56, January,
Classe di Science Fisiche, Matematiche e Naturali, Vol. 85, 1958.
14
pp.41-52, 1950-1951. Miele, A., Minimality for Arbitrarily Inclined Rocket Tra-
7
Miele, A., and Cappellari, J. O., Jr., Approximate Solutions jectories, Jet Propulsion (in publication).
T h e r m a l B u c k l i n g of Solid W i n g s . . .
(Continued from page 580)
distribution differs materially for some of the cases sional Stresses, Journal of the Aeronautical Sciences, Vol. 23,
considered. Also, the shape of the cross section is not No. 6, p. 603, June, 1956.
4
Bisplinghoff, R. L., Some Structural and Aeroelastic Consider-
very significant, as long as the wing is thin. T h e ations of High Speed Flight (Nineteenth Wright Brothers Lecture),
curves of Fig. 8 are t h u s a good indication of the end Journal of the Aeronautical Sciences, Vol. 23, No. 4, p. 289, April,
effect in general. I t should be pointed out t h a t for 1956.
5
wings of parabolic or wedge shaped section the curves Singer, Josef, Thermal Buckling of Solid Wings, PIBAL Report
are not applicable to aspect ratios below 2.5, since then 408, Polytechnic Institute of Brooklyn, October, 1957.
6
Singer, Josef, Anliker, M., and Lederman, S., Thermal
the torsional analysis ceases to be valid. Stresses and Thermal Buckling, Chapter 1, WADC T R 57-69, 1957.
T h e results obtained indicate t h a t t h e effect is three 7
Timoshenko, S., Theory of Plates and Shells, pp. 303-307,
to five times as large as t h a t mentioned b y Budiansky McGraw-Hill Book Co., Inc., 1940.
8
and Mayers. 1 T h e error in neglecting the end effect Tricomi, F. A., Vorlesungen tiber Orthogonalreihen, pp. 34,
can therefore be significant even for larger aspect ratios, 166, 177, and 235, Springer-Verlag, Berlin, 1955.
9
Klosner, J., and Forray, M., Buckling of Simply Supported
and the end effect factor should therefore be included Plates Under Arbitrary Symmetrical Temperature Distribution,
also in rough calculations. However, since X can be Republic Aviation Corp., Report No. E-SAM-15, April, 1956.
10
taken from Fig. 8 in most cases, no additional work re- Heldenfels, R. R., and Roberts, W. M., Experimental and
sults. Theoretical Determination of Thermal Stresses in a Flat Plate,
NACA T N 2769, 1952.
11
Horvay, G., The End Problem of Rectangular Strips, Journal
REFERENCES
of Applied Mechanics, Vol. 20, No. 1, p. 87, March, 1953.
1 12
Budiansky, Bernard, and Mayers, J., Influence of Aerody- Horvay, G., and Born, J. S., Tables of Self-Equilibrating
namic Heating on the Effective Torsional Stiffness of Thin Wings, Functions, Journal of Mathematics and Physics, Vol. 33, No. 4,
Journal of the Aeronautical Sciences, Vol. 23, No. 12, p. 1081, January, 1955.
13
December, 1956. Mendelson, Alexander, and Hirschberg, Marvin, Analysis of
2
Vosteen, Louis F., and Fuller, Kenneth E., Behavior of a Thermal Stresses in Thin Plate with Spanwise and Chordwise
Cantilever Plate Under Rapid Heating Conditions, NACA R M Variations of Temperature and Thickness, NACA T N 3778, 1956,
14
L55E20C, July, 1955. Timoshenko, S., and Goodier, J. N., Theory of Elasticity,
3
Hoff, N. J., Approximate Analysis of the Reduction in Tor- (2nd Ed.), pp. 401 and 425, McGraw-Hill Book Co., Inc., 1951.