Professional Documents
Culture Documents
6, 1973
589
entropy that are higher than those of the oncoming flow. The
conditions for isentropic and nonisentropic jumps are formulated
and evaluated next, and a survey of the discontinuities which
satisfy conditions for isentropic and also for selected nonisentropic
jumps is given. Up to this point, the analysis is concerned only with
the state distribution along the control surface. Jumps of the state
in the interior require the occurrence of centered compression
waves. Sample computations show that, in most cases, flow fields of
this character can be generated by the choice of the nozzle shape.
In some cases, no nozzle contours exist which generate the optimizing state distribution along the control surface as determined
by the present analysis. It would then be necessary- to include from
the very beginning conditions for the realizability of the flow field.
I. I n t r o d u c t i o n
This paper presents in its first sections the theory and some numerical results about the optimization of thrust nozzles of prescribed length
in a flow with swirl. More details can be found in internal reports of
the authors (Refs. 1-2). The remainder of the report is devoted to a
special question which arises in this context. Initially, it is assumed that
the state distribution along a certain control surface is continuous.
However, the character of the equations shows that one may encounter
optimum flow fields for which this assumption is not satisfied. It is
true that this difficulty will not appear if only the length of the nozzle
is prescribed, but it may arise if length and exit radius are given and
if the opening angle of the nozzle is rather large. At least from a theoretical
point of view, one would like to know what kind of flow patterns to
expect under these circumstances. The answer may be of use for problems where constraints other than length and exit radius are prescribed
This problem has been explored rather thoroughly by Shmyglevskiy.
Nevertheless, the reader may find the presentation given here worthwhile.
It is simpler, mainly because we use a simpler form of the variational
formulation but also because of a more natural choice of the dependent
variables. Moreover, a considerable amount of numerical details has
been provided. The analysis is not quite complete; there exist nozzle
configurations where the whole flow pattern, rather than merely the
flow at the control surface, must be included in the analysis, even if only
the length and the exit radius are prescribed. The nozzle shapes where this
would happen are rather unrealistic and a discussion of such cases would
hardly be justified unless it is required for a specific practical purpose.
590
2. B a s i c E q u a t i o n s
Considered are axisymmetric flows; let the x-axis coincide with the
axis of symmetry, and let y be the distance of a point of the flow field
from this axis. In a meridian plane, the velocity components in the
x- and y-directions are denoted, respectively, by u and v. Because of the
presence of swirl, a velocity component normal to the meridian plane is
admitted; it is denoted by v t . Let p be the pressure, p the density, T the
temperature, i the enthalpy, and s the entropy. T h e outside pressure is
denoted by p , . We define
(dp/dp)s=cons t ~
a s.
Furthermore, we introduce
q~ = u Z + v ~,
0 = a r c t a n (v/u),
(1)
or
u=qcos0,
v=qsin0.
(2)
(3)
(4)
== ypu,
O4~i~x= --ypv.
(5)
~b, defined in this manner, represents the mass flow through an axisymmetric surface whose intersection with a meridian plane extends from
the axis to a point x, y and which is bounded by the intersection with
two meridian planes which form a dihedral angle 1. This definition
avoids the occurrence of factors 2~r in some of the formulas.
T h e streamlines have the form of rather steep helices with varying
diameter. A surface ~b ---- const in the three-dimensional space is naturally
axisymmetric, a streamline that has one point in c o m m o n with such a
surface lies completely within it. All streamlines of one surface 4~ = const
can be brought into coincidence by rotation around the axis of symmetry.
Since all streamlines lying in a surface 4s = const are equivalent, one
has, for an inviscid flow,
s -----s(~b).
(6)
591
For short, we shall also call streamlines curves arising by the intersection
of a surface = const with a meridian plane. Applying the law of
moment of momentum with respect to the axis of symmetry to a particle,
one fmds
yv, = ~ ( ) .
For later convenience, we set
(7)
i + q? + [gl()/y ~] - g~() = 0.
(9)
= o,
(lo)
(1t)
Aside from these equations, one will have relations describing the
thermodynamic properties of the medium. The system of differential
equations is hyperbolic for q > a.
3. O p t i m i z a t i o n P r o c e d m ' e
At first glance, it seems as if the solution of the variational problem
must the computation of the flow field; one would then be confronted
with a two-dimensional variational problem. But, for nozzles for which
either the length or the position of the endpoint is prescribed, the
number of independent variables is reduced to one if one chooses the
characteristic surface that runs through the endpoint of the nozzle as
control surface for thrust and mass flow. The flow equations are then
compressed into the compatibility condition for this characteristic, and
the endpoint of the controI surface is directly given by the formulation
of the variational problem. This observation was made independently by
Guderley and Hantsch (Ref. 3) and Nikol'skij (Ref. 4). This approach has
592
(12)
fldy,
K
F
~RtSTICS
AB
Fig. 1.
T h r u s t nozzle.
593
(t3)
with
(14)
where derivatives along the control curve with respect to y are denoted
by a dot. Equation (14) is considered as one of the constraints of the
problem. Point F lies on the same streamline as the throat contour; this
determines the value of ~ at point F. At the boundary of the region I
and II, that is, at point K, ~b is continuous. The contribution of the
control curve to the length of the nozzle is written as
L = |
1"VF
.f~@,
(15)
~K
with
f~ = cot %
(16)
g, = g~(),
g~ = g~(,/,).
(17)
Inserting this equation into the differentiated form of (9), one obtains
(l/p) dp -4- qdq q- (gl'/2y ~ -- ga' q- Ts') de -- [gl()/y ~1 dy ~= 0.
(18)
@t~q =
-pq,
@/~y == pg~/y.a.
(19)
594
(20)
fyF
UK
(21)
(22)
0 =
~.
(23)
Now, h is rewritten as
h = q cos(0 --
)/cos
(24)
(25)
The variation 3~b must vanish at the points K and F. One is led to the
equation
dh/dy ----y[(Op/~) cot(0 + c0 tan ~ -- s'(Op/&)q~ sin 0 sin 2 a/sin (0 q- a) cos ~].
(26)
Inserting (23) into (14), one obtains
&b/dy = ypq sin a/sin(0 + a).
(27)
Equations (24)-(27), together with (3), (9), (17), are the equations
governing the problem. The essential variables are h, q, 0, ~b. If the
position of point K is considered as known, then q, 0, ~, h [because of
(24)] are kno~,a at this point, i.e., one has initial conditions at point K.
T h e integration is terminated at the value y, to be denoted by Ye, for
which ~ assumes a value which is given by the mass flow through the
throat. As one computes the flow conditions along the control curve,
595
(28)
For a derivation, we refer to the literature, for instance ReL I, Eq. (4.18).
The problem can be brought into a form which is more suitable for
numerical work. A starting point is given by the observation that the
equations valid for the control surface imply that the compatibility
condition for a left-going characteristic is satisfied. This can be demonstrated by elimination of h from the system. The compatibility
conditions can be taken as a substitute for (26); and the only equation
in which h is still present, namely (24), can be disregarded for h is
merely an auxiliary quantity. The resulting equation contains the
derivatives of q and 0. One of these derivatives can be eliminated by using
the differentiated form of (25). Finally, we consider ~ rather than y as
independent variable. For details of these manipulations, see Ref. 1.
One finally arrives at the following system:
(d log
q/d~)
sin(2c~ - - 0)/sin ~ ~ s i n 0
+ (1/sin ~ cos
c~)(dc~/d~) +
[sin(0 - - a)/sin
+ O!p)(aplas)s' =
(29)
o,
8o9/~z,/6-5
(30)
(31)
(32)
596
(33)
4. E q u a t i o n s U s e d in t h e N u m e r i c a l E v a l u a t i o n
In the computations, the problem has been specialized to an ideal
gas with a constant ratio of the specific heats. A reference state is characterized by a subscript zero. Let
H .... exp[--(s -- So)/R].
(34)
One has
p/p - RT,
y = c~/c~ = const,
(35)
i = c~T = [7/(~' - - 1)] R T ,
~2 =
~,(p/p) =
(7 -
1)i.
Let
P ---- po Hp,:~,
(36)
with
p. = ( r/To)'/(,-'~
= (i/io).I,-~>.
597
(37)
T h e l e n g t h of a line l e m e n t is
(38)
(39)
(40)
T h e c o m p a t i b i l i t y c o n d i t i o n s f o r t h e characteristic lines r e a d
cot a(1/q)(dq/dlj) dO/dt~ (sin 0 sin ~/y)(1 47 ga/y2a 2)
(~ot
(I/y2)(dgl/d) 47 d&/d](d/dl,)
O. ( 4 l )
(42)
(43)
Pa
(1/q)(dq/d){2 sin a cos a ~ cos 0 - - sin 0(sin 4 a 47 cos 4 a) -- [(y - - 1)/2] sin 0}
47 (sin ~ sin O/poHp~sqy2)
{sin(0 - - ~) cos a ~(1 47 gl/y2a ~) 47 [(y - - 1)/2](gx/yZa ~) sin(0 47 ~)}
47 (1/q2)[d&/a4, -- (dg,/a)(1/2y=)]
{cos ~ ~(sin 0 -- 2 sin ~ cos ~ cos 0) 47 [(7 - - 1)/2] sin 0}
(45)
(46)
598
599
teristics is chosen to take the place of IK. Along the axis of the nozzle,
a quotient sin 0/3, is encountered in the characteristic conditions which
must be evaluated by some suitable interpolation. The method which
we have used is described in Ref. 2.
The flow field in region II can be computed from the flow conditions
along the characteristic I K taken from region I and the conditions along
the characteristic KF, which are to be determined by the optimization
process. So far, we do not know which position K will assume if a
certain value for the length of the nozzle is assigned. We choose some
point of field I as point K, and then solve the system of optimization
equations (42)-(46). The computation terminates at the value of ~b
which corresponds to the wall streamline in the nozzle throat. The
results will satisfy- all the necessary conditions of the problem, except
(47). Solutions in which (47) is disregarded can be interpreted as solutions
for nozzles for which the length and the exit radius of the nozzle are
prescribed. The integration of (44) immediately gives the thrust. This
value of the thrust can be compared with the maximum thrust that can
be obtained for a nozzle of infinite length and infinite diameter (assuming
that the outer pressure is zero). At the exit surface of such a nozzle, one
would have i = 0 and for most of the outflowing mass y = Go. Then,
from Bernoulli's equation (9),
q = [2g2(~b)]~,
and hence,
M = f [2&()]~ de.
Let us call nozzle effectiveness the ratio between the actual thrust and the
maximum attainable thrust. The result for arbitrarily chosen positions
of K can be used to determine charts showing the nozzle effectiveness in
dependence of the position point F. For a family of points K which lie
on the same left-going characteristic, the corresponding points F will also
lie on a smooth curve. Along curves of this kind, one can interpolate for
different values of the nozzle effectiveness. From curves of constant
nozzle effectiveness, one can determine the endpoints of the nozzles for
which the thrust is a maximum if only the length is prescribed; they are
given by the point where these curves have a vertical tangent. Actually,
this fact is used merely as a check; a direct characterization of these
optimal nozzles is given by (47). For the tentative curves KF described
above, we have computed the difference of the left- and right-hand sides
in (47). Then, it is possible to interpolate for those positions of point
K where the criterion (47) is satisfied. For these position of K, the
state along K F has also been computed.
600
These data are used to compute the flow field in region II. Of
technical interest is the shape of the streamline IF. To find it, one
determines points of the flow"field for which ~ assumes the value pertaining to the throat contour. This is done by interpolation along the
characteristic lines.
6. Results
The evaluation of nozzle shapes has been carried out for flows in
which the functions g l , g 2 , / / a r e given by
g~ = 452~b~,
//=
1.
L
2=0
2, CORRESPOND TO ~ =.55734
3.
=.80627
84%
82%
il \
78 ',~
%2"~---I
{~ \
~"="5
--T-
THROAT
RADIUS
I.
2.
3~
4.
5.
6.
7.
e,
9,
Fig. 2.
.X .................
io.
601
J O T A : V O L . 12, N O , 6, 1973
1
J
I ........................... ~ ......
I ....
I.
2.
5,
5.
4,
Fig, 3.
--1
I
6.
8.
9.
I0,
~"
,I,,
I.
2.
5.
4,
5.
6.
7.
Fig, 4,
8.
9.
I0,
602
ii
JOTA:
VOL.
12, N O .
6, 1973
I.
2.
3.
4.
5.
6.
7.
8.
Fig. 5.
9.
I0.
N o z z l e s h a p e s for f12 = 0 . 8 0 6 2 7 .
of a given length are obtained for the points of curves of constant nozzle
effectiveness which have a vertical tangent. The nozzle effectiveness
increases slightly if swirl is present, at least for the initial conditions
shown here. A similar observation, although in a different setting, was
made by Shmyglevsky and Naumova (Ref. 9). Because of the reduction
of the mass flow through the nozzle in the presence of swirl, the total
thrust obtained is somewhat lower for nozzles with swirl than for nozzles
without swirl. Thus, the presence of swirl is only advantageous with
respect to fuel consumption, but not with respect to nozzle weight.
The actual shapes for nozzles whose outer radius is allowed to vary
are shown in Figs. 3, 4, 5. The nozzles are very similar for different
values of the swirl. If we had superimposed these three figures, the
nozzle shapes would practically coincide.
7. Further D i s c u s s i o n of the Solutions
The factor of dq/d~b in (45) can vanish, and then this derivative
will be infinite. Of course, one may defer the discussion of such a
singularity until it occurs in a problem of practical importance. But
the phenomenon shows that the class of flow fields admitted so far is too
603
_z
Fig. 6.
604
0
>Y6
\,
,g
Fig. 7.
Y='0D
605
..#"
o
Fig. 8.
a~
j/
Curves of q versus y tbr the state along
the control characteristic in axisymmetric isentropic flows without swirl.
606
= yoq
sii~(~ - O)/sin ~.
(48)
(49)
YK
(50)
607
(5t)
(52)
with
p. =
(iiio)~l<~-~>.
(53)
Then,
c~F/~H = y{p~s + [sin(p -- 0)/sin ~o]pq0 If cos 0 -- e0 cos(00 -- %)/cos ~d}F can be increased if dF/dor < 0. For an ideal gas, one finds that
tan 0 > sin(2~)/[2y -- 1 -- cos(2c0].
(54)
This is the curve VPWQU in Fig. 6. We shall call it the curve B~.
A true maximum of the thrust requires that the solution lie below the
curve B~ and below the curve B1 .
According to these discussions, the solutions obtained so far are
useful only if they lie below the curve B 1 of Fig. 6. If point K in Fig. 1
should lie outside this region, then it is likely that the assumption of
continuity of the data along the control characteristic is no longer tenable,
and thus one is led to explore whether the conditions for an optimum
nozzle can be satisfied by a control curve along which the flow conditions
are discontinuous. But, before we carry out such a discussion, it appears
desirable to define the limits of the previous approach, at least approximately. For this purpose, a problem has been studied where the solution
can be obtained with a minimum of work, namely best nozzles in twodimensional isentropic flow. For nozzles with fixed endpoint, the
solutions are simply given by nozzles which would ultimately produce a
parallel supersonic jet, but chopped to whichever length is desirable
608
D
o
J
@
I
c~
>~-
~.oo
Fig. 9.
,1;o
~:oo
~'oo
i~.~0
x
~.oo
:~.oo
~0
~oo
2~.~o
Optimum nozzle shapes for plane isentropic flows. The curves cut off at any
point give optimum shapes for nozzles terminating at the cutoff point. Curve C:
is the locus of nozzle endpoints which, for outside pressure zero, have maximum
thrust if only the length (not the exit diameter) is assigned, Curve B1 is the left
boundary of the region which has no discontinuities along the control characteristic.
609
interpolation within this field, we have determined the points for which
the nozzles are best if only the length is assigned. For such points, the
condition (28) is satisfied. This is the curve C 1 . Moreover, we have
drawn a curve which corresponds to the curve B 1 of Fig. 6, again denoted
by B I . This is the curve which bounds the region where the solutions
correspond to a m a x i m u m of the thrust (provided that we do not admit
changes of entropy), at least in the small. A n y point of the portion of the
xy-plane that is covered by these curves and ties to the right of B 1 can
be taken as the endpoint of the nozzle, and the curve running through
it gives the best nozzle which has this assigned endpoint. For a nozzle
endpoint lying to the left of the curve B1, the continuous solution is not
a m a x i m u m of the thrust, even in the small, and the previous procedure
must be modified. For points to the right of this curve, the procedure
holds, even if the nozzle contour intersects the curve B 1 .
One comes to the conclusion that optimum nozzles for which the
flow conditions along the control characteristics are discontinuous
correspond to cases where the opening angle of the nozzle exceeds by far
the angle which would be desirable for m a x i m u m thrust at the available
length. Thus, the practical importance of the following discussions is
limited. Nevertheless, it may be welcome if we clarify some of the basic
questions.
8. l u m p C o n d i t i o n s
We assume now that, at some point K of the control curve A K F , a
discontinuity occurs (see Fig. 10). T h e control curve of the unvaried
flow is a characteristic. T h e portion A K in Fig. 10 lies in region I;
the portion K F lies in region II. We assume that the flow fields I and II
are defined in a neighborhood of the control curve, but we postpone
the question of how a j u m p of a certain kind can be physically established.
F
.i K!
610
x,v" -- XK = 3X
and
YK" - - Y K == ~Y"
M =
foK~Ady @
f~dy
YKH
COS
O)KI - - (Dq
cos
O)Kll] - -
~x[(~q
[fl + h(y)g ~- h.f2] dy + Sy[(p + pq2 cos 20)g~ -- (p ~, pq2 COS~ 0)g~l]
YKI1
-
(p + pq2 cosa 0)K~ _ (p + pp2 cos2 0)zix -- h[(pq cos 0)Kx -- (Pq cos 0)/u] =-- 0,
(55)
(pq~ cos 0 sin 0)~c~-- (#qZ cos 0 sin 0)KH -- h[(pq sin 0)gI -- (pq sin 0)Kix] = 0.
(56)
J O T A : V O L . 12, N O . 6, 1973
611
(57)
(58)
where
(59)
(60)
(61)
612
5"0V
tlaLl
*~
~iii,~i!lli
/:li;ll!rj
2"0t
l/ !lwr~
IL IJ';ll
//! '!Ill tr
7N
],O n
4:.
al'-~!
,fl~I, i
Fig. 11.
/,
:2
, i , ! !~
i l
L.
2
i i
q
5
vertical lines and solid horizontal lines has been imposed. This grid
is mapped by the jump conditions into another grid of the q0-plane.
In Fig. 11, dashed or solid lines are mapped into dashed or solid lines,
respectively. We refer to the portion of the curve B 1 which lies to the
right of its maximum as its right part. Points of the right part of B 1 are
mapped into themselves. In this manner, one can identify the dashed
and solid curves with their maps. The lower points of the left part of B 1
map into rather distant points of the q0-plane. They can be identified
by the symbols used for these points. In its upper portion, the left part
of B 1 is mapped into itself. For details of this kind, the plotting of the
curves (61) is very helpful. Figure 11 extends up to 0x = ~r; actually,
0, = =/2 is the outside limit for values that are of practical interest; the
wider range has been included to give a survey of the whole field of
curves. The dashed line of Fig. 14 in the field below the curve B 1 gives
approximately the locus of the points that arise from a state I for 0i = ~r/2.
For nonisentropic discontinuities, we have chosen the state q , ,
011 along the portion of the curve B~ which lies underneath of B 1 . If the
jump would lead to points that lie below this curve, then the thrust could
be improved if a smaller value of the entropy change would have been
6t3
chosen; if the points lie above this curve, a larger entropy change could
be advantageous. There may arise specific technical problems where
other possibilities should be admitted; but, for a first survey, the choice
made here is probably sufficient. Plotting the expression (61) versus ql
.2 1 I
/
0
Fig. 12.
[
1.5
2.5
614
one recognizes that, for a given entropy change, each state qH, 0u
chosen along B 2 can be reached from two different states q~, O,.
Figure 12 gives a survey of the results. By varying 17, one obtains,
for each point qii, 0tl, two curves for the state q~, Ox from which this
point can be reached. By varying the position of the point along B 2 ,
two families of curves are generated. Only one of them is shown in
detail, because the curves of the other family overlap rather strongly.
The cross-hatched region in Fig. 12 shows one of the regions from
which suitable nonisentropic jumps lead to the curve B 2 . The boundary
of the region for isentropic jumps is given by the curve B1. The
cross-hatched region lies entirely within the region of isentropic jumps.
For a given state qi, 0~, one can therefore find one solution for an
optimum nozzle without entropy change and a second one with an
entropy change. Of course, the endpoints of these nozzles will be
different, but it is conceivable that, even for the same nozzle endpoint,
two different solutions of the optimization problem (in the small) may
exist. Challenging as such speculations may be, it seems to be best to
defer a discussion until the problem arises in a more realistic technical
setting.
9. D i s c o n t i n u i t i e s in t h e F l o w F i e l d
So far, our discussion has been restricted to the state along the
control surface. The question arises of now state distributions which
have jumps along the control surface can be generated in an actual flow
field. Flow fields with discontinuities can exist only if qx > qH ; for, in
the interior of a flow, sudden expansions are impossible. A sudden
isentropic compression is brought about by centered compression waves.
We ask whether such a configuration will generate the discontinuities
required here, namely, a transition from a state qt, 0i to the state qu,
OH that belongs to it. Our discussion is restricted to the immediate
vicinity of the discontinuity; therefore, the results are the same for plane
and axisymmetric flows and even for flows with swirl. The oncoming
flow has the state I. A family of centered right-going compression waves
coalesce at point K (Fig. 13) and form a shock downstream of K. The
streamline that runs through K separates the streamlines that do not
pass through the shock from those that pass through the shock and,
therefore, experience an increase in entropy. The velocity direction
and the pressure in the parts of the flow field adjacent to this separating
streamline are the same (the velocity has a jump). From K, some leftgoing Mach waves (or a left-going shock) will emerge. The discontinuity
615
centered e o m p r e s s i o n ~
K~
velocity vector in
the oncom~i.
ng ~
*I~
do~stre~
~
X ~ I
of the shock
shock
5/
Fig. 13.
sin/~.
(62)
Let
~2
~,~
= [2/(r + O]{(a? + [(r -
U/2]wgO}.
(63)
= as ,
(64)
(65)
(66)
616
For an ideal gas with constant ratio of the specific heats, one has
P2/Pl' = H(a2~/a12) 1/(~'-1)
Therefore,
17 = (w,aiw,~2)(a12ta~2) la~-1).
(67)
0a = 0a -- 01.
(68)
Let
F r o m the shock relations just derived and the further condition that the
vdocity component tangential to the shock remains unchanged, one
obtains
qd = q? cs~ fi + w,,2,
(69)
tan 0 z = (qi sin/~ cos fi - - wn2 cos fi)/(qx cos2/~ q- wn2 sin/~),
0a = 0, q- 0a.
(70)
(71)
These formulae allow us to compute the state 3, if state I and/~ are given.
State 4 is connected to state 3 by the equations
G = 04,
pa " - p ~ .
(72)
(74)
State 2 and 4 lie on the same right-going characteristic. The characteristic conditions are
J(q) q= 0 = const,
(75)
where the upper sign applies to left-going and the lower sign to rightgoing characteristics, and
J(q) =
cot c~(q)(dq/q).
(76)
617
For an ideal gas with constant ratio of the specific heats, one has
(77)
where
v =
cot ~ =
[(q* -
~)/{1 -
[(r -
1)/(r +
1)lq~}P.
(78)
(79)
and
J = 3 arctan(v/3) -- arctan v.
(80)
F = 0.
(81)
F can be evaluated as a function of/~, and one must seek the value of fi
for which F vanishes.
State 5 and state I lie on the same left-going characteristic, and
state 5 and state II lie on the same right-going characteristic; hence,
Hence,
1
05 = ~[J(qzx)
- J(ql) + 0ii ~-I or].
(82)
Of course, one can also determine J(qs) and qs. I n order for the perturbance which propagates f r o m K to the left to be an expansion fan,
one must have
05 -
o4 > o .
(83)
I n order for state II to lie within this expansion fan, one must have
05 -- 01, >~ 0,
(84)
e,, -
(85)
o4 > 0 .
618
Fig. 14.
619
has a subsonic region downstream of the shock. The shock itself arises
upstream of the ideal point of impingement, and part of the necessary
turning of the streamlines is brought about in the subsonic field.
For points marked by m, a value of/~ can be found, but the conditions (84)-(89) are violated. In some of these cases, the left-going
perturbation which starts at point K is a (weak) shock, and therefore no
characteristic in which one encounters the state II exists. In other cases,
there is a left-going fan of expansion waves, but it does not contain
state II. But these states can occur only in flow fields of a rather uncommon character. In principle, phenomena of this character indicate
that the variational problem can no longer be solved by studying the state
distribution along the control characteristic alone; the flow conditions
the control characteristic are now limited by the requirement that a flow
field generating these conditions must exist. Nevertheless, one would
expect in many cases that the actual flow field is still rather well approximated by the configurations considered so far, and that the modifications which must be admitted are only minor. For instance, the
right-going compression waves which are responsible for the jump might
2
I
o
Fig. 15.
620
not be completely centered; then, one would find, instead of the weak
shock in the region II, a sequence of left-going compression waves,
which do not coalesce within the flow field, and one of these compression
waves might be suited as the control characteristic. Again, the authors
believe that detailed discussions should be postponed until they occur
in examples of actual technical interest.
An example for a nozzle with an isentropic discontinuity is shown
in Fig. 15. The basic problem is the same as in Fig. 9. The line I F is one
of the characteristics of the expansion fan that starts at point I; it represents the boundary of region I. The nozzle contour between I and H
has been constructed in such a manner that it generates compression
waves centered at point K. The state at KI is given by q~ = 2.108,
0~ = 0.6632. The state at KH is then computed using the formulas of
Section 8; one finds qH = 2.0266, 0ix = 0.5645. For these values, the
states 3, 4, 5 have been determined using the procedure of Section 9.
Only the state 5 is important for nozzle shape. The left-going perturbation that starts at K is extremely weak, but the state at KII corresponds
indeed to one of the expansion waves of this perturbation. The flow
conditions are the same all along KF, since we have the special case of
an isentropic plane flow. The nozzle shape has been found graphically
by means of the method of characteristics, taking as starting points the
state along the chosen characteristic IK, previously computed in connection with Fig. 9. If one would transfer this particular nozzle shape
into Fig. 9, one would find that it ends at a point F slightly to the right
of the curve B~, that is, in a region for which solutions which are
continuous along the control characteristics are also possible. Thus, there
are two candidates for the best nozzle. Of course, it is not difficult to
decide, in a specific case, which of the two solutions gives the larger
values of the thrust, but the example shows the difficulties which one
would encounter in a general theory.
(86)
62t
Since
sin(50 - - 0)/sin 50 = cos 0 - - sin 0 cot 50,
one finds the following alternate expression:
(87)
where
A = ~ (pq2 __ 00%2) sin 0 cos 0 + %[cos(0 o - - %)/cos o:d(pq -- 0o%) sin 0
+ poqo2[--sin 0(cos 0 - - cos 0o) + sin0 o tan %(sin 0 - - sin 0o)].
The
(88)
f o r m o f A s h o w s i m m e d i a t e l y t h a t A ~ 0 f o r q = q o , 0 = 0o ,
50 = 500 O n e h a s
F~ == - - ( 1 / s i n 2 50)A,
(89)
F ~ ( q o , 0 o , % ) = 0.
(90)
I n c o m p u t i n g F~o , w e set q -~ q o , P = Po
(91)
Moreover,
/Tq = _ (1/sin 2 50)Aq = - - ( 1 / s i n s 50)
(92-2)
D i f f e r e n t i a t i n g (86), o n e o b t a i n s
Fo = (1/sin ~)(--pq~ cos(~o - - 20) + pq qo[cos(0 o - - so)/cos %] cos(9 - - 0)}.
T h i s e x p r e s s i o n v a n i s h e s f o r q = q o , 0 = 0 o , 50 = 500, as o n e r e c o g n i z e s
immediately. Hence,
(93)
F o r t h e c o m p u t a t i o n o f F q o , w e s e t i m m e d i a t e l y 0 = 0 o , 50 -----500 T h e n ,
622
Hence,
(94)
0o .
I t is t h e n o b v i o u s t h a t Fq = 0 f o r q =
qo, 0 =
0 o , 9 = 9o H e n c e ,
Fqq(qo, 0o, %) = (p0/sin %)[--2(sin 0o/sin % ) ( & / d log q) -- cot % cos(00 -- %)].
(95)
Next, we c o m p u t e
26F = F~q(3q) 2 + Foo(30) 2 + F~(3~) "~+ Fqo3q30 --r 2F~o6p30 + 2F~&p3q, (96)
with
(97)
(98)
ac -- b 2 > O,
or, m o r e specifically,
cos2(0o - - %) -- [cot % sin(0 o - - %) -- (sin 00/cos %)(d~/d log q)]2 > 0.
(99)
(100)
(101)
(102)
623
(103)
(104)
or
After substituting the relation (102) for an ideal gas, one obtains
tan 0o < sin(2%)[1 + cos(2%)]/[7 + cos2(2%)].
(105)
(1/sin ~ sin 0)[sin(2~ - 0)/sin ~ + (sin 0/cos o~)(da/dlog q)](d Iog q/d~b).
11. A p p e n d i x B: S p e c i a l P r o p e r t i e s o f t h e C o n d i t i o n s f o r
Isentropic Discontinuities
First, we collect some results from the variational treatment of
continuous solutions. The equations which we shall use are direct
8o9/I2/6-7
624
P = p(q, Y, ~).
(106)
(107)
flo - - hf~o = 0.
(I08)
T h e y are satisfied if
h(q, O, y, ) = q cos(O -- cO/cos c~,
9 = 0 + ~.
(109)
(110)
(111)
(112)
We introduce
Here,
In (25), we found
--Yf4(q, O,y, ~b) = h = const.
(113)
625
entropy changes are not admitted. At the same time, B 1 is the curve for
which dy/dq = dy/dO = 0. For a solution of the variational equations,
one obtains, from (26), (109), (113),
l eh/aq ah/ao
~A/Oq of~/ao[
= o.
(114)
(ahlae)(d41do) + ahlaO = o.
(115)
(116)
Ea~f~feq~
(117)
(118)
626
Similarly,
[8~fl/Oq80 - h(8~f~/~qSO)](d~/dO) -t- [(~fl/80 2 -- h(8~fa/802)]
50(Cgfl/~50])Kt I --
(119-1)
(Ofa/e50)x,] = 0. (119-2)
(120)
This equation holds for any choice of 50. In this appendix, the j u m p
conditions are used in the forms (1 19) and (120).
In the following discussions, the state II is considered as given,
and we ask for which values of qz and 01 Eqs. (119)-(120) are satisfied.
T h e system (119)-(t20) has one trivial solution, namely, q~--~ qH and
0 t = OiZ. We shall show that, for this solution, these equations can be
satisfied up to terms of the first order by a suitable choice of dqffdOt.
I f we had a single condition, this would correspond to a zero of the
second order. In these discussion, 50 is arbitrary, it can be considered
as constant, and h is constant, too, for we consider the state II as given.
One then obtains, for the first-order terms in (120),
[(efl/Oq) -- h(SA/eq)]r,n dq~ + [oS~t80 - h(eA/OO)]x , dO, = O,
but because of (t08) the coefficients of dql and dot vanish separately.
Therefore, we are free to choose dqffd01 in such a manner that (119) is
satisfied up to the first order. One has specifically
[8~'f1/Sq850 -- h( ~ f a/3qc~50)]xu( dq~/dO,) + [82f~/80850 -- h( O2fa/80850)]xu .... O.
627
For cases where state II corresponds to a point of the curve B 1 , one can
use (117) to rewrite this equation as
(y/sin2 q~)[(~fa/Oq)(dqt/dOz) -- ~f4/~O] + (Of3/~q~)i(~h/~q)(dqr/dO,)- Oh/~O] = O.
This equation is t h e n satisfied if one chooses
dqI/dO, = d4/dO.
(121)
References
1. GUDERLEY, K. G., and TABAK, D., On the Determination of Optimum
Supersonic Thrust Nozzles of a Given Length for a Flow with Swirl: Theoretical
Part, Office of Aerospace Research, United States Air Force, Report
No. ARL 66-0013, 1956.
2. GUDERLEY,K. G., and BREITER, ~ . C., On the Determination of Optimum
Thrust Nozzles of a Given Length for a Flow with Swirl: Numerical Results,
Office of Aerospace Research, United States Air Force, Report No. ARL70-0161, 1970.
3. GUDERLEY,K. G., and HANTSCH,]~., Beste Formenfuer Achsensymmetrische
Uberschallschubduesen, ZFW, Vol. 3, pp. 305-314, I955.
4. NIKOL'SKI, A. A., Concerning Bodies of Revolution with a Duct which
Possess l]/[inimum Wave Drag in Supersonic Flow (in Russian), Collection of
Theoretical Works on Aerodynamics, Oborongiz, Moscow, USSR, 1957.
628