A.

B E R N S T E I N 1
Senior Analytical Engineer,
Pratt & Whitney Aircraft,
East Hartford, Conn.

W. H. H E I S E R 2
Assistant Professor,
Department of Mechanical Engineering,
Massachusetts Institute of Technology,
C a m b r i d g e , Mass.

C. H E V E N O R 3
Analytical Engineer,
Pratt & Whitney Aircraft,
East Hartford, Conn.

Compound-Compressible Nozzle Flow

A one-dimensional theory based upon fundamental flow relationships is presented for
analyzing the behavior of one or more gas streams flowing through a single nozzle.
This
compound-compressible flow theory shows that the behavior of each stream is influenced
by the presence of the other streams.
The theory also shows that the behavior of compound-compressible flow is predicted by determining how changing conditions at the
nozzle exit plane affect conditions within the nozzle. It is found that, when choking of
the compound-compressible flow nozzle occurs, an interesting phenomenon exists:
The
compound-compressible flow is shown to be choked at the nozzle throat, although the individual stream Mach numbers there are not equal to one. This phenomenon is verified
by a wave analysis which shows that, when choking occurs, a pressure wave cannot be
propagated upstream to the nozzle throat even though some of the individual streams have
Mach numbers less than one. Algebraic methods based on this
compound-compressible
flow theory are used to demonstrate the usefulness of this approach in computing the
behavior of compound-compressible flow nozzles.
A comparison of the compoundcompressible flow theory with three-dimensional computer calculations shows that the
effects of streamline curvature on nozzle behavior ca?i be disregarded for many practical
nozzle configurations.
Test results from a typical two-flow nozzle show excellent agreement with the predictions from the theory.

j\l INTERFACE

Introduction
M IcODERN propulsion engines often exhaust several
different streams of gas side-by-side through a single nozzle, Fig.
1. These flows can exhibit sizeable compressiblity effects and
they will be referred to here as compound-compressible nozzle
flows. The purpose of this paper is to provide, for the first time,
a simple method to predict the behavior and clarify the understanding of such flows.
A one-dimensional analysis similar to that used in single-stream
compressible flow problems is applied here to compound-compressible flow problems. The great advantage of this type of approach is that it provides physical insight into the nature of the
flow.
Mixing between the various streams is not considered in the
development of the basic theory, but its effect on compoundcompressible flow behavior will be discussed. It will be shown
that mixing often has a negligible effect on the flow behavior.
The usefulness of the compound-compressible flow theory is
demonstrated by comparing its predictions with both threedimensional computer calculations and experimental results.
The basic approach used to develop the compound-compressible flow theory will be to determine how changing conditions at
the nozzle exit plane change conditions within the nozzle. This
will be seen to be the heart of the matter and all results obtained
in this paper are presented in this light. Note should be taken
of a pioneering contribution to compound-compressible nozzle
flows made by Pearson, Holliday, and Smith [l]. 4 Their results
are consistent with the general conclusions arrived at in this
paper.
Now, Propulsion Staff Engineer, Commercial Airplane Division,
The Boeing Company, Renton, Wash.
2 Now, Head, Turbine Research and Technology Group, Pratt &
Whitney Aircraft, East Hartford, Conn.
3 Now, Computer Specialist, Gerber Scientific Instrument Company, South Windsor, Conn.
' Numbers in brackets designate References at end of paper.
Contributed by the Applied Mechanics Division for publication
(without presentation) in the J O U R N A L O F A P P L I E D M E C H A N I C S .
Discussion of this paper should be addressed to the Editorial Department, ASME, United Engineering Center, 345 East 47th Street,
New York, N. Y. 10017, and will be accepted until October 15, 1967.
Discussion received after the closing date will be returned. Manuscript received by ASME Applied Mechanics Division, November 15,
1966. Paper No. 67APM-L.

FLOW 3

i
2
1
1
FLOW 1 !
i i;
1 /

SEPTEMBER

ii
1

1
1
1
i

INLET
PLANE

Fig. 1

Schematic d r a w i n g of a x i a l l y symmetric compressible-nozzle flow

One-Dimensional Compound-Compressible
Nozzle Flow Theory
The development of one-dimensional compound-compressible
nozzle flow theory follows that of Shapiro [2] for single-stream
flow. The most important alteration is that the fluid static
pressure is chosen as the dependent parameter because it can vary
only along the nozzle in one-dimensional flow, whereas all other
fluid properties can also change from stream-to-stream across the
nozzle.
This analysis is sufficiently general to include any arbitrary
number of streams designated by the integer n. For example,
at anj' position in the nozzle,

dx

5 4 8

~ t=
^l dx

(1)

where A is the total flow area, / l , is the flow area of the ith stream,
and x is the axial nozzle position coordinate. In single-stream
one-dimensional theory, dA /dx is arbitrarily small and this carries
over into the present case where all dAJdx are arbitrarily small.
The transverse pressure gradients caused by streamline curvature
can then be neglected and this leads to the conclusion that static
pressure is only a function of axial position.
It is also assumed that the flow in each stream is stead}', adiabatic, and isentropic and that each fluid is a perfect gas with constant thermodynamic properties. Note that these assumptions
exclude mixing effects.

Transactions of the A S M E

1967

Copyright 1967 by ASME

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

to Fig. 2, the consequences of the choice of inlet pressure can now

be examined.

n STREAMS

D T V _
r0'H oi '
i'
R.

If the inlet pressure is sufficiently high (curves a and b), equation (6) will yield values of M , 1 small enough that (3 will be > 0
everywhere in the nozzle and p will therefore change in the same
direction as A throughout, equation (3). In particular, both p
and A will have their smallest values at a geometric throat where
A reaches its minimum. Note that the integration also shows
what the back pressures must be to maintain these flows.
At the same time, the differentiation of
with respect to p
yields

KNOWN

d_
dp

Fig. 2

Compound-compressible flow in a nozzle of fixed geometry

Each stream may then be separately treated as a single-stream

one-dimensional flow (Shapiro [2], Table 8.2). Consequently,

da;

_ Ai ( J L _
A,- V M , j

i\
(In p)
/ da;

(2)

where A:, is the ratio of specific heats, M , is the Mach number (M,= VJy/kiRiTi)
for the ith stream, and p is the fluid static pressure. Equations (1) and (2) may be combined to yield

P ^ w [(1 -

M'a)i +

1 +

M, 2

> 0
(7)

which shows that (3 always changes in the same direction as p.

Therefore, for curves a and b, /3 will also change in the same
direction as A and will also have its minimum value at the throat.
As the inlet pressure is decreased, the value of /3 at the throat
will also decrease. In fact, when the inlet pressure is chosen to
be sufficiently small, /3 reaches zero at the throat. When this
occurs, equation (3) is indeterminant and no longer serves to
determine the axial pressure gradient at the throat. Under this
condition, application of L'Hospital's rule to equation (3) yields

dx

(In p)

dA
dx

(In V )

dx

1
M.
/3 dx

i -k ( - 2
hi
< \m,

d2A
(3)

dx2
M,-2)2 + 2

where

(8)
k< \ M , 2

The term /3, the compound-flow indicator, will subsequently be

shown to be significant in determining the nature of the flow (i.e.,
whether it is compound-subsonic or compound-supersonic).
The behavior of compound-compressible flow in a nozzle of fixed
geometry can be most profitably examined by regarding the inlet
pressure as an independent variable, Fig. 2. If the stagnation
pressures (poi), the stagnation temperatures (7V), and the gas
properties are constant and known for each stream, the mass
flows (w,) are functions only of the local pressure and the local
flow area:

Wi

Ajpoi
=

VToi

Co.)

G, - l ) _i Co,)

^H,

(5)

Thus, for any given value of p at the inlet plane, where the A ,
are known, equation (5) may be used to determine the corresponding values of w t . With the w { fixed by the inlet pressure,
it can be seen from equation (5) that the local A{ are functions
only of the local p and known quantities in the remainder of the
nozzle. Since the local M , are also functions only of the local p
and known quantities, namely,
l- 1
M, =

kj -

L(?)" -

(0)

it follows directly that the local value of j3, equation (4), is a

function only of the inlet pressure, the local pressure, and known
quantities. Therefore, equation (3) can be integrated in principle
from inlet to exit for any chosen value of inlet pressure. Referring

Journal of Applied Mechanics

The geometry of any throat is such that d'A/dx' is always > 0 .

Therefore, d(ln p)/dx will be either the positive or negative root
of a real number.
Curve c represents the choice of the positive root while curve d
represents the choice of the negative root. Comparison of
curves c and d reveals a familiar single-stream compressible flow
situation: The geometric throat is a saddle point for two isentropic solutions in the divergent section of the nozzle. N o back
pressure between curves c and d can correspond to an isentropic
flow. It is anticipated that those back pressures which do not
correspond to isentropic solutions, such as that of curve e, may
be reached by means of compound shocks initiated at some point
on curve d.
The behavior of the flow along curve c is similar to that of
curves a and b; i.e., A, p, and (3 will reach their minimum values
at the throat. Note that the positive root of equation (8) is
chosen only when the back pressure corresponds exactly to that
of curve c.
The implications of the choice of the negative root of equation
(8) will now be considered in detail (curve d). Since equation (7)
has shown that j3 always changes in the same direction as p, dfl/dx
must also be negative at the throat. Accordingly, /3 must decrease from positive to negative as it passes through zero at the
throat. Furthermore, with /3 negative entering the divergent
section (where dA/dx is > 0), simultaneous examination of equations (3) and (7) shows that the local values of both p and /3 must
continue to decrease through the divergent section. Note that
the integration of equation (3) still shows what the back pressure
must be to maintain this flow.
No isentropic solutions exist for inlet pressures corresponding
to values less than that of curve d because /3 would reach zero
upstream of the geometric throat. This would result in an infinite
axial pressure gradient, equation (3).
SEPTEMBER

1967

549

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Some interesting conclusions can be made by examining the influence of back pressure, p, on the inlet pressure. A back pressure greater than that of curve c will affect the pressure at the
inlet plane and will thus influence the flow rates of the individual
streams. A n y back pressure less than that of curve c will affect
neither the inlet pressure nor the flow rates. This condition will
be referred to as compound-choking.
Under such conditions, the
nozzle geometric throat controls the behavior of the flow.
Also, since dp/dx is always < 0 for curve d, Bernoulli's equation shows that a continuous acceleration of the flow takes place
throughout the nozzle. Applying this to equation (3), it can be
seen that, for continuous acceleration of the flow, |3 > 0 whenever
dA/dx < 0 and (3 < 0 whenever dA/dx > 0. Thus, for continuous acceleration of the flow in a single-stream convergentdivergent nozzle, examination of /3 reveals that the flow must be
subsonic in the convergent section, sonic at the throat, and
supersonic in the divergent section, equation (4). In the following section, it will be shown that for compound-compressible flow
an analogous situation exists: The flow must- be compound-subsonic in the convergent section, compound sonic at the throat,
and compound-supersonic in the divergent section. It will also
be shown that these regimes are differentiated by the compound-flow indicator, /3.

Compound Waves
The compound-choking phenomena just described can be explained by examining the effects of small pressure disturbances
on the flow. A diagram of such a disturbance is shown in Fig. 3.
It is consistent with one-dimensional theory to take the flow area
as constant in wave calculations. If a weak plane pressure disturbance is imposed on the flow, Fig. 3(a), this disturbance cannot
propagate at different absolute velocities in each stream without
violating the condition that the static pressures at the stream
interfaces be equal. Therefore, the wave must be continuous
and must travel as a single compound wave, Fig. 3(b).
Although
the wave is not necessarily plane, the pressure rise across it cannot
vary from stream to stream.
As indicated in Fig. 3(6), the absolute terminal velocity in the
upstream direction of the compound wave is designated by a . It
follows directly that: a > 0 corresponds to compound-subsonic
flow; a = 0 corresponds to compound-sonic flow; a < 0 corresponds to compound-supersonic flow.
An analytical expression for the compound wave velocitj^, a ,
can be derived by treating each stream separately as a flexible
tube and conserving mass, momentum, and entropy across the
compound wave in the frame of reference of the compound wave.
It follows that
AAf

A^

A-i

kiPi

(9)

M=+m,.Y

A VhRiTi

where A signifies the change across the compound wave.

the flow area is constant,

Y , A/1, = A.-t = 0
i=i

Since

(10)

across the wave. And since pt = p and A p t = Ap, equations (9)

and (10) may be combined to yield the desired relation for
the compound wave velocity a\ namely,

= 0

i = 1 Ki
A VktRiTi

Equations ( 4 ) and ( 1 1 ) may be combined to yield

550

SEPTEMBER

1967

(11)

3a INITIAL

PLANE PRESSURE

DISTURBANCE

FLOW

FLOW

cy
F<AP
PiAP

FLOW

P -A P

FLOW

P AP

' / / / / / / / / / / / / /// / / / / / /
3b UPSTREAM

MOVING

p Al
p A2

Fig. 3

WAVE

I ! A3 AA3
i A2 AA2
a -'
AA|

P A.I
p A2

COMPOUND

AA2
r 1 A2
,
A3 *AA3

A3

IN TERMINAL

STATE

P A P
P AP
P A P
P AP
PIA P

Evolution of a one-dimensional compound w a v e

"

0 - E

i
f

M,2

(vfe

+ M') _

(12)

Examination of equation ( 1 2 ) shows that the compound wave

velocity a , and the compound flow indicator j3, must always have
the same sign. Therefore /3 > 0 corresponds to compound-subsonic flow; P = 0 corresponds to compound-sonic flow; /3 < 0
corresponds to compound-supersonic flow.
Nolo that compound-choking can only occur at the nozzle minimum area for it is there only that (3 can equal zero. Note also that
small-amplitude compound waves cannot move upstream in the
compound-supersonic region. One would therefore expect that
compound-shock waves could arise and cause the steady compound-compressible flow to be nonisentropic in the compoundsupersonic region. B y the same reasoning, compound-shock
waves are not expected in the compound-subsonic region.
Several conclusions may be drawn from this analysis:
1 The concept of flow choking at Mach one is no longer valid
in compound-flow analysis. Indeed, when compound flow is
choked, the individual stream Mach numbers at the throat will
not be equal to one (except for the unique case where the stagnation pressures of all the streams are equal and the Mach numbers
of all the streams are unity). Rather, compound-choking is determined by the compound-flow indicator and can occur only
when j8 = 0 at the minimum nozzle flow area.
2 Not every stream need have a Mach number < 1 in order
that the flow be compound-subsonic and not every stream need
have a Mach number > 1 in order that the flow be compoundsupersonic. In fact, equation (4) shows that the various streams
influence f3 in proportion to their flow areas, which agrees with
intuitive reasoning.
3 The compound-flow regimes determined by j3, compoundsubsonic and compound-supersonic, are analogous to the subsonic and supersonic flow regimes encountered in single-stream
nozzles. In fact, the usual single-stream results are obtained
when n = 1 in the foregoing equations.

Computational Procedures tor Compound-Compressible

Nozzles
Despite the need for differential equations in drawing conclusions about one-dimensional compound-compressible flows, the
solutions for the behavior of these flows will lie seen to require
only algebraic computations.
Although the equations in the preceding sections were derived
for n flows, to illustrate the application of this analysis, only
two-stream convergent-divergent nozzles will be considered in
this section. This will not only result in the development of less
involved equations but will also clarify the role of the geometric
throat and the phenomenon of choking.

Transactions of the A S M E

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

As has been shown, the behavior of compound-compressible

flow is determined by the relationships of k{, Rh A,-, wit Tm, po;, and
p. Equations (1) and (5) can be combined to yield

uu Vr~*i / M

i=1

Poi

12

k <

kj

Ir> \ki - 1/ L

\pj

ki-V
b

/
\PoJ
=

(13)

1 = 1

Using equation (13), the following expression may be written for

two streams at any point in the nozzle:
fcl+1- A

wt

VT,
ir, VT,

( V ) * '

\ p j

1-ii
V'l
Tm ^ /h

w, V t

u *

Under choked flow conditions, it has been shown that the flow
behavior is determined b y the nozzle geometric throat where
ft = 0. Equations (4) and (6) can be combined to yield the following equation for choked conditions:
v2

Thus, for unchoked flow, equation (14) may be applied at the exit
plane. Typical solutions of equation (14) for unchoked flow are
presented in nondimensional terms in Fig. 4 as the curved lines
of different A , * / A ^ \ t . Although these curves were generated for
particular values of k, R, and temperature-corrected mass flow
ratio (i2 V T m / W i Vt\>\), similar solutions can be generated for
any other values of these parameters.

( p . )

i"i'/i

[R>k\ \A-2 1 /

\Po./

( h

\A-1 p j

1\

Po2\ JA-i -

C02)

'ft

/p_\
L

l)

(a-2 -

-11

l-h

C02)

h-r
hi

-l

1
where
Wi

fa+r

IC^)'-'

VT01
Poi

<

These are the fundamental equations for solving two-stream

compound-compressible flow probems.
When unchoked flow exists, it has been shown that the flow behavior is determined by the back pressure, Fig. 2, and
= p CI it.
0.6

K,,K, = 1.4
R, ,R, = 53.3 FTLB, SLUG-R
'

= 0.5

AT'wiV%"=

0.5

-315

0.4
Al'/ATHR0AT = 0 5

0.4
P02

/Pffll

0.226

0.431
0.4

0.3
0.2

__

0.3

0.2
1
0.1

t
4

i
5

p o.

l
6

l
7

.
8

/Pco

fcl

fa

1 -

1J

A'i -

(")

-1
h

(16)

\p0./

Equations (14) and (16) may be simultaneously solved, by trialand-error, to determine the relationship between P02/P01, "-'2 VTm/
' 1 VT0i, Ai*/Athroat and pthroat/poi at choking for any given gas
properties. The relationsliip between P02/P01, W2 VT\n/wi V'Tqu
and /liV-lehroat is shown in Fig. 5. It can be seen that, for any
given value of w2 VTm/wt VT\u there is a unique P02/P01 corresponding to each .Ai*/ylthroat. These values appear as horizontal
lines in Fig. 4 because the back pressure does not affect the
choked solution.
Referring to Fig. 4 for any given combination of Ai*/yi 8 I u and
A]* / A throat in a single nozzle, the intersection of the corresponding
lines indicates the onset of choking (curve c in Fig. 2). For
values of poi/pa smaller than that at the intersection, the flow
behavior is given by the unchoked curve. For values of poi/p
greater than that at the intersection, the flow behavior is given
by the choked straight line. Therefore, the entire flow behavior of a
compound-compressible nozzle can be described by a single line.
For example, the dashed line in Fig. 4 represents a nozzle with
-41 * / . 1 throat = 0.431 and A S / A ^ n = 0.226. (These are the actual dimensions of the test model described in the experimental
section of this paper.)
The behavior of Mi and M 2 at the throat for choked flow as a
function of w2 VT^/Wi VTm
and fixed -4i*/Athroat and gas
properties is shown in Fig. 6. Note that neither stream is sonic at
the nozzle throat.

Fig. 4 Typical relationship of flow parameters in a convergent-divergent

nozzle

-1.4
K|. K2 = I*

R,. R2 = 53.3 FT-LB/ SLUG - :R

S T R E A M

R,, Rj = 53.3 FT-LB.SLUG- =R

0.431

I N

M A C H

N U M B E R

P02/P01

A T

T H R O A T ,

1.0

Relationship of flow parameters during choked

Journal of Applied Mechanics

0.1

0.2

0.3

'

0.4

0.5

W 2 v % /w,

W2yTo2/Wlv/Toi
Fig. 5

2.5

flow

Fig. 6

Stream Mach numbers at the nozzle throat during choked flow

SEPTEMBER

1967

551

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Additionally, it is important to note here that the solutions

presented previously are independent of the individual flow areas
at the inlet plane. In practice, however, it is possible for these
areas, if not properly designed, to cause large three-dimensional
effects at the inlet plane, due, for example, to independent choking of the primary stream. Nevertheless, it is still either the
nozzle exit area or geometric throat, not the inlet areas, which
will control the flow behavior. Thus, even if these three-dimensional effects cause stagnation pressure losses at the inlet plane,
the theory still gives the correct solution provided that pM and pa
are measured downstream of the loss. In any case, this situation
can be avoided by designing the proper flow areas for the inlet
plane. This can easily be done by solving equations (1) and (5)
simultaneously at the inlet plane for the flow parameters corresponding to the one-dimensional solution.
The procedures outlined in this section are particularly suited
to problems where w2 ^/TM/W, y/T0i is specified. However, the
theory can be applied equally as well to generate solutions for
any suitable set of given conditions. The methods presented
here may also be extended to cases of more than two flows.

Three-Dimensional Two-Stream Nozzle

Computer Calculations
One of the basic assumptions of the one-dimensional theory is
that the effects of streamline curvature are small enough to be
neglected. This can be an important restriction in the application of the theory to actual nozzles. Therefore, to demonstrate
the effect of such curvatures, a three-dimensional two-stream
nozzle solution will be compared with the one-dimensional theory.
To do this, an axially symmetric two-flow nozzle computation
was carried out. A brief description of the computation is given
here to justify its use in evaluating the one-dimensional theory.
The basic idea behind this calculation can best be understood
by recalling that the curves in Fig. 2 were obtained by an integration of equation (3) through the nozzle. Here a numerical
integration includes three-dimensional effects.
Some important assumptions are made about the flow before
this solution is applied; namely,
1 The system must be symmetrical with respect to its center
line.
2 The two flows are isentropic and do not mix but are in contact along a slipline.
3 The static pressure is free to vary along the slipline but
must be equal across it at every point.
4 The primary, or inner, stream must be everywhere supersonic while the secondary, or outer, stream is not restricted.
This allows a method of characteristics to be used for the primary
stream calculations.
For any one problem, the temperature-corrected mass flow ratio,
the ratio of specific heats for each stream, the gas constant for
each stream, the coordinates of the wall shape, and the primary
stream inlet Mach number are considered to be specified.

NOZZLE

S E C O N D A R Y
P

WALL

S T R E A M

02,T02,w2,r2,k2

S U P U N E

MAC'H

"

LINES

P R I M A R Y

Fig. 7

552

S T R E A M

Construction of a typical slipline point in three-dimensional flow

SEPTEMBER

1967

The flow field of the primary stream is constructed using

standard numerical forms of the method of characteristics equations for axially symmetric flow. The secondary stream is calculated using one-dimensional, isentropic flow relations. The flow
areas of the secondary stream are measured along a projection
normal to the slipline, Fig. 7. The assumption that this flow
is one-dimensional and has no streamline curvature effects is not
a serious restriction. It should be noted that the secondary flow
passage occupies only a small portion of the total radius at any
section (relatively small mass flow rates and annular geometry).
Therefore, the streamline curvature effects are small compared
to those of the primary stream, where they have been accounted
for.
The secondary and the primary streams meet along the slipline.
The calculation of a succession of points along this slipline involves
an iteration procedure incorporating the calculations of the
flow properties of each stream simultaneously, Fig. 7. Point C
illustrates a typical slipline point to be calculated. Point B, the
previous slipline point, and point A, the previous point on Mach
lines A-B and A-C, are known. The conditions of point C are
that it be on the Mach line A-C and that the static pressure at C
be the same in the primary and secondary streams. These conditions are satisfied through a trial-and-error solution involving
the location of point C.
As was noted, the program is performing a numerical integration equivalent to the integration of equation (3). For any set of
given conditions, the results can be presented in terms of the
Pai/Pn required to drive the flow. A reduction in p02/poi corresponds to operation at a decreased back pressure p. The term
p will correspondingly decrease until pm/poi reaches a minimum
value. Any value below this minimum will cause the slipline
iteration to converge on a secondary area ratio, A2/A2* < 1,
which is physically meaningless. This minimum value of pm/poi
corresponds to choked flow. All values of pm/Vn greater than
the minimum value correspond to unchoked flow. In short, the
three-dimensional solution closely parallels the one-dimensional
solution, Fig. 4.
The three-dimensional computation for both choked and unchoked flows usually presents little difficulty, and the integration
through the geometric throat is generally smooth. Difficulties
may arise when the accelerating secondary stream approaches
Mach 1. This is due to the usual mathematical sensitivities encountered in computing one-dimensional flow properties near
Mach 1. However, the transition through this region can be
made by special techniques.
This study was undertaken to find the effects of streamline
curvature on the behavior of a compound nozzle. The first calculations were performed with geometry corresponding to the
basic nozzle used for the experimental portion of this paper. The
flow fields in these cases showed definite three-dimensional behavior, especially in the initial expansion region. However, the
calculations showed no important coalescing of Mach lines, so that
the isentropic assumption was essentially valid.
T o further study three-dimensional effects, the nozzle geometry
was modified. A series of nozzle shapes was generated by applying a scale factor to only the axial coordinates of the basic nozzle.
The resultant series then consisted of nozzles which were identical
one-dimensionally but had severe variation in the slopes of the
contour. These geometries are shown in Fig. 8.
The results from this series of computations are presented in
Fig. 9. Curves a and b demonstrate the compound-choked flow
regime.
The term pm/poi is plotted versus Z*throat/-Dthroat, the
length from the inlet to the throat divided by the diameter of the
throat. Notice that for both curves the agreement with onedimensional results is excellent for relatively large Lthrot/-Dthrot
ratios. For the smaller values of Lthroot/Duiroat, the three-dimensional effects influence pce/poi. This is not surprising since in this
range the nozzle contour slopes are most severe. Also, the nozzle
throat is close enough to the inlet to be in the influence in the
initial expansion of the primary stream.

Transactions of the A S M E

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

THE

B A S I C

N O Z Z L E

A N D

T Y P I C A L

EXPERIMENTAL

V A R I A T I O N S

ONE

W 2 A 2

AEXIT

1H

S H O R T E N E D

NOZZLE

NOZZLE

THEORY

W 2 y T 0 2 / W , y T | 1 , = 0.209

P02/P01

E L O N G A T E D

N O Z Z L E

"-EXIT

B A S I C

BASIC

"-THROAT

W , / T 0 , = 0-315

RESULTS

DIMENSIONAL

N O Z Z L E

DIMENSIONS :
A'r' A THROAT

= 0.431

M / *EXIT

- 0226

w 2 yf 0 2 /w l v /r 0 i = 0.052
Pfl2 /Poi

U H R O A T / D T H R O A T = 0.374
L E X I T - " E X I T = 0.695

Kj=K2= 14
Rl=R2=53.3
Fig. 8

N o z z l e s used in t h r e e - d i m e n s i o n a l

10

10

Comparison

of

compound-compressible

flow

theory

with

e x p e r i m e n t a l results

DIMENSIONAL

AXIALLY

6
Poi/Pco

comparison
Fig.

ONE

o-

SYMMETRIC

0.6
W j ^ / W . / f o , =0.052

2 V

/ T

0 !

/ W , Y F O ,

0.315

0.4

P02/P01

C O M P O U N D
C H O K E D

W 2 y T o 2 / w , y T m =0.052

P02

/ W W i v / T o i =0.315

C O M P O U N D
P01

U N C H O K E D

P01/R0

275

Pm/P 4 0
0

0.4

0.8

1.2

LTHKUAT'DTHROAT

Fig. 9

Comparison

of

one-dimensional

0.4

08

1.2

LTHROAT/DJHROAT

and

three-dimensional

results

Curves o and d demonstrate the compound-unchoked flow

regime. Again, agreement with the one-dimensional results is
excellent. The discrepancies for short nozzles are not so pronounced here since the exit conditions, and not the throat conditions, are dominant. Some departure from the one-dimensional
line is seen for the longer nozzles. It is felt that this may be due
to accumulating inaccuracies resulting from the finite difference
techniques used. For long nozzles, these errors may become
significant by the time the exit plane is reached.
The agreement between the one-dimensional and the axially
symmetric solutions is excellent for the range of nozzle variation
examined. Three-dimensional effects have little influence on the
level of P02/P01 for the nozzle geometries considered.

Experimental Results
Extensive test programs conducted 011 a wide variety of nozzle
types and geometries have shown excellent agreement between
the one-dimensional compound-flow theory and experimental results. These test programs have included convergent, cylindrical,
and convergent-divergent nozzles with both two and three
streams.
The success of the one-dimensional theory is not surprising.
The previous section indicated that the behavior of compound
flow nozzles is reasonably insensitive to three-dimensional effects.
For nozzles with small wall frictional effects, and fairly undistorted inlet flow, only the effects of mixing can cause the onedimensional model to be inaccurate when applied to real nozzles.
It is reasonable to assume that, since the flow is turbulent, mixing
is confined to a shear layer between adjacent streams which grows
with axial position at an angle of less than 8 deg [3]. It is therefore clear that the shapes of the
and the nozzle length are the
major factors which determine the degree of mixing. The tendency of the mixing will be to pump the low-velocitj' streams and

Journal of Applied Mechanics

to retard the high-velocity streams. The extent of the influence of

mixing depends primarily on the flow rates of each stream. The
behavior of streams with proportionately low rates will be greatly
affected by mixing while those with higher flow rates will experience only small effects. Note, however, that in the important
case of choked Uow, mixing effects downstream of the nozzle throat
can exert no influence whatsoever upon the behavior of the flow.
Thus, for many nozzle applications, mixing will influence only a
small portion of the flow.
Because of the space limitations of this paper, a fully comprehensive comparison of one-dimensional theory with experimental
data from a wide variety of compound-flow nozzle types and
geometries is impractical. According^, the basic nozzle previously discussed was selected for comparison. Being convergentdivergent, it is a representative two-stream nozzle in that it can
display both the choked and unchoked regimes of compound
flow.
Tests were run over a wide range of temperature-corrected
mass flow ratios while varying pm/p^ from approximately 2 to 10.
This allowed the nozzle to exhibit both choked and unchoked
behavior at each mass flow ratio. Fig. 10 is a comparison of experimental results with predictions based upon the one-dimensional compound-flow theory. The choked flow regime is the
straight portion of the theoretical lines and, as expected, occurs
at the higher poi/paThe
unchoked flow regime is the curved
portion of the lines which occurs at the lower primary nozzle
stagnation pressure ratios.
It can be seen that the one-dimensional theory shows excellent
agreement with the experimental data, particularly in the choked
flow regime. Correlation with experimental data in the unchoked flow regime is somewhat less accurate. However, in view
of the mixing effects previously discussed, this was to be expected.
During choked flow, all mixing downstream of the nozzle throat
can have no effect 011 the flow. Thus the effective mixing length
for choked flow is merely the distance from the inlet plane to the
throat. On the other hand, during unchoked flow, all mixing
downstream of the nozzle throat will have a very definite effect
on the flow behavior. With unchoked flow, then, the effective
mixing length is the entire length of the nozzle.
As anticipated, we also observe that mixing has little effect at
the higher mass flow ratios but exerts increasing influence as
mass flow ratio decreases.
Furthermore, the important effect
of mixing is to pump the secondary flow and therefore reduce the
required pm/poiIt is noted that the model tested was not an ideally designed
nozzle; i.e., the primary stream was independently choked and
slightly underexpanded at the inlet plane. However, the model
used in the experiments was operated sufficiently near to its
isentropic design conditions that stagnation pressure losses in the
supersonic stream were probably not important, and shock losses
SEPTEMBER

1 9 6 7

/ 5 5 3

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

are impossible in the subsonic secondary stream. This was indicated in the previous section, where the three-dimensional solution showed no significant coalescing of the primary stream Mach
lines under conditions for which the tests were run. Although
Poi/Pos is measured upstream of the inlet plane, these arguments justify our neglect of total pressure losses in all calculations.

Concluding Remarks
A new one-dimensional theory describing the behavior of compound-compressible nozzle flows has been developed and its
implications have been examined from a number of viewpoints.
The theory yields simple algebraic methods for calculating the
operation of compound-compressible nozzles. Comparison of the
algebiaic results with those of three-dimensional flow field computations indicates that the effects of streamline curvature are
not important for many practical nozzle configurations. Comparison of the algebraic results with experimental data for flows
with unimportant mixing effects shows that the theory can

554

SEPTEMBER

1 9 6 7

accurately predict the behavior of real devices.

A complete definition of the range of applicability of the simple
theory requires a great deal of experimental experience. For example, little information is available about the rate of growth of
the mixing zone between the high-velocity streams, and no information is available about the detailed nature of compoundshock waves. Furthermore, no predictions about the thrust or
nozzle efficiency can be made when the flow is choked until the
losses caused by a compound-shock wave are known. Consequently, it appears that the compound-compressible nozzle provides a number of interesting and important areas of research.

References
1 Pearson, H., HoUiday, J. B and Smith, S. F., "A Theory of
the Cylindrical Ejector Supersonic Propelling Nozzle," Journal of the
Royal Aeronautical Society, London, Section 62, 1958, p. 746.
2 Shapiro, A. H., The Dynamics and Thermodynamics of Compressible Fluid Flow, Ronald Press Co., New York, 1953.
3 Sclilichting, H., Boundary Layer Theory, 4th ed., McGraw-Hill,
New York, 1960, pp. 598-600.

Transactions of the A S M E

Downloaded 05 Feb 2013 to 171.64.160.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm