NATIONALADVISORY COMMTITEE

FOR AERONAUTICS
—.

I TECHNICAL NOTE 2604

A GENERAL THEORY OF THREE-D= NSIONAL FLOW IN

I

1
SUBSONIC AND SUPERSONIC TURBOMACHINES OF
(
AXIAL- ,RADIAL-, AND MIXED-FLOW TYPES

By Chung-Hua Wu
1

I Lewis Flight Propulsion Laboratory

Cleveland, Ohio

Washington .
January 1952

Awlml
WXM?W”.Li.miwmv .
JUFLam.

----- . . .. . .. .. ..
 TECH LIBRARYKAFB,NM
Iv
NACA TN 2604 lllMM!llmmluM
0CIL571JU

#
TABLE OF CONTENTS

. Page

SUMMARY . . . . . . . . . . . . . . . . . . . . . . . ...0. ..1

INTRODUCTION
. . . . . . . . . . . . . . . . . . . . . . . . . . .1

SYM801S. ● . . . . . . . . . . . . . . . . . . . . . . . . . . ● 4

BASIC AEROTEERMODYNAMIC REL&ITONS. . . . . . . . . . . . . . . . 8

POl?ENTIiKEQUA!l?ION
FOR THREE-DIMENSIONAL FLOW T8ROUGHROTATING
BLADE ROW.. . . . . . . . . . . . . . . .“. . . . . . . ● . .
14

FOILOWINGFLUIDFIOW ONREIATDE sTREAM mAcEs . . . ;. . . . 16

s~ StreamSurface ofFirstHnd . . . . . . . . . . . . . . . 16
s~ StreamSurface ofSecondKind . . . . . . . . . . . . . . . 17
Relations among Relative Velocity of Fluid, Coordinates of
Stresm Surface, and Normal to Stream Surface . . . . . . . . 18

. EQUATIONS GOVIIRNINGFHJIDl?UJW ON S1 SURFACE FOR AXIAL-FLOW AND

AXIAL-DISCHARGE MIXED-FIQIJTURBOMMxINES . . . . . . . . . . . 19
Flowalong General SI Surface . . . . . . . . . . . . . . . . 19
Equations of continuity and motion . . . . . . . . . . . . . 20
Prticipalequation . . . . . . . . . . . . . . . . . . . . . 22
Procedure of”solution. . . . . . . . . . . . . . . . . . . . 26
Flow along Surface of Revolution . . . . . . . . . . . . . . . 28
Flowalong Cylindrical Surface . . . . . . . . . . . . . . . . 30

EQUATIONS GOVERNING FLUID FI13’W

ON S1 SURFACE FOR RADIAL-FKXWtD
RADIKG-DISCHARGE MIXED-FLOW TURBOMACHINES . . . . . . . . . . . 31
FlowalongGeneral S1 Surface........;. . . . . . . 31
Equations of continuity and motion . . . . . . . . . . . . . 32
Principal equation . . . . . . . . . . . . . . . . . . . . . 32
FloITalong a Surface of Revolution . . . . . . . . . . . . . . 34
FlowonRadialPlane . . . . . . . . . . . . . . . . . . . . . 36
Alternative Form of Equations for Flow along Surface of
Revolution . . . . . . . . . . . . . . . . . . . . . . . . . 37

EQUATIONS GOVERNINGFIJJID FLOW ON S2 SURFACES . . . . . . . . . 38

Equations of Continuity and Motion . . . . . . . . . . . . . . 38
Principal Equation for Case with Vur Given . . . . . . . . . 41
Principal Equation for Case wi~h W~z Given . . . . . . . . 44
Procedure of Solution. . . ~ . . . . . . . . . . . . . . . . . 46
Directprobla . . . . . . . . . . . . . . . . . . . . . . . 46
. Inverse problem . . . . . . . . . . . . . . . . . . . . . . 47

..
 NACA TN 2604

Approximations Involved in Through-Flow Theory . . . . . . . . . 48

CIR~IAL VARIATION OF FLUID PROPERTIES BY USE OF P3W3R

SERIES . . . . . . . . . . . ...00 . ..*.. . . . . . . . 49

STEFS FOR COMPLETE SOLUTIONS OF TEREE-DIMENSIONAL DIRECT AND

INVERSE HKJBIEM8 . . . . . . .. *O. ●.0... . . . . . . . 53
Direct Problem . . . . . . . . . . . . . . . . . . . . . . . . . 53
Axial tmbomcbines with nontapered straight walls . . . . . . . 53
Axial turbomachines with tapered or curved walls . . . . . . . 54
Radial- and mixed-flow type turbomachines with curved walls . 55
Inverse Problem . . . . . . . . . . . . . . . . . . . . . . ..55
Conditions prescribed on mean stream surface . . . . . . . . . 55
Boundary conditions for mean stream surface . . . . . . . . . 56
Determination of blade shape . . . . . . . . . . . . . . ...56

GENERAL METHODS OF SOLVTNG PRINCIPAL EQUATION . . . . . . . . . . 58

Elliptic Case . . . . . . . . . . . . . . . . . . . . . . . . . 59
Finite-difference form of principal equation . . . . . . . . . 60
Boundary conditions . . . . . . . . . . . . . . . . . . . . . 62
Solution of finite-difference equations .. . . . . . . . . . 64
Gen=al table for evaluation of density from ~-derivatives : 65
Hyperbo13.cCase . . . . . . . . . . . . . . . . . . . . . . . . 67 .
Changes of $-derivatives along characteristic curve . . . . . 68
Changes bf fltid velocity and direction along characteristic
curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

CONCLUDING REMARKS . . . . . . . . . . . . . . . . . . . . . ...73

lmFEFumcFs. . . . . . . . . . . . . . . . . . . . ● ● . . . . ● . 73

,
.,t

-. .
 w,
. NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS

TECHNICAL NOTE 2604

.
A GENERAIJTEEORY OF T9REX-DIMENSIONAL FLOW IN

SUBSONIC AND SUPERSONIC TURBOMACHINES OF

AXIAL-, RADIAL-, AND MIXED-mow TYTEs

By Chung-Hua Wu

SUMMARY

A general theory of steady three-dimensional flow of a nonviscous

fluld in subsonic and supersonic turbomachines having arbitrary hti and
casing shapes and a finite mmiber of blades is presented. The solution
of the three-dimensional direct and inverse problem is obtained by
investigat@g a appropriate combination of flows on relative stream
surfaces whose intersections with a z:plane either upstream of or some-
where insMe the blade row form a circular arc or a radial line. The
equations obtained to describe the fluid flow on these stieam surfaces
show clearly the several appro~tions involved in ordinary two-
dimensional treatments. They a~o lead to a solution of the three-
dimensional problem in a mathematically two-dimensional manner through -
iteration. The equation of continuity is cotiined with the equation of
motion in either the tangential or the radial direction through the use
of a stream function defined on the surface, sad the resulting equation
is chosen as the principal equation for such flows. The character of
this equation depends on the relative magnitude of the local velocity of
sound and a certain cotiination of velocity components of the fltid. A
general method to solve this equation by both hand and high-speed
digital mchine computations when the equation is elliptic or hyperbolic
is described. The theory is applicable to both irrotational and rota-
tional absolute flow at the in?& of the blade row and at both design
and off-design opmations.

INTRODUCTION

The problem of three-dimensional flow in turbomachines of tial-,

radial-, and mixed-flow types is treated in references 1 to 19. Because
of the enormous mathematical difficulties involved in the problem, Lorenz
(reference 1) ftist intr~uced the idea of an infinite number of blades
of infinitesimal thickness in order to follow the flow on a given surface.

.
2 NACA TN 2604 .

.
Bauersfeld (reference 2) added to the theory the condition of integra-
bility for the blade surface that must be satisfied in the inverse, or
design, problem. The theory is further clarified and strengthenedby - .
the works of Stodol.a(reference 3), vonhlises (reference 4), and Dre@s
(reference 5), and is the basis of many recent investigations on axial-,
radial-, and tied-fluw compressors and turbines.

For incompressible flow, Rud.en(reference 6) proves that the

through-flow solution obtained under the assumption of an infinite num-
ber of blades gives a circumferentiall.yav~age value of the fluid prop-
erties, provitid the deviations of the fluid properties from their cir-
cumferential averages are small. In ref=ence 7, ‘l!raupel points out the
osci~tory nature of radial flow in a multistage turbomachine ati gives
solutions of the three-dimensional potential flow through inclined
stationary blades and also of the rotational flow through a homogeneous
stage of identical nontwisted blades for an incompressible fluid and an
infinite nmiber of blades boundedby cylindircalwalls. Meyer gives a
detailed treatment of three-dimensional potential flow in a stationary
blade row, for an incompressible fluid and cylindrical bounding wall.,
in reference 8, where the solution for an infinite number of blades is
extended to a finite nmiber of blades by the vortex+amd-source method of
Ackeret, which is originally given for two-tiensional flow (reference 9). .
In reference 10, a 13-nearizedsolution for an incompressible fluid and an
infinite ntier of blades for a prescribed loading and cylindrical
bounding wa~ is obtained by Marble, and is used later to investigate
the problem of mutual interference of adjacent blati rows and off-
design operations (reference H) . Siestrunck and F~bri (reference 12)
also obtained a linearized solution for incompressible flow, and the
method is extended to compressible flow. For genwal wall shapes,
Spannhake (reference 13) examines the flow through diffuser and impeller
by the use of bound vortices for blades. The incompressible through
flow in a mixed-flow impeller is treated by Gra’volos(reference 14).
In reference I-5,Wisllcenus examines the influence on the meridional
flow of the blade force and nonuniform circulation along the blade span.

For compressible flow, Reissner (reference 16) gives a blade-design

method in which the extension from an ~inite n~er of blades to a
finite nunib& of blades is accomplished by the use of a power series in
the circumferential direction, and the terms in the series sre deter-
mined by a comparison of the equations for an infinite number of blades
and a finite number of blades. (In ref=ence 5, Dreyfus gives a method
of designing water turbines of thin blades, in which the solution for
an infinite nuniberof blades is extended to a finite nuniberof blades by
the use of a power series, the second term of which is determined from
the equations of continuity and irrotational absolute flow and is
explicitly given.) In reference 17 the compressible flow problems in .
axial turbomachines having an infinite nuniberof blades are treated, and
both the direct and inverse problems are considered. Methods for
.

——. .
 NACA TN 2604 3

limiting solutions for zero and ~inite blade-row aspect ratios and a
step-by-step methd of solution, as well as a simpler method based on an
. approxte knowledge of the shape of the streamline, for a finite blade-
row aspect ratio are given. Unaware of the,work of Traupel at the time,
the authors of reference 17 also emphasized the oscillatory nature of
radial flow in multistage machines EUM3suggested the use of a simple
sinusoidal form of the streamlhe as a first approximate solution. Their
methods are derived for compressible flow, however, and are also extended
to the case where both the hub and casing walls or either is tapered.
Reference 18 gives a general through-fluw theory for both direct and
inverse problems and for subsonic or supersonic flow in turbomachines
having arbitrary hti and casing shapes. The supersonic through flow in
rotating impellers having a prescribed flow along the cqsing and pre-
scribed blade shapes is treated in reference 19.

A general theory of three-dimensional flow in subsonic and super-

sonic twbomachines of axial-, radial-, and mixed-flow types for a
finite number of thick bl.adesof finite thickness has been developed at
the NACA Lewis laboratory and is presented herein. Both the direct and
inverse problems .me considered. The theory is applicable to either
irrotational or rotational absolute flow at the inlet of a blade row
and to both design and off-design operations.

In the section BASIC AEROTBEEWODYNMCC REIA!IIONS,the motion and

energy equations for the unsteady flow of a non~scous compressible fluid
in a rotating blade row are expressed in terms of the velocity components
and of two basic thermodynamic properties of the fluid, namely, entropy
and a modified total enthalpy for flow in rotating blade rows with change
in radial CM.stancefrom the machine axis. Estimated entropy changes due
to shockwaves (in the case of supersonic flow), heat transfer (in the
case of a cooled turbine), or viscous effect can be easily accommodated
in the calculation. The equations obtained show clearly the condition
under which the flow through blade rows can be treated on the basis of
irrotational absolute flow.

In the following section, a general potential equation is obtained

for steady three-dimensional compressible flow through rotating or
stationary blade rows when the absolute flow caa be taken as irrota- “
tional. The methods of solution for both subsonic and supersonic flows
sre briefly discussed.

A shnpler method of solvtng the three-dimensional irrotational

(absolute) flow, which is also applicable to rotational (absolute) flow,
is obtained by considering fluid flows on a number of relative stream .
surfaces whose intersection with a z-plane either upstresm of or some-
where in the blade row form a circular uc or a ratial line. Equations
governing the flow on these surfaces are obtained in the next four sec-
tions. Through the use of a stieam function defined on the stream

-.. .

— —__... —-— .- .. ..—.— —. ..+- ——— —-

4 NACA TN 2604

surface, the equations of continuity and nmtion for fluid flow on these
surfaces are conibinedinto one principal equation. The character of the
principal equation is dependent on the relative magnitude of the local
velocity of sound and a certain conibinationof velocity components.

The process involved in solving the direct and inverse problems by

this approach is d.escribd in the section SITE’SFOR COMN2EU!ESOLUTIONS
OF THREE-DIMENSIONAL DllllWTAND INVERSE PROBLEMS. In the inverse prob-
lem, besides the blade-thickness distribution determined by blade
strength and other considerations, either the tangential velocity, a
relation between the tangential and axial velocity, or one other rela-
tion is prescribed on a mean stream smface about midway between two
blades. The last section gives a general method of solution of the
principal equation when it is elliptic or hyperbolic.

SYMBOLS

The following synibolsare used in this report:

a velocity of sound
.
B ,b integrating factor for continuity equation for S2 and S1
surfaces, respectively
.
differentiation coefficientused to multiply function value
at point j to give the mth derivative at point i based
on nfi degee polynomial

C,c nonzero tm on right-hand side of continuity equation for

S2 and S1 surfaces, respectively

specific heat of @s at constant pressure and volume,

respectively

differentiation with respect to time following relative

motion of fluid particle

Pq mth derivative of ~

F,f vectors having the unit of force per unit mass of fluid

G,g given function of WJWz on S2

H total enthalpy per unit I@SS Of fluid, h+b~

.
h static enthalpy per unit mass of fluid, u + p/p

.. .
N!WA ‘IN2604 5

I modified total enthalpy for flow h rotating blade row with

change in radial distance from machine axis,
122
h+&-Zor or H - o(Vur)

J,K,L,M,I17 coefficients of first- and second-order derivatives in the

principal equation

k thermal conductivity

L distance along streamlbe

1>9 orthogonal coordinates on surface of revolution

M mass flow betiieenmean stream surface and one surface of

blade

N n@er of blades

n unit vector normal to relative stream surface S

P static presswe

Q heat added to fluid particle along its path per unit As

per tit time

(1 w q~tity on re~tive stre~ s-ace S

R gas constant
FL remainder term of mth derivative at point i obtained by
using nth degree polynomial

r radius vectar

s~ relative stream surface passing through fluid particles

X on a C&C@ arc upstream of or midway in blade
row

S2 relative stream surface passing through fluid particles

- on a =~1 or c~ed he ~stre~ of or midtiy in
blade row

s entropy per unit mass

s* s/R

T static temperature

— _.— .— ---- .—.——

 NACA TN 2604

time

velocity vector of blade element at radius r

interval energy per unit mass

absolute velocity of fluid

velocity of fluid relative to blade, v-u

independent variables

distance along turbomachine axl.s

1 Wu
arctan ——
r Wz

ratio of s~ectiic heats

average value of T for the temperature range involved

.
grid spacing

equal to landr for S1 and S2 surfaces, respectiwly

independent variable z or r for S1 smface and z for

S2 surface

independent variable qandrfcm S1and S2 surfaces,

respectively

angular distance of fluid particle measured with respect to

s~tionary radial Mne

slope of characteristic curves, V

2

tanu

equal to r anal forslandsz surfaces, respectively

absolute vorticity, VX V
.
fluid density

.
w
NACA TN 2604 7

generalized variable used for general density table

angle between tangent of streamline or boundary wall in the

meridional plane and axial direction

radial, axial, or angular thickness of stream sheet

velocity potential

generalized variable used for general density table

angular distance of fluid particle measured with respect to

radial line on rotating blade

angle between w and axial.direction

stream functions defined on relative stream surfaces S2 and

respectively

angular velocity of blade

subscripts:

c casi.ng

e exit

h hub

i inlet

2 meridional component

m mean stream surface

o lower limit of integration

r,u,z radial, circumferential, and axial components

s isentrbpic

T total state

O)c components in ~- and ~-direction, respectively

1 on S1, or in front of rotor

2 on S2, or behind rotor

——. .—.— --- —. ——- — -

 —

8 NACA TN 2604

superscripts:

a,b~. ..k refa to points a,b, . . . k, respectively

BASIC AEROTEERMODYNAMIC REIATIONSl-

The three-&huensional flow of a nonviscous, compressible fluid

through a turbomach5ne is governed by the folluwing set of basic laws
of aerothermodynsmics* From the principle of conservation of mattm,
the equation of continuity is

ap
~+v”(pw) = o (1)

or

VOW.+*. O (la)

For a blade rotating at a constant ~ velocity a) about the

z-axis, Newton’s second law of motion gives

DW
~-02r +2@XW= - $VP (2)

Because the boundary wal& are surfaces of revolution and the relative
flow can be appro~ted as being steady in mmy cases, it is convenient
to we a relative cylindrical coGd.inate systan r, ~, and z with
q measurd with respect to the rotating blade (see fig. 1). By use of

#’#+(w”v)w”#+ *W2 - wx&Xw)

the scalar forms of the equation of motion (2) in the axial, circum-
ferential, and radial directions can be expressed as

air *r Wu awr awr WU2

~+wr F+-- W+wz ~-7-02r-~lu= - ~
~xa (2a)
.
awu awu WU Mu Mu Wrwu la
~+Wr~+~~+W~ ~+~+Z21Wr=-_&~q (2b)

(2C)
.

l-someof the rel&ions given in this section have been given in refer- .
ence 18. They me repeated here for completeness ~d easY reference
for the following developments.
 NACA TN 2604 9
. .

The first law of thermodynamics may be written

Du
Q (3)

where u is related to the temperature T by

(4)

and Q is given by the following equation if only conduction is

considered:

Q = # V.(kVT) (5)

For the ranges of tempmature and pressure encountered in ordinary

turbomachines, p, p, and T of ;he gas are accurately related ~y the
following equation of state:

p=RpT (6)

Although the flow of the gas through the turbomachine is completely

defined by the preceding equations together with the lmown variations of
~ and k with temperature and the given boundary and initial condi-
tions, it is found more convenient in references 17 and 18 to e~ress
the state of the gas in”terms of the entropy and the total enthalpy or -
a quantity I of the gas, besides its velocity components. These
quantities are defined as follows:

T ds = dun+ p d(p-l) (7)

12
H=h+ZV (8)

I=h+~w2- ~U2=H--u(Vur) (9)

and

h=u+pp-l (lo)

From equations (10), (4), and (6) is obtained

dh= (c@)d Y=~d T= (n)

where T is equal to #cv and is a function of tempemture. Another

, expression for dh is o&t.ainedby using equations (10) and (7), so that
dp
w= ~+Tds (ha)
.

——
. . . . . -.— --——— —— ——— — _.—. —— ———
 . —.—— .—. —.

“lo NACA TN 2604

By the use of equations (7), (4), and (6),

1
—dlnp —— T dlnp (u)
0 : = T-1 y-l

and
s
%)= T-1
~dln T-dlnp (12a)

can be obtatied, and the equation of continuity can be written

v“w++++p-&; =o (us)

Equation (I-3]can be expressed in a slightly different form. From the

definition of the local velocity of sound (reference 20),

(14)

By the use of equations (12) and (6),

&F=r; =@T (M)

Substituting this relation into equations (12a) and (13), with the use of
equation (n), results in

(12b)

V.w+$g Ds
-mR=o
(1.3a)

Frm equations (9) and (ha),

$i7p + +2 -m2r =VI -TVs

The equation of motion (2) can then be written

WX$7XW) -t-
2(LIXW= -vI i-TVS (15)
%-

An alternative form of equation (15), which imolves the vorticity of

the absolute motion, is obtained as fol.lows: with the z-axis para~el
to al,

.——_— —
 NACA TN 2604 SL

V= W-1-w Xr (16)

hence

vxv=vxw.tvx(uxr) (17)

But

vx(~x.r) = (r.V) w -(U.V) r -1- r) - r(v.aj)~ 2aI

f,u(v.

therefore

VXV=VXW-I-2LJ (17a)

This relation can also be seen from the following qressions of rela-
tive and absolute vorticity expressed in terms of the rotating and
stationary cylindrical coordinates r, cp, z and r, e, z, respectively:

)
(18)

1 awr
-FaQ7
avu
-&-

a~. avz
(vXv)u=
+--aF- (19)

and the relation

Using equation (17a) results in the alternative form of equation (15)

-Wx(vx v) = -. + TVs (1.5a)

%!

. ..— —.—. -.— —— —- .-—

I-2 NACA TN 2604

By use of equations (2), (9), and (l-la),

It may be emphasized that the preceding equation is a consequence

of the equation of motion (2) and the thermodynamic relations (4), (6),
(7), and (lo). For steady relative flow, the rate of change of I
along the streamline is seen to be proportional to the rate of change
of entropy along the streamline.

The energy equation (3) can be used to express the rate of change
of entropy along the streanikbe by the use of equation (7) as follows:

(21)

The preceding equations lead to several important general consider-

ations: If the blade rows are not placed too close together and no
trai13mg vortices are shed from preceding blade’rows (or where these
effects can be neglected), the fluid properties at a fixed point rela-
tive to the blade can be taken as constant with respect to time. Con-
sequently, according to equations (20) and (21), the quantities s and
I of the gas remain co~~t along the Streml i~e for adiabatic flow.
The invariance of I means that the rate of change in total enthalpy
along the streamline is equal to the anguhr speed of the blade multi-
plied by the rate of change in angular momentum (about the machine axis)
of the fluid particle along its streamline, which is the well-known
Eulm turbtie equation usually derived under less general conditions.
In a cooled turbine where the heat transf= may be large, the rate of
change of s and I along the streamline can be corrected by equa-
tion (21) for an estimated value of Q. Againj fcm steady relative flow,
equation (15a) shows that either when gradient I and.gradiLent s both
vanish or when the difference between VI and TV s vanishes, the
absolute vorticity either vanishes or is parallel to the relative
ve10city.

For the flow through a stationary blade row u = O, W beccmes V ,

I becomes H, and equation (13a) becomes

-V X(VXV)=-VH+ TVS (15b)

%
.
which agrees tith similm relations previously obtained by Vazsonyi
(r@erence 21) and Hicks, Guenther, and Wasserman (reference 22). It
.

—-. .
NACA TN 2604 33

is interesting to see that, for relative flow in a rotating blade row,

V%(V X V) becomes WX(V XV) and H becomes I.

If it is assumed that the fluid enters the inlet guide vanes of a

turbomacldne with uniform H and s and zero vorticity and that the
flow is adiabatic, s does not vary in the inlet guide vanes and p
is then a function of only p, according to equation (12). Conse-
quently, by virtue of Kelvin’s circulation theorem, the absolute vor-
ticity will remain zero in passing through the Met guide vanes and the
flow in the inlet guide vanes can be treated on the basis of _ota-
tional absolute flow.

If the guide vapes impart a radial variation of tangential velocity

of the fluid in a z-plane downstream of the vanes similAr to that in a
potential vortex, that is, inversely proportional to the radius, the
circulation is constant along the blade span and the fluid matitai.nsa
umiform s and H and a zero vorticity of absolute flow entering the
following rotor-bkie row. If the rotor-blade row is situated f= away
from the inlet guide vanes, the fhid enters the rotor with a urdform
I in the circumfaential tiection, as well as in the radial direction,
and the flow through the rotor blades can again be treated on the basis
of zero absolute vorticity and steady relative flow. If the rotor is
close to the guide vanes, however, vortices are shed from the inlet
guide vanes because of periotic variation in circulation caused by
unsteady flow, and the flow downstream of the stator and through the
rotor blades should theoretically be treated on the basis of rotational
flow.

If the guide vanes impart a radial variation of tangential velocity

of the fluid at a z-plane downstream of the vines not inversely propor-
tional to the radius, the circulation varies along the span of the guide
vanes, vortices are shed ltromthe trailing edge to the fluids downstream
in the direction of the exit velocity, and the fluid enters the follow-
ing rotor blades with a uniform s and H but a nonuniform I and a
nonzero value of absolute vorticity. Consequently, the flow through the
rotor-blade row can no longer be treated on the basis of zero absolute
vorticity, even if it is far apart from the preceding guide vanes.

From the preceding discussion, the choice of s and H or I

as the two basic thermodynamic variables of the gas besides its veloc-
ity components is apparent. Compressor and turbine rotors are usually
designed to impart or wibtract the same amount of energy to or from the
gas radially; hence H is usually ra&M_ly constant throughout the
machine if the inlet flow is uniform (except in the boundary layer along
hub and casing walls). If the circumferential velocity of the gas
upstream of the blade row is zero or varies inversely with radius, I
is then constant throughout the machine. These facts till be utilized
in the following developments.

... — .—— —.— .—_. ..— .—-— —- — ———--—--———

14 N4CA TN 2604

POTENTIAL EQUATION FOR THREE-DIMENSIONALFLOW THROUGH .

ROTATING BLADE ROW

.

Consider first the special case of steady relative flow where the
fluid upstream of the blade row is free of vorticity and is uniform in
I and S. The adiabatic flow through the blade row is then relatively
steady and absolutely irrotational and is most conveniently treated by
the use of a velocity potential @ based on the zero absolute vorticity
and rehted to the relative velocity components through equation (16) as
fOllows:

ao
s =Vr=wr

(22)

For steady isentropic flow, the continuity equation (13a) becomes

.

From equations (9) and (22),

‘=l+u@ a“-’[r’7+
2 aF (=?7’ (w..
(24)

(25a)

(25b)

(25c)

By the use of equations (22) and (25a) to (25c), the continuity equa-
tion (23) may be mitten .

.
.

——. .——
3V
NACA TN 2604 15 -

(26)

Equation (26) is then the three-dimensional potential equation for

isentropic flow in a rotating blade row. It is seen from this equation
and equation (16) that the three-dimensional flow through a rotatzlng
blade row cannot be treated by a three-dimensional flow through a sim-
ilar stationary blade row with the ssme inlet condition relative to the
blade row, as in the case of two-dimensional flow on a cylindrical sur-
face, because the difference between the absolute and relative vorticity
2@ does not enter into the two-dhensional flow on a cylindrical surface
but does enter into the three-dimensional case.

Equation (26) is very s~lar to the or~ three-dhensional

potential equation for flow past stationary objects, except that both
relative and absolute velocity components are involved in the coeffi-
cients of @ derivatives and that @ is directly defined by the abso-
lute velocity. The real difficulty h solving this equation lies in
the fact that all the velocity components change @?eatly in passing’
through a turbomachine and, consequently, the equation cannot be 13n-
earized and yet give a good approximate answer. For supersonic relative
velocity, the method of characteristic swfaces (refmences 23 and 24)
may be used to solve equation (26), with the initial conditions not
given on a characteristic surface. For subsonic relative flow, the
equation is more conveniently written in the form

(26a)

and can be solved by Southwell*s relaxation method (reference 25) or

other numerical methods using the differentiation formulas obtained in
reference 26 to take care of the unequal grid spacings near the blade
surfaces and the curved hti and casing walls. The last three terms in
equation (26) are computed from the & values or velocities obtained
in the previous cycle and kept as constants during each improvement of
Q values, and the whole process is repeated until the desired accuracy
is obtained. Because a three-dimensional stream function cannot be
defined, the use of velocity potential results in a boundary-value prob-
lem of the second hd, which is more difficult to handle in the calcu-
lation than the first kind. The boundary condition to be satisfied is
that the relative velocity nmmal to the moving blade is zero, or

-.. + ..—-- .—— -—— .— — —. .- —.———

16 NACA TN 2604

?
(27)

where n is the unit normal vector at the boundary surface, and that,
at inlet and exit stations far away from the blade, the velocity is
parallel to the bounding hub and casing walls, which, in the case of
the -al machine, means that

Vr = =Oat z=+m (27a)

%

In both the subsonic and supersonic cases, the solution is

extremely the-consuming. Furthermore, this direct approach to the
three-dimensional problem requires that the absolute velocity at the
inlet to the blade row be irrotational and the flow be adiabatic. In
actual machines, the flow entering the blade rows is always rota-
tional, which is caused by a nonuniform total enthalpy and entropy at
the inlet of the machine, by entropy change caused by shock waves or
heat transfer, or by the effect of boundary layers along the hti and
casing walls. Some other approach to the problem, which is simpler to
handle and is also applicable to rotational inlet flow, is therefore
desirable. One approach is suggested in the folluwing sections.
.

FOLLOWING FLUID FLOW ON REIATDE STREAM SUFU?AC13S

In order to solve the steady three-dimensional flow, with either an

irrotational or rotational absolute flow at the inlet, in a relatively
simple manner, an approach is taken to obtain the three-dimensional
solution by an appropriate cotibation of mathematically two-dimensional
flows on essentially two diff=ent kinds of relative stream surface
(figs. lto 3). The ftist kind of relative streem surface is one ~~hose
intersection with a z-plane either upstream of the blade row or midway
in the blade row forms a circular arc (fig. 1). The second kind of
relative stream surface is one whose titersection with a z-plane either
upstream of the blade row or somewh~e inside the blade row forms a
radial line (fig. 2). These two ldnds of re-tive stream surface will
be hereinafter designated stream surfaces S1 and S2, respectively.

s~ Stream Surface of First Kind

In figure 1 is shown a stream swface of the first kind formed by

fluid particles lying on a circular src ab of radius oa upstream
of the blade row. It is usually assumed in ordinary two-dimensional
treatments (for example, references 27 to 30) that the stream surface
thus formed is a surface of revolution. in the following development, .
the surface will be allowed to take whatever shape it should have in
order to satisfy all the equations governing the three-~ensional flow.
.

—
NACA TN 2604 17

E most cases, the deviation Of the surface from a surface of revolution

is not large, and it is satisfactory to consider s~ smfaces formed by
fluid particles originally l@ng on a circular arc upstream of the blade
row. If the rotationality of the inlet absolute flow is large, if the
blade is designed for a velocity diagram quite different from the free-
vortex type, or if the bbde length is long in the direction of the
through flow (radial- and mixed-flow machines), the Mst of the surface -
may be quite large, resulting h v- large circumferential derivatives.
If this effect is found during calculation or lnmwn from experience, it
is more satisfactory to consider S1 surfaces formed by fluid particles
originally lying, in front of the blade row, on curves inclined to the
circular arc in a direction opposite to the twist of the surface. In
this way, the intersection of the S1 surface with a constant z-plane
about midway in the flow path is a nearly circular arc, and the total
twist of the surface willbe about equally distributed toward the
upstream and downstream directions (fig. 3). If this distribution of
the twist of the stream surface is still not enough, it maybe necessary
to divide the complete flow path tito a few shorter paths and consider
an S1 surface for each of them. Under these conditions, S1 surfaces
formedby fluid particles originally lying on the hti or casing walls
upstream of the blade row should not be chosen in order that the compli-
cation arising from the possibility of fluid particles leaving the wall
and flowing along the blade surface maybe avoided. In such cases it
is better to consider the S1 surface a short distance from the hti’and
casi~j otherwise, for an appro-te solution the fluid can be con-
sidered to follow the hub and casing walls, which are surfaces of rev-
olution, and the calculation is thus much shpler than that for a
general surface.

S2 Stream Surface of Second Kind

A stream surface of the second kind is shown in figure 2. The

most important surface of this family is the one about midway between
two blades dividing the mass flow in the channel into two approximately
equal parts. This surface is designated the mem stream surface
(s2,m) For blades with radial elements, such as the one shown in fig-
●

ure-2) it is convenient to consider a mean streaulsurface formed by

fluid particles originally lying on a radial line ab upstream of the
blade row if the twist of the surface is not expected to be large.
Otherwise, the radial line is chosen about midway in the passage with
the fluid particles originally starklng out frcuna curved Une upstream
OE the blade row such as shown in figure 3.

The mean stream surfaces for tial-flow gas turbines designed on a

free-vortex velocity Uagrm are shown in figures 3 and 4. The radial
element of the mean stream surface (fig. 4) is.chosen accordingly as the

—— z — ——-
18 NACA TN 2604

stator iy designed to aline the blade sections radially at the leafing

edge, trailing edge, or somewh=e between. Inasmuch as the rotor-blade
sections are.usuaQy alined radially at or nesr the center of gravity
of the blade sectiom’, the r-al position of the mean relative stream
surface is chosen at the same position (figs. 3 and 4) . The continm-
tion of’the stream swface outside the blade row is not shown. The mean
stream surfaces for the inlet stage of a multistage axial compressor
designed on the principle of a symmetrical velocity diagram at all radii
are shown in figure 5.

Both of these two Hnds of stream surface are employed, in general,

in the solution of the three-dimensional problem. The correct solution
of one surface d?ten requires some data obtainable from the other, and,
consequently, successive solutions between these two are involved. Yet,
the solution of each surface is mmageable with the present mathematical
technique and tiomputationalfaci13.ties. h many practical cases, and
especiald,yin the inv~se problem, howev=, this iteration may not be
required if only an approximate solution is required or if the prescribed
values lead to a satisfactory blade shape. These points will be dis-
cussed in the section next to the last (pp. 53 to 57).

Relations among Relative Velocity of Fluid, Coordinates of

Stream Surface, and No-1 to Stream Surface

In general, the coordinates of the stream surfaces and their differ-

entials are related, respectively, by the folladng equations:

S(r,q,z) = O (28)

=0 (29)

Rather than use the three partial derivatives of S with respect to

the coordinates, it is convenient to consider the unit vector n norml
to the surface, which is related to S by

4 % nz 1
(30)

The vector n is, of course, perpendicular to the relative velocity W,

so that
.
n.W=O

—-
 NACA TN 2604 19

or

.. (31)

By using equation (30), equation (29) can be written

~dr+~rd~+nzdz=O (29a)

The vectors n and W are shown on S1 and S2 surfaces in fig-

ures 1 and 2.

EQUATIONS GOVERNING FLUID FLOW ON S1 SURFACE FOR AXIAL-FLOW AND

AXIAL-DISCHARGE MIXED-FLOW TURBOMACECNES

If the fluid motion on S1 is followed, equations (28) and (31)

can be used to e13minate one of the three cowdinates. For axial.-fluw
figs. 6(a) to 6(c)) and -al-dischsrge mixed-flow turbomachines
[fig. 6(d)), it is convenient to express r intermsof~ and z.
For radial-flow and radial-discharge -d-f low turbomachines (figs. 6(e)
and 6(f)), this system will encounter cliff
iculty at the ~t where the
rate of change of fluid state tith respect to z becomes infinite. It
is therefore necessary to eliminate z and to consider r and q as
the two independent variables.

Flow ~0~ General S1 Surface

For axial- and mixed-flow turbomacMnes, ~ qmtity q on the

S1 surface is considered a function of 9 and Zj ttit iS,

The change in q along S1 due to a small change in g while z is

held constant is (see fig. 1)

aq
dq=~~d~+$~qd~

From equation (29a),

%?
%=-~
.

— . . .—. —.-— -—— . —.. .—— ——— — -

20 NACA TN 2604

.
hence

Similaxl.y,for Wp = 0,

With a bold partial derivative sign used to denote the rate of

change of any quantity q on S1 with respect to cp or z, with the
other kept const.azit,
the preceding relations give

(32)

.
With the~~elations (31) and (32), the rate of chmge of q along a
streanilineon-,,S1 is

Dq ‘U~q+W” aq
z= (33)
‘=
Dt Tacp

Equations of continuity and motion. - When the fluid motion is

followed along the stream surface and equations (31) and (32) are used,
the continuity equation for steady relative motion becomes

1 ~(P%) + a(Pwz)
= p C(q),
z) (34)
r
-~ az

where

(35)

For rotational steady relative mq$ion, the equations of motion (14)

in the radial, circumferential, and “&ial directions are

——
NACA TN 2604 21

(36)
Relations (9), (16), (31), and (32) along the relative stream Bur-
face S1 can be used to reduce equations (36) to the fol.lowing:

(37)
The last term in each of the preceding three equations is propor-
tional to the components of the normal vector and therefore can be
expressed as a component of a vector that is parallel to n and has
the dimension of force per unit mass. If this term is defined as

the preceding equ&ions can be written

wu awr awr WU2

—— -l-w —- —-2u.Wu=fr (39a)
r ap z az r

a%
-w —’ + Wu
r az

—. —. —.—
 22 NACA TN 2604

Similarly, the equations of motion in the form of equations (2) can be

written
wu awr awr v~
——+wz ~-— = f’
r
r a~ r

Wu awu awu WrWu 1 ap

——+wz ~-t-y +ZWWr=— —-+f; (40)
r a~ pr acp
1
Wu awz
TV +-w
awz
zK=-—
laP+f,
—
Paz z J
where

f’=-—–
nap
+p&Fn

Because this vector f is Paranel to n, it is

the relative velocity of fluid, or

f rwr + fuwu + fzwz = o (41)

.
By the use of equation (41) and equations (39), it can be shown that for
steady flow on an S1 surface,

(41a)

,. which @ees . Th=ef ore, for the present probla of

with equation (20),
steady relative flow on a stream surface, tie relation (41a) can be
taken eith= as one of the equations of motion or to represent the rela-
tion given by equation (41). In oth= words, th=e are only four inde-
pendent relations among equations (39a), (39b), (39c), (41), aad (41a).

Just as in the case of the continuity equation, either set of the

preced3mg equations of motion is ~ressed in terms +of,the special par-
tial derivatives with respect to the two independent variables cp and
z. The effect of radial pressure gradient is taken into account in all
these equations by the f term, which is neglected in the ordinary two-
dimensional treatment on a surface of revolution. Equations (28), (31),
(34), and (39) or (40), however, lead to a possibility of correctly
solving the threedlmensional flow of fluid psrticles on an S1 surface
in a mathematically two-dimensional manner.

l%ficipal equation. - The equations of continuity and the equation .

of motion h the cticumfmential direction can be ccmbined into a prti-
cipal equation through the use of a stream function ~ as fol.1.ows:
.
Ftist, if a variable b is introduced such that
4V
NACA TN 2604 23

Dinb n= W=
—=-c+—— (42)
Dt ?rr

or

in which the integration is performed along a stresailineon the S,

&
surface, then the continuity equation (34) can be written
a(bpwu) ~(bpWzr)
-1- =0 (34.)
acp az

The preceding equation is the necessary and sufficient condition that

there ed.st a function ~ with

#q. rbpWz (43a)

a+
z= -bpWu (43b)

The difference in ~1 at two points j and k on the ‘1 surface is

JJ
k k
*k-vJ= d*= bp(Wzr dp - Wu ~)
J J

In particular, the difference in W at two potits j and k on the

constant-z plane at the inlet where the fluid state is uniform is
k
~--+j=bipiwz,i ~jr@
r

These two equations show that, physically, the integrating factor b

can be interpreted as proportional to the local radial thickness of a
thin stream sheet whose mean surface is the stresm surface considered
here. The continuity equation (34a) can also be obtained by consider-
ing the mass flow going into an element of such a stream sheet as shown
in figure 7. By equating to zero the mass flow going into the element,
which is defined by two -.1 phes d~ apart and two normal pknes
dz apart (see fig. 7(a)), and letting dp ahd dz approach zero,
there is obtained

-.. —.. .. —.. — ._ —— -. —.- —— —

24 NACA TN 2604

~(TPwu) ~(Tpw~)
(34b)
aql + az ‘0

where T is the radial thickness of the streem sheet. From equa-

tions (34a) and (34b), it is apparent that b is proportional to ‘r,
and the d3.ff=ences h ~ at two points j and k as @ven by the
two equations preceding equation (34b) are proportional to the mass flow
across any We joining the two points. @ actual computation, only the
ratio b to bi or T to Ti is important (a cliff erent imitial value
mounts to a clifferent cons@nt multiplied of the relation between ~
and IUaSSflOW). In the following, b will be retained in the equation,
but in actual calculation it is simpler to evaluate the ratio T to
‘i than to evaluate the ratio b to bi, both from the data obtained
on the S2 surface to be discussed later. Although the evaluation of
this ratio requires, in general, calculations on the S2 surfaces, a
means is nevertheless provided to determine correctly the flow on a
general S1 surface through iteration.

I?rmnequation (43), .

- bp m
aw”=25L.Aw?Q2k
~zz az 32
(44b)

The third terms in the preceding equations can be expressed in terms

of h through the use of equation (12b):

(43)

wh~e s* = s/R. But frcm equations (9) and (43),

(46)

Then from equations (45) and (46),

Wza% 3_)
(
_I ‘U a WZ2 %
(bp) –—-–
r2 a+ r~az ‘——‘4
NACA TN 2604 25

e ___
WZ2 ~
az2 ) ‘+

Substituting the preceding two equations idm equations (44a) and (#b)
and adding yield

WUZ’+ (F~
WZ2)] awz - ~~
awu =
[(
bp a2 -
)

5Z
[(
a I+
~2r2 - w 2
2
r
)(
+az alnb
—-
az
~B*
._
)1
az
’22 ‘z
r 4
3*
32
(47)

Substituting the precediig eqtition into equation (3%) and dividing by

az give the principal equation for the determination of fluid motion
along a general S1 surface:

()
1 -—
‘U2
a2
1 a%
——-z~w+
r2 a~
‘UwZ a%
1
()
-—
WZ2
@+:?!
az . azz
+M~=o

where (48)

a2-(Wu2t02r2)nu
r
-.)+* [u+n.s+fu+l,yh’]
z

The equation of the characteristics of the differential equation (48) is

(reference 31)

_. —- —.—— _________ ___—————. .-—

———.—-
26 NACA TN 2604

from which

~g=- ~uwz + (WU2+ WZ2 - a2)

(50)
& a2-Wz2 a2 -WZ2

Equation (50) shows that the chmacteristics are real when

WU2 +1~z2 >a, in w~ch ease the method of ~~a~teristi~s fOr ~o in&-
pendent variables (references 20, 30, 31, and 32) can be applied. When

r Wu +Wz <a, the characteristics are imagimry, and it is more con-

venient to solve the equation by re~tion (references 25, 33, 26, and
29) -d matrix methods (references 26 and 29) in the following form,
which is obtained by substituting equations (44) into equation (3%):

L&+&- 1 alnbp
)
‘Ug+alnbpa$ +
.2* azz [(
7 aq ‘GW a~ ~ 1

l~+Tas Wr awr Wu
g –—+fu+— —- Wr y+ti =0 (48a)
[ -Facp r acp r aq ( )1
F Fq

Procedure of soltiion. - It may be noted that equation (3%),

instead of (39c), is chosen to form the principal equation (48) or (48a),
becawe fu is, in genewal, much smaller than fz. The various quan-
tities appearing in equation (48) or (48a) are to be computed frbm
other equations given earner. With the introduction of the stream
function, three are altogether seven basic independent relations - one
energy equation (21)3 three equations of motion, (48) or (48a), (39a),
sad (39c); two equations between ~ derivatives and fluid propaties,
(43a) and (43b); and the orthogonal relation between W and f, equa-
tion (41) or (41a). On the other hand, there are ten basic dependent

to define the flow and the shape of the surface. In general, the vari-
able b is to be evaluated according to equation (42a) or from the var-
iation in the radial thickness of stream sheet using the data obtained
in the solution of S2 swfaces and is therefore considered as given
here. E during the complete solution of the three-dimensional flow the
shape of an S1 surface is taken as the one obtained by joining corre-
sponding streamlines obtained on S2 swfaces of the preceding cycle,
two relations between the n- (or f-) components are given by equa-
tions (29), and there are now altogether nine equations to be solved to
NACA TN 2604 27

find the nine udmowns. Alternatively, the variation of Wr may be

considered as known from the S2 solutions of the previous cycle, and
the remaining eight variable~, which determine the flow on and the shape
of the S1 surfaces, can be determined from the seven preceding rela-
tions given and the following additional relation: Because fr, fu,
and fz, respectively, are proportional to
as m as
x’Fr# - @ ‘ffie
integral surface S, they satisfy the following equation (reference 34):

f.vxf=o (51)

which may be written

(51a)

By using equations (31) and (32), equation (51a) becomes simply

(51b)

This equation can be used to give fu by integrating along a constant

~ line:

:= ELO+J:$J3= (51C)

Ifat Z=zo, fu=O, then

z
fr f
fu. — a ~dx
(51d)
r J Tip()
fr
‘o
In this case, then, the shape of the S1 surface is determined after
the f-components (or n-components) are obtained in the solution. In
either case, equations (21) and (41a) are invariably to be used first to
determine the change of s and 1. If the flow is isentropic, s
and I remain constant along its streamlines on the surface. (For
such a case and for a uniform ~t condition, p in the continuity
q
equation may be replaced by h and, consequently, the p and z
derivatives of s, as well as those of I, will not be involved in
the equations (46) to (48)). In case of heat transfer or shock, the
changes in s and I can be estkted by whatever method is available

..— ——.— --. .—.— .—— ——. — .. . ——. ————- .-— —.— ——-.
 28 NACA TN 2604

.
ad used in the calculation. For supersonic flow, all the equations are
used to compute the fluid state at each point and the solution is carried
downstream step by step. For subsonic flow, iteration over the whole .
domain is necess&y. The details of these computations will be given in
the last section. In general, the solution of the flow on the general
S1 surface is very laborious, and is to be used in the final stages of
calculation of the complete three-dimensionalproblem or when a high-
speed computing machine is available.

If the flow is such that it may be assumed to take place on a sur-

face of revolution (at the hub and casing walls or other radii), the
equations are considerably simplified as follows:

Flow along Surface of Revolution

When the S1 swface is a surface of revolution,

%l=fu=o (52)

Let

‘z fz wr
-—= -—=I~=tana=A (53)
% ‘r z

where A is a given function of z. (For aconical flow surface, A

is simply a constant.) Equation (35) now gives

(54)

Whether c can be taken as zero will be determined by the relative

xtu~ of the three t~ on the right side of the equation. h
general, for nonnegligible c, equations (43) now become

a-= rbpWz (55a)

5
aq
z= - bpWu (55b)

Because Wr is now related to Wz by equation (53), the three velocity

components.can be solved simultaneously as follows: By use of the rela-
tions (52) and (53), equation (39b) can be changed to
.

.
NACA TN 2604 29

Instead of eqution (46),

~2r2 1
h =1 +7-– ~ (bp)-2 [@~2)(*%$+(#] (,7)

should be written. Then

(58)

Combining the preceding equations with equations (55) and substituting

the resulting equation into equation (56) give the fo13.owingprincipal
equation for the flow on a surface of retiolution:

where

~2_#_wawz2&2
M=-
alnb
az
+as*+l
——
az ~2 ( “ar+
_—
az
r
r
A+wz%g
)
For this equation, the chmact=istics are real or imaginary when the
resultant relative velocity W is supersonic or subsonic, respectively.
For the stisonic case, it is again better to use the following form
obtained by differentiating equations (55) and substituting the result-
ing equations into equation (56):

..—. . ..—— .——. —..— —.— — —

30 NACA TN 2604

With ~ given and b d.etermhed &rom data obtain& on the S2

surface, there are now the six independent relations equations (21),
(59) or (59a), (55a), (55b), (53), ~d (41.a)for the determination of
the six main variables in ~, Wu, Wz, Wr, s, and I. The f-components
are not involved in the calculation. If the flow is adiabatic with
Uniform I and S, the equations are further simplified.

Flow along Cylindrical Surface

If the flow nesr the walls of an axial-flow turbomachine can be ,,

considered to take pkce on a cylindrical surface, then

nu=nz=fu=tfz=wr so (60)

Equation (35) now gives

aw~
(61)
C=-F

which is relatively small. (If c is negligible, b can be taken as

1 everywhere.) For flow without change in radial distance, the quantity
I can be replaced by ~ ~ = h + 1%2). The equations governing the
cylindrical flow are then (compare reference 29)

a-= rbpWz (62a)

%
(62b)

(63)

(64)
iv
NACA TN 2604 31

and

where

or

In general, the circumferential derivatives of ~ and s are to

be determined by the inlet flow and equations (63) and (64).. For aiiia-
batic fluw with uniform ~ and s upstream of the blade row, these
derivatives are equal to zero everywhere, making the problem much sim-
pler. The main clifferehce between this simplified case and the ordinary
two-dimensiotil flow on a cylindrical surface is the inclusion of the
factor b in equations (62) and (65) (in general, b is a function of
~ and Z). If the velocity diagam is such that there is considerable
radial gradient in the radial velocity or conside=ble variation of the
distance between the adjacent streamHnes, the facta b is not
negligible.

EQUATIONS WVERNING FLUID Fl13WON S1 SURFACE FOR RADIKG-FLOW

AND RADIAL-DISCHARGE MIXED-FLOW TURBOMACHINES

Flow along General S1 Surface

For turbomachines with radial discharge, r smd Cp are considered

as the two intipendent variables; that is,

..— _ .—— —— —.—.

 NACA TN 2604

lw_=+$G$g (66)
5 Zxp z

Lq=~7r~+Yaq Wu ~
Dt

Equations of continuity and motion. - By the use of these relations,

the equations of continuity and motion become

(67)

wh-e
wr awu awz
~l=-_ 1
nz (%~+%1~+%~ ) (68)

and
WG2 Wu awr awz
-— -wu#+-— —w 22bWu=- ~+ T~+f: (69a)
r r- z%V -

Wrwu awu Wr awr Wz awz 1 aI T as

—+wr~--— —- —+tir=— —-+— —+f:
r r acp r aq racp r~

(69b)
Dwz awz wu awz
=wr =+-— = f!
z (69c)
. K r acp

with

(70j
“=-$@ T3n=-$$~n

‘, Principal equation. - If a variable b‘ is introduced such that

!’,’
D@b* —
=-cl (71)
Dt .
NACA TN 2604 33

or

‘ (71a)

in which the integration is performed along a streanilineon the surface,

then the continuity equation (67) can be written
.-
a(b ‘pWrr) E@ ‘pwu)
i- =0 (72)
ar aq .
and a stream function> ~ can be defined on the surface with

(73a)

alf
— = rb’pWr (73b)
acp

Here b~ can be int~reted as’,thethickness (in the z-direction) of

the stream sheet whose mean surface is the S1 surface considered. The
continuity equation (72) can again be obtained by equating the mass flow
into and out of an element of the stream sheet as defined by two -al
planes d~ angle apart and two cylindrical surfaces * distance apart
as shown in figure 7(b). As before, the difference of ~ at any two
points on the S1 surface is equal to the mass flow across any Mne
connecting these two points. By the uqe of the preceding two equations
and the relation

~2r2 l$~z 1
h=I+~ -—- Z(b’p)-z
2 RWW1 ’74)
the principal equation for the flow of this surface is obtained from
equation (39b):

u 1
Wr2
‘F
a2$
Z?’-2
WrWu
—. 1 ~z~
—
a2 ‘a’aQ’+
Wuz

1- Tr2@
()
“
@!k+Ma&+g~
r~

(75)

where

M=-—
ahbt
ar
+**+
ar
1
~ [
-~+wz=+
awz a2-(Wr2iWu2)
r
a2r2+Wr2
r 1

.
--—. — —— ..———
34 NACA !E’?
2604

w
1
( )
z

N=-;~++:~+ - :&+w zaj+

T
ar

U2-(wr%uq wz awz
.&trz ( 1 aI +T~s+f,
.—
- r a~ raq u
-wr+y~
)

This equation is seen to be hyperbolic or elliptic when m ‘s

~eater or less than the ‘speedof sound, respectively. For the elliptic
case, it is preferable to use the following form:

(b1p)2 la~+Tas+f, +Wz~Wz

—— -w u
13$ [ ‘Ta~ r~g u r aq r~ +2!.D =
1
0 (75a)
r
F TV

The integrability condition (51) is now written

.
$@= &&) (76)

hence

fg
fz = q--
fu

r) ~.ro
+Jro$ix+
(76a)

The procedure of solxzLngthe principal equation with the various terms

in it determined by other flow equations is the same as that in the
previous system.

Flow along a Surface of Revolution

For the special case of flow on a surface of revolution, equa-

tions (52) and (53) hold (with h considered as a function of r) and
the expression of c’ reduces to

(77)

— —
 NACA TN 2604 35

Furthermore, equations ‘(73)become

a$ -b ‘pWu
z= (78a)

(78b)
.
and equation (6%) becomes

(79)

Using the relation

(80)

gives the corresponding

(81)

where

a2-W2-w 2_Wz2~2r2
alnbl +~s*+l~I r WZ2 ~
M=- -— —
ar ar .7 -s+ r Adr )
(

—— –————— —.. -. ___ —— _ __

 NACA TN 2604
36
.
Flow on Radial Plane

For the special case of flow on a radial plane,

(82) .
%’%= fr=fu=wz=o

and equations (77) to (81) reduce to

awz
c’=- (83)
K

(84a)

(84b)

(85)

(86)

where
a2-W2-Wr24%2
aInb’ as’ 1 aI+
M=- ~+~+~-z r )
(

(
Ialnb’ 1 as* I a2-% aI “a2-W2 as a2-W2 h
N=-– -7&+ w2 %--— a2 Wr
r~+~=+~ r )

(86a)
.
NACA TN 2604 37

Alternative Form of Equations for Flow along

Surface of Revolution

The equations given in the preceding sections are obtained for

turbomachines to avoid an infinite value of the partial derivative with
respect to z. Difficulty stiLl exists in using either of the systems
in the case of a mixed-flow type machine with an axial inflow and a
radial outflow. For solutions of general s~ surfaces, this difficulty
can be avoided by dividing the machine at the middle of the flow path
and using the first system at the inlet portion and the second system
at the exit portion. If the S1 surface can be appro-ted by a sur-
face of revolution, it is convenient to use a set of orthogonal coordin-
ates 2 ~d ~, where 2 is the arc length of the generating line of
the surface of revolution in the meridional plane and ~ is the usual
cylindrical angle (fig. 8). Because
Wr
—LQLsinu (87)
WZ dl

and

(88)

then, for use with the first system,

aq=secaaq (89)
az z

and, for use with the second system,

(90)

By use of the preceding relations, the equation of motion in the circum-

ferential direction as given by either equation (56) or (79) for the
two systems, respectively, becomes in both cases

which agrees with the results obtained in references 29 and 30 in a

different manner. The subsequent equations given in these two ref~-
ences can be modified and used for such surfaces. (The last term on the
left side of equation (79a) represents the rotationality of the absolute
flow and is not included in reference 29, which is derived for irrota-
tional absolute flow.)

.——— .. —:. .— .——— . —.—. —._ _

38 NACA TN 2604

By comparing the integp?atingfactor b used herein and the thick-

ness of the stream filament of revolution T used in these two refer-
ences, it is seen that the two play exactly the same role in the con-
tinuity rektion. Although b is obtained mathematically as an inte-
grating factor, p~ically it may then be visualized as the thiclmess
of the stresm filament in the r or z-direction for the two systems,
respectively. The use of b herein is, of course, more general in that
it varies two-dhensionally over the surface in the general case, where,
as in references 29 and 30, T is considered a function of 2 only.

EQUATIONS GOVERNING FLUID FLGW ON S2 SURFACES

In the preceding section, it was shown that the determination of

the flow on S1 surfaces requires a knowledge of the radial variation
of the velocity components. This knowledge can be obtained by following
the fluid motion along relative stream surfaces of the second kind, S2.
On S2, the relations (28) to (31) also hold. These relations, how-
ever, will now be used to elhinate the independent variable ~j that

is, any quantity q on S2 is now considered as

q= q~r, z,q(r,z)]

Accordingly, on S2

(91)

and along a streamline on S2

_=wr~+T7z~
Dq
(92)
Dt

Equations of Continuity and Motion

Equations (30) and (91) are used to change the continuity equa-
tion (35) to

.= p C(rjz) (93)

. -.

— —
6V
MACA TN 2604
39

where

(94)

For general rotational motion, the equations of motion (l@) h the

three perpendicular directions sre

wu
-y--~] [
a(vur)
+wz(g-g)=-g+,~
,

In following the motion on S2, equations (95) are reduced to the fol-
lowing form by~~ equations (9), (16), (31), and (91):

Wu a(v~)
--~+Wz~~-~=-~+T~+Fr
r (96a)

Wr a(v&) Wz ~(Vur)
1
D(Vur)
—— + ——=FU or Fur = —
r r az (96b)
[ Dt
awr 3WZ Wu ~(Vur)
—-— -—— __aI , as
~+,z (96c)
-Wr[ ~z ar ) r az = a’ +

where F is a vector having the unit of force per Unit mass of gas
defined by:

(97)

A similar result is obtained for the equation of motion in the form of

(2):
VU2
w
r%+wz%-y=-;%+%
)
Wr a(vur) +Wz a(v~)
r
‘T ‘—
r az = ‘u (98)

awz awz l&p+F

w —i-w
z az ‘T=-~az z
J

. — —— ——..—.. .
40 RACA TN 2604

Because the vector F is normal to the S2 surface,

FrWr + FUWU + FZWZ = O (99)

By the use of equations (99) and (96), it can be shown for steady flow
on an S2 surface that

(99a)

This result is the same as that obtained for the S1 surface. Again,
for the present problem of steady relative flow on the S2 surface, the
relation (99a) can be taken either as one of the equations of motion or
to represent the relation (99). In other words, there are only four
independent relations smong equations (96a), (96b), (96c), (99), and
(99a). In the following development, it is found convenient to use
equations of motion in the form of equation (96), not only because
aI/& is zero in many design problems (whereas ?P~ # O), but also
because equation (96) leads to a form capable of a rigorous solution
for both stisonic and sup=sonic flow and shows clearly how the various
design factors affect the three-dimensional motion in general. (See
equations (106) to (3.14)that f,ollow.)

In a manner analogous to the S1 surface, the continuity equa-

tion (93) is put into the form
~(rBpWr) ~(rBpWz)
+ o (loo)
ar 8Z =

by the use of an integrating factor B, which is related to C by the

follh?wingequation:

DlnB=w ?lnB+w~lnB ~
r z (101)
Dt ar az =

or

(10EL)

Equation (100) is the necessary and sufficient condition that a

stream function $ exist and

rBpWz (102a) .
.NACA TN 2604 41

alf=
- rBpWr (loizb)
az

The difference in Y at two points j and k on the S2 surface is

u’
Ufk-lfj=‘d+-=krBp(Wz
j L1
h - Wr dz)

Similar to’the flow on the S1 surface, the preceding equation

indicates that B is proportional to the angular tbiclmess of a thin
stream sheet whose mean surface is the stream surface S2 considered
herein and whose variable c~cumferential thiclmess is equal to rB.
Indeed, if the mass flow tito and out of the element of such a stream
sheet (cut between two planes normal to the z-axis, and a distance dz
apart ad between two cylindrical surfaces dr apart (fig. 7(c))) is
equated to zero and the tistances dr and dz approach zero as a limit,
the following equation is obtained:

a(TpWr) + a(TPWz) = o
(100a)
ar az

Comparing this equation with equation (100) and considering the mass
flow relations show T to be proportional to rB. This proportionality
means that B can be physically interpretedas a quantity which is pro-
portional to the angular thickness of a stream sheet whose mean surface
is the S2 surface considered herein. With this interpretation, B is
immediately seen to be closely related to the angular distance between
two neighboring blades. In actual calculation, only the ratio rB to
(rB)i or T to Ti is important, and it is also easier to obt%in the
variation in rB from the distance between adjacent streamlines
obtained on S1 surfaces than to evaluate B/Bi by equations (10la)
and (94) using data obtained on S1 surfaces.

Principal Equation for Case with Vur Given

In the solution of flow on an S2 surface, the continuity equations

and the equation of motion in the radial direction are conbined to form
the principal equation. The principal equation will now be obtaiaed for
two main groups of present designs in which a certain desirable varia-
tion of the angular momentum of the fluid Vur and of the ratio of
relative tangential and axial velocity are prescribed on the S2,m sur-
face, respectively. These equations can also be used for the solution
of a direct problem, in which the same information obtained on S1
solutions of a previous cycle is used as lmown values in the S2
solution.
42 NACA TN 2604

For the first group, the following equation is considered known:

Vur = G(r,z) (103)

Among this group of designs are the free-vortex design (in which G is
stmply a function of z), the more general “solid-body rotation” design,
the nsymmetrical velocity dia@wn at all radii” design, and others (foz
example, see references 17 and 18).

From equations (102) -d (45),

awz ah + ~s* -—~ln B

azw + 1 1 —— —~Y
rBp —=—
ar
ar2 ( -—-—
r a2 ar ar ~r
)
ar
(104a)

awr a% + 1 ah as* ~ln B ay

rBpw=— (104b)
-~z+=-— az —az
az2 ( )

But from equations (9) and (101),

(105) .

Differentiating with respect to r and z gives

a2- (Wr2+Hz2)
—s
(
ah

I+
a2 ‘r ~-%2)+ (wr2+w~2)(*+
*-~)-
a% a%
(r@)-lWZ— - H=~
( 3r2 )
&’-(Hr2+WZ2J
ah_ ~ 1 + &&-wuz
z
+ (Hr%z2) yBlnB - ~
ad -
~z a’ ( ) ( )

,rM,-l(wz&wrfE)

Substituting the preceding equations into equations (104) and adding

give

-z WU2 as*
az-— aqf
[(
a I-
-m2r2
z
)(
-a
2 aln B
)1 az z
(106)

-— .—
 J!?ACA
TN 2604 43

Substituting equation (106) into equation (96a) and dividing by

a2 yield the following principal equation for the fluid flow on sur-
. face S2:

‘r2 a2y -2 ‘rwz a2g WZ2 a2Y a~

1 —~+ l-—— +N%+M~=O
() a2
-—~ a2 () a2 az2

(107)

where /

a2-(Wr2+Wz2) Wu ~(Vur)
-g+T ~+ Fr+FT
a~z2 [ 1

From the coefficients of the second derivatives, the principal equation

is seen to be hyp~bolic or elliptic when the meridional velocity

“Z=m is greater or less than the speed of sound, respec-

tivel.y. For the elliptic case, it is again convenient to write the
principal equation in a slightly cMfYerent form. From equation (101),

(108)
awr azyf aIn BpbY
rBp —
az ‘~- az z

Stistituting into equation (96a) results in

(107a)

——. ..— .-. —— .-—. — —-— — —

44 NACA TN 2604

With the variation of Vu or Wu prescribed by the designer in an

inverse problem or taken from the previous S1 calculation in a direct
problem, the meridional velocity components are determined by equa-
tions (107) and (107a). (Other equations sme used to determine various
terinsinvolved in the coefficients M and I?. )

Principal Equation for Case with Wfilz Given

In the second group of designs, the following relation is pre-

scribed en an S2,m surface (for exaqle, see references 17 and 18):

Wu
~ = dr)z) (109)
.
In order to result in blades with the mean blade surface composed of all
radial elements (for high-speed rotation), it may be desirable to spec-
ify a mean S2 surface consisting of all radial elements. Then
WU
r m(z) (I-1o)
~= .

Similarly, in order to obtain a cooled turbine rotor blade with minhum

twist, the following function maybe specified on S2,m:

Wu
~= gz(z) (111)

In application to direct probl~, one of the preceding relations is

obtained from the S1 soltiion in the previous cycle and is considered
as given in the S2 solution. In both inverse and direct problems,
with the relation between Wu and Wz given by these equations, all
three veiocity components are to be conbined into the main terms of the
principal equation as follows: Substituting relation (109) into equa-
tion (96a) gives

(112)

Instead of equation (105),

.
(m)

—..
 NACA TN 2604 45

should now be written. Differentiating with respect to r and z,

ccmibiningwith equation (104), and substituting into equation (112)
give the following form of the principal equation:

(114)

where

This equation is hyperbolic when the relative velocity is supersonic,

elMptic when the relative velocity is subsonic. For the subsonic case,
a form of this eqmtion more convenient for computation is obtained ’by
substituting equation (108) into (lIL2):

(ma)

It may be noted that for both groups, equation (96a).rather

than (96c) is chosen to obtain the principal equation of the present
problem, because Fr is always much smaller than Fz in axial
.
machines and I?r is zero or nearly zero on S2 surfaces for high-speed
centrifugal and mixed-flow @ellers whose mean blade surfaces are
usually composed of all radial elements. (For low-speed centrifugal
impellers, equation (96c) can be used to form the principal equation
in a similar manner.)

. . ..— .—— — —- —. -.
46 NACA TN 2604

Procedure of Solution

Although the equation of motion (96a) is chosen to form the prin-

cipal equation, other equations are to be used to obtain the various
terms involved in the principal equation. As in the case of general
S1 surfaces, there are ten basic variables to define the flow and the
shape of the S2 surface. They ~: V, B, Wrj Wu) Wz) Fr> Fu) Fzj sj
and I (or p). B is considered given. (b the direct problem, B
is evaluated directly fra the tistances between adjacent streamlines
or according to equation (100a) using the value of C obtained on S1
surfaces; in the inverse or design problem, B is estimated (refer-
ences 29 and 35) from the blade thiclmess as desired from blade stress
and other considerations.) On the other hand, there are seven indep-
endent relations b one ener~ equation (21); three equations of
motion, one of equations (107), (107a), (114), or (l14a), and equations
(96b) and (96c); the orthogonality relation between W and F, equa-
tion (99a); and the ~o eqqations relating ~ and velocity, (102a)
and (102b).

Direct problem. - In the dfrect probiem, two alternative procedures

may be used. If the shape of the S2 surface (detemined from the
data obtained on S1 surfaces) is considered as given in the present
S2 solution, two additional relations between the n- or F-components
completely define the problem. The procedure of calculation is as
fOllows:

(1) Use equations (20) and (21) to determhe the variation of

S and I.

(2) Compute Wu from the orthogonality relation as follows:

Wu=-
(:Wr+=w ~z
)
(3) Compute Fu from equation (96b).

(4) Solve the principal equatim.

(5) Obtain Wr and Wz frcnnequations (102).

.
If only the tangential velocity or the relation (109) is taken
from the S1 solutions of the previous cycle and is considered as
given in the present S2 solution, one more relation is available
between the F-components such as that which exists between the
f-components on the S1 surface:

FoVXF=O (115)
 7V
NACA TN 2604 47

.-
Writing equation (U.5) in scalar form and using the relatio~ (31)
and (91) give
.

(l15a)

By integrating along a constant z-line, equation (l15a) provides ~he

following relation to determiriethe value of Fr to be used in the
principal equation from the values of Fu and Fz:

(llsb)

IfFr= 0, at 20

(IJ5c)

The procedure of calculation is as follows:

(1) Use equations (21) and (99a) to determine the variation of s

and 1.

(2) Compute Fu and Fz from equations (96b) and (96c).

(3) Compute Fr from equation (11~) or (125c).

(4) Solve W from the principal equation.

(5) Compute Wr and Wz from equations (102a) and (102b).

Inverse problem. - In the inverse or design problem of a finite

nuniberof thick blades, in addition to the blade-thiclmess distribution
or its equivalent B, either equation (103) or (109) is prescribed on
a mean stream surface ‘2,m. lt may appai that still tiother rela-
tion can be prescribed on the meam surface. !llhe-differentid-sof the
coordinates of the surface are now governed by

Frdr+Fur@+Fzdz=O (116)

—.—.———. ——.—. .— .—. ——— .—-. ——— - -—

 48, NACA TN 2604

however, and, ti order that this differential equation will.lead to an

integral surface of the form represented by eqqation (28), F must
satisfy the condition of integrability as given by equation (115)
‘ (reference 34). An expression shilar to equation (3Ma) for the case
of an inf~te number of blades was first ~ointed out by Bauersfeld
(reference 2) ina discussion of the Lorenz paper (reference 1). h
effect, it restricts the freedom t~t the designer has in prescribing
the velocity components of the fluid on the surface. Hence, in the
tiverse problem of a finite number of thick blades, in addition to the
bide thiclmess distribution or its equivalent B, the designer can
specify only one relation on the mean stream surface, which relation may
be either the tangential velocity as givenby eqgation (103), the flow
angle between the tangential and axial velocityas givenby equa-
tion (109), the axial veIocity, or any other reasonable relation that
~ lead to a solution of the set of equations.

U the preceding consideration, the hub and casing shapes sxe also
prescribedby the desi~er in the inverse problem. Alternatively, the
prescription of the hub shape canbe replacedby a prescription of
another relation at the casing, thereby fixing the shape of and the flow
along S2,m at the casing entirely. l?heflow is then extended to the
hub and the last streamlhe gives the hub contour (reference 19).

@roximations 12nvolvedin Through-Flow Theory

When the equations previously derived in reference 18 for a lsrge

number of thin blades are compared with the corresponding equations
derived herein along a stresm surface, the two are obviously exactly
the ssme if the ordinary derivatives used ti reference 18 are replaced
by the present ysrtial derivatives following the stream surface, and
if B is equal to 1 or if the variation of B along the flow path
is zero. b the interpretation of the though-flow solutions as the
flow aiong a mean stresm surface (which divides mass flow into two equal
parts circumferentially)or as the flow along the-mean channel surface
(geometrical mesa), the first difference can easily be ramvedby shply
interpreting the values obtained in the solution as those along the sur-
face rather thsn in the meridional plane. The second condition, however,
is satisfied only when the circumferentialvariation of alllthe velocity
components approaches zero, or when the cticumferential derivative of
the tangential velocity and the ratios of nr and nz to nu approach
zero (see equation (94)).

Besides the use as a limiting solution in general and to give cer-

tain trends ~~herethe contribution due to the finite number of blades
is shall or constant, the through-flow calculation should be properly
mo~iedby the factor B in its application to actual turbomachines

.
 NACA TN 2604 49

of a finite numiberof thick blades. As B can be physically inter-

preted as the ratio of the local sagular thickness of the stresn sheet
to its inlet value, a good approximate value can be obtainedby solving
the two-dimensional flows on a nuder of.stream surfaces of revolution
starting at ~erent radii at the inlet. For-the subsonic flow in the
turbine cascade reported in reference 29 and for the supersonic flow in
two hqnilse bladings investigated in reference 30, the reductions in
angular thickness from the inlet value along the mean streamline are
seen to be a chordwise average of 4 and 9 percent more tk the reduc-
tion in the channel width, respectively. Also, in the stisonic case,
the influence is extended a certain distance outside the blade row. The
inclusion of this factor B, even if it is approximate, should give a
much closer answer than that obtained with B taken as 1.

h this interpretation of the infinite nurher of blades solution

as the solution of through flow along a particular stream surface
between two adjacent blades, the distributed ‘%odj force” F has a
definite meaning, as given by equation (97). (For an infinite number
of blades, F becomes the blade force.) For blades with large turning
and large radial twist, as in a free-vortex turbine, the influence of
the radial component of F on the flow is not negligible.

CIR~IAL VARIATION OF FLI%EDPROPERTIES BY USE OF POWER SERIES

In general, the blade-to-blade variations of fl@d properties are

to be obtained from calculations on S1 surfaces. When the twist of
the S1 surface is large, some other method of obtaining the blade-to-
blade information is desirable, For subsonic irrotational absolute
flow, this information can be obtainedby etiending the solution
obtained on the mean stream surface in the circumferential tiection by
the use of power series (without the consideration of the shape of the
S1 flow surfaces). The various derivatives involved in the series are
obtained from the flow condition on the mean stresm surface. The higher
the eolidity and the thinner the blade sections, the fewer are the terms
required for a given accuracy. Results obtained in references 29 and 36
indicate that only three terms in the series will be required to give
sufficient accuracy for high-solidity turbines and centrifugal com-
pressors.

The series method will also be used h one of the two methods of
the inverse solution in which-the flow obtained on the mean stream sur.
face is extended out circumferentiall.y.

Denoting the absolute vorticity VxV by ~ and using the rela-

. tions (16), (91), and (97) in equations (19) give

. . . _— . ———.-– —— —.—— —— .——.—— . . —

50 NACA TN 2604

From the preceding equations,

This equation means that the apparent vorticity, which is obtained by

cliff
erentiattig the velocity on the mean stresm surface with respect to
the coordinates, is not zero even if the absolute vorticity is zero or
tangent to the mean streem surface. Substituting equations (l17a)
and (U7 c) into equation (94) results in

Wbstitut3ng equation (IJ9) into equations (l17a) ad (137c) gives

(120)
Fz ~(Vur) “Fr
——+C+ ~(r-~ q
Fur az

(121)

The first derivatives of h or p, and p can be obtained as fallows: ‘

From equation (97),
.

NACA TN 2604 51

Or from equation (9),

with * how+ can be obtained by using equation (Mb):

(123)

The second derivatives of the fluid properties with resyect to ~

can be obtained in a stilar manner. Differentiating the continuity
equation (1) with respect to Q and diviting by r give

1
~ (124)

Eqpations (91) are used to change equation (124) to

(X2-4a)

Differentiating equations (l17a) and (IL7c) with respect to cp and

dividing by r result in

(125)

(126)

—.. . —— . .. ——
 52 NACA TN 2604

Substituting equations (125) and (126) into (124a)‘and noting that F

is perpendicukr to -W give

..
~ a2Wu FU2 Fr a a(vur) Fz a a(vur) 1 ~ a(p~lrr)
—— —.
~z=-~ [ a=-+ Fur’z-+P’2’r~+

(127)

Equation (103) is to be used in equations (101) and (102) to obtain the

second derivatives Of Wr ma w’. The second derivatives of h and
p are obtained frmn equations (9) and (123) as

(128)

.. la2h p= 1 a2h I a2s

(129)
7 a$ a2r2 aq? ‘~a$

Stiar formulas can be obtained for higher-order derivatives. At

a fixed value of r and z, the velocity components, h, and pata
;Ghortangular distance away from the mean stream surface S2 can then
‘beobtained by a Teylor series:

(130)

An alternative way to obtain density is to use equation (145) (to be

given subsequently) titer the other fluid ~roperties are determined.
Obviously, the preceding equations are most useful when the flow is
isentropic with vorticity equal to zero. Otherwise, the variation of
vorticity along the mean stream surface has to be determined first.
At present no such method is available. It appears, however, that the
method of Squire and Winter (reference 37) and Hawthorne (reference 38)
may be generalized to compressible flow for the variation of vorticity
.along a mesm stream surface in turbcnnachines.
 , NACA TN 2604 53

.
STEPS FOR CCMPIETE S(ILUTIONSQE’THREE-DIMENSIONAL

DIRE(W AND INVERSE PRQBLEMS

In general, the solution of the three-dimensional direct and inverse

problem involves the use of both 61 and S2 surfaces. In the direct
problem, starting with assumed flow surface, the solution is obtained
through the successive (alternate) use of the two kinds of flow sur-
face, although a satisfactory apprmte solution maybe obtained in
one or two complete cycles. The use of an appro-te method of solu-
tion to get a good starting value on each surface will shorten the length
of computation. For inverse problems, the process is usually shorter.
The calculation will start on the S2 m surface on which either a con-
dition on the fluid velocity or the shpe itself is prescribed and an
esthated value of B for a desirable blade thickness distribution is
used. After the solution on the S2,m surface and its shape are obtained,
the blade coordinates are obtainedby extending the solution circumferen- “
tially either by the series method or by the method given in reference 35
using the variation of the distances between the streamline obtained in the
S2,m surface. Because it is important only to obtain the right order of
magnitude and the right kind of variation (three-dimensionally)of the
blade thiclmess, the first solution may give satisfactory results. The
velocity distribution on the blade surface is controlled directly by the
one relation specified on the S2,m surface and the variation of B.

Suttable procedure is stisequently suggested for the solutions of

Mrect and inverse probl- with either irrotational or rotational inlet
absolute motion, at design or off-design flow conditions, for turbo-
machines hav5mg various wall configurations (fig. 6).

Direct Rroblem

Axial turbomachines with nontapered straight walls. - In this type

of machine, it is desimble to start the computation on S1 surfaces,
because with short axial blade length, the total deviatim of the S1
surface from the cylindrical surface is relatively small, especially
along the hub and casing walls.

The fofiowing steps are therefore suggested:

(a) h the initial calculation, the flow surfaces are assumed to be

cylindri&l and the set of equations (60) to (65) derived for cylindrical
,. flow or the approximate method given in reference 39 can also be used to
obtain the streamlines and circumferentialvariation of fluid state on
S1 surfaces at three or more radii.

.>..,.

.—..— _ —— J-_ ———. -..—..

54 NACA TN 2604

(%) From the data obtained in step (a), an S2 stream surface about
midway between two blades is constructed by connecting the streamlines
which divide mass flow on the S1 surfaces in the same percentages. The
direction ntiers of the surface and the ~ and Wz at the surface
obtained in step (a) give the starting value of Wr by use of equa-
tion (31). The factor B is evaluated either directly from the angular
distances between streamlines obtained in (a) or according to equa-
tion (lO&) with C evaluated from the information obtained in (a). Its
value at other radii is obtainedby interpolation or by proportioning
according to the channel-width ratio. Calculation of the flow on this
surface is then made by the use of equations (91) to (115). For subsonic
fluw with tirotational inlet flow, the solution obtained on the S2,m
surface is easily extended circurnferentially”byseries expansion using
ec$xations(117) to (130). The values obtained can be further adjusted
to fit the given blade (reference 39) and csmbe used in a more accurate
second calculation on S1 surfaces in the neti step. For subsonic flow
with large rotationality at the inlet and supersonic flow with signifi-
cant check causedby the blade entrance angle, it is more desirable to
obtain the information cm circumferential variations by’the use of two
or more S2 surfaces at or near the two blade surfaces.

(c) The radial variation of fluid state cmputed from the solution
obtained in step (b) or the variation of the radial distance between
streamlines is used to determine the factor b and used in the principal
equation (48) for a more accurate determination of S1 surfaces and the
flows thereon. The general equations (32) to (51) should now be used
for the S1 surfaces located between hub -d cas~, if not at or near
these walls.

(d) The calculation of S2 -

surfaces can again be repeated and so
forth.

~ the inlet flow is quite rotational, so that the S1 surfaces

along the walls and the S2 surface nesr the blades may turn around the
corners, these surfaces should be chosen at a short tistance from these
boundary walls as shown in figure 3. By the use of these two kinds of
surface, the secondary flow caused by a rotational inlet piofile or by
the t- of the blades is included in the complete solution.

Axial turbcmachines with tapered or curved walls. - The steps

involved here are quite shilar to those of the preceding case, except
that for the initisl calculation of S1 surfaces along or near the
tapered or curved wall, either equations (52) to (59) are to be used, or
equations (13’) to (23?) given in reference (35) cm be used in the
manner given in reference (39).
8V
NACA TN 2604 55

Radial- and mixed-flow type turbomachines with curved walls. - In

this type of machine it is not desirable to start the computation on the
S1 surfaces because the flow surfaces near the walls may deviate con-
siderably from surfaces of revolution because of the long flow path. On
the other hand, the solidity of the blade is very high and the blade
section is umiformly thin. As a result, the shapes of the S2 surfaces
are closely related to the blade shape and the factor B can be esti-
mated relatively accurately from the blade thiclmess distribution. The
fol.lowimgsteps are therefore suggested:

(a) The computation is begun on the S2,m surface. For subsonic

irrotational inlet flow, computation need be made only on a mean S2
surface and the solution can be extended out circtierentially by equa-
tions (117) to (130). The approximate method given in reference 40 can
also be used in the initial calculation. For subsonic flow with rota-
tional inlet profile and for supersonic flow it may again be more
desirable to compute two or more S2 surfaces between the blades.

(b) The data obtained in step (a) may be used to make calculations
for three or more S1 surfaces between hub and casing walls.

(c) The processes (a) and (b) can be repeated until the desired
accuracy is reached.

Inverse I&oblem

Conditions prescribed on mean stream surface. - h the inverse or

design problem it is most convenient to consider a mean stream surface
of the S2 kind about midway between two neighboring blades to be
designed (figs. 3 to 5). From the results developed previously for
such surfaces, it is seen that in addition to the factor B, the
designer can specify only one relation among the fluid properties on
that surface, which can be either a velocity component, a relstion
between two velocity components, or one other reasonable condition.
The factor B essentially control~ the blade thickness distribution,
whereas the relation specified on the surface essentially controls the
mean camber surface of the blade. From a consideration of strength and
Mach number in general, and the requirement of coolant passage in the
case of cooled turbine blades, the designer always has a very good idea
of what kind of blade-thickness distribution he wants. With thiS thick-
ness distribution, the ratio of pitch minus circumferential thickness
of blade to pitch can be obtained. After correcting this ratio with
scnneknown relations between this ratio and B (such as those given in
references 29, 30, and 35), especially near the leading and trailing
edges, it can be taken as the factor B in equation (101). Then fram
the type of velocity diagram or a certain feature of blade shape

..
. . ,. ,,
_.—z _.—._ — -—.—.—. —. —.— —— -
56 NACA TN 2604

.
desired, a relation along the mean stream surface S2 m can be pre-
scribed and coordinates of the mean surface and the f~ow on that surface
caQ be solved at the same time by equations (101) to (115). It may be
noted that in this process, the designer still has, in general, a little
freedum in choosing the value of Z. in equation (IJ5c). For a rota-
ting blade, Z. is usually taken somewhere nesr the center of gravity
of the blade section, whereas for the stationary blade, the position of
Z. can be utilized to control the magnitude and distribution of Fr
in the most favoreble manner.

Boundary conditions for mean stream surface. - In the solution of

this S2,m surface, the boundary conditions are a little different for
subsonic and supersonic flow. For subsonic flow, not only the varia-
tions of the stresm function at stations far upstream and downstream
are given, the meridional contours of the hub and casing walls are also
given (these contours canbe determinedly approximate calculations
from bide row to blade row such as given in references 17 and 41). For
supersonic flow, the variation of the stream function and its nom@
derivative is prescribed on an initial curve, which is not a character-
istic curve. Then either the hfi and casing contours are prescribed,
or only the casing contour bti with one more velocity component along
the casing is prescribed. In the second case, the flow is extended
.
toward the axis of the machine and the hub contour is determinedly the
shape of the last streamline for the required mass flow.

Determination of blade shape. - For subsonic irrotational flow,

the solution obtained on the mean stream surface canbe extended out
circumferentiallyby using equations (117) to (130). The blade sur-
face can be then deterndned as follows:

(a) The position of the mean streb surface is first determined

by solving the circumferential.coordinate as a function of the axial
coordinate at several r@ii. With the circumferential coordinate
measured from the radial element of the surface chosen at Zo, equa-
tion (116) gives, at a constant r:

z
% - (rqJ 2=20 =

J’L)
Zo
Fz
Fmb
(131)

(b) 5e blade coordinates (r,Q ) will first be chosen at one

station 20 as follows (see fig. 9): The mass flow passing through
the Z. plane between the mean stream surface and the tentaiive suction
or pressure surface is computed as follows:

——. ———
NAOA TN 2604 57

J
‘c
p Wzr dr dq (132)
J--h

rc
Mp .
J’%2
Pm J %
pWzrdrdT (133)

Because of the inaccuracy in B for the blade-thichess effect, the

mass flow obtained will be a little different from that required. The
blade coordinates ~ and ~p as functions of Z. and r are modi-
fied until the mass flow checks. It is not hportant that ~ and Mp
are a ~ttle different frmn one-half the reqyired mass flow as long as
their sum is eqwl to the total mass flow, but once the division is
chosen, it should be maintained at other z-stations.

(c) The blade coordinates obtained at z = Z. are extended

upstream and downstream according to the velocity components evaluated
at the surface. For exsmple, for a short distance z - Z. away, the
changes in the blade surface coordinates r and Q are

T4r
r=ro+
(z)w— o
(z Zo) “ (134)

(135)

After r and 9 are thus obtained, the total mass flow may be checked
againby equations (132) and (133).

When the blade coordinates are obtained close to the leading and
trailing edges, they can be faired in according to some standard shapes.
A blade shape is therefore obtained in which the three-dimensional flow
of the fluid is considered. The right kind of three-dimensional blade-
thiclmess distribution is obtained”and a good knowledge about the flow
on the blade surface is also available at the same time. !llhe~tita
obtained in the solution can also be used directly for a m?re accurate
and detailed determination of the velocity variation around the nose of
the blade, for a relatively qtick check of the series approximation, or
for improvement, if necesssmy, of the inverse solution t~oughout by
the method given earlier for solving the direct problem. This process
seems to be the quickest way of establishing some standard.three-
Mmensional flow variations for typical designs of blades from which a

———.——- ———— — ..———.

58 NACA TN 2604

good appro-te method for routine design calculations can be estah-

lished and of providing a basis on which the viscous flow along the
blade surfaces and hub and casing walls can be analyzed. The results
given in references 29 and 35 indicate that for blades of high
soliddty, three terms in the series give sufficiently accurate
results.

For subsonic flow with vorticity, the circumferential extension

cannot be accurately made at the present because of lack of adequate
methods for the determination of vorticity variation along the mean
stresm surface (S’2,m). An estimate of this variation can be made,
however, and the solution can be cheeked later. An alternative method
is to use the shapes of the streamlines and the distances between them
obta3ned in the S2,m solution and to de~ign the blades with the
assumption that the flow surfaces are surfaces of revolution by the
method given in reference 35. Iiwanuch as the rotationality of inlet
flow is usually serious only in later stages of a multistage compressor
where the hti-tip radius ratio is high, this assumption is reasonably
good.

For supersonic flow, the flow in the mean stream surface S2,m is
also determined first. E the shock due to the entrance wedge angle
is small, an appro~te solution of the blade shape can also be
obtained by.the series method neglecting the finite jumQ across the
shock or using an estimated value. The hprovement of the flow varia-
tion for the resultant blade is then more important than that in the
subsonic case. Iacal modification of the blade shape can also be made
if the velocity Ustributfon on the blade obtained is unsatisfactory.
Abetter method is to use the shape of the streamlines and the dis-
tances between them obtained in the S2,m solution and to design the
blades assuming flow surfaces of revolution according to the method
given in reference 30.

The processes described here for the three-dimensional solution

have been and are now being usedto analyze the compressible flow
through a nuniberof compressors and turbines. Some of the results
obtained are given in reference 42.

GENERALMETHUDS OF SOLVIHG PRINCIPAL EQUATION

In the solution of the S1 surface for the direct problem and of

the S2 surface for both the direct and inverse problems, the main
calculation is the solution of the principal equation} which is a
second-order, nonlinear partial &lfferential equation in two independent
variables. The case when the prticipal equation is elliptic willbe
considered first.
NACA TN 2604 59

mptic Case

A common form of the principal equation is written as follows:

M2(%p)2 = O \&--,

@ equation (136), ~ and b are used for both S1 and S2 surfaces;

rj denotes ~ for the S1 surface and r for the S2 mrface;
~ denotes z or r for the S1 surface, S@ z for the S2 — sur-
face. The values of q, (, J, and K for each tidividual case are given
in the following table:

Case Surface Coordinates J K L Equation

m
%
1 S1 (general) Q z —1 (48a)
r2 w
2 S1 (surface of z
l+N’ o (59a)
9 r2
revolution)
3 S1 (cylindrical 1
‘T z 0 (65a)
surface) ~

4 S1 (general) ‘T r 0 (75a)

5 S1 (surface of 0 (81a)
revolution)

6 S1 (radial 0 (86a)
I plane)

7 S2 (Vur given) --1 (107a)

r

a
1(
S2 J: specified
)
1
~+g~ ag (l14a)

_ ———.-—__
—— -—.- -——-
60 NACA TN 2604

The equation is nonlinear even in the case of incompressible flow. In

the numerical computation, it is convenient to rewrite the equation in
the following form:

(137)

where

alnbp Qr+nnl)@. M1bp - M2(bp)2

N =K
aq q a~ a~
(138)
ZnJ

and is evaluated frm any approximate solution at the start of the cal-
culation and from the values of $ and p obtained in the previous
cycle during the calculation. For simple boundary shapes for an S2
surface and s3mple functions of J, K, and L, it is possible to find
a Green’s function G(q,-~, ‘~,y) with its proper characteristics so
that “the solution of the problem can be written h the fol.lowingform
(for example, see reference 10):

if (%!.) = j’~ G (,, C, x,y)I?(x, y)axay (139)

If the boundary wall is arbitrarily curved, it is necessary in this

method to use the technique of conformal transformation to render the
given boundary into a shupler one, such as cylindrical. Because this
process wUl involve a numerical solution of the Mplace equation with
the given boundary shape, it may be better to solve the given equa-
tion (137) directly with the given shape by the numerical method.
Furthermore, this method will be the only choice in the general case
where J and K are complicated functions, which makes the task of
obtaining the proper Green’s function very difficult if not impossible.

Finite-difference form of principal equation. - b order to solve

the given equation (137) directly, the general numerical differentiation
formula for ftist s& second derivatives with the function value given at
unequally spaced grid potits using second- and higher-degree polynomial
representation as given by reference 26 is used to give the finite- .
difference expressions conveniently and accurately at the grid point
near the curved boundary. H the value of any quantity q on the
stream surface under investigation corresponding to a nuniberof values

.—
NACA TN 2604 61

of one of the independent variables x not equally spaced, denoted by

Xo, the mth derivative of q (on
the x . xi may be written

(140)

The differentiation coefficients B and the coefficients of the deriva-

tives in the ftist or second remainder term have,been explicitly
expressed in reference 26 in terms of the spacings between the successive
grid points by ushg Polynomials of the second, third, and fourth degree
for general nonunifomn spacing throughout and for the special case near
a tapered or curved boundary where only the first or last spacing is
&Lfferent from the others. For the speciaL case, these coefficients
have also been computed for different ratios of the distance between the
boundary and the nearest point and the other spacing, from 0.1 to 1.29
in intervals of 0.01, and are given in reference 26. For spacing ratios
lying between these tabul&ted intervals, B can be obtained from the
values tabulated by applying interpolation formulas given in refer-
ence 43, or by the direct use of the formulas. Differentiation.coeffi-
cients B for equal intervals using various degrees of polynmnials are
given earlier by Bickley in reference 44.

In the present fluid-flow problems, a large region must be covered

in order to get to the boundary conditions which are always given at
stations far upstream and downstream of the blade row. In order to
reduce the labor of computation, it is desirable to attempt to reduce
the number of grid points reqpired for a given accuracy by using a
degree of polynomial higher than the customary second. A study of the
expression of the remainder terms (see reference 26) and actual experi-
ence in the present problem show that, in most cases, the use of the
fourth-order polynomial will reduce the necessary number of grid points
to less than one-fourth that required by the second-order polynomial.
Near the leading and trailing edges of the flow on surfaces of the S1
‘kind,the variation of ~ is such that accuracy is obtained most
effectively by using small spacimg there. In such case, the grid
pattern shouldbe chosen at these regions first, and either be kept
constant or be continuously increased toward the inlet and exit stations.

With the grid pattern and the order of polynomial representation

selected, the coefficients B at each point canbe obtained from refer-
ences 44 and 26 for equally and uneqwlly spaced points. Then the dif-
ferential equation (137) at any grid point whose ~ value is @ is
replaced by the following algebraic equation:
 .
62 NACA TN 2604

where @ and @ denote the values of ~ on the surface considered

corresponding to the grid points along constant E and constant TI
lines,’respe&ively. ‘(See-figs. 10 ~d il.) It =houldbe noted t&t,
in accordance with the defhition of the special partial derivatives,
~ values are those on the surface S; whereas the grid spacings
involved are just the distances along the q- and ~-coordinates.

Boundary conditions. - In flow on surfaces of the first kind, the

flow picture is as shown in figure 9. Arbitrarily assigning a value
VI on the SUCtiOn SUrface, the ValUe ~11 on the pressure surface of
the next blade is determined fran the mass flow passing between them.
These two values are used as the end values in equation (141) for grid
points next to the boundary. Outside the blade region, however, the
position of the Mvid@g streamline is not known. Enstead, there is
the condition that the flow repeats itself or the ~-value increases
by vll-~l when CII increases by an amount equal to the pitch angle
(2x divided by nuder of blades). It is then convenient to draw any
two parsllel lines up to the leading and traikhg edges of the blade and
consider only the grid points lying between the two reference ties.
For the ~-derivative at a point ~, for example, the required ~b .
value is obtained from ~f, which is a pitch angle awsy from ~b
(fig. 10), as

~b . ~f (142)

This relation is used between the inlet station 1-1 and the leading edge
of the blade and between the exit station 2-2 ‘andthe trailing edge of
the blade when the S1 surfaces are assumed to be surfaces of revolu-
tion. For the general S1 surface where its deviation from the surface
of revolution is considered, mtiication has to be made in places such
as shown in the exit portion of figure 9. Because of the twist of the
flow surface, the dividing line from station 1-1 to the leading edge of
the blade beccmes two separate Mnes from the trailing edge of the blade
to the exit station 2-2, acccmrpaniedby trailing vortices. Although the
flow still repeats itself circumferentia13yevery pitch angle, the use
of equation (142) for the derivative at a point close to these lines
will give inaccurate results. b these cases, it is better to use the
end-point differentiation formulas to evaluate the derivatives.
Dv
~CA ~ 2604 63

At the station 1-1 sufficiently far upstream of the blades, the

flow condition can be taken as uniform and the flow angle, equal to the
given inlet angle. For the point h, the ~ value at point i
upetream can be obtained from the given flow angle as follows:

@=*13 +5’ty~(@ - w)
v

= ~-qti.%)ti
(

Thus, the required ~ value upstream of station 1-1 can be replaced by

the values on that station, and only the ~ values downstream of
station 1-1 will be involved in the finite-difference expression (141).

An alternative method to take account of the inlet condition is

as fouows: If the first station 1-1 is chosen sufficiently far from
the blades, the variation of the stream function upstream of the
station 1-1 is linear in the circumferential direction. The value of
the stresm function, however, depends on the inlet single. If solutions
for a range of inlet angle are desired, they canbe obtainedby speci-
fying a number of sets of linearly varying stresm functions upstream of
station 1-1 as fixed boundary values. The slope of the streamlines
obtained in the solution at the inlet then gives the value of the inlet
angle. If, however, the solution for a certain specific inlet angle
is desired, the streamline obtained in the solution must be adjusted
according to that inlet angle, for exsmple, as jk in figure 10 is
adjusted to position gk, thereby obtaining an improved set of boundary
values of the stream functions to be used in the neti calculation.
This method is, of course, not so accurate and convenient as theprevi-
ous method for obtaining a solution for a given inlet angle, but is
desirable in the matrix solution because the inlet angle is then not
involved in the matrix factorization, thereby making the same matrix
factors usable for a range of inlet angle and Mach number.

At the exit station far downstream of the blade, the same methods
can be applied. For a blade having a sharp trailing edge, the Kutta-
Joukowski condibion can be used and the correct exit angle far down-
stream is the one that gives the flow at the trailing edge satisfying
that condition. For round trailing edges, either the position of the
stagnation point is assumed or some available empirical rule for the
exit angle is used. If the calculation is made ~o ccmpare with certain
experimental results, the measured exit angle may be used.
.

—.—— _..—.. .— —— -—.— —.— —— .

64 NACA TN 2604

In flow along surfaces of the second kind, the boundary walls

extend all the way to the inlet and exit stations with the ~ values
given on the wslls (fig. 11). Across the inlet and exit.stations, the
flow is considered to be uniform and parallel to the walls so that the
required $ value outside the station can be obtained by an eqyation
dmilar to equation (143). For the inlet station where the axial
velocity is radially uniform and there is no radial or tangential
velocity, $ varies as the square of the ?diu.s. For the exit station
with a certain radial gradient in fluid state, -theradial variation of
~ canbe dete~edfrcm the corresponding radial variation in axial
velocity and density.

Solution of finite-difference equations. - With the grid system

and the degree of polyncunialrepresentation chosen and the boundary
conditions taken into account, the problem rema~ is the solution
of the set of linear algebraic equations (141) written for ti interior
grid points. For a small nuniberof solutions with a given blade, the
best method is the relaxation method (references 25, 33, 45, and 36).
A modification of this method tivolving the use of higher-order differ-
ences is suggestedby Fox (reference 46). Formulas and tsbles of
coefficients obtained in reference 26 enable the direct use of higher-
de~ee polynomials for problems with curved boundaries (reference 29).
For the present flow problems, it is necessary to include a large
danain to get to the boundary conditions that are given at places far
from the blades, and the use of higher-degree polynomials whenever it
is applicable greatly reduces the numerical work..

If a nuniberof cases are to be solved for a given geometry (ssme

blades for S1 surface and same hub and casing shapes for S2 sur-
face), it is advantageous to solve the problem on a large-scale digital
computing machine. If a high-speed digital machine is available, the
shnultaneous equations may be solved by Liebmann’s iterative process,
which is the most s~le to set up. For quicker results or when only a
relatively slow-speed machine is available, the matrix process discussed
in reference 26 is most suita%le. In a calculation of the S2,m sur-
face for a gas turbine and in a calculation of the S1 surface of
revolution for a centrifugal compressor, the coefficient matrices
(about 400and 200 interior grid points for the twoprobl~, respec-
tively, and the fourth-degree differentiation formula we used) were
factorized into the lower and upper triangdar matrices on an IBM CPEC!
and an IEM 604, respectively, in about 60 hours. The determination of
~ for a given set of values of N took only 2 hours on the CPEC for
the gas-turbtie problem. The gas-turbine problem was also worked out
on an Univac; the factorization took only I-1minutes and the determina-
tion of ~, 2.5 minutes. The increasing availability of these high-
speed brge-scale digital calculating machines will render the suggested.
methd of solving the three-tiensional-flow problem a practical one.
 NACA TN 2604 65

General table for evaluation of density from $-derivatives. -

After the ~ values are obtained at the end of each cycle of calcula-
tion, the velocity components are evaluated from the derivatives of ~
with respect to the coordhates, after the density is obtained as
follows: From equations (46), (57), (74), (80), (85), (105), and (113),
the relation between h or p and ~-derivatives can be put into a
common form as

~2r2
h= I+ Y- X-+.(bp)-2 k
[[*%s+(:*)!l (144)

The quantities representedby X, as well as by C, ~, and ~ for dif-

ferent cases, are given in the following table:

Case Surface nates k z--

Equation
c
—
.< 1 S1 (general) z 1 1 (46)

2 S1 (surface of z (1+ w) 1 (57)

. revolution)
3 S1 (cylindrical z 1 1
surface)
—

4 S1 (general) r 1 (74)

5 S1 (surface of r 1 , (80)
revolution)
6 S1 (radial r 1 (85)
plane)
—
7 S2 (Vur given) r z 1 r (105)
8 r z l+gz r (1.13)
‘2 k ‘iven)
—
Nith the t-derivatives evaluated, if an exact determination of h
or p from the”preceding equqtion considering the variation of specific
heat with temperature is desired, the Keenan and @y gas tables (refer-
ence 47) can be used. With two or three readings of h and p (or its
reciprocal, specific volume), the correct value of h or p satisfying
equation (144) is found. lbr most cases where the temperature range
~nvolved is not too large the use of an appropriate average value ~f Ty
r> maY give accurate eno~ res~ts” With the use of ~ aver%e T)
the density at any point in the flow field canbe related to the inlet
total value by equation (12a) as

..— ———.-
~— —..——-—, —— .——. —
66 NACA TN 2604

1
L
~Fi sf,
i-s*I + %-x [-(%$+(++7]=s; ;,-.”
&=% () e =
{
Hi - 2(bP)2 Hi
}
e

(145)

In order to make out a general table for the calculation of density from
the ~-derivatives, the preceding equation is rewritten as
2

22= 1-~7-1 (146)

() F
where -——1

S*-s;,i)
.

The functional relations between ~ and @ are given in table I for 7

equal to 1.4 and 4/3, respectively. Emm the given inlet condition and-
the given X values, the vari@ion of

(2 ~)-l(b~,i)-2

is “fkrst or table as a func-

tion of table is prepared
for the ~iation of
.—

as a function of
~2r2
~~2
~
(
FT2-X);’
- ). Anytime during the calculation, from the
—
2
value of — - at each poimt (in general, X changes during
\ 2X )
successive improvements between S1 and S2 surfaces),

— .
WA TN 2604 67

(,~) ‘%@ ‘2
is read from the first ~aph or table and is combined with

and the entropy factor to obtain $’. The value of Z is then read frcm
tables I or II. After the value of

is read from the second curve or table, the density ratio is obtained.

Hyperbolic Case

In the hyperbolic case, the main problem is the solution of the

followfng principal equation, written in a common form for the two kinds
of flow surface:

with the initial condition that ~ and,its normal derivative are given
on a curve which is not a characteristic curve. From equation (147),
the equation of the characteristic curve is

.(v*)2_2K@/)+L=o ( 148)

The slopes of the characteristic curves are

(149a)

.——
— ——-— —
— __.. -—. .—— ——. —
68 NACA TN 2604

(149b)

The coefficients J, K, L, and v and the independent variables

o and ~ for the eight cases considered are given in the table on the
following page. Using these values of J, K, and L, Al ~d A2 are
also expressed in terms of the velocity components. Except for cases 2,
5, and 8, they can also be expressed in the usual trigonometric form,
tan (Xkp). The values of X and p are also given.

Changes of ~-derivatives along characteristic curve. - When the

reference point on the q~-plane moves along the tige of the charac-
teristic curve in the q~-plane corresponding to a small change in ~,
d~, the change in q iS dq =$d~. Because of these two smaU
changes, the change of any quantity q on the s~face iS (fig. 12)

(150)

or
(151)

Hence along ~

(152)

(153)

l?romequations (152) and (153),

(154)

.2
(155)
70 NACA TN 2604

Substituting equations (154) and (155) into equation (147) gives

d aq ~~afl+ 182
(J.A12. 2KA1”+L) ~—
J~~+@-J*l)vd~aq v a~2 +M*+:%=O

(156)

By virtue of equations (149a) and (14%), eqution (156) becomes

(157b)

Starting from two points a and b a short distance apart on ~he

initial curve, equations (149a) and (149b) give the tangent to the
characteristic curves at these two points and eqpations (157a) and
(157b) give the new value of M/a~ and aV/aq at the point of inter-
section C of the two tangent lines (fig. l.1) . The auxiliary equa-
tions corresponding to the p&rticular problem are then used to deter-
mine other pertinent quantities at the point C. This process is to
be carried step-by-step downstream.

Changes of fluid velocity and direction along characteristic

curve. - When the characteristic curve hits the boundary wall, it is
more convenient to express equations (157a) and (157b) in terms of the
magnitude of the fluid velocity and the flow direction. In order to
do this, the definitions of $-derivatives are first put in a common
form for all cases as

(158)

(159)

where c equals 1 and r for the S1 and S2 surfaces, respectively.

By the use of equation (45),
vOv
NAC!ATN 2604
71

(160)

(161)
Substilzrktagequations (160) and (161) into equation (156) yields

(162)

Let

Wq=T’Tsinx
(163)
w< =Wcos)(

and

~2r2 ~2r2
h =1+ T-– ;(w~2+Wq2+ T?~2)=I+ T-- ; (W2 + WL2)

(164)
Where w~ iS equal to Wr, Wr, O, Wz, Wz, O, Wu, and Wu for cases 1
to 8, respectively. By the use of equations (163), (164), and (144),
equation (162) can be written

dlnb stiX dlnc

~+A2cosi -Sinx ~-% COSX*) +&*+
(

= o [165)

. ..———._ ———— —z . ———-

72 M.ACATN 2604

A similar expression can be obtained for the change in w and X

along ~ by replacing ~ by Al in the preceding equation. For
cases 1, 3, 4, 6, and 7, A can be written as tan (X & p), where p is
equal to sin-l ~,
a through purely trigonometric transformations as
fOllows:

* _ K ~~~ti
J=

. sinx cos Xksin I.L

cosI.L sin 2X A sin 2p
Cos2x- sin2 ~ = Cos 2x+ Cos 2~

=tan (X+I.L)
or tan (X-W)

For these cases, eqyation (165) can then be written (compare refer-
ence 30):

1
F

:M sinx -Ncosx=o
(166)
JL Cos x - sin X 1

where the minus and plus signs on the second term and subscripts 2
and 1 for A in the last two terms are used along characteristics AI
and A2, respectively. Equations (165) and {166) are most useful when
the characteristic hits the boundary wsll. For a direct problem, the
slope there is known from the given blade shape and for an inverse or
design problem, either the desired turning at the boundary or the
velocity on the boundary is prescribed. With either dY.or dw known,
dw or dx is evaluated from equation (162) or (165) (only one charac-
teristic equation is used at the wall). For convenience of setup in
ltACATN 2604 73

calculation, this system can also be used for interior points. Except
that more terms are involved in the present problem and that w takes
different meanings in different cases, the procedure of calculation is
very much the same as ordinary two-dimensional flow described in refer-
ences 32 and 30.

CONCUJDING REMARKS

A general theory of steady three-dimensional flow of a nonviscous

fluid in subsonic and supersonic turbomachines having arbitrary hub and
casing shapes and a finite number of thick blades is presented. The
solution of the three-dhensional direct and inverse problem is
obtained by investigating a combination of flows on relative stream
surfaces whose intersection with a z-plane either upstream of or some-
where inside the blade row form a circular arc or a radial line. The
equations obtained to describe the fluid flow on these stream surfaces
show clearly the several approximations involved in ordinary two-
dimensional treatments. They also lead to a solution of the three-
dhensional problem in;a mathematically two-dimensional manner through
an iterative process. The equation of contin~”ty is combined with the
equation of motion in either the tangential or the radial direction
through the use of a stream function defined on the surface, and the
resulting equation is chosen as the principal eqtition for such flows.
The character of this equation depends on the relative magnitude of
the local velocity of sound and a certain combination of velocity tom- .
ponents of the fluid. A general method to solve this equation by both
hand and machine computations when the equation is eUiptic or hyper-
bolic is described. The theory is applicable to both irrotational and
rotational absolute flow at the inlet of the blade row and to both
design and off-design operations.
\
{
Lewis Flight Propulsion Laboratory
National Advisory Committee for Aeronautics
Cleveland, Ohio, July 13, 1951

REFERENCES

1. Lorenz, H.: Theorie und Berechnung der Vollturbinen und fieisel-

pumpen. V.D.I. Zeitschr., Bd. 49, Nr. 41, Okt. 14, 1905, S.
1670-1675.

2. Bauersfeld, W.: Zuschrift an die Re&ktion. V.D.I. Zeitschr.,

Bd. 49, Nr. 49, Dez. 9, 1905, S. 2007-2008.

3. Stodola, A.: Steam and Gas Turbines. McGraw-IH.11Book Co., tic.,

1927. (Reprinted, Peter Smith (New York), 1945, pp. 990-992.)

4. v. Mises, R.: Theorie der Wasserr&der. Zeitschr. f. Math. und

~yS., Bd. 57, 1909, S. 1-120.

—.— -— - ——— —— .———-—- ——. –-

74 NACA TN 2604

5. Dreyf’us,L. A.: A Three-dimensional Theory of Turbine Flow and Its

Application to the Design of Wheel Vanes for Frsncis and Propeller
Turbines. Acts Polytechnicaj Mech. Eng. Ser., vol. 1, Nr. 1, 1946.

6. Ruden, P.: Investigation of Single Stage Axial Fans. NACA TM 1062,

1944.

7. Traupel, Walter: New Gener& Theory of Multistage Axial Flow Turbo-

machines. Navships 250-445-1, Navy Dept. (-S. by C. W. Smith,
Gen. Elec. Corp.)

8. Meyer, Richard: Beitrag zur Theorie feststehender Schaufelgitter.

Nr. U, Mitteilungen aus Inst. f. Aero. (Z&ich), 1946.

9. Ackeret, J.: Zum Entwurf dichtstehender Schatielgitter. Schweiz.

Bauztg., Bd. 120, Nr. 9, Aug. 29, 1942, S. 103-108.

10. Marble, Frenk E.: The Flow of a Perfect Fluid through an Axial
l?urbomachinewith Prescribed Blade Loading. Jour. Aero. Sci.,
VO1. 15, no. 8, Aug. 1948, pp. 473-485.

Il. Marble, Frank E.: Some Problems Concerning the Three-Dimensional

Flow in Axial Turbomachimes. I.A.S. Reprint No. 182, 1949.

12. Siestrunck, R., et Fabri, J.: ficoulementstourbillonnaires clans ‘

les machines axiales. Office National d’Etudes et de Recherches
Aeronautiques. Pub. 45, 1950.

13. m~ , W.: The Three-Dimensional Theory of Turbines and Pumps

for Incompressible Fluids. R.T.P. Trans. No. 1568, British
M.A.P. (From Forschung, Vol. 8, No. 1, Jan.-Feb. 1937,
pp. 29-34.)

14. Gravolos, Fausto G.: The Flow of an Inviscid, Incompressible Fluid

through a Turbo-Machine with Arbitrary Hub Shape. Special Tech.
Rep. No. 3, Rensselaer Poly. Inst., Aug. 17, 1948. (I?7-ONR
336-T.O. No. 1, NR 061-014.)

15. Wislicenus, George F.: Fluid Mechanics of Turbomachinery. McGraw-

Hil.1Book Co., hC., 1947.

16. Reissner, Hans: Blade Systems of Circular &rangement in SteadY~.

Compressible Flow. R. Courant Anniversary Volume, Interscience
Pub., tiC., 1948, pp. 307-327.

17. Wu, Chung-Hua, and WoHenstein, Lincoln: Application of Radial-

Eq@librium Condition to Axial-Flow Compressor and Turbine Design.
NACA Rep. 955, 1950. (Formerly NACA TN 1795.)

——
 NACA TN 2604 75

.
18. Wu, Chung-Wa: A General Through-Flow Theory of Fluid Flow with
Subsonic or Supersonic Velocity in Turbomachines of Arbitrary Hub
and Casing Shapes. NACA TN 2302, 1951.

19. Goldstein, Arthur W.: Axisymmetric Supersonic Flow in Rotating

Ihpellers. NACA TN 2388, 1951.

20. Courant, R., and Friedrichs, K. O.: Supersonic Flow and Shock
Waves. lnterscience Pub., Inc., 1948.

21. Vazsonyi, Andrew: On Rotational Gas Flows. Qwrt. Appl. Math.,

vol. III, no. 1, April 1945, pp. 29-37.

22. Hicks, B. L., Guenther, P. E., and Wasseman, R. H.: New Formul-
ationsof the Eqpations for Compressible Flow. Quart. Appl. Math.,
vol. V, no. 3, Oct. 1947, pp. 357-360.

23. Ferrari, C.: Determination of the Pressure Exerted on Solid Bodies

of Revolution with Pointed Noses Placed Obliquely in a Stream of
Compressible Fluid at Supersonic Velocity. R.T.P. Trans. No.
1105, British M.A.P.
.
24. Moeckel, W. E.: Use of Characteristic Surfaces for unsymmetrical
Supersonic Flow Problems. NACATN 1849, 1949.

25. Southwell, R. V.: Relaxation Methods in Theoretical Physics.

Clarendon fiess (Oxford), 1946.

26. WU, Chung-Hua: Formulas and Tables of Coefficients for Numerical

Differentiation with Function Values Given at UnequaUy Spaced
Points and Application to the Solution of Partial Differential
Equations. NACATN 2214, 1950.

27. Weinig, F.: Die Str&ung um die Schaufeln von Turbomaschinen.

Johann Ambrosias Barth (Leipzig), 1935.

28. Keller, C.: Axialgebl&e mm Standpunlrtder Tragfl&eltheorie.

Gebr. Leeman & Co. (Zurich), 1934.

29. Wu, Chung-Hua, and Brown, Curtis A.: Method of An&lysis for Com-
pressible Flow Past Arbitrary Turbomachine Blades on General
Surface of Revolution. NACA TN 2407, 1951.

30. Wu, Chung-Hua, and Costilow, Eleanor L.: AMetho3 of Solving the
Direct and Inverse Problem of Supersonic Flow along Arbitrary
Stream Filaments of Revolution in ‘lurbomachines. NACA TN 2492,
1951.

—.— —.—- .-— .—.—. — . . . .. .

76 NACA TN 2604

31. Webster, Arthur Gordon: Partial Differential Equations of Mathe-

matical Physics, Samuel J. Plimpton, cd., G. E. Stechert & Co.,
2d cd., 1933, Pp. 239-241.

32. Ferri, Antonio: Elements of Aerodynamics of Supersonic Flows.

The MacmilJan Co., 1949.

33. Enmmns, HowardW.G The Numerical Solution of Compressible Fluid

FIo~?Problems. NACATN 932, 1944.

34. Ince, E. L.: Ordinary Differential Equations. Dover Pub., 1944.

35. Wu, Chung-Hua, and Brown, Curtis A.: AMethod of Designing Turbo-
machine Blades with a Desirable Thickness Distribution for Com-
pressible Flow along an Arbitrary Stream Filament of Revoluti~.
NACATN 2455, 1951.

36. Stanitz, John D.I Two-Dimensional Compressible Flow in Turbo-

machines with Conic Flow Surfaces. NACA Rep. 935, 1949. .
(Formerly NACA TN 1744.)

37. Squire, H. B., and Winter, K. G.: The Secondary Flow in a Cascade
of Airfoils in a Nonuniform Stresm. Jour. Aero. Sci., vol. 18,
no. 4, April 1951, pp. 271-277.

.38.Hawthorne, Willism R.: Secondary Circulation in Fluid Flow. Gas.

Turbine Lab., M.I.T., May 1950.

39. WU, Chung-Hua, and Brown, Curtis A.: A Theory of the Direct and
Inverse Problems of Compressible Flow Past Cascade of Arbitrary
Airfoils. I.A.S. Preprint No. 326, Jan. 1951. ~

40. Hamrick, Joseph T., Ginsburg, Anbrose, and Osborn, Walter M.:
Method of Analysis for Compressible Flow throu@ Mixed-Flow
Centrifugal hpellers of Arbitrary Design.” NACA TN 2165, 1950.
..
41. Wu, Chung-Hua~ Sinnette, John T., Jr., and Forrette, Robert E.:
Theoretical Effe@”of Inlet Hub-Tip-Ratius Ratio and Design
Specific Mass F@w on Design Performance of Axial-Flow Cmu-
pressers. NACATN2068, 1950.

42. WU, Chung-Hua:-“A Gener&l Theory of Three-Dimensional Flow with

Subsonic and Supersonic Velocity ’inTurbomachines Having Arbitrary
Hub and Casing Shapes. Part II. Paper No: SO-A-79, presented at
the Ann. Meeting of A.S.M.E. (New York), Nov. 27-Dee. 1, 1950.
.
43. Anon.: Tables of Lagrangi.anInterpolation Coefficients. Columibia
Univ. Press., 1944.
.
 NACA TN 2604 77

44. Bickley, W. G.: . Formulae for Numerical Differentiation. The Math.

Gazette (London), vol. XXV, no. 263, 1941, pp. 19-27.

45. Ehmons, Howard W.”:The Theoretical,Flow of a Frictionless, Adia-

batic, Perfect Gas Inside of a Two-Dimensional Hyperbolic Nozzle.
NACATN 1003, 1946.

46. FOX, L.: Some Iinprovementsin the Use of Relaxation Methods for
the Solution of-Ortinary and Partial Differential Equations.
Proc. Roy. Sot. (London), vol. 190, no. A1020, ser, A, June
1947, pp. 31-59.

47. Keenan, Joseph H., and Kaye, Joseph: Gas Tables. .JohnWiley& Sons,
hlC., 1948.

— ..———— .-.—.—- —
78 NACA TN 2604

~
TABIXI - m62RALDm6mY!cAmE

(a) ; -1.4

=s=’
%
0.0011.oQo5007
.002
.003
.004
.OQ5

0.006
$
1.001@327
1.001S361
1.0020109
1.0025171

1.0030246
z
0.99949955
+ .99899630
.99849616
.99799314
.99746922

0.99698452 I
.057
.058
.059
.060
1.0309453
1.0315371
1.0321307
1.0327263 7
1
z
0.97053890
.96996357
.96942708
.96886954
.96831077

0.96773086
Scp

0.111
.112
.112
.114
.I.u m E
1.06595=
1.0666704
1.0673882
1.0681087
1.0688323

0.11(1.0695582
z
0.93812555
.93749672
.93686627
.93623430
.93560073

0.93496548

-1--1
.007 1.0035335 .99647894 .96718963 .lli1.0702863 .93432857
.008 1.0040438 .99597249 .96662757 .llc1.0710192 .93369008
.009 1.0045555 .99546516 .96606419 .~: 1.0717540 .93304993
.010 1.0050687 .99495686 .065 1.0357332 .965499611 .la 1.0724918 .93240806

0.o11 1.0055833 0.99444770 0.066 1.0363404 .96493392 0.121 1.0732326 0.93176447

.012 1.0060993 .99393768 .067 1.0369496 .96436702 .122 1.0739763 093111924
.013 1.0066168 .99342669 .068 1.0375609 .96379885 .122 1.0747230 .93047232
.014 1.oo71357 .99291466 .069 1.0381742 .96322948 .124 1.0754728 .92982361
.015 1.0076561 .99240207 .070 1.0387695 .96265894 .1= 1.0762256 .92917321
0.016 1.0081779 0.99188M4 3.071 1.0394069 40.96208713 0.126 1.0769815 0.92852106
.o17 1.0087012 .99137386 .072 1.0400263 .96151415 .127 1.0777405 .92786714
.018 1.0092260 .99085634 .073 1.0406478 .96093991 .12e 1.0785026 .9272U49
.019 1.0097522 .99034199 .074 1.0412714 .96036442 .12s +
1.0792678 .92655410
.020 1.O1O26W .98982460 .075 1.0418971 .95978768 .13C
I

=0.98930629

7
O.of?l 1.0108093 3.076 1.0425249 0.95920970 0.131
.022 1.o113401 .98878706 .077 1.0431548 .95863049 .132
.023 1.03.3.8724 .98826690 .078 1.0437869 .95804996 .133
.024 1.0124063 .98774573 .079 1.04A4212 .95746812 .134 1.0831424 .92323964
.025 1.0129417 .98722365 .080 1.04!33576 .9568831361 .135 1.0639272 I .92257118
1 1

0.026 1.0134786 0.98670066 ).081 1.0456962 0.95630069 .136 1.0647153 .92190089

.027 1.0140171 .98617666 .082 1.0463370 .955nm3 .137 1.0855068 .92122868
.028 1.0145572 .98565167 .083 1.0469W .95512608 .130 1.0863017 .9205%58
.0’29 1.0150989 .98512568 .084 1.0476253 .95453976 .139 1.08710cm .91987858
.0s0 1.0156421 .98459881 .085 1.0482729 .953954206 .140 1.0879018 .91220061
1
0.031 1.0161869 0.98407094 3.086 1.0489227 0.95335910 0.141 1.0887071 0.91852069
.032 1.0167333 .96354210 -t-.087 4..0495748 --i.95276678 .142 1.0895160 .91783875
.033 1.0172814 .98301!ZL7 .088 1.0502292 .9521731.1 .143 1.0803284 .91715487
.034 1.01765U :98248128 .089 1.0508859 .951.57609 .144 1.0911444 .91646898
.035 1.0163825 .98194932 .090 1.0513449 .95098174 .145 1.0819640 .91578111

0.036 1.0189354 0.98141649 ).0911 1.0522063 0.146 1.0827873 0.91509116

.037 1.0194900 .98088260 .092 1.0528701 .94978478 .147 1.0936143 .91439916
.038 1.0200462 .98034775 .093 1.0535362 .949KM28 .148 1:0944450 .91370512
.039 1.0206042 .97981176 .094 1.0542048 .94858228 .149

H
.040 1.021.1638 .97927482 .095 1.0548758 .94797890 .150
I

0.041 1.0217252 0.97873675 ).096 1.0555492 0.94737413 0.151

.042 1.0222882 .97819773 .097 1.0562251 .94676788 .152
.043 1.0228530 .97765759 .098 1.0569034 .94616026

I
.153 1.0986557 .91020326
.044 1.0234194 .977U652 .099 1.0575842 .94555119 .154 1.0995095 .90949646
.045 1.0239876 .97657433 .100 1.0582675 .94494067 .155 1.1003674 .90878737
I
0.046 1.0245575 0.97603112 ).101 1.0569534 0.9443286 0.156 1.1012282 0.90807618

-1-
.047 1.0251282 .97348680 .102 1.0596418 .9437K13 .157 1.1020950 .90738280
.048 1.0257027 .97494137 .103 1.0603327 .9431002 .138 L.102965O .90664708
.049 1.0262780 .97439465 .104 1.0610263 .94248371 .158 L.1038391 .90592613
.050 1.0268550 .105 1.0617225 .94186569{ .160 L.1047174 .90520888

4
0.051 1.0274338 0.97329872 ).106 1.062421.3 0.943.24619 3.161 L.1055999 0.90448633
.052 1.0280145 .97274893 .107 1.0631227 .94062520 .162 L.1064866 .90376151
.053 1.0285970 .9721.9805 .108 1.0638268 .94MI0264 .163 L.1073777 .90303426

S&&l
.0s 1.0291813 .108 1.0645336 .9393785 .164 L.1082731 .90230468
.055 1.0297674 .110 1.0652431 .93875285 .165 L.109173O .90157261
llV
HACA TN 2604 79

TABLE I- unKRALmm3rrl mBIE-ccmtinwd

“

(a) ; = 1.4- @mcllnid

-&
X2-1-S
() X2

~
* z 517 z 5q
4 z

T
z
).166 L.1100772 0.90063825 0.221 L.1683271 ).85592468 ).276 1.2556050 0.79642881
.167 L.1109859 .9W10143 .222 L.1695813 .B5500683 .277 L.2576894 .79510887
.168 L.1118991 .899362M3 .223 1.1708M4 .85408445 .278 1.2598004 .79377654
.169 L.1128170 .89862035 .224 1.172U64 .85315759 .279 L.2619389 .79243139
.170 1.1137394 .89787611 .225 L.1733975 .85222612 .280 L.2641055 .791073!ZL

1
).171 1.1146665 0.89712932 0.226 1.1746880 3.85128987 ).281 L.26630f2 0.78970153
.172 1.I.I.55983 .89638W3 .227 1.1759880 .85034881 .282 1.2685268 .78831602
.173 L.1165348 .89562815 .228 L.1772974 .84940305 .203 L.2707833 .78691623
.174 1.1174762 .89487364 .229 L.1786165 .84845240 .284 L.2730n5 .78550184
.175 1.1184225 .89411649 .230 1.1799455 .84749677 .285 L.2753927 .78407223
).176 1.1.193737 0.89335670 0.231 L.1812846 3.84653605 3.286
.177 1.1203298 .89259431 .232 1.1826339 .845570ZL .287
.178 1.12H91O .89182915 .233 L.1839935 .84459923 .288 1.2825638 .77968831
.179 1.1222573 .89106126 .234 L.1853636 .84362300 .289 1.2850276 .77819340
.180 1.1232287 .89028064 .235 L.1867445 .84264136 .230 1.2875299 .77668099
. ).181 1.1242053 0.88951724 0.236 1.1881362 0.84165435 D.291 1.2800725 0.775M023
.182 1.1251871 .88874108 .237 1.1895390 .84066180 .292 1.2826566 .77360066
.183 1.1261742 .88796209 .238 1.1908530 .83966370 .=3 1.2952840 .77203146
.184 1.1271667 .887M022 .239 1.1923785 .63865987 .294 1.2979563 .77044196
.185 1.1281647 .88639540 .240 1.1938156 .83765030 .295 1.3006755 .76883127
I

T
).186 1.1291681 0.88560773 0.241 1.1952645 0.83663490 0.286 1.3034429 0.76719893
.187 1.1301771 .88481708 .242 1.1967255 .83561351 .297 1.3062609 .76554385
.188 1.1311917 .88402346 .243 1.1981988 .83458605 .298 1.3091314 .76386526
.189 1.1322120 .88322682 .244 1.1996846 .83355242 .299 1.3120571 .76216195
.190 1.1332381 .88242709 .245 1.XKKL631 .83251255 .300 1.3150400 .76043314

).191 1.1342699 0.88162438 0.246 1.2026946 0.83146628 0.301 1.3180832 0.75867745

.192 1.1353077 .88081848 .247 1.2042193 .83041353 .S02 1.3~1890 .75689398
.193 1.1363514 .88CO0948 .248 1.2057574 .82935423 .303 1.3243612 .75508102
.194 1.1374012 .87919724 .249 1.2073092 .82828823 .304 1.32760ZVL , .75323774
.195 1.1384571 .87838180 .250 1.2088750 .82721539 .305 1.3309163 .75136205
I
).196 1.1s95191 0.87756318 0.251 1.2104551 0.82613556 0.306 1.3343067 0.74945288
.197 1.1405873 .87674131 .252 1.2120497 .82504868 .307 1.3377787 .74750779
.198 1.1416619 .87591607 .253 1.21.36591 .82395460 .308 1.3413356 .74552558
.199 1.1427428 .87508748 .254 1.2152836 .82285323 .309 1.3449840 .74350327
.200 1.1438304 .87425W3 .255 1.21.69235 .823.74434 .310 163487277 .74143951

2.201 1.1449245 0.87342004 0.256 1.2185792 0.82062763 0.311 1.3525748 0.73933065

4
.202 1.1460252 .87258116 .257 1.2202510 .81950353 .312 1.3565299 .73n7505
.203 1.1471327 .87173873 .258 1.2219392 .8183n32 .313 1.3606032 .73496814
.204 1.1482470 .87089276 .259 1.2236442 .81723102 .314 1.3648001 .73270804
.205 1.1493681 .87004328 .81608251 .~~: 1.3691338 .7303888C
.260 1.2253663
0.206 1.1504963 0.86919011 0.261 1.2271059 0.81492559 0.316
.207 1.1516316 .86833324 .262 1.2288634 .81376010 .317

I .7204347:
.208 1.1527741 .86747265 .263 1.2306393 .81258578 .31&
.209 1.1539238 .86660835 .264 1.2324338 .81140261 .31$ 1.3880508
.210 1.1550810 .86574015 .265 1.2342475 .81021027 .32C 1.3932489 .ZL774684

0.211 1.1562457 0.86486808 0.266 1.2360808 0.80900860 0.321 1.3986833 o.n4958u

.212 1.1574179 .86399217 .267 1.2379342 .80779738 .32i 1.4043617 .71206727
.863L1.228

I
.213 1.1585978 .26@ 1.2398081 .80657644 .3’2 1.4103403 .7090487:
.214 1.1597855 .86222840 .269 1.2417032 .80534543 .324 1.4166227 .7059042E
.215 1.1609811 .86134046 .27C 1.2436198 .80410428 .32: 1.4233067 .7025892t
1

0.216 1.16 ZL847 0.860M843 0.80285263 0.32[ 1.4303770 !0.699U63$

-1
0.271 1.2455586
.217 1.1633964 .65955226 .272 1.2475200 .80159036 .32
.218 1.1646164 .85865183 .2Z 1.2495048 .80031705 .32{
.219 1.1658447 .85774718 .274 :.X151~; .7990326C .3E
.220 1.1670816 .85683812 .27: .7077366C .33(
.

--~—————”–
80 NACA TN 2604

m 6(p
i
0.001 1.0005006 0.99949965 0.056 1.0303226 0.”97056900 O.1.11 1.0657881 0.93827281
.002 1.0010026 .99899840 .057 1.0309112 .97001565 .112 1.0664991 .93764730
.003 1.0015060 .99649626 .058 1.0315016 .96946044 .113 1.0672128 .93702025
.004 1.0020107 .99799333 .059 1.0320939 .96890409 .114 1.0679283 .93639158
.005 1.0025168 .99748952 .060 1.0326880 .96834868 .KL5 1.0686486 i .93576130
0.006 1.0030242 0.99698492 0.061 1.0332840 0.96778814 0.116 1.0693706 0.93512951
.007 1.0035330 .996.47944 .062 1.0338819 .96722646 .I.I.7 1.0700954 .93449612
.008 1.0040432 .99s97308 .063 1.0244817 .96666766 .lle 1.0708230 .93386115
.009 1.04345548 .99s46585 .064 1.0350835 .96610563 .119 1.0715534 .93322461
.G1O 1.00S678 .99495775 .065 1.0356872 .965%249 .120 1.0722867 .93258641
I
O.o11 1.0055822 0.99444879 D.066 1.0362928 0.96497824 0.121 1.0730223 D.93194656
.012 1.0Q60980 .99393896 .067 1.0369003 .96441288 .122 1.0737620 .93130508
.013 1.0066152 .99342827 .068 1.0375089 .96384622 .123 1.0745040 .93066196
.014 1.c071339 .99291663 .069 1.0381215 .96327838 .124 1.0752489 .93001723
.015 1.0076541 .99240404 .070 1.0387350 .96270845 .125 1.0759968 .92837079

0.016
.o17
.018
.019
.020

0.021
.022
.023
1.0061756
1.0086985
1.ca92230
1.0087490
1.0102764

1.0108053
1.01H357
1.0118676
0.99189070
.99137651
.99086123
.99034513
.98982813

0.98931021
.98879136
.98827159
2.071
.072
.073
.074
.075

).076
.077
.078
1.0393506
1.0399681
1.0405876
1.0412093
1.0418331

1.0424509
1.0430867
1.0437167
0.96213924
.96156796
.96099550
.96042169
.95984664

0.95927043
.95869308
.95811440
0.126
.127
.128
.128
.130

0.131
.132
.133
1.0767477
1.0775017
1.0782587
1.0790187
1.0797818

1.0805481
1.0813175
1.0820901
7
3.92872267
.92807278
.9274ZL22
.926768CQ
.92611304

).92346261
.92479776
.92413746
.024 1.0124010 .98775090 .079 1.0443488 .95753449 .134 1.0828659 .92347538
.025 1.0128358 .98722830 .080 1.0449830 .95695337 .135 1.0836449 .92281152

0.026 1.0134723 0.98670679 ).081 1.0456194 0.95637093 0.136 L.0844271 ).92214590

.027 1.0140104 .98618318 .082 1.0462580 .95578720 .137 1.0852126 .921.47843
.028 1.0145499 .98565876 .083 1.0468987 .95520226 .138 1.0860015 .92080904
.028 1.0150910 .98513335 .064 1.0475416 .9!361603 .139 1.0867937 .92013783
.030 1.0156336 .98460705 .085 1.0481868 .95402842 .140 1.0875693 +.91946473
0.031 1.0161778 .98407975 ).086 1.0488341 0.95343963 0.141 1.0883882 ).91878982
.032 1.0167236 .96355148 .087 1.0494837 + .95264948 .142 1.0891906 .91811285
.033 1.O17Z71O .98302222 .088 1.0501355 .95225607 .143 1.0899965 .91743414
.034 1.0178200 :98249189 .089 1.0507896 .95166530 .144 L.0908058 .91675347
.035 1.0183707 .98196069 .090 1.0514460 .95107I29 .145 1.0916187

T
).036 1.0189229 0.98142863 ).091 1.0521047 0.95047575 0.146 L.0924351 ).91536619
.037 L.0194767 .98089539 .092 1.0527657 .94987897 .147 1.0932551 .91469960
.038 1.0200322 .980361ZL .093 1.0534280 .94926087 .148 1.0940787 .91401103
.039 1.0205894 .97982597 .094 1.0540947 .94868137 .149 L.0849059 .91332050
.040 L.02U482 .97928978 .085 1.0547628 .9480K146 .150 1.0957368 .91262792
I
).041 L.0217087 0.97875255 ).096 1.0554332 0.94747825 0.151 1.0965714 ).91193332
.042 1.0222708 .97821438 .097 1.0561061 .94687456 .152 1.0974098 .91123662
.043 1.0228346 .97767518 .098 1.0567813 .94626958 .153 L.0982520 .91053704
.044 1.0234001 .97713494 .089 1.0574580 .94566314 .154 L.0890979 .90983706
.045 1.0239673 .97659369 .10Q 1.0581391 .94505533 .155 L.0999477 .90913413
I
).046 1.0245363 0.97605131 ).101 1.0588217 0.94444608 0.156 L.1(2138014 1.90842808
.047 L.O251O7O .97550792 .102 1.0595068 .94383538 .157 L.1016591 .907721.82
.048 1.0256794 .97496352 .103 1.0601944 .94322324 .158 L.1025207 .90701245
.049 1.0262336 .97441802 .104 1.0608846 .94260959 .159 1.1033863 .90630090
.050 1.0268285 .97387151 .105 1.0615773 .94199452 .160 L.1042559 .90558719

A
).051 L.0274072 0.97332392 ).106 1.0622726 0.161 L.1051286 1.90487125
.052 1.0279867 .97277523 .1o7 1.0629704 .162 L.106(K)74 .90416808
.97222556 I

A
.053 1.0285679 .108 1.0636702 .94014041 .163 ..1068894 .9034S263
.054 1.0291510 .971674n .109 1.0643740 .93951938 .164 L.1077756 .90270890
* L.0287359 .9ZL122791 .110 1.0650797 .93889687 .165 ..1086660 .90198491

———__
 NACA TN 2604 81

,.

TABI.EI -

. (b) F . 4/5 - Concluded

1
Q--l w
617 2.
T

F
0.166 1.1085607 0.90125759 0.221 1.166928S 0.85695068 0.276 1.2514571 0.79906854
.167 1.1104597 .90052795 .222 1.1681568 .85604946 .277 1.2534457 .79780081
.168 1.1113630 .89979602 .223 1.1693937 .85514399 .278 1.2554579 .79652213
.169 1.1122708 .89906163 .224 1.1706391 .8%23424 .279 1.2574944 .7952321.7
.170 1.1151830 .89832489 .225 1.1718931 .85332015 .280 1.2595556 .79393081

0.171 1.1140997 0.89758574 0.226 1.1731558 0.85240170 0.281 1.2616424 0.79261762

.172 1.11502J.O .89684410 .227 1.1744273 .85147884 .282 1.2637555 .79129230
.173 1.1159468 .89610006 .228 1.1757079 .85055140 .283 1.2658957 .78995450
.174 1.1168773 .89535350 .222 1.1769977 .84961933 .284 1.2680636 .78860398
.175 1.1178125 .89460442 .230 1.1782966 .84868275 .285 1.2702602 .78724028

0.176 1.1187524 0.89385283 0.231 1.1796049 084774148 0.286 1.2724861 0.78586320

.177 1.1196971 .89309868 .232 1.1808228 .84679540 .287 1.2747424 .78447222
.178 1.1206466 .89234197 .233 1.1822504 .845844% .288 1.2770289 .76306702
.179 1.1216010 .89158266 .234 1.M35879 .84488866 .289 1.2793498 .78164705
.180 1.1225603 + .89082Q74 .235 1.1849354 .84392786 .290 1.2817028 .7802U207

0.181 1.1235246 0.89005617 0.236 1.1862830 0.84296207 0.291 1.2840902 0.77876149

.182 1.1244940 .88928887 .237 1.1876610 .84199111 .292 1.2865130 .77729490

41
.183 1.1254685 .88853.887 .238 1..1890394 .84101502 .283 1.2889726 .77581168
.184 1.1264481 .88774618 .239 1.1904285 .84003365 .294 1.2814700 .77431144
.185 1.1274329 .88697075 .240 1.1918285 .63904689 .295 1.2940068 .77279347

0.186 1.1264230 0.88619250 0.241 1.1s32395 0.6380%72 0.296 1.2965843 0.77125722

.187 1.1294184 .88541146 .242 1.1946617 .83705705 .297 1.2392041 .76970200
.188 1.1504192 .88462758 .243 1.1960954 .83605371 .298 1.3018675 .768%2732
.189 1.1314255 .88384078 .244 1.1975406 .83504.476 .289 1.3045765 .76653228
.190 1.1324372 .88305117 .245 1.1989976 ,.83403@33 .300 1.3073325 .76491635
4
0.191 1.1334545 0.88225862 0.246 1.2004666 0.83300943 0.301 1.3101379 0.76327843
.192 1.1344774 .88146313 .247 1.2019479 .83198282 .302 1.3129943 .76161793
.193 1.1355061 .88066458 .248 1.2034416 .83095017 .303 1.3159043 .75993369
.194 1.1365405 .87986306 .249 1.2049480 .82991133 .304 1.3188697 .75822502
.195 1.1375807 .87905851 .250 1.2064673 .82886623 .305 1.321.8935 .75649059

0.196 1.1386269 0.87825081 0.2511.2079997 0.82781478 0.306 1.3249779 0.75472957

-1
.197 1.1396790 .87744035 .2521.20%5455 .82675663 .307 1.3281263 .75294044
.198 1.1407372 .87662610 .253 1. 2H1OW .82569224 .308 1.3313412 .75112225
.199 1.1418015 .87580897 .254 1.2126783 .82462101 .309 1.3346266 .74927324
.200 1.1428719 .87498870 .255 1.21426S .82354292 .310 1.3379854 .74739231

7
0.201 1.1439486 0.87416515 0.256 1.=58677 0.82245790 0.311 1.3414224 0.74547734

--l
.202 1.1450317 .87333827 .257 1.21748U .821.36576 .312 1.3449408 .74352715
.203 1.1461212 .87250807 .258 1.2191161 .82026642 .313 1.3485465 .74153913
.204 1.1472172 .87167452 .259 1.2207632 .81915969 .314 1.3522432 .73951195
.205 1.1483197 .87063762 .260 1.2224259 .81804549
I
} .315 1.3560384 .73744224

0.206 1.1494289 0.86999727 0.261 1.2241045 0.81692372 0.316 1.3599361 0.73532867

.207 1.1505449 .86915339 .262 1.2257994 .81579417 .317 1.3639457 .73316702

1
.208 1.1516677 .86830602 .263 1.2275U.O .81465665 .318 1.3680720 .73095568
.209 1.1527975 .86745504 .264 1.2292396 .81351105 .319 1.3723269 .72868935
.210 1.1539342 .86660054 .265 1.2309855 .81235725 .320 1.3767161 .726S6617

7 T
0.21.1 1.1550780 0.86574240 0.266 1.2327492 0.81119501 0.321 1.3812551 0.72397923
.212 1.1562280 .86486057 .267 1.2345310 .81002421 .322 1.3859506 .721.52644
.213 1.1573873 .86401501 .268 1.2363314 .80884462 .323 1.3908236 .71899844
.21’4 1.1585530 .86314567 .269 1.2381508 .80765606 .324 1.3958818 .71639304
.2J.5 1.1597262 .86227249 .270 1.2399896 .806458361 .325 1.4011550 .71369691

0.216 1.1609069 0.86139552 0.271 0.80525133 0.326 1.4066513 0.71090824

--l
1.2418463
.217 1.1620953 .86051462 .272 1.2437273 .80403478 .327 1.4124156 .70800691
.218 1.1632916 .85962969 .273 1.2456273 .80280835 .328 1.4184546 .70499260
.219 1.1644958 .05874075 .274 1.2475485 .80157204 .329 1.4246420 .70183220
.220 1.1657060 .85784776 .275 1.2494916 .80032551 .330 1.4315749 .69853139

—z —.
.. . ..—
82 NACA TN 2604

...— .

--~

....,’” :> :.:

. . ; .. . . ...

/
\,,v,, /

Figure 1. - Relative dream surface S1. .

 z

.,
Figure 2. - Relative S* time ~.
84

Figure 5. - Intersect@ S1 and S2 surfmes in a blade rm.

,.,
NACA TN 2604 85

STATOR ROTOR
r
t

---

Figure 4. - Mean stream surfaces for -al-f low gas turbtie.

_ —.. - _ ._ .—. — ———— -.

_. —.— —
 NACA TN 2604

INLET GUIDE VANE ROTOR STATOR

r r

,t

w,

,h
I w,

—
12V
NW2A TN 2604
87
.

. ..

(e) ~tfShMf3ckge
*ed-ti @peuer.

A=ial-, radial-
, and UL&Sd-flm turbcqnactis .

—— _. —._ _.
—— -—— — --— —— ..__ .... .
88

dmz)
pwz +
aZ.

T+a~

Pwu

/1
I
!3Wr I
1 ,

(a) ~ surface with q and z as (b) ~ surface with r and q ae

Independent variables. independent variables.

a(Pwr)
Pwr + *

T+?

pwz

j_pwrj

{c)% surface with r and z as

@pend.ent variables.

Figure 7. - Elements of streem sheet.

N4CA TN 2604 89

(1)

-1
1
=5=
Figure 8. - Orthogonalcoordinatesfor surfaceof revolution.

_ .—— — —— —- .—-. —. - ———- .—

90 NAC!ATN 2604

Figure 9. - Relation between mean stream surface and

blade stiaces.
NACA TN 2604
91

Figure 10. - Grid ayatem and boundary condltims. for general ~ surface
(elliptic case).

.—.
.— —._ —._._ ____ . . . _ . . —. ____
 .
2 ‘1
n-r

I I
1
I

I
I

\l
11 2

c-z
Q
co

~
*
.&i.3Egstmd~ o.mditima for gmmral S2 emrte.ca (elliptlo case).
Figure Il.

.-
.-
rr

/
/ -“
$ /“
\ /

.—— ——— . 1
. I ---
/
/

‘w
/
/
0

=s=
/.

Figure 12. - Characteristics-ystemfor hyperboliccase.

NACA -Iangley Field, V&

_—. . ._ — --- —————— –—-