A Single-Parameter Theory of Vortex Turbo-Machines Flow in
A Single-Parameter Theory of Vortex Turbo-Machines Flow in
A Single-Parameter Theory of Vortex Turbo-Machines Flow in
, " :~,
MINISTRY OF AVIATION
1. Introduction.
In the design of an axial-flow compressor or turbine it is usual to proceed in three steps. First the
overall conditions for each stage are determined by specifying the axial and radial velocities of the
flow at some design radius. Then the radial variations of these velocities are considered, and finally
blading is specified which will realize the desired angles of flow. This paper is concerned with the
second step of this process. The flow is therefore considered to be axi-symmetric, and as is usual the
effects of viscosity and turbulence will be neglected. Consequently the theory will be in error near
the inside and outside walls of the machine, but may be expected to hold over a substantial proportion
of the blade height.
A large number of methods has been proposed for calculating these flows, so that it may be useful
to enumerate the criteria by which it is suggested that these methods should be judged. Firstly the
method must be sufficiently accurate.' The accuracy required is not high, since the changes in axial
velocity resulting from the three-dimensional effects which are to be calculated correspond to
changes in the flow angle of a few degrees only. Discrepancies of 10% in the theory will not
therefore give rise to measurable effects. The Second criterion is that results should be easily
obtainable. The advent of digital computers means that volume of arithmetic is no longer the primary
consideration, and ease of programming becomes more important. Thirdly, the theory should be
conceptually simple, so that the qualitative effects of changes in t h e design can be easily visualized.
The simplest type of theory assumes that the flow is in 'radial equilibrium'i It is now recognized
t h a t except for certain special designs this is very inaccurate.
3
(87503) A2
Profiles for various values of the parameter A are shown in Fig. 5. At the design radius, R, the axial
velocity is always V~, and the shape of the curve at this point is controlled by the parameter A.
This of course is far from being the only family of profiles which satisfies the required conditions,
but it has been chosen both for algebraical simplicity and since it gives profiles which are similar
to those used in practice.
The design radius R is chosen as the root mean square of the hub and tip radii,
R e = ½(te + he). (2)
= 7 r p V n ( # - h e)
from equations (1) and (2). This is independent of the parameter A {the choice of design radius in
equation (2) has been made to this end}. Since it is being assumed that p, t, and h are constants,
this shows that VR is a constant throughout the length of the machine, and is determined by the
mass flow. The problem of finding the flow in the machine is therefore reduced to finding the way
in which A varies along the length of the machine.
The continuity equation allows a stream function ¢ to be defined so that
a¢
rv = - (3)
and
a¢
ru-- ax" (4)
Substituting for v from equation il) in equation (3) gives
Or
Integrating this with respect to r, and choosing the constant of integration so that ~b is zero at the
hub gives
¢ = _ ½v (,,e_he) + g g e (te--re)(re--he)" (5)
If this is now substituted into equation (4) an expression for the radial velocity is obtained as
VR d?t
u = 8rR~ (# - r 2) (r ~ - h e) dx" (6)
The stream function may also be used to compute the radial displacement of the streamlines.
It will be assumed in the theory that this radial displacement is small. Consider in particular a
streamline which lies at the design radius at an axial station where the axial-velocity profile is
specified by A (point A in Fig. 4). Farther downstream the streamline lies at B where the radius
is (R + 3R), and A has increased to (A + 3A). For the point D, which is the point on the design radius
at the second axial station, the stream function is given by equation (5) as
VR
CD = CA + ~ (te-Re)(R~-he) ~A.
Also from equation (3) for the second station
o~
¢~ = CD + U~ , R = ¢ ~ - R Y d e R .
But since A and B are on the same streamline ~bA = CB, and eliminating the stream functions gives
1
~R = ~ ( # - R ~) (R2-hZ)3A.
This may be written
a2~ 2
~R = W 3~, (7)
where k is a constant which depends mainly on the aspect ratio of the blades and is given by
k- (t~-h=) (8)
4~/2 Ra"
Equation (7) shows that, for small displacements, the radial displacement of a streamline near the
design radius is directly in proportion to the change of A, and this is probably the easiest way to
visualize A. )t is also simply related to the slope of the axial-velocity profile at the design radius,
since differentiation of equation (1) gives at the design radius
R(0v)
a= ~ ~ ~=. (9)
p 3r 3r
)
1 0P=_0 P+½u s +v(1+/3~')Tr + v~l~r" (13)
a = pvR--- (is)
This is the basic differential equation for the variation of A within a blade passage. The quantities
/3n, (O/3/ar)n, and W R / V R are presumed to be known functions of x, since they are determined by
the blade geometry.
It will also be shown that provided the displacement of the streamlines is small, G is a constant
in any one blade passage. For, considering the two streamlines AB and CD in Fig. 4 which are
always close to the design radius, the change in stream function 8¢ from one line to the other is
constarit and, at the design radius, the axial velocity VR is constant, so that from equation (3) the
distance between the two streamlines 3R is also constant. For flow without loss through stators P is
constant along a streamline. The corresponding condition for rotors is that ( P + ½pf~2re) is constant
along a streamline. Now" since the streamlines AB and CD are deflected by the same distance, the
term ½pg2Zre will change by the same amount for each of them, and so 8P, the difference in stagnation
pressure between the two lines, is constant. Therefore, since 8P and 3R are both constant, G must
also be constant.
If there are losses in the flow, the assumption that G is constant will still hold, provided that the
losses are the same at all radii.
Before considering methods of solving equation (15) within the blade rows, the flow in the
annular ducts which form the inlet and outlet to the machine, and also in the gaps between the
blade rows, will be considered.
4. F l o w in an A n n u l a r Duct.
For flow in an annular duct the angular momentum, wr, and the stagnation pressure are constant
along streamlines. At the design radius the separation of the streamlines is constant, and therefore
3
&~7(wr)R = constant = oJ VR (17)
and
G = constant. (18)
~o is a constant which determines the amount by which the flow departs from free-vortex conditions.
Referring to Fig. 4 in which points A and D a r e at the design radius, the angular m o m e n t u m at B
is given by
(wr)B = + O (wr)3R.
But since points A and B are on the same streamline, (wr)B = (wr)A, and using equation (7) for the
displacement of the streamlines, and also equation (17),
~°VRa~k~ 32t.
(wr)A = (wr)D + R
This may be written
~oVRaeh 2 ~ = wr + -~ ~ ,
wr+ ~ ~ . D
so that at the design radius and providing the displacement of the streamlines remains small
RVR + Tr
or, using equation (9)
~¢.I). (20)
Similarly equation (19) gives
coa2k 2
fir + - ~ 5 - Z = constant. (21)
Eliminating fiR, (Ofi/Or);v and G from equations (16), (18), (20) and (21) then gives
a~ d2A
= constant = - A~.
k2
This may be written
k,2a ~ dZA (22)
where
_ 1
(23)
k'~ k2
T h e solution of equation (22) is
where the upper sign refers to the inlet duct and the lower sign to an outlet duct. It is seen that Ao~
is the value of A far away from the machine.
Differentiating equation (25) gives
k,a dA
dx = + (a°--A°°)e±X]k'a" (26)
Differentiating again and using equations (11) and (23) gives for the static-pressure gradient
( k~a~°J°~
g = - 1 R z ](A0-A~o) e~xlk'~+fiR 2- (27)
T h e result obtained in this section may be compared with the exact solutionfor flow in an annular
duct when Ao~ = 0 and co = 0 (Bragg and H a w t h o r n O ) .This exact solution has the form of a series
in which each term varies exponentially with x in the same way as in equation (24) but with a
different value of k for each term. T h e s e values of h are the roots of
where J and Y are the usual Bessel functions. T h e axial-velocity profile corresponding to the lowest
root has no maximum or m i n i m u m for h < r < t (although the slope is zero at r = h and r = t) and
is therefore comparable to the profiles assumed in this paper. T h e values of ka/t calculated from
equation (8) and for the lowest root of equation (28) are compared in Fig. 6. T h e agreement is
seen to be very close.
W h e n ~o is not zero the exact linearized solution for flow in an annular duct has been given by
Marble 13. T h e effect on the rate of decay of disturbances agrees exactly with equation (23).
VR / = fir 2 (29b)
where m is a mass-flow parameter defined by V~If~R which is constant in any blade passage.
Thus any analysis applied to equation (16) for the blade passage of a rotor row will apply for the
blade passage of a stator row, if 1/m is put equal to zero. The remainder of this section will therefore
be devoted to the flow through an isolated rotor row of blades.
The values of fir and (3fi/3r)R at entry to and exit from a blade passage are determined by the
flow conditions at inlet and exit. However, we can assume any reasonable variation of fir and
(8fi/3r)n within the blade passage itself. In this analysis it has been assumed that both fin and (3fi/3r)R
vary linearly with x through the blade passage, so that fR and (3f/3r)n at any point within the
passage are given by
(30)
J
This assumption is very close for the central streamlines in a blade passage, and is exact for a row
of infinitely thin blades designed with a parabolic centre-line, working at zero incidence. A more
accurate variation could be obtained by analysing the pressure distribution round the cascade blade
and putting in the correct variation of mean flow angle.
Substituting equation (30) in equation (16) yields
dZA
dx 2 (L+Px+Qx2)A = W+ U x + Vx ~, 0 ~ x <~ a (31)
where
k2a~L = 1 + f r o ~
k~aap = 2 f m ( f i n ~ - f m )
k a%o =
k2a 2 W
k2a 3U = R o. _ R
8B
k2a4 V .=
This is a linear differential equation for A, the solution of which will contain two arbitrary constants,
one of which is determined by the conditions at entry to the passage and the other by the conditions
at exit from the passage.
If we assume that the isolated blade row is preceded by an inlet duct and followed by an outlet
duct, it follows from equations (25) and (26) that the solution of equation (31) at inlet to the blade
passage must satisfy the equation,
Similarly at outlet from the blade row, the solution must satisfy
Cl= ka (dA)
dx _ k(A®a_ ;t~,) '
C~ = Ai - A~u.
Thus the value of A at exit from the blade, row is given by
The values of A ~ and A~a are calculated from equations (16) and (34) and the same method of
solving the problem is used as was described in the previous paragraph. It is interesting to note
from Fig. 7 that (for a multi-stage compressor) the values of )t obtained using this assumption agree
very closely with the values obtained using the assumption of a linear variation of tan ~. Since this
assumption leads to a vast saving of arithmetic and seems to be quite accurate it is recommended
10
for use when a digital computer is not available. However, since equation (35) contains A, which is
the unknown which we are trying to find, the physical meaning of the assumption in equation (36)
must always be dubious, and when a computer is available the assumption of linear variation
of/~R and (a~/ar)R is preferred.
- fi2v + ar =
so that
= + --,
gl
and
afi~ 3fl~ f~ D.r 3v
Or Or + v v ~ Or"
T h u s at the design radius, putting r = R and m = VR/t2R,
1
fi8 = - fi~ + - - (39)
m
and
11
axial length of the gaps between the blade rows is usually very small compared with the axial length
of the blades, so that they are usually neglected. If they are not neglected the solution within each
gap is given by equation (24) as
A = A~ + A e ~Ik'~ + B e -xI~'~,
where
Now (m/R), given by equation (20), will in general be small compared with (l/ha) which is itself
small. Since x is also small, (x/k'a) will be much less than unity so that within the gap, equation (24)
may be written
)t Ao~+A l+~7~a +B 1--
and the effect on A of neglecting the modification to 1/ka will not be significant. Thus within the gap
between blade rows it will be assumed that k' is equal to k.
At exit from the compressor the air will normally flow through a set of outlet guide vanes before
entering the outlet duct. These guide vanes will remove the swirl component of velocity so that ~o
becomes zero in the outlet duct and k' assumes the constant value k. Consequently, the problem is
linear within the whole compressor and a method of solution similar to that for a single blade row
followed by a row of outlet guide vanes can be used. Two initial guesses are made for the value of A at
inlet to the first row of blades, and two complete solutions using these initial guesses are then
obtained. The solutions at entry to the outlet duct are scaled linearly to fit the boundary conditions
at that point expressed by equation (26) in the form
The scaling factor b thus derived is used to calculate the correct solution at that point and thus at
every point within the compressor.
The special case when there are no outlet guide vanes (and thus k' is not equal to k in the outlet
duct) can still be treated in the same manner, since k' only occurs in the solution when the scaling
factor b is being calculated at exit from the final blade row. This is given by
1
b=
1 - +k-~(dA)~
)~ 1 (A,._A~)
(dh)~ 2 + ~ a l (A_A~o~)
where the subscripts 1 and 2 refer to the two initial solutions. Since (dA/dx) and (h-A~o) will be
very large at this point for any initial guess which is not the exact initial value of A, a small change in
the small number 1/k'a will not significantly affect b, and so k' may be assumed to have the value k.
Since the solution for a multi-stage compressor is basically a series of solutions for individual
blade rows, the problem of the flow through a multi-stage machine lends itself very readily to an
efficient digital-computer solution. A programme for the solution of equation (16) in any compressor
has been written for use on the EDSAC 2 computer of the Cambridge University Mathematical
Laboratory. Details of this programme are given in Appendix I. This programme has been used
to obtain the flow patterns through several compressor designs, the results of which are shown
in Figs. 7 to 14.
12
7. Flow in a Multi-Stage Compressor--Identical Repeating Stages.
In the previous section the flow through a general multi-stage compressor has been considered,
and it has been shown how the problem can be solved numerically. A numerical solution, of
necessity, involves dividing each section into small intervals, and solving the equation step by step.
However, for the case of a set of identical' repeating stages a method of solution has been evolved
which will provide values of A at the leading and trailing edges of all the blade rows, without having
to solve the differential equation at any intermediate points within the blade rows. It will be assumed
that the compressor under analysis consists of a number of identical stages which may be preceded by
several stages which do not form part of the repeating group. The compressor will be terminated by a
row of outlet guide vanes followed by a duct. The axial gaps between the blade rows will be
neglected.
Consider a row of rotor blades. The linearity of equation (16) permits the solution at the trailing
edge (station 2) to be written in terms of that at the leading edge (station 1) in the form
As = alas + a2 a dx 1 + aags + a4
where a s to a12 are constants and a is the axial-chord length of the blade row.
Similarly for a stator row the solution can be written
(=
da hb~)Sx ~ a4+ b 6 . (a~)3+bTga +bs (42)
=
(43)
A ~
as
a5
~2
a6
a3 ]
a7 D -
a4]
a8
X3.
Thus by combination of equations (44) and (45), there follows for a stage
-1 </~< 1
-1</~2< 1
and
/x8 < - l o r / x 3 > + 1 .
However, in practice it has usually been found that ~1 and/~2 are complex conjugates and will be
denoted by ~ _+ i~. In this case the modulus of the complex roots must be less than unity. For
complex conjugate eigenvalues'the eigenvectors will also be complex conjugates, which will be
denoted by A + i I I .
Using these definitions, the solution at entry to the first repeating stage can be written as
Since (1)X' must necessarily be real, it follows at once from equation (52) that Yl and 72 are complex
conjugates, Which will be represented by p + is.
Now, if A0 is the value of ;~ at entry to the compressor,
then
X0 = ~o
[!:1
and it can be seen that the value of X after all the stages which precede the repeating stages (i.e. the
value of X at entry to the first repeating stage) may be expressed in the form
where P is a 3 by 3 matrix and Q is a 3 by 1 matrix devised from the matrices which define the blade
rows preceding the repeating stages.
Therefore, equating equations (52) and (54) the matrix equation
is obtained.
15
This can be written in the form
or
1¸1-1
E-o- • (56)
This equation determines A0, p and ~. Hence, the solution at the front of the compressor and the
constants 71 and y~ have been found, so that there remains only Y8 undetermined.
T o find 7a, equation (33) expressed at exit from the outlet guide vanes is used. Since co is zero in
the outlet duct, k' has the constant value k and equation (33) can be expressed in the form
dA 1
a Zx' = k (g-/32)" (57)
If J and K are respectively the 3 x 3' and 1 x 3 matrices governing the flow through the outlet
guide vanes, the matrix X at exit from these blades is given by Xe, where
= J(r3 3 E + M) + K
which can be written as
X~ = Tatar(J v s) + (JM+K). (58)
If equation (58) is now written in terms of the matrix components in the form
16
This matrix method of solution is considerably quicker than the numerical method, and is
particularly advantageous for a compressor having a large number of identical repeating stages.
The main disadvantage of the method is that it is only possible to obtain solutions at the leading
and trailing edges of each blade row, and thus no indication is given about the form of the solution
within the blade passages themselves. The method of solving equation (16) described in this section
has also been programmed for use on the EDSAC computer, and details of this programme are
given in Appendix II. Results obtained by this method have been found to agree exactly with results.
obtained by the longer numerical method of solution.
G1 = G2, G3 = G~.
Equations (39) and (40) relate the flow angles at stator inlet to the conditions at rotor outlet.
Since the pattern is periodic, similar equations relate the conditions at rotor inlet to conditions at
stator outlet, namely
1
/31 = m /34,
and
{a/3q = [a/34 1 a2
Rkar]R -Rk&]n+m m"
If equation (16) is written for each of stations 1, 2, 3 and 4, there are then sufficient equations.
for the unknowns to be found, and the result for A is
2 / Off2
(~2 +/3~) - {R
m- \ -~r + R -'-~h
Or/
A= (60)
/32 +/34
The corresponding result for the gradient of stagnation pressure relative to stators is
4 2 2 ~ 2
1 - 2/32 - 2/3~+ ~/34 + m/34 - --/32/34 +
G = 2(/32 +/3~) m
+ R~ -2-/342+/32/34-~/34 +
17
(87503) B
For the special case of 50o//o reaction at all radii
Yr, Yr "
'Then
1 R 3fi
A- 1 (62)
mE p ar
and
1 Rg~ 1
G3 - I +-- (63)
rnp fi ar 2m ~
These results are particularly useful in the preliminary design calculations for a compressor.
8. Discussion of Results.
8.1. Model Compressor.
To illustrate the method of using the theory presented in the foregoing sections, it has been
:applied to calculate the flow through a hypothetical ten-stage compressor both in the 'as-designed'
condition and for conditions when various modifications are made to the compressor geometry and
to tile mass-flow rate through the compressor.
The machine consists of a row of inlet guide vanes followed by ten identical stages, the axial
gaps between the blade rows being ignored. It is presumed that the blading has been designed by
assuming constant axial velocity throughout (i.e. A = 0), equal work done at all radii, and 50%
reaction at all radii. The rotor blades are therefore mirror images of the stator blades.
The information relevant to the specification of the compressor is summarized below.
Constant tip radius t 12in.
Constant hub radius h 6in.
Angular velocity of rotor f2 1208 r.p.m.
Axial-chord length of all blades a 1.6086in.
Exit air angle relative to blade c~ 30 °
Stage work AH
0.4
(Blade speed at design radius) 2 (f~R) ~
500//0 reaction at all radii.
Hence,
Design radius R 9.4868in.
Blade speed at design radius f~R 100 ff/sec
Mass-flow parameter m 0.5196
Aspect ratio 3.73
Blade twist at exit 3B/Or 0.1420.
T h i s value of 3fi/Hr is obtained by putting A = 0 in equation (62).
18
As-designed conditions.--The vortex flow for the hypothetical compressor under as-designed
conditions is shown in Fig. 7. It is seen that the flow pattern can be divided into three distinct
phases. First there is a basic periodic variation which persists right through the compressor; then
superimposed on this basic flow are the other two phases. These consist of disturbances at inlet and
exit caused by the conditions existing in the inlet and exit ducts respectively. Each of these phases
of the flow will be considered in turn.
After the first four stages the variation in A consists entirely of the basic periodic variation, called
the regular flow. This regular flow has a wavelength equal to the axial length of a single stage and is
of very small amplitude. This result verifies the assumption used by Traupel ~°, Wu and Wolfenstein22,
and Schnittger 17, that the streamlines in a plane containing the axis of the machine vary in a periodic
manner.
The regular flow comprises a radially inward deflection of the air through the second half of a
rotor row and the leading half of a stator row, and a radially outward deflection through the latter
half of a stator row and the leading half of a rotor row. These flow variations are small, and can
normally be neglected for design purposes, since the overall effect is to increase by approximately 1 c
the air angles on to the leading edges of the blades at the stator tip and the rotor root, and to decrease
by the same amount the air angles on to the leading edges at the rotor tip and stator root.
Now consider the disturbances at inlet to and exit from the compressor. As would be expected,
the effect at inlet propagates downstream much farther than the effect at exit propagates upstream.
As the air enters the compressor it is deflected radially inwards through the inlet guide vanes and
the first-stage rotor, the maximum deflection occurring just before the trailing edge of these rotor
blades. The radial deflection then gets progressively smaller, and through the third-stage rotor there
is almost no variation of the flow in the radial direction. Through the third-stage stator there is a
small outward radial deflection as the disturbance at inlet dies away and the flow settles down to
the regular flow. The inward deflection of the air causes a maximum increase of 7½° in the air angle
on to the tip and a decrease of 4½° on to the root at the leading edge of the first row of stator blades,
compared with the design condition of zero deflection. As the air leaves the compressor it flows into
an annular duct, the effect of which is to cause an outward flow of air in the last stator row, but it
has negligible influence on the flow farther than one blade chord upstream.
Also shown in Fig. 7 are values obtained by the matrix method of solution described in Section 7,
using the assumption of a linear variation of A~ mentioned in Section 5. The agreement between the
two solutions is close.
Effect of varying the massflow.--Fig. 8 shows the effect of varying the mass-flow rate through the
hypothetical compressor. All the solutions exhibit the same three characteristics, namely, an initial
disturbance which settles down after the fourth stage to a regular periodic flow, followed by an
outward deflection of the flow in the stator row of the final stage.
As the mass-flow rate is decreased from the design flow rate the maximum inward deflection of
the air in the inlet guide vanes and first rotor row is decreased, but a large outward deflection is
induced in the second and third stages. After the fourth stage the periodic flow is such that more
air is going through the tip section than the root section. This will tend to stall the root section more
than the tip. It will be noticed that as the mass flow is decreased so the amplitude of the periodic
variation increases. This is caused by the air having greater difficulty in trying to attain radial
equilibrium conditions for the smaller mass-flow rates.
19
(87503) 13.
For mass-flow rates greater than the designed flow rate the maximum inward radial deflection is
increased, and the flow eventually settles down with more going through the root section than the
tip. It appears from Fig. 8 that as the mass-flow rate is increased the amplitude of the periodic
variation decreases, so that the flow through a stage becomes more uniform although radially along a
blade row there is more air going through the root than the tip.
Modifications to the inlet guide vanes.--The considerable inward flow set up in the first few stages
of the hypothetical compressor running under as-designed conditions is a departure from the
design assumptions and is therefore very undesirable. Consequently a simple method of reducing
this deviation from the zero-slope design condition was sought. Two methods were tried, the first
being a complete removal of the inlet guide vanes and the second a modification to the exit air angles
of these blades. The results of these investigations are shown in Fig. 9.
It is seen that removing the inlet guide vanes has the effect of replacing the inward radial deflection
by an almost equal outward deflection. In this case the maximum deflection occurs just before the
trailing edge of the first row of stator blades. As in the as-designed case, the flow settles down to a
periodic variation after the large deflection, and becomes identical to the original flow pattern after
the fifth stage.
The reason for the inward deflection of the flow is that a non-constant stagnation-pressure profile
is required in the later stages in order that equilibrium may be attained. This stagnation-pressure
gradient is given by equation (63). Since the stages are designed for constant work, the first stages
must depart from their design conditions in order to produce the required stagnation-pressure
profile. If therefore the inlet guide vanes are designed so that the required stagnation-pressure
gradient is produced in the first stage of the compressor, the departure from the design axial-velocity
profile 'will be reduced. Consequently the air angles out of the inlet guide vanes were modified to
produce this effect. This was done by changing the twist on the trailing edges of the guide vanes,
keeping the exit angle of 30 ° at the design radius.
The rise of stagnation pressure in the first stage is
AP
- ar(w -w0
P
or
AP
- Q r ( Q r - / ~ 2 v - p~v), (64)
P
where the subscript e denotes conditions at the trailing edge of the inlet guide vanes. Since
G-pVR 2\ar]R
(aP t ,
2 1 + 8o + R
G-m~ m -37 + R 8r]" (65)
The hypothetical compressor under analysis has been designed for 50% reaction at all radii.
Consequently,
and
0t3
Or Or
20
and the value of (a/ge/ar) can be found using equations (62), (63) and (65). In this case the value is
0. 04057, compared with the as-designed value of 0- 14200, and the twist on the inlet guide vanes is
considerably reduced.
Using the modified inlet guide vanes, the resulting flow shown in Fig. 9 is seen to possess a
much smaller inward deflection in the first two blade rows. As in the two previous cases the flow
assumes a periodic form after the first four stages, and again becomes identical with the original
design flow after the fifth stage.
Alternative design.--Another possible basis for the design of a compressor is to make the stagnation-
pressure gradient zero in the regular flow. This would have the advantage of giving nearly constant
stagnation pressure at outlet from the compressor. The twist on the blades has therefore been
altered to achieve this condition. The required twist is obtained by putting G = 0 in equation (63)
giving
Op fi 1 _1+ =0.25470.
Jr = r ~m~
Comparing this with the value of 0. 14200 for the constant-axial-velocity design, it is seen that the
constant stagnation-pressure design requires blades with a much higher degree of twist.
The design value of ~t is given from equation (62) as
1
A= - 1. 852.
2m 2
This design axial-velocity profile is in fact the mean of the radial equilibrium profiles before and
after a rotor row, assuming constant stagnation pressure with radius.
Fig. 10 shows the computed flow for this compressor. It is seen that the design condition is
reached after about 3 stages.
9. Conclusions.
The comparison of the results of Wu's ~5 calculation and Horlock's 6 experiment with the present
theory shows good agreement, and it is suggested that the present theory is quite sufficiently accurate
for design and performance calculations on axial-flow turbo-machines. It is also suggested that
greater accuracy of the calculations is not warranted unless an adequate means of calculating the
effects of the boundary layers on the inside and outside walls of the machine is included. The
limitation to a one-parameter family of axial-velocity profiles leads to a great simplification of ideas
about how these flows behave. An illustration of this is t h e example showing how the first stage of
a compressor may be designed to do varying work with radius, in order to achieve a more nearly
constant axial velocity in the first few stages.
The calculation may very easily be programmed for digital computers. EDSAC 2 takes about
five minutes to calculate a ten-stage compressor. An alternative assumption, which is recommended
if no computer is available, has also been given, and this method appears to give equally accurate
results, although the physical basis for the assumption is not entirely satisfactory.
The calculations for a model compressor with identical stages show how the flow consists of a
regular pattern which repeats from stage to stage, together with perturbations at the front and the
back of the compressor. The flow fluctuation in the regular repeating pattern is small and can
probably be neglected in design, and the outlet perturbation is also small, but the inlet perturbation
should certainly be included in design calculations. The effect of varying mass flow has been studied,
22
and also the effect of modifications to the design of the inlet guide vanes, and to the basis of design
of the vortex flow in the compressor. It is found that if a compressor is designed to have any constant
axial-velocity profile, and is designed to do constant work with radius, then the design profile will
be achieved after the first few stages. There will however be substantial deviations from design in
the first few stages, and the stagnation pressure will vary with radius at exit.
Further work now being undertaken is to include the effects of taper of the inside and outside walls
of the machine, and also the effect of compressibility. These are known to be effects of importance
in turbines and in the first stages of axial compressors. Other effects which could be taken into
account but which are probably of minor importance are the effects of the radial force exerted by
the blades, the effect of taper of blade thickness, and the accurate representation of the blade loading
working from the pressure distribution round the blades, which could be obtained from cascade
tests or by calculation.
10. Achnowledgment.
The basic ideas of this report are taken from internal reports written in 1951 and 1952 by one of
the authors (D. S. Whitehead) while he was working with Messrs. Rolls-Royce Ltd. Grateful
acknowledgment is made to Messrs. Rolls-Royce Ltd. for permission to publish this report.
Grateful acknowledgment is also made to Dr. M. V. Wilkes, who made the computing facilities
at the Cambridge University Mathematical Laboratory available to the authors.
23
NOTATION
r Radial co-ordinate
ig Axial co-ordinate
0 Tangential co-ordinate
Zt Radial component of velocity
Axial component of velocity
W Tangential component of velocity
a Axial chord of blade
g Static-pressure gradient
G Stagnation-pressure gradient relative to the blade
h Hub radius
k Constant defined by equation (8)
m V~/~2R, mass-flow parameter
P Static pressure
P Stagnation pressure
R ~_~_r~2± / , ~ I 2 design radius
t Tip radius
o~ Air angle relative to the blade
tan
P Air density
co A measure of rotation in the flow
Angular velocity of rotor
Matrix Notation
A
3X3
X 31} matrices for flow through rotor row
D
B
3x3
x 31} matrices for flow through stator row
E
24
NOTATION--continued
Matrix Notation--continued
C 3x3 =BxA
H 3xl=BxD+E
M 3xl =(1-C)-IH
P
3 x 3 tmatrice s defining flow through all stages preceding repeating stages
Q 3xl )
X 3 × 1 = (~, a & / d x , g}
Imaginary part of/~1 and ~2, and also used for the stream function
Prefixed subscript
(n) Conditions at entry to the nth repeating stage in a multi-stage compressor
25
REFERENCES AND BIBLIOGRAPHY
2 A . D . Carmichael and J. H. Horlock Actuator disc theories applied to the design of axial compressors.
A.R.C.C.P. 315. April, 1956.
3 B. Eckert and G. Korbacher .. The flow through axial turbine stages of large radial blade
length.
N.A.C.A. Tech. Memo. 1118. April, 1947.
4 W. R. Hawthorne and J. Ringrose Actuator disc theory of the compressible flow in free-vortex
turbomachinery.
A.R.C.R. & M. 3346. September, 1962.
8 J . H . Horlock . . . . . . Some actuator-disc theories for the flow of air through an axial
turbo-machine.
A.R.C.R. & M. 3030. December, 1952.
26
REFERENCES A N D BIBLIOGRAPHY--continued
27
APPENDIX I
A programme for the EDSAC computer has been written which will solve the basic differential
equation {equation (16)} for any compressor consisting of rotor rows, stator rows and small lengths
of duct. A Runge-Kutta-Gill numerical method has been used to furnish the solution, which consists
of the values of A and (dA/dx) everywhere within the compressor under consideration. Two initial
guesses are made for the value of A at inlet to the compressor, and using these values two complete
solutions are evaluated simultaneously throughout the whole compressor. These solutions are then
scaled to satisfy the exit condition given by equation (26), and the scaling factor so obtained is used
to provide the exact solution everywhere within the compressor.
To enable the solution to proceed, certain data about the compressor under consideration must
be provided, and this nmst be done in a definite format. This consists of a block of general data
about the compressor, followed by small data blocks for each of the compressor components,
arranged in the order the components appear in the machine.
The general data block must supply the following quantities: hub radius, tip radius, axial velocity
at entry, angular velocity of rotor, Runge-Kutta-Gill interval.
Each component data block must consist of--the axial-chord length, fin at exit, (~/Or)n at exit and
the number of values per section required in the output.
28
APPENDIX II
This solution consists of two phases. The first of these is the calculation of the matrix coefficients
for the individual blade rows, which coefficients are then used in the second phase to obtain the
values of ;~, (dk/dx) and g at the leading and trailing edges of all the blade rows comprising the
compressor.
Twelve matrix coefficients have to be evaluated for every blade row which does not constitute
part of a repeating stage (as defined in Section 7), and also for an isolated rotor row- and isolated
stator row which go to form one of the repeating stages. These coefficients are obtained simply by
solving equation (16) in the isolated row for the three different cases obtained when two of the
variables )~, (dA/dx) and g at inlet are put equal to 0 and the third is made equal to unity, and for
the fourth case when all three variables at inlet are put equal to zero. The solution through the blade
row is obtained using the Runge-Kutta-Gill numerical integration mentioi~ed in Appendix I.
The matrices evaluated in the first phase of this solution are now used in the second phase to
evaluate A, (dA/dx) and g at all the blade interfaces. The algebraic method based on the Caley-
Hamilton theorem is used to set up the characteristic equation of the matrix C, the dominant
eigenvalue of which is then extracted by Newton's method. To give an idea of the magnitudes of
the errors in the method the imaginary parts of all the results (which should be zero) are printed in
the output in addition ):o the real parts. In all the compressors thus far tested, the largest imaginary
part recorded has been of the order 10-11.
In order to evaluate the matrix coefficients the following information about each blade row must be
supplied--the axial chord, the axial velocity at the design radius at the leading edge of the blade,
the hub and tip radii,/3R and (Ol~/Or)R at exit from the blade and also at exit from the blades of the
preceding row, and the angular velocity of the compressor rotor. Additional data required for
the second phase of the solution are the number of blade rows before the repeating stages and
the number of repeating stages in the compressor.
29
(0 (a'~
///////////
DIRECTION OF
ROTATION
ROTOR
_O-r
- - 2
I1
t~
ST
4 t~
4-
7v-
2-0 2.0
I-8 1-8
v
%
I-G I-G
1"4 1.4
B
A I-2 1.2
f D
I-C 1,0
R R+~R
R I 0-@ 3-8
04 O.G
I
I
I 0.4
Fie. 4. Displacement of streamlines near design radius. I...,._RANGE O F 0 " 4 H U B / T I P ,~ I
I RATIO C O M P R E S S O R I
I I 0.2
[ I
I RADIUS ; I
0L DESIGN RADIU R
J O
0.2 0"4 O.G C~8 I-0 1'2 1"4 16 1.8
0-2 0.2
~ P R E S E N T TH ~OR'Y
BESS~L F U N IT ION J
O-I O.I
O I I I I I I f I O
O o.i 0.2 o.s 0.4 o-s o.c~ 07 o-s 0.9 I'O
HUB/TIP RATIO ["i/t
+O'I
.,,<
iii
/
_.J J
o COMPRESSOR DESIGNED FOR )~=O
I~ c/
D- -0.2
>- /
U
jO -0-4
4: -0.to
x
<
IL
O
ill -0-6
Q.
o
o~
I.G~IFT~TE SECOND THIRD i =OURTH FIFTH 51NTH SEVENTI EIGHTH NINTH TENTH
STAGE STAGE ! 8TAG E STAGE STAGE STAGE STAGE STAGE._ STAGE
AXIAL DISTANCE
32
+3.C
k/k/
\ /
Vw
ILl
-1
E
o
Iv
n r~= O ' 4
I-
+1'0
W W W
o
o
_1
[U
>
5<
o v v
<
IL
O
Ifl
n
3tD
-I.0
SECOND THIRD IFOURTH FIFTH ISINTH iSKVNNTH EIGHTH NINTH TENTH
L STAGE STAGE STAGE STAGE STAGE STAGE STAGE ;~;~ I~T~GE
FIG. 8. Vortex flow in ten-stage compressor. Effect of varying the mass flow.
0.8-
O-GI - / \
/ NLET G U I D E
/ \ / VAN ES REMOVED
0'4-
/ COMPRESSOR D E S I G N E D F O B
, \
\
_1
+o-2 - il I \,
,\
"// ~ \"-/f
\ //Tx. /
°t. \.j/
d
>
-0-2 -
I "~'INLET GUIDE VANES
<
Ic
O
- '-i- \ /
/
A S DESIGNED
i
I
DESIGNED TO GIVE
REQUIRED S T A G N A T I O N -
PRESSURE GRADIENT
13-
ILl
9 -0'~it
- 0 .t
i.c=.v F I R S T
STAGE SgCT£2
£ T?'T2& FOUIk~'H
STAGE
FIFTH
STAGE
SIXTH
STAGE
NINTH
STAGE
TENTH
ST~,G E
AXIAL DISTANCE
L
FIG. 9. Vortex flow in ten-stage compressor. Effect of modifications to inlet
guide vanes.
33
+1.O
/ 0 i
IA T
f I __: LA-
LU
_J
r
Y £ONSTANT-AXIAL-M EL0ClTY DESIGN
(DESIGN ",/ALOE OF N =0 )
K
0
IX
Iz
-I'0
P
u
0
_J
UJ i
<
)< -2.C CONISTANT -TO-J'AL-HEAD DESIGN
< (DESIGN VALOE OF . : k = - ) - S 5 2 )
IL
o
IIJ
{3-
o, i
t/)
-3.0 t FIFTH SIXTH SEVENTH EIGH'TH ~TA;H TENTH
FIRST SECOND ,-II
I.G~' STAGE I STAGE T~ " STAGE STAGE STAGE STAGE I STAGE
AXIAL DISTANCE
0.0
~<
IU
J
O.B
0.4
)~ U S I N G PREDICTED
/
f
o
0.3
//
)-
i- )~ U S I N G OBSERVED {~"~
O 0-2
O
_J
IIJ
>
I
J +0'1
<
<
IL
o
UJ
0 // /-
/OBSERVED 1~
13-
3 -0")
J
(D
-0"2
I.G.V. ~ =ROTOR- -STATOR : ' ' ~ DUCT l ;' O.G.V
1-275 ~ 1.225 ~ 1"22S M 9"390" 1"5001
34
• I
V"
200 -
I
_r = 0 - 9
190 --
~<
ILl -O'2
._1
-200 if_
IBO-- o
-0.3
r
~t - 0-7616 >-
0 -0.4
\
o
~q
-19o
._J
tlJ
\
>
~1 - 0 . 5
<_
2O0-- X
.<
uJ -O-G -
I
F-
r ~0.6
It_
0 -0,7
IU
fi_
190- O
0~ - 0 . 6
I N L E T GUIDE ~AF
VANE~.
0,7-
0.(o
- - YORTEX-FLOW THEORY
0.5 ------ wU -FHEOI~'f
~ -,.o
0"4
I
>- tl.l O . 7 _
~w
-J&
>mFo.e
-]w r =0.8
t
~< o.~
0.9
O.B
Y
0.7
36
(87g03) Wt. 64/1857 K.5 8/63 Hw.
Publications of the
Aeronautical Research Council
Special Volumes
Vol. 1. Aero and Hydrodynamics, Aerofoils, Controls, Flutter, Kites, Parachutes, Performance, Propulsion,
Stability. i26s. (post 3s.)
VoL II. Aero and Hydrodynanaics, Aerofoils, Airscrews, Controls, Flutter, Materials, Miscellaneous, Parachutes,
Propulsion, Stability, Structures. I47S. (post 3s.)
Vol. III. Aero and Hydrodynamics, Aerofoils, Airscrews, Controls, Flutter, Kites, Miscellaneous, Parachutes,
Propulsion, Seaplanes, Stability, Structures, Test Equipment. 189s. (post 3s. 9d.)
Index to all Reports and Memoranda published in the Annual Technical Reports
I9O9-X947 R. & M. 2600 (out of print)
Printed in England