C.P. No.

868

LIBRARY
YAL AIRCRAFT. ESTABLISH.MEhu
BEDFORD.

MINISTRY OF AVIATION
AERONAUTICAL RESEARCHCOUNCIL

CURRENT PAPERS

Turbulent Boundary Layer Theory

and its Application ’
to Blade Profile Design ”
BY
D. J. I. Smith

LONDbN: HER MAJESTY’S STATIONERY OFFICE

/
1966

Price 15s. 6d. net

 U,D,C. No. 532.526.4:621.438-253,5,001.1

C.P. No.868

March, 1965

Turbulent boundary layer theory and its

application to blade profile design

- by -

D. J. L. Smith

SUMMARY

Five methods of predicting the incompressible, two-dimensional

turbulent boundary layer have been applied to flow conditions considered

to occur over the suction surface of turbo machine blades and the measure

of agreement between the separation criteria and boundary layer charac-

teristics assessed. The methods considered were those due to Buri,

Truckenbrodt, Stratford, Maskell and Spence.

All of the criteria could be brought into tolerable agreement pro-

vided that a value of -0.04 was used for Buri's criteria, and that for

Truckenbrodt and Spence's methods the position of separation was determined

by the condition that local skin friction coefficient is zero. It was

additionally necessary in the methods of hiaskell, Truckenbrodt and Spence

for the calculation of the shape parameter to be started with a value of

1.4.
All of the criteria except Spence's were sensitive to Reynolds num-
ber and showed that an increase in Reynolds number delays separation,

Stratford's method was extremely easy to apply, was the simplest of

the five and predicted the lowest pressure rise to separation.

To assist in the design of blade profiles, envelopes of suction

surface velocity distribution have been constructed to give separation at

the trailing edge; these are considered to be conservatively based.

-.1*".T.a .**, . .>L-..U..-, I--.c* *.. ..,.__"
_ -. Y_-I,-.Il--*I-.___I_I-. -.=. .rY--P__

,Replaces N.G.T.E. M.395 - A.R.C.26 961

 -2-

CONTENTS
Page

1 .o Introduction

2*0 Flow models 6

2,l Surface velocity distributions

State of boundary layer
Reynolds number

3.0 Methods of analysis

Buri 7
Truckenbrodt 8
3.3 Stratford 11
Maskell 13
Spence 15

4.0 Results of comparison 17

4.1 Momentum thickness 17

4.2 Shape parameter, skin friction coefficient and
position of separation 17
4.3 Application to turbomachinery design 22

500 Conclusions 23

References 26

Notation 23

Detachable abstract cards

No. Title

I The prediction of the characteristics of the

turbulent boundary layer 31
 ILLUSTkATIOI'?S

Fig, No, Title

1 Flow models

2 Momentum thickness Type A flow model

3 Boundsry layer characteristics Truckenbrodt method..

Type A flow model Ht = 1.3, I.4 and 1.7

4 Boundary layer CharacteristicsTruckenbrodt method.

Type B flow model Ht = 1.3, I.4 and I07

5 Boundary layer characteristics Truckenbrodt method.

Type C flow model Ht = 1,3, I.4 and 1.7

6 Boundary layer characteristics Spence method.

Type A flow model Ht = 1.3, lo4 and 1.7

7 Boundary layer characteristics Spence method.

Type B flow model Ht = 1*3, I.4 and 1.7

0 Boundary layer characteristics Spence method.~ .

Type C flow model Ht = 1.3, 1.4 and 1.7 l

9 Boundary layer characteristics Kaskell method.

Type A flow model Ht = 1.3, 1.4 and 1.7

10 Functions f,(H) f',(H) and fs(H,Q) in shape

parameter equation

11 Comparison of Truckenbrodt, Naskell and Spence

methods. Type A flow model Ht = lo4

12 Comparison of 'Zruckenbrodt, bIaskel1 and Spence

methods. 'Type B flow model Ht = lo4

13 Comparison of Truckenbrodt, Maskell and Spence

methods. Q-pe C flow model Ht = lo4
t
14 Velocity gradients for separation at trailing
edge

15 Envelopes of type B velocity distribution for

separation at trailing edge

16 Envelopes of type C velocity distribution for

separation at trailing edge
 -4-

I-0 Introduction

The design of a turbine blade profile has conventionally followed a

rather arbitrary pattern whereby certain geometrical parameters such as
trailing edge thicla?ess, maximum thickness/chord ratio and leading edge
radius, have been selected in the light of earlier elcperience* The sec-
tion profile has then been constructed, either by using a number of circu-
lar arcs or by laying out an arbitrary thickness distribution along a sim-
ple camber line (usually parabolic). The position of adjacent blade sec-
tions is chosen to conform to some simple aerodynamic loading criterion
such as that of Zweifel', the passage geometry at the outlet being adjus-
ted to satisfy the g 3 outlet requirements by for instance the rules of
Ainley and Mathieson 9 o At the jnlet the,blade geometry is chosen to
satisfy an incidence requirement O

In the case of compressors it is usual to use standard aerofoil

sections on circular or parabolic arc camber lines, the amount of camber
being determined by the air deflection and by current incidence and devia-
tion rules3. The pitch/chord ratio is chosen to satisfy a loading crite-
rion (e,go, that of Howel13) for the required deflection.
It is clear, however, that these methods are not necessarily ideal
aa they possess no means for differentiating between the effects of many
possible variations in blade shape* In practice, empirical restriction3
havg been placed on such features as the form of the blade channel shape,
and, in the case of turbines, blade back curvature, but design rules of
this type cannot command,a very high degree of confidence in their
application,

In many instances, it has 'been possible to obtain a good turbine

efficiency using such very elementary design rules, due to the predomi-
nantly accelerating nature of the flow in a reaction blade design. There
are, however, regions, such as rotor blade roots, where considerable areas
of diffusing flow occur and where past empirical design practices may not
have avoided separation of the boundary layer and thei-efore resulting in
. less than optimum efficiency.

It Seems possible that a more fundamental approach to blade profile

design might enable the aerodynamic loading of compressor blade sections
to be increased above conventional values without incurring losses due to
separation of the boundary layer, Also it would provide the distribution
of heat transfer over the blade surfaces which is particularly important
in the case of high temperature turbines,

There is also a requirement to minimise the number of blades in a

turbo machine, to reduce blade cooling requirements in a hot turbine, to
reduce engine coats and to optimise efficiency,

During recent years, the quest for higher efficiency and for more
economical use of blading has encouraged an increasing amount of interest
in the problem of blade profile design. This means that a much more pre-
cise assessment of permissible aerodynamic blade loading is required and
this is only attainable by detailed consideration of the flow conditions
over the blade surface, As a first step towards this, various methods
to surface pressure distribution have been and
z; ,re,~~~i~~v",;~;~dt!?~~, 7, Honever, the question then arises as to what
is the optimum pressure distribution which should be aimed at in design,
 -5-

For the present it will be assumed that attention is restrioted to

blade rows in which the exit Mach number is low enough for the peak sur-
face velocity on the suction surface of the blade to be below a Mach num-
ber of 1.0. With the peak velocity restricted, any attempt to increase
aerodynamic loading may require that the diffusion gradient near the
trailing edge on the suction surface should increase, a condition which
may cause separation of the boundary layer and increased loss.
It is commonly assumed that optimum two-dimensional performance
will correspond to a blade for which th boundary layer is just stable,
lee., near to separation, and Stratford 8 has demonstrated a diffusing flow
in which the turbulent boundary layer is critical at all points, In a
turbo machine however, the precise flow conditions at all blade sections
can only be defined approximately and it is thought that a safer basis for
blade design would be to ensure that if separation is encountered it will
be progressive from the trailing edge,

Both Swainstong and Allan 10 have given consideration to the use of

pressure gradient as a design limitation, following the reasoning that
optimum performance is likely to be achieved when the gradient is just
insufficient to cause separation of the boundary layer. This approach is
of course highly sensitive to the state assumed for the boundary layer, in
particular the position of transition, and for this Swainston and Allan
assumed an incompressible, two-dimensional fully turbulent boundary layer
with the aim of ensuring that design is conservatively based. Allan also
considered the case of a mixed laminar turbulent boundary layer.

There remains, however, the problem of predicting separation of the

turbulent boundary layer and for this a number of empirical methodT,aJiS
available. The method used by Swainston was that of Truckenbrodt 9
which has a very complex derivation and involves calculating the variation
of the shape parameter, H, (ratio of displacement thickness to momentum
thickness) separation occurring when H = 1.8 to 2.4. Using a value of
H= 1.8, Swainston deduced the envelope of pressure distribution having a
constant pressure over the forward portion of the blade and a linear pres-
aure gradient over the rear portion for separation to occur at the trail-
ing edge, the Reynolds number based on blade surface length and outlet
velocity Be being 3.5 x IO'. Allan made use of a simpler analysis due to
Burii1,13 and deduced the envelope of velocity distributions having a
linear velocity gradient, instead of a linear pressure gradient, for a
Reynolds number of 2 x IO'.

It was thought desirable that the various quite distinct methods of

predicting the behaviour of the incompressible, two-dimensional turbulent
boundary layer should be examined with a view to assessing the measure of
agreement between the various criteria for separation and boundary layer
characteristics, and this Memorandum presents a comparison of five methods.
Also considered is the application of turbulent boundary layer
theory in assessing the pressure distributions over the auction surfaces
of turbo machine blades which would give separation at the trailing edges,

The paper may be read without reference to the Appendix, which

contains a detailed summary of the five methods investigated.
 -6-

2.70 Flow models

The flow within a turbo machine is complex in nature and at the

present time the characteristics of the precise nature of the flow over
the surfaces of the blades is a matter for speculation, However, to pro-
vide a common basis for analysis three f1ov.r models, whosesurface velocity
distributions could be considered to give a simplified representation of
distributions associated with the suction surfaces of turbo machine blades,
were selected. The models are sho<Jn diagrammatically in Figure 1.

2.1 Surface velocity distributions

To compute th,0 incompressible boundary layer characteristics and

position of separation the velocit,y distribution at the outer edge of the
boundary layer is required, Therefore, simple velocity distributions, in
the sense that the computations were made easier as vii11 become evident on
reading Section 3,0, nere chosen.

The distributions are shown non-dimensiona1l-y in Figure l(a) as

velocity ratio V plotted against distance x where V is the ratio of velo-
city-at outer edge of boundary layer to the velocity at the trailing edge
and x is the ratio of surface distance, measured from the leading edge
stagnation point, to total surface length. For all three distributions
it was assumed that the velocity rose from zero at the leading edge stag-
nation point to a definite value over an infinitely small distance. The
distributions are referred to as type A, B and C,

Tme A - the velocity decreases linearly with surface length from

V = V, at the leading edge to V = 1.0 at the trailing edge.

Q-peB- the velocity is constant, v = To, over the first 60 per

cent of the blade surface followed by a linear decrease to
trailing edge.

TQpe C - the velocity increases_linearly over the first 60 per-cent

of blade surface from V = 0.5 at the leading edge to V =
TO at the 60 per cent station, follolfed by a linear
decrease to the trailing edge.,

2.2 State of boundary layer

In order to _oredict the behaviour of the turbulent boundary layer
the position of transition must be known.
The flow in the boundary layer as it develops from the leading edge
stagnation point is initially laminar. The laminar boundary layer is
very sensitive to disturbances in the presence of a positive pressure gra-
dient iOeo, pressure increases in the direction of flo;i, and will readily
separate or become turbulent,
It is frequently assumed that design will be conservatively based
if the boundary layer is taken as being fully turbulent (momentum thick-
ness zero at leading edge). However, J, H. Prestonlb has shown that for
a circular pipe and a flat plate the minimum ;ieynolds number, based on
momentum thickness 8, for turbulent flow is Ito = 320 and suggests that in
the case of flow with a favourable pressure gradient the minimum value
will decrease and for flow v:ith an adverse gradient trill increase.
 -7-

In view of this it was assumed in the present analysis that the

boundary layer was fully turbulent, having a momentum Reynolds number
Be of 500 (Figure l(b)) at the leading edge, This approach although not
strictly correct should help to ensure that design is conservatively based,
since a laminar boundary layer grows at a slower rate than a turbulent
layer, Thus, if in practice a length of laminar layer occurs and is fol-
lowed by transition to iurbulent flow, it is believed that the momentum
thickness at the position of,the start of the turbulent layer would be
less than if the flow had been fully turbulent.

‘2-o3 Reynolds number

The present methods of analysis for the turbulent boundary layer
are based on data from experiments conducted at Reynolds numbers which
were very much higher than is associated with the flow within a turbo
machine. In the present study it eras assumed that these methods could be
applied to flows where the Reynolds number, Be, is low and representative
of turbo machines0 The Revnolds number, based on outlet velocity and
blade surface length, range" examined for all three flow models was Re =
2 x IO5 to 1 x 10 , the aim being to assess the effect of Reynolds number
not only on the position of separation but also the measure of agreement
betcfeen the various methods of analysis, . I
3.0 Methods of analysis
---_..__- --_
There are in existence several semi-empirical methods of predicting
the characteristics of the incowp~ossible, tt-Jolcij.,e,ls.i.onal turbulent bound-
ary and the methods considered TJere thos due to B~rill9~3, .
Truokenbrodt"j12, Stratford'7, Maskell 18 and Spencelpj

3.1 Buri

For floby along a flat plate in the a3sence of a pressure gradient

the 1/7th po;Jer law for the velocity profile in a turbulent boundary layer
may be considered an approximate empirical relation, To specify the
velocity profile in the presence of.a pressure gradient Buri chose, in
analogy to IL Pohlhausen's approximate method for laminar boundary layers,
a form parameter I? = $ I&j4 dxdV and assumed that the shearing stress at the

wall a, and the shape parameter H are functions of I' alone.

=w Q"
Thus, = .fi(F) .
7

displacement thickness 6"

andH = =-F = f2 03
momentum thickness

Experimental data were used to cotiirm the analogy, and the results were
moderately satisfactory.
The position of separation involves the calculation of I' over the
surface and using the momentum integral equation Buri was able to shorn
that

$ (BR&) = A-E .
 -8-

where A and B are empirical constants. . .

A unique critical value of I' is then assigned to the point of

separation ijhich corresponds to the condition of local skin friction
,
coefficient Cf
=w =
= +- 0. According to the curve Buri drew through the
2pva
experimental points the value of i'critical is -0.06.
The main advantage of this method is that it is fairly straight-
forward to compute and does not involve further arbitrary assumptions
regarding the value of the shape parameter at transition which can grossly
affect the conclusions in some other methods. However, this method
involves a knowledge of the velocity gradient, g, which may prove to be
difficult to assess from measured pressure distributions.

Both the empirical relationship and rcritical were derived from

very limited early experiments of Nikaradse and Burl *II J3, *The experi-
ments of Nikaradse nere for flow in converging and diverging channels hav-
ing flat walls and of rectangular cross section. Buri's experiments were
for flow in converging channels and of circular cross section. In the
case with an adverse pressure gradient (divergent channel) the boundary
layer was very thick and extended as far as the centre of the channel, the
Reynolds number based on momentum thickness Q ranging from 3000 to 9000,
For the flow with a constant and favourable pressure gradient (convergent)
the range of Be was 500 to 3000,,
Hovrarth'lC has applied Buri's criterion using a value of I'critical =
-O,O6, to a measured pressure distribution over a circular cylinder at a
Reynolds number based on diameter of 2.12 x IO'. In vien of the assump-
tions in the calculation (i.e., position of transition, conditions at
transition), and the experimental difficulty in locating the separation
point the result may be considered satisfactory,
It is north mentioning here that Howell 15 has made use of Buri's
parameter in analysing compressor cascade resultso By assuming the
velocity distribution in the boundary layer on a cascade blade is linear
at separation Hovel1 found fairly good correlation between diffuser and
cascade test results.

It was found necessary, in the present analysis, to adopt a more

conservative value of rcritiOal for types A and C flow models than had
been suggested previously in order to yield results which compare favour-
ably with predictions by more recent methods. The validity of such pro-
cedure is obviously open to suspicion. On the other hand, the original
experiments defining rcritical (with adverse pressure gradient) involved
boundary layers extending as far as the centre of the channel, This could
have produced secondary flows and thus destroyed the two-dimensionality
of the flow assumed in von K.&&n's classical derivation of the momentum
equation rihich in conjunction with the measured velocity profiles, Buri
used to calculate the tirall shear stress 71;50

302 Truckenbrodt

This method has a very much more complex derivation than that pro-
posed by Buri,
 -9-

Location of the position of separation involves the calculation of

momentum thickness 8 and shape parameter HO Unlike the other investi-.
gators who used the momentum integral equation for 8, Truckenbrodt used
the energy integral equation, For calculating H both the momentum and
energy integral equations were used0
.
The calculation hinges upon semi-empirical relationships between

W energy dissipation in the boundary layer D and Reynolds num-

ber based on momentum thickness X0

D 0.56 x 10'~
pv"= -se"

(ii) wall, shear stress l;w, shape parameter H and Rs

-=
TW 0,246
Cf = 0.676H 00266
+.M IO %

(iii) a unique relationship between H and a parameter z

1.269H
if=, _ o,g7g ; where H = -7energy thickness
momentum thickness

The momentum thickness is given by

$ (BR;) = A - EX'

eR: dV
where I' = - V -dx and the constants Aand B depend . on .the empirical

relationship for the energy dissipation, In arriving at this equa,tion

it was assumed that the shape parameter has little effect on the growth
of the momentum thickness and was taken as being constant and equal to
1 .I+.

Using the above relationships and the momentum and energy integral
equations it can be shown that
 - 10 -

Truckenbrodt succeeded in transforming this equation by introducing

a shape factor L, which is related to the shape parameter H (see nota-
tion), so that it could be integrated and thus obtained an equation for
the variation of shape parameter0

The final separation criterion is to ascribe a critical value to He

Unfortunately this is not known with any certainty but difficulty can be
avoided by calculating the variation of local skin friction coefficient Cf
and applying the condition Cf = 0 at separation as in Maskell's method
(Section 3.4).

The advantages of this method are firstly that the calculation of

shape parameter H is not grossly affected by the value of H at transition
and secondly that no derivatives of the velocity distribution with respect
to distance along surface are needed, in contrast to the Buri, Maskell and
Stratford methods, The calculation, however, is long and laborious when
carried out by hand using an electrical desk machine, particularly if com-
mencing from a measured pressure distribution,
The equation for the energy dissipation (Rotta) has a very complex
derivation; for details see References 12 and 20,
The relationship for the shear stress 7;vf was obtained by Ludweig
and Tillmana for flow under the influence of both adverse and favourable
pressure gradients by means of a simple instrument developed by Ludweig2Z0
This instrument enables the wall shearing stress to be determined by a heat
transfer measurement, The experimental apparatus consisted of a channel
of rectangular cross section, one c~all being used as the flat test plate
on which the boundary layer measurements were performed, and the other wall
adjustable to give the desired pressure distribution. The instrument was
calibrated by setting the apparatus for flow with uniform static pressure,
the calibration sheering stress being determined by the Schultz-Grunon
friction law for plate flow

%f = 334
(log:o;)1.83e

This law, nhich is in close agreement with others for plate flow, was
chosen because it was obtained from measurements in the same experimental
configuration, Four different test series were carried out, constant
pressure, moderate pressure rise, strong pressure rise and pressure drop,
the range of Reynolds number RC being IO' to 4 x IO'. The formula was
also checked for disturbance in the boundary layer by carrying out two
tests at constant pressurez-

(i> with a turbulence grid consisting of metal strips upstream

of the measuring section to increase the free stream turbu-
lence
(ii) with a continuous square section strip placed just downstream
of the leading edge of the test plate, crosswise to the
direction of flow.

The relationship between H and

A ?i was determined by Weighardt23 from
the velocity profile law the numerical constants being
 - 11 -

adjusted to give agreement with experiment, These experiments were for

flow with constant, favourable and adverse pressure rises at high values
of Re and the experimental configuration eras similar to that of Lud1iei.g
and Tillman, *

There appears to be only one independent experimental pressure

distribution to which this method has been applied, this being for the
flow over the suction surface of an N.A.C.A. isolated aerofoil'1~12~~ on
which separation of the turbulent boundary layer occurred0 The test was
carried out in a low turbulence two-dimensional wind tunnel, the pressure .
distribution being similar in shape to type A flow model of the present
analysis, at a Reynolds number based on the blade chord of 2.64 x ?C?o
The agreement between the experimental and calculated boundary layer
momentum thickness and shape parameter was very good, separation occurr-
ing when H = 2,2,

3.3 Stratford

Stratford's criterion for separation of the turbulent boundary

layer results from an approximate solution to the equations of motion and
requires a single empirical factor0
The method assumes that the boundary layer in a pressure rise may
be divided into two distinct regions, namely the inner and outer regions0

In the inner region, the inertia forces are small so that the velo-
city profile is distorted by the pressure gradient until the latter is
largely balanced by the transverse gradient of' shear stress.

In the outer region the pressure rise just causes a lowering of the
dynamic head.profile, and the losses due to the shear stress are almost
the same as for the flow along a flat plate*

. A parameter B is incorporated in the first term of a series expan-

sion representing the whole inner layer profile obtained by mixing length
theory, with the higher terms omitted; B is assumed to represent the
effect-on.the.separation criteria of the higher terms, It is also used
to represent any effects which the pressure rise might have on the mixing
length. -

The final-equation contains a parameter In' which is the flat plate

comparison profile at the position of separation, the relevant Reynolds
number R, being that based on the peak velocity and the distance to the
point of separation. Stratford found using the data of Schubauer and
Klebanoff26 a8d the results of his ovm experiment with'continuously zero
skin friction that n = logiORs but suggests that the criterion is not
sensitive to the value of n,
The criterion, which is obtained as a simple formula applying
directly to the separation point, was developed for pressure distributions
in which a sharp pressure rise starts abruptly after constant pressure for
a distance xoo The distance x is measured from a pseudo origin which is
the point where the turbulent boundary layer would have zero thickness,
After simplification the criterion, at a Reynolds number of the
order of 106, is given by a simple formula
 - 12 -

where

However, as in the case of Buri, the method involves the calculation of a

derivative $$ which
- may not be easily or accurately obtained for some
experimental distributions.
The parameter B was obtained from an experiment described in
Reference 8. In this experiment the turbulent flow was maintained just
at the separation condition (71J = 0) throughout the pressure rise and it
was found that B was independent of Cp and had a value of 0.66, Tie
condition r~\~= 0 is added because Stratford found that analysis of further
experiments shoned that B varied somewhat with the value of dap immedi-
ately prior to separation; p is the static pressure. dxa

Reference 24 (von Doenhoff and Tetervin) contains data for three

experiments on E.A.C.A. isolated aerofoils tested in a low turbulence,
two-dimensional wind tunnel at a Reynolds number based on blade chord of
the order of 2 x IO6 m The pressure distributions Rere of the same form
as the type A flon model used in the present analysis. One of these
three tests was that used as a test case by Truckenbrodt which, it will be
recalled, showed good agreement between the experimental and calculate
separation point. A fourth test was that of Schubauer and Klebanoff2 %,
conducted in the same sirind tunnel, in which the flow passed over an aero-
foil-like section at an angle of attack of O". The pressure distribution
over approximately the first 60 per cent of the surface was favourable,
and was followed by a pressure rise leading to separation; the Reynolds
number based on surface length has the extremely high value of 2.8 x IO'.

Stratford points out that the pressure distributions allowed some

range of interpretation as regards 9 and the effect on the theoretical
dx
prediction of the position of separation could be as much as 25 per cent,
However, it vas found that the criterion, using B = 0.66, resulted in the
calculated separation points being upstream of those experimentally obser-
ved for all test cases0 From a close examination of the results Stratford
found that the error in B increased as dap increased, ranging from zero
d3 dX2
when -d$ maximum negative to 20 per cent when large and positive and sug-
gested a modification that would halve the error:-

B = 0.66; when $$ < 0

2
B = 0*73; rrhen$-$) 0
 - 13 -

The advantage of Stratford's method is that it is extremely simple to

apply since it does not involve graphical integration as do the other
methods or a step-by-step solution for the shape parameter as does
Maskell's method. In fact it was found, using an electrical desk machine
that whereas the methods of Maskell, Truckenbrodt and Spence took the
author at least half a day to apply and Buri about one hour, Stratford's
method took only half an hour. This, of course, does not apply neces-
sarily to all types of flovi since for the models used in the present analy-
sis the boundary layer was assumed to be turbulent over the entire sur-
faoe, Horrever. the method demands the calculation of a derivative
which may not be easily obtained for some experimental pressure
distributions. Also if one is interested in other boundary layer
characteristics, such as local skin friction coefficient, then an alter-
native method would have to be used.

It was found that of the methods considered in this Memorandum

Stratford's criterion predicted the lamest pressure rise to separation;
dCp- was known exactly in the present analysis, A possible reason for this
dx
is that the factor B was determined from test distributions at Reynolds
number Be ranging from 2 x IO6 to 2 x IO7 whereas for the flow models the
Reynolds number was very much loner and it may be that B varies somewhat
when Be < IO'.

3.4 Maskell

Maskell's method is based upon a large amount of experimental data

not only for flat plate and channel flow but also for flow over isolated
aerofoils.

The position of separation involves the calculation of the momentum

thickness and shape parameter, the equations for which have been made more
general than before by making them fit flat plate data very closely and by
the use of some limited data for favourable pressure gradients*
The equation for the momentum thickness was derived from the momen-
tum integral equation, in a manner similar to that of Buri, making use of
the Ludweig and Tillman relationship for skin friction:-

& (6R;) = A - BI'

where I' = -eRi -dV

Vdx

and A, B and n are empirical constants, n being determined to make the

solution correct for zero pressure gradient,

The approach used to find an equation for the shape parameter H

was that of selecting the probable parameters affecting the variation of
H, and using experimental data to find an equation connecting them, This
approach, which has been used by other investigators24, was well suited to
the nature of the available data. The form of the equation is:
 - 14 -

for zero and favourable pressure gradients

Ht f(%)

for unfavourable pressure gradient

mdH
eRa dx = %(I';H)

eR: qH dV
where I'* = -V e o- dx

and m and q are empirical constants, The form of the function G(I':H) was
determined by plotting experimental values of BBCm gdx (which were themselves
determined by differentiating curves of H to obtain dK)
dx against i'" for
particular values of H. Maskell found that the points could be approxi-
mated by two straight lines and plotted the slopes, intercepts on the axis
I'* = 0, and intersections against H and found that

G.(F;H) = @(O,H) + r(H)l?* for I'* > I';

= s(H) + t(H)I'" for I'* < rf

where r, s and t are linear functions of H and for H < 1.4 the function
Q(O,H) satisfies the flat plate equation H = f(RC)O

The position of separation is dttermined by the condition that the

local skin friction coefficient Cf = Z = 0, the distribution of Cf being
+pTra
calculated using the Ludweig and Tillman lava, described in Section 3.2 and
which is a function of 0 and H, This law cannot, in fact, give Cf = 0
explicitly and so the procedure adopted is to extrapolate to zero the curve
of Cf against surface length x by assuming that once the rapid fall in Cf
dCf does not decrease in magnitude.
has started the gradient dx

If transition to turbulent flow occurs in an adverse pressure gradi-

ent a value has to be chosen for the shape parameter H at transition, Ht,
and it was found that the degree to which the growth of H and therefore
position of separation were affected by Ht depended on the Reynolds number.
The calculation of momentum thickness involves no more computation
than other methods. However, the calculation of shape parameter in an
adverse pressure gradient is a step-by-step process which is both long and
laborious, the interval between the points dictating the accuracy0 Once
 - 15 -

again the method depends on the calculation of a velocity gradient, and

when experimentally determined pressure distributions are employed it
becomes susceptible to the same sources of error as the methods of Buri
and Stratford,

The empirical relationships for momentum thickness and shape para-

meter were derived from experiments conducted at a Reynolds number of the
order 2 x IO'. Four of these experiments were in fact used by Stratford
and one by Truckenbrodt as test cases.

Unfortunately all of the available experimental data was used in

deriving the empirical relations 30 that no independent comparisons with
experiment are presented, However, the comparisons with the data show
that the boundary layer characteristics and position of separation can be
predicted with reasonable accuracy for practical purposes,

The result of applying this criterion to the flow models of the

present analysis was that the pressure rise to separation was much higher
than that according to Stratford's criterion. This was surprising in
view of the number of common test cases for which these two methods have
been demonstrated to be in agreement with experiment. However, it must
be remembered that the velocity distributions for all of these cases were
similar to those of the type A flow model only, except one which was simi-
lar to that of type C, and 'the Reynolds numbers were very much higher than
the range considered in the present study.

3.5 Spence

This method involves the calculation of the momentum thickness 0

and shape parameter H,

The equation for momentum thickness was derived from the momentum
integral equation aa in the methods of Buri and iViaskel1, the difference in
the solution being the assumption of the 1/5th power law for the skin
friction coefficient,

In determining an equation for the variation of shape parameter,

Spence made use of the momentum and energy equations aa did Truckenbrodt,
Using theseequations it can be shown that

ndH = Q(H)I' - e(H)

eRe dx

.where I' = - -ORi -dV

vdx

In arriving at this expression assumptions were made regarding

(0 distribution of shear stress within the boundary layer

(ii) velocity profile in a variable pressure

(iii) wall skin friction coefficient,

 - 16 -

The above equation is of the same form as that used by Truckenbrodt

and Maskell, the solution of which varies in the choice made for functions
Q(H) and $(H). Spence chose relationships such that the equation could
be integrated directly and so avoided the calculation, as did Truckenbrodt,
dV
of the velocity gradient duo
a
To ensure that good results are given for a flat plate with zero
pressure gradient, i,e., I = 0, the function q(H) was determined from the
momentum equation assuming

(i> one-fifth power law for the wall skin friction coefficient
c = 0.0176
f 1
Re"
(ii) Cole's relationship for the shape parameter

where E; = -uT
v f: log, RC + constant
C2? C I, are constants/and U, is the friction velocity.
For the function G(H) a quadratic was chosen.

For the case of a thick boundary layer and using a value of Ho =

Id+ where V = V,, Spence shows this to be in good agreement with the func-
tions used by Maskell.

The final separation criterion is to ascribe a critical value to H,

which as Spence points out, is not known with any certainty. To overcome
this the position of separation can be determined by the condition Cf = 0,
the distribution of Cf being again calculated using the Ludneig and
Tillman law.

There appears to be only one case for which this method has been
demonstrated to be in agreement with e eriment and that is for the flow
over an isolated aerofoil-like section* "s at a Reynolds number of 2.8 x
10'. It is in fact one of the distributions that was used by Stratford
and Maskell as a test case0 The pressure distribution was favourable
over the first 60 per cent of the surface followed by an adverse pressure'
rise leading to separation, transition to turbulent flow occurring near the
leading edge. The comparison between the calculated and experimental
distributions of momentum thickness and shape parameter was good, separa-
tion occurring when H = 2.6. In the calculation of H a value of lo4 was
assumed at the position of the peak velocity.

The advantage of this method over those of Maskell and Truckenbrodt,

which also involve the calculation of H, is that the distribution of H over
the surface is more rapidly calculated. Another advantage over Maskell's
method, which is however shared by Truckenbrodt's method, is that the solu-
tion does not involve the calculation of the first derivative of velocity
dV
with respect to surface distance -0dx
The result of applying this method to the flon models was that the
velocity gradient to separation was grossly affected by the value assumed
for H at transition whereas Truckenbrodt's method is not; a change from
I,3 to 1.4 grossly affects the conclusionso Also the growth of shape
parameter was very little affected by a change in Reynolds number, Re
 - 17 -

* .
resulting in the pressure rise to separation being constant with Re,.the
other methods-showing
. that en increase in.Re ‘delays separation.
,I *
4.0 Results of comparison
1 -:.
11 '4.1 * 'Momentum A thickness, l
,: .*
-All of the'methods.e&cept Stratford's,involve the calculation.of
the:Lmomontum thickness, the equations for which 'are all of the same form.
. .. 1
.I ,: I
.. .* .J” ,- .._
. . ’ 8:
+6-- ,a
k

g = 1 Tc d'; + constant .,,.(I)

ikb
Ta: s

Xt- a)
?
I
- *. 1 ‘. 1

a l+b
where constant = e'v at transition and A, a, b and c are empirical
constantp. The...abl
t j below.gives the values of the constants a&cording.
to the various.methods and Figure 2 show the variation-of momentum thick-
ness for the type A flow model at a Reynolds number of 2 x, 105, and a velo-
city gradient.,of. --Oe5; _ a. - .
.. ; .'I *- - I
. - 1. ._ 1. . .
. . . .. ' 6 '
,.. ..I . " ;. . . ,.,..,.. ,':... " : " ,...-"_ : -.. .. ".. -. _ - .;;
. :..I ..A
1 . : .
a b p1 c ; . A*‘. ’ ; .. :
.,, ._. .d .
II. ..v .. I... . : : . .. ... .. . . . ”1,* . . .. . ,
. .: ./ . . : *. -. . . “.
;: -Buri . ii 3a4 a': 0.25 . ’ 4.0’ i 0.016 i, .

'I Truckenbrodt i 3.0 .’ 0.1667' ; 3@333 i 0.0076~ .’

‘il.
.
:: . ,:* . :
’_" Maskell L ; 3.632: 0.2155-i 4.2 - 0.01173 . ’ ’
r . i. .i
I
.'- Spence.! j 3.5 _I 0.20 i Lt.0 ; 0.0106 I;..
. 11 .
. -
. . . . ,T.. . .’ I . . . ... . ‘; .. . ~. . . . . ..I. ..: ..” :‘...,I .
_/ -.’

It may be seen that three of the methods show good agreement but Ir '
Truckenbrodt's gives somewhat smaller values of 0' especially towards the '
trailing edge.. This measure .of agreement was found for all Reynolds . . .
numbers investigated (i,e,,.Re = 2 x IO' to 1,x lO?).and also for types
B and C flow models. ‘
t ..
402 Shape'parameter,'
. coefficient and position skin friction
.I. of separation , .
. .. ,
. . 6. .
The meihods.of. Truckenbrodt, Maskell and Spence also involve the
calculation of the shape parameter (ratio of displacement thic‘kness to
momentum thickness) and in.order to solve these equations,a:knoxledge is
required of the initials (transition) shape parameter. .For types B and C
f1a-J models transition occurs.in a constant pressure and favourable pres-
sure gradient respectively and for such flow conditions Maskell's method
uniquely determines the shape parameter as a function of the momentum
Reynolds number. However, for type A flow model transition occurs in an
 - 18 -

adverse pressure gradient and so a value for the transition shape para-
meter, Ht, has to be assumed in this method. To solve the equations of
Truckenbrodt and Spence a value for Ht has to be assumed for all flow con-
ditions.
The local coefficient of skin friction is particularly important
for two reasons. Firstly it is a measure of the velocity gradient at the
surface and therefore, the stability of the boundary layer, and secondly
to investigate blade temperature distributions which may be required in
the stress analysis of turbine blades the distribution of heat transfer
coefficient is required which, using Reynolds analogy, is related to the
skin friction coefficient. According to Ludweig and Tillman the skin
friction coefficient is given by

z 40581H -o*aaa
Cf .= ,T--$ = 0.246 e % . ..*(2)

On examining this equation it may be seen that if the initial value of H

affects the distribution of shape parameter it will also influence the
local value of Cf and hence the position of separation (given by Cf = 0)
and local heat transfer coefficient. In view of this results are presen-
ted (Figures 3 to 9) showing the distributions of shape parameter and skin
friction coefficient for a range of Ht = I,3 to 1.7 at Reynolds number of
2 x IO' and I x IO'. It is seen that the velocity gradient is varied as
well as the Reynolds number and the reason for this is that it is consi-
dered desirable to assess the effect of Ht on the position of separation,
Therefore, the gradients were selected such that separation occurs at the
trailing edge for Ht = 1.4 which is the commonly assumed value.

Figures 3, 4 and 5 show the distributions according to

Truckenbrodt's method of analysis for types A, B and C flow models res-
pectively. The shape parameter (and therefore the displacement thickness)
and skin friction coefficient are not grossly affected by the value of Ht,
only in a region close to the transition point are the differences in H
and Cf significant, particularly when transition occurs in constant pres-
sure or pressure rise regions at low Reynolds number.

The characteristics H and Cf using Spence's method are shown in

Figures 5, 7 and 8. It was found that the calculation of H was very lit-
tle affected by a change in Reynolds number or in ether words to a change
in the distribution of momentum thickness and so the shape parameter is
shown for the low Reynolds number only, For types A and B flow models,
Figures 6 and 7 respectively, the distribution of H and Cf are grossly
affected by the value of Ht in particular the range of 1.3 to 1*4, over
the entire surface for both Reynolds number. In applying this method to
type C model, Figure 8, it was found that the shape parameter dropped off
rapidly to a value of approximately 1.2 at the start of the pressure rise
and thereafter remained approximately constant, This resulted in very
high values for Cf and thus indicated no separation point. By assuming a
value for H of 1.4 at the start of the pressure and continuing the computa-
tion of H in the normal way beyond this point the method predicted separa-
tion. However, as in the case of A and B flow models the pressure rise to
separation is especially sensitive to the initial value of H.
 - 19 -

As mentioned earlier Maskell's method required a knowledge of the

transition shape parameter for type A flow model only and Figure 9 shows
the variation of H and Cf for three values of Ht. At the low Reynolds
number of 2 x IO5 the characteristics H and Cf are grossly affected by all
values of Ht whereas for the higher Reynolds number of I x IO' the change
of H and Cf between Ht = I.4 and I.7 is relatively much smaller,

The equations for the shape parameter, H, are of the form

= fl (H)r + fi (H) . ..Spence and &IaEkell .0.(3a)

= fl‘(H)p + f,(H,Ro).,. Buckenbrodt . ..(3b)

n+l (TRe)n o a
where I' = 8
7 a;;

The table below gives the form of the functions f,(H), f,(H) and fc(H,B)
and the values of the exponent n

:, . .I’ .” . .._ ., -. . . ::. ” : _. :I . . ,. I .:. :. -‘: ... .. I. ..-. .._.. . ._ - _“” . . “.“.I

: Function f,(H) : Functions f2(H) and fc (H,Q) : n :i

.
!, ..: . ;.-' . . ,.;.. .,, . " .. ... . :."'.:'.:.", .- ..". Y".' "..E ..I.. .:'.: ...." ':...'.'L- - . .. L-p :..' iIi
i$ruckenbrodt'L2.6383
i
Ii
!i
H(H-1) (H-0,379)'
:I
0.32.!+5e-10661H Q-o*"" ; l/G
. ;i
;/
I!
!:!
I,
ii
; H(H-0.379)

; (H-O.379)2
-0.02329
1 ;
. .
f;
ii
il
~;

.I /
j.
ii . . . . .... .. . . . . ,J. . . ..*.... . .. . . ... . .. . . ... , . . . .. . . . .. . . . .. . . ... . . . ... .. . , .. .. .. .. . ,. . . . . . ..;
i,
3'
!i$askell :(0.32;0.3H)e1'"81H i (0.01485 .- 0001399H)e-1'661H~ '1
i: jI
:i 2.1
for large i? i for H < 1.4 ;!
:I
11 ‘1
8. p.15 (1-2’H)e1*681H : 0,07Y6 - 00054H
If iiI
i! 'bar
f small p i for H > 1.6 i 0.268i
I! ;j
I: :i
I tabulated for H = 1.4 :.

' to 1.6 (see Appendix) I

!i
i' i
1‘i ! H
= f(Re) for constant pressure and pressure :
II drop i:
jI . .. ..,,.. . ... ;I. . .. .,,. .. . .. . .... ... . .... .. .. ;. ...... .,...........,.... ..... . ........... ..... . ... ......i...... .... ..j!
i
ISpence ,19.524 (H-l 21) (H-I.) ; -0eOO307 (H-1)' ; l/5 j;
I.I1 i ir
 - 20 -

It is worth mentioning here that Truckenbrodt was able to integrate

the above equation for H by introducing a shape factor, which is related
to the shape passeter (see Appendix), and thus avoided the calculation of
the derivative 2 o Spence also integrated the differential equation for
H,

In the present study the integrated equations were used. Figure 10

shows a plot of the above functions for a range of H = 1.3 to 1.7. The
value of Truckenbrodt's function fo(H,Re), is shovdn for Ro = 1509 500 and
6000 these values being found in the present study for the type C model
just downstream of transition, at the leading edge for all three flow
modals and the maximum value respectively0 It is to be noted that in
Uaskell' s method the above equation fcr H only holds for flow with pres-
sure rise. For the other flow conditions, i.e., constant pressure and
pressure drop, the shape parameter is a function of the momentum Reynolds
number.

Considering Spence's function f,(H) it ma.y be seen that its value

is very small and negative and remains approximately constant with H.
The reason for this is that Spence made use of the one-fifth power late for
skin friction to deduce the form of f,(H) and this law shows that the skin
friction is independent of H, Therefore, as a first approximation,
Spence's equation shows that for flow with pressure rise and pressure drop
H is given by the first term of Equation (ja).

The value of f,(H) changes rapidly with H and as a result the calculation
of H was grossly affected by the initial value of shape paremeter (Ht) for
the above flow conditions, Also on examining the above approximate equa-
tion it may be seen that for a given velocity distribution the calculation
of H is not affected by a change in Reynolds number, Re, as was found in
the present study.

The value of Naskell's function f,(H) and Truckenbrodt's function

f,(H,Re) varies rapidly with H and as a result the distribution of H is
sensitive to a change in Reynolds number as will be seen later.
Truckenbrodt's function varies from positive when H < 1.4 to negative and
large when H = 1.7 resulting in the distribution of H not being grossly
affected by a change in the value of Ht' Kaskell's function for large l?
is small, negative and approximately constant and equal to the value of
Spence's when H < 1 o4 and for H > 1.4 drops off rapidly to large and nega-
tive values when H = 1.70 Therefore, it is to be expected that the cal-
culation of H for cases where I? is large will be particularly sensitive to
values of Ht in the range 1.3 to 1.4. However, for small r the function
varies rapidly for all values of H and so it is not obvious that the
growth of H will be independent of the initial value for this case0

As mentioned earlier Spence's method when applied to type C flea

model did not predict a rapid rise in the region of pressure rise. For
this model the first term of Equation (3a) was large and negative over the
first part of the surface due to the velocity, V, being small and a posi-
tive velocity gradient in this region, Therefore, H dropped rapidly to a
value of approximately 1.2 at the start of the pressure rise (x = 0.6) and
 -.21 -

for this value of H the function f,(H) is very small resulting in a const-
ant value for H of approximately 1.2 in the region of pressure rise,
However, Truckenbrodt's method also showed a drop in H to 1.2 just down-
stream of the transition point at the low Reynolds number but downstream
of this region the calculation showed a rise in H to 1.4 at the start of
pressure rise and in the region of pressure rise H rose rapidly (Figure 5).
The reason for this was that in the region where H = 102, the momentum
Reynolds number was small (Re = 150) resulting in the second term
(Equation 3b) being sufficiently large and positive in comparison to the
first term for the calculation to give positive 2 and, consequently, a
rise in the shape parameter*

In the case of constant pressure H is given by, according to the

methods of Spence and Truckenbrodt

..'.(5b)

It is easily seen from the plot of f,(H) and fc(H,&)'why the former of
these methods is very sensitive to the value of Ht for such flow condi-
tions and the latter approximately independent of Ht.

It is not cbvioua from plct of the functions f,(H), fs (H) and

fc(H,Q) that the above three methods will be in agreement regarding the
calculation of H for any particular value of Ht. The generally accepted
value for the transition shape parameter for flow at high Reynolds number
is I *4, which is the flat plate or constant pressure value for H, although
there is experimental evidence24 that Ht can be as high as 1.8 in SOme
cases. A comparison of the distributions of H and Cf using the methods *
of Truckenbrodt, Maskell and Spence, is shown in Figures II, 12 and 13 fcr
types A, B and C flow models respectively at a Reynolds number, Re, of
2 x IO' and 106 and a value of 1.4 for HtO The velocity gradient S is
-0.5, -0.75 and -I,0 for A, B and C models respectively.

Truckenbrodt's criterion for separation is that the shape parameter

H, takes the value of 1.8 to 2.4 and it may be seen that this results in a
wide range for the position of separation, Maskell's criterion is that
the skin friction coefficient, Cf, is zero at separation and in view of
the good measure of agreement Maskell foun?, between experimental and
theoretical positions of separation it was decided to adopt this criterion
for comparing the above three methods.

Considering type A flow model, Figure II, it may be seen that at

Re = 2 x IO' the methods of Truckenbrodt and Spence are in reasonable
agreement regarding H but Maskell's deviates greatly from these methods
giving very much lower values over the last 40 per cent of the surface,
As regards the distribution of Cf significant differences occur over-the
last part of the surface and the position of separation varies from x =
O,& according to Spence to L = 1.0 according to Maskell. However, at
 - 22 -

the Reynolds number of -lethe methods of Maskell and Truckenbrodt are in

good agreement but Spence deviates greatly from these methods, predicting
the same values for H as at the lower Reynolds number for the reason men-
tioned earlier.

Turning to type B flow model, Figure 12, at the high Reynolds num-
ber all three methods are in tolerable agreement regarding both H and Cf
but at the lower Reynolds number whereas Maskell and Spence are in good
agreement Truckenbrodt shows very much higher values of H and consequently
lower values for Cf in the region of pressure rise,

For type C model, Figure 13, all three methods are in tolerable
agreement in the region of ressure rise, However, in the region of
pressure drop (2 = 0 to 0~6 P there are significant differences in the dis-
tribution of H and Cf at Re = 2 x 105. Truckenbrodt's method shows very
low values of H in this region compared to Maskeli's and the reason for
this is that the value of Truckenbrodt's function fs(H,Re) is strongly
dependent on the momentum Reynolds number, Re, which drops rapidly from
500 at the leading edge to 150 at Z = 0.2.
Various methods for relating blade shape to surface velocity dis-
tribution are clprently being examined and the question arises as to what
is the optimum velocity gradient which should be aimed at in design.
Figure 14 shows, for the flow models considered in this study, the varia-
tion of adverse velocity gradient, p with Reynolds number, Re, for sepa-
ration at the trailing edge (j; = l.Oj using, the criterion Cf = 0 and a
value of Ht = I ,4, in the above three methods. Also shown are the gradi-
ents using the separation criterion of Buri and Stratford, Buri's cri-
eR@s dv
terion for separation is that a parameter I' = - V -dx = -0.06 at separa-
tion but in view of the limited experimental data from which this value
was derived the velocity gradients were also calculated for I' = -0-04.

Stratford's criterion predicted the lowest pressure rise to separa-

tion for all flow models except type A at high Reynolds number. However,
it must be pointed out that the pressure rise to separation according to
Stratford is likely to be from 0 to 10 per cent too low since asp is small
and negative, dxa

Spence's method showed that the critical velocity gradient was

independent of Reynolds number and had the same value for types B and C
flow models for the reasons mentioned earlier. However, the other methodc
of analysis showed that the effect of increasing Reynolds number is'to
delay separation,

4.3 Application to turbomachinery design

The Mach number over the suction surface of a blade may be as high
as unity and so the applicatio of incompressible boundary layer theory is
questionable. Van Driest11t2g has shown that for flat plate flow i.e.,
zero pressure gradient, the effect of Mach number on the local coefficient
of skin friction is small up to M = 1.0 and can be neglected, the ratio
-'fM= 1 being O,Y3, but there appears to be no evidence available for flow
CfM= 0
under the influence of pressure rise.
 - 23 -

Without reliable experimental measurements of the boundary layer

development and separation relating to flows typical of those within turbo
machinery it is not possible to comment confidently on the particular vali-
dity of any of the methods considered in this Memorandum in such an appli-
cations In a typical turbo machine both Reynolds number and turbulence
differ substantially from those appertaining to experimental data on which
each of the five methods has been based, and against which each has been
tested in varying degree,,

Under these circumstances preference leans naturally to the use, as

a guide, of the method which is simplest to compute or which yields the
most conservative solution, particularly in the lower range of Reynolds
number. Stratford's method would seem to combine both these attributes
commendably, so far as provision of a convenient criterion for separation
is concerned.

However, if the boundary layer characteristics such as displacement

thickness and skin friction coefficient are required then it is suggested
that 'Pruckenbrodt's method be used since it does not involve the calcula-
tion of local velocity gradients g as does Maskell's and is not so sensi-
tive to the shape parameter at transition as the methods of Maskell and
Spence.

From the results shown in Figure 14 it was considered possible to

construct two envelopes of velocity distributions for separation at the
trailing edge

(a> distributions of type B having a constant velocity over the

forward portion of the blade followed by a linear decrease
to trailing edge and,

(b) distributions of type C having a favourable velocity gradient

over the forward portion followed by a linear decrease to the
trailing edge, ~

Figures 15 and 16 show the critical envelopes for types B and C according
to Stratford's criterion and it is suggested that until definite eqeri-
mental data become available for the flow conditions over the surfaces of
turbo machine blades the envelopes for a Reynolds number of 2 x IO" should
be used as a limiting criterion in design.

500 Conclusions
Five methods of predicting the behaviour of the incompressible,
two-dimensional turbulent boundary layer have been applied to three basic
types of velocity distribution, selected to represent the family of dis-
tributions associated with turbo machine blades, and the measure of agree-
ment between the separation criteria and boundary layer characteristics
assessed, The methods considered were those due to Buri, Truckenbrodt,
Stratford, Maskell and Spence,

The velocity distributions that were analysed were type A - linear

decrease of velocity from leading t\> trailing edges of the blade, type B -
constant velocity over the first 60 per cent of the blade surface followed
by a linear decrease to the trailing edge and type C - linear increase of
velocity over the first 60 per cent of the surface followed by a linear
decrease to trailing edge.
 - 24 -

The precise flow conditions over the surfaces of a blade in a turbo

machine is a matter for speculation, but for the present study it was
assumed that the boundary layer flow was fully turbulent with a momentum
Reynolds number of RC = 500 at the leading edge.

The methods of Truckenbrodt, Maskell and Spence provide the growth

of the shape parameter and to solve these equations an assumption has to
be made with regard to the transition (i,eo, initial) value Hte Spence's
method was extremely sensitive to the value of HtO Sven a small change
from I,3 to 1.4 grossly affects the distribution of H and, therefore, the
position of separation, Truckenbrodt's method was very little affected
by Hte Maskell's method only reqtiires the initial value of H when transi-
tion occurs in an adverse pressure gradient and it was found that the
extent to which the growth of H and pressure rise to separation were
affected by Ht depended on the Reynolds number.

The equations for the momentum thickness were of the same form and
all of the methods were in good agreement except Truckenbrodt which showed
smaller values.
Spenoe's method showed that the pressure rise to separation was
independent of Reynolds number whereas the other methods showed that the
effect of increasing Reynolds number is to delay separation.
All the methods could be brought into tolerable agreement regarding
the position of separation provided that

(0 Buri's criterion was taken as I'critical = -0.04.

(ii) In applying Spence's metho,d the calculaticn of shape para-

meter started at the position of maximum velocity, if transi-
tion occurred upstream of this point.

(iii) For the methods of Truckenbrodt and Spence the Ludweig and
Tillman law was used to calculate the variation of local skin
friction coefficient Cf and the position of separation was
given by the condition Cf = 0 and not by a predetermined
value of shape parameter H. Since this law cannot yield
explicitly Cf = 0, the point of separation was obtained by
linear extrapolation from the steepest negative gradient of
the Cf curve0

(iv> For the methods of Truckenbrodt, Maskell and Spence the vari-
ation of shape parameter was calculated using an initial
value of H = 1.4e
On reviewing the methods, al.1 of nhich derive from experimental con-
ditions somewhat removed from the environment within a turbo machine,
Stratford's was simplest to apply, predicted the lowest pressure rise to
separation, and is therefore preferred as a conservative design criterion.,

On this basis envelopes of critical suction surface velocity dis-

tributions (i,e., distributions which yield separation conditions at the
trailing edge) were constructed which, it is believed, are conservatively
based and might be used as a limiting criterion for turbo machine blade
design,
 - 25 -

If the boundary layer characteristics such as displacement thick-

ness and skin friction coefficient are required then it is suggested that
the method of Truckenbrodt be used as it does not involve the calculation
of local velocity gradients 2 and is not grossly affected by the value
of the Shape parameter at transition.
 - 26 -

REFERENCES

No, Author(s) Title, eto,

1 0. Zweifel The spacing of turbo machine blading

especially with large angular deflec-
tion, The Brown Boveri Review Vol, 32,
No. 12 1945

2 D. G. Ainley A method of performance estimation for

G, C. B. Mathieson axial-flow turbines.
R, & M. 2974, December, 1951

3 A, R, Howell Fluid dynamics of axial compressors,

Design of axial compressorse The
Institution of Mechanical Engineers,
Proceedings Vol. 153, 1945 pp. 441-462

4 J, D. Stanitz Design of two-dimensional channels with

prescribed velocity distributions along
the channel walls*
N.A.C.A. Report 1115, 1953

5 J. D. Stanitz Application of a channel design method

L, J, Sheldrake to high-solidity cascades and tests of
an impulse cascade with 90' of turning.
N.A.C.A, Report 1116, 1953

6 Chung-Hua Wu A general theory of three-dimensional

flow in subsonic and supersonic turbo
machines of axial - radial and mixed -
flow types,
N.A.C,A. TN.2604, January, 1952

7 E. Martensen Berechnung der Druckverteilung an

gitterprofilen in ebener potentialstro-
mung mit einer Fredholmschen Integral-
gleichung,
Archive for rational mechanics and
analysis. Vol. 3, No. 3, 1959

8 B, S. Stratford An experimental flow with zero skin fric-

tion throughout its region of pressure
rise.
Journal of Fluid Mechanics Vol, 5,
Part I, PP.1 7-35, 1959
9 M. J. C. Swainston The boundary layer characteristics of
some hypothetical turbo machine blade
pressure distributions,
Ae9.C. 23,568, February, 1962
IO W. K. Allan Theoretical analysis of the performance
of cascade blades,
A.R.C. 23,061 July, 1961
 - 27 -

--l3FERENCES (cont'd)
J&* Author(d Title, etc.

11 H. Schlichting , Boundary layer theory.

McGraw-Hill Book Co. Inc., New York,
555-590, 1960

12 E. Truckenbrodt Ein Quadraturver fshren zur Berechnung

der laminaren und turbulenten Reibung-
sschicht bei ebener und Rotationssymmet-
nischer Stromung.
Ingenieur - Archiv 20, 211-228 .I952

13 A.Buri _ A method of calculation for the turbulent

(translated from the boundary leyer with accelerated and
'German by M. Flint) retarded basic flow.
R.T.P. Trans. No. 2073. Issued by the
Ministry of Aircraft Production

14 L. Howarth Note on the flow past a cirouler cylinder,

Proceedings of the Cambridge Philosophical
Society Vol. 5, Port 4, 1935

15 A. R. Howell The present basis of axial flow compressor

design. Part I - Cascade theory and
performance.
R. 6'. Ma 2095, June, 1942

16 J. H. Preston The minimum Reynolds number for a turbulent

boundary layer end the selection of a
transition device.
Journal of Fluid Mechanics, Vol. 3,
part4, 1958 PP~373384

17 B. S. Stratford The prediction of separation of the

turbulent boundary layer.
Journal of Fluid Mech,anics, Vol. 5,
Part I, 1959 PP.I-16
18 E. C. Maskell Approximate calculation of the turbulent
boundary layer in two-dimensional
incompressible flow.
A.R,C.l4 654, November, 1951

19 B. Thwaites Incompressible Aerodynamics.

Oxford at the Clarendon Press, 1960

20 J. Rotta Schubspannungsverteilung und Energie-

dissipation bei turblenten Grenzchichten.
Ingenieur-Archiv 20, 195-207, I952

21 H. Ludweig Investigation of the wall shearing stresses

I?. Tillman in turbulent boundary layers.
N.A.C.A. T.M. 1285, May, 1950
 - 28 -

Author(sl Title, etc.

22 H. Ludweig An instrument for measuring the skin

. (translated by friction coefficient of turbulent boundary
Sylvia We Skan layers.
the Aerodynamics B.R,C.12 991, March, 1950
Division, N,P.L.)

23 IC. Wieghardt On turbulence friction layer for rising

VI. Tillmen pressure,
N.A.C.iL T.M. 1314, October, 1951

24 E. A. von Doenhoff Determination of general relations for

N. Tetervin the behaviour of turbulent boundary
layers. --
I"r.R.C. 6845, F.M. 597 and N.A.C.A. R772,
~qx-il , 1943

25 G. B. Schubauer Investigation of separation of the

P. S. Klebanoff turbulent boundary layer.
N eA. C.A. TN,21 33, August, 1950

26 ~ E. R, Van Driest Turbulent boundary layer in compressible

fluids.
Journal Aero. Sci., Vol. 18, No. 3,
March, 1951 pp.l45-I 60
 - 29 -

NOTATION

D energy which is converted into heat in the laminar boundary layer

H boundary layer shape parameter

= displacement thickness
momentum thickness

energy thickness
parameter = momentum thickness which is related
ii to H

L parameter which is related to H by

L = &r where gp I* 1.73

s

M Mach number

Re Reynolds number based on velocity at outer edge of boundary

layer and momentum thickness
=- V0
v

Re Reynolds number based on velocity at trailing edge and blade sur-

face length
v2e
=-
V

R Reynolds number based on maximum surface velocity and surface

distanoe

U velocity within the boundary layer

z
UT . friction velocity = JZ
\I p
V velocity at outer edge of boundaxy layer

ii ratio of velocity at outer edge of boundary layer to velocity at .

trailing edge

ji. an equivalent distance defined by Equation (21)

K ratio of distance measured along blade surface (from leading edge

stagnation point) to'total blade surface length

CP incompressible pressure coefficient = 1 -($ k 1 -g-j

T
Cf local coefficient of skin friction = 2X-
&pv2
 - 30 -

e blade surface length

P static pressure

t energy of the turbu+ent motion per unit time for a turbulent bound-
, ary layer
X distance measured along surface of blade from leading edge stagna-
tion point

Y distance normal to surface of blade

P velocity gradient = -

6 boundary layer thickness

8 momentum thickness of boundary layer

e' ratio of momentum thicla?ess to blade surface length

s* displacement thickness .of boundary layer

tP* energy thickness of boundary layer

P density

V kinematic viscosity

%W shearing stresses at the blade surface

I? parameter =S
V kn8 dx
dV

r* parameter = ii0 emH kn8 zdV

ri parameter defined by Equation (36~) in Appendix

Subscripts

t conditions at the transition point

0 maximum conditions and position of maximum conditions

a conditions at the trailing edge of the blade

 - 31 -

APPENDIX1

The prediction of the characteristics of the turbulent boundary layer

This Appendix presents a summary of the five methods used to pre-

dict the behaviour of the incompressible, two-dimensional turbulent bound-
ary layer.

The working equations have been made non-dimensional by dividing

the velocity at the outer edge of the boundary layer V by the velocity at
the trailing edge Va and the distance measured along the blade surface x,
from the leading edge stagnation point, by the surface length e.-
Buri method

In a manner analogous to K, Pohlhausen's approximate method for the

laminar boundary layer, Buri i1913 chose a parameter I' for predicting the.
behaviour of the turbulent boundary layer. It is assumed that the shear-
ing stresses at the wall 7;w and the shape parameter H are function p alone.
Thus

.0..(l)

where 1

%i dV . ...(2)
RO = F and I' = --
Vdx

or in non-dimensional terms
5 1
- 5;-
r d;;
d?
r = Bv Be4
v

Nikuradse and Buri II,13 have carried out a series of experiments on the
flow in convergent and divergent channels and using these results Buri was
able to show that the above assumptions are reasonable.
Using Equations (I) and (2) and the momentum equation for steady
motion we get
 - 32 -

hence

-&(eR$= o...(4)

Buri found that the right hand side of Equation (4), which on the above
assum:3tions is a function of T only, was, approximately, a linear function
A - BI'. Equation then becomes

eR; dv
$ ('3Ri) + B --V dx = A

This is a linear equation of the first order for BR: whose integral is

V'%R~ = A[ VBdx + constant

Xt

or in non-dimensional terms
s1
1
=--ii
Iii
zs A
ev Rez
. ..0(5>

2t

where the constant is evaluated from the momentum thickness at the transi-
tion point,

The values of A and B from Nikuradse's measurements which were for

decelerated flow, are A = 0,0175 and B = 4.15; the values from Buri's,
which were for accelerated flop: were A = Oe01475 and B = 3.940 TO
include both cases Schlichting 11 suggests that A = 0.016 and B = 4.0.
This value of B implies that H is constant and equal to lo4 in
Equation (4).
Separation of the boundary layer occurs when the local coefficient
of skin friction is zero, ioeo, when S = 0. The curve that Buri drew
through the experimental points gave a value of I'critical = -0.060,

From Equations (3) and (5) we get, substituting the values for A
and B

dv =
r = -0006 at
z

It is interesting to note that if separation occurs for any fixed value of

1 then for a fully turbulent boundary layer, i.e,,
-', momentum thickness
 - 33 -

8 =OatG = 0, the condition of separation is independent of Reynolds

number Rez if the velocity distribution does not vary with Ree o

Truckenbrodt method

E. Truckenbrodt11T12 made use of the momentum and energy integral

equations for predicting the position of separation of the turbulent
boundary layer,

The equation for the variation of momentum thickness was obtained

from the energy equation which may be written in the form

6
Id
72.2 I=-$ T'$dx = 2D+t *o..(7)
di --r

The quantity on the right hand side of Equation (7) represents the dimen-
sionless work done by the shearing stresses 'G. In the caze of the lami-
nar boundary layer the work done by the shearing stresses is equal to the
energy which is converted into heat D (dissipation). For the turbulent
boundary layer there is a further contribution to the work done which is
the energy of the turbulent motion per unit time, t. This is usually
small compared to D and may be neglected.
D can be
Truckenbrodt shows, using the results of Rotta, 21 that -
PV3
expressed, approximately, as a function of Reynolds number R.c only,

Thus

D 0.56 x 10-a o...(8)

pv3= 1
Ri

Assuming that all velocity profiles form a one-parametex family then

Weighardt24 shows, using the velocity law (u/V) = (~/6)~

a = H1,269 H
- 0.379 odY>

where g = !$? and the numerical constants were adjusted to give agreement
with experiment,

Combining Equations (8) and (7)

where n = 6

assuming E is constant and equal to a mean value then we get

 - 34 -

3. i
ii de eRi
-- dV 1.12 x lO-a
Rez+3 v dx =
ii

or 1
& $ (*$)+ 3 !$ g = le.12 _
x 10-a *

( ). H

or
1.
ORi dV 1.12 x 10-a n+l
& (en!) + (3 + %, v-F& = o- n
ii

This is a linear equation of first order for 6i$ whose integral is

X
2 3+3 3 ,+ a
o- n+ln
1.12 x iO-a
OR; .vT. = VT dx+
s
Ti Xt

constant

Truckenbrodt assumed a mean value of g = 1.72, which corresponds to H 2

l,L+ giving, putting n = 6
X
10 i ‘2

6 V3 R; = C V3 dx + constant

where the constant is evaluated from the momentum thickness at the transi-
tion point and C = 0.0076, or in non-dimensional terms the momentum thick-
ness is given by
6

v dL + I:onstant

The equation for the shape parameter H was obtained from the momentum and
energy integral equations.

The momentum equation may be written in the form

7;
g+(H+2)$g = 4 . . . ..(I?)
PV
 - 35 -

Replacing 6** in the energy Equation (7) by fi0 and from this equation
subtracting Equation (II) multiplied by B we obtain, multiplying through
1
by R;
i -
.eRfJ dH
= r#)r + f,(E) .oe.(12)
8Z

where
fl(@ = (H - I)& f,(g)
i
eRi dV
and I' = vz

The shearing stresses at the wall zw, using the results of Ludweig and
Tillman22, can be expressed as a function of Re and H.

Thus
z
W o.,23 ,0-0.d78H Re-O.=e
a-(13)
pva=

Truckenbrodt transformed Equation (12) into such a form that it could be

integrated, by introducing a shape factor L. This factor is related to
the shape parameter H thus

aFi
L(B) = L(H)
s f,=
E&l

where the lower limit of integration was chosen to make L = 0 correspond

to the case of zero pressure gradient, ioer, flow over a flat plate, giv-
ing Hp = 1.73 and H = 1.4.

Introducing this relationship into Equation (12) we get

1:
ORe
edxdL = r - K(L) . . ..(14)
f2 03
where i<(L) = -- =
K(g)
flm

The function K(L) can be represented with a satisfactory degree of

accuracy by the linear relation

K(L) = a(L - b) .,..(15)

 - 36 -

The numerical values are

a = 0.0304 and b = 0,07 log,,Ro - 0.23

Combining Equations (14) and (15) we obtain a linear differential equation

for L which Truckenbrodt integrated giving

where
000076

b = 0.07 loglo (Re 7 e') - O,23

and

It is to be noted that the new variable S occurs in the equation for momen-
tum thichess (10).

According to work of Ludweig and Tillman the shear stress at the

wall IYWdecreases as the shape parameter increases but never vanishes com-
pletely. Truckenbrodt assumes that separation occurs Flhen H = 1.8 to 2.4
which corresponds to L = -0,13 to -O,l8,

Stratford method

The separation criterion due to Stratford'7 results from an appro-

ximate solution to the equations of motion, The method assumes that the
turbulent boundary layer in a pressure rise may be divided into two dis-
tinct regions, namely the inner and outer regions.

In the inner region, the inertia forces are small so that the velo-
city profile is distorted by the pressure gradient until the latter is
largely balanced by the transverse gradient of shear stress.

In the outer region the pressure rise just causes a lowering of the
dynamic head profile, and the losses due to the shear stresses are almost
the same as for the flow along a flat plate.

The criterion is developed initially for pressure distributions in

vJhich a sharp pressure rise starts abruptly at the position x = x0 after
constant pressure for a distance xoo
 - 37 -

A parameter B is incorporated in the first term of a series expan-

sion representing the lyhole inner layer profile, obtained by mixing length
theory, and the higher terms omitted; the factor B was assumed to repre-
sent the effect on the separation criterion of the higher terms, It is
also used to represent any effects which the pressure rise might have on
the mixing length, The velocity profile has therefore been over ideali-
zed as regards to shape and good agreement with experimental profiles
would not be expected,

The oriterion for separation is, applying directly to the separa-

tion point

dCp Yi = 3' I x 0.41 B (n - 2 )&(n-a) R6

cpJh-a)
(X
dx) 7
(n + l)z(n+L) (n + 2)1
1 . . ..(17)

For Cp < nn+2

where the Reynolds number X is bas'ed on the local value of distance x and
n+2
the peak velocity V,. The limitation Cp Q n+l results from the join
of the inner layer with the outer layer reaching the edge of the boundary
layer when using the idealized velocity profiles.

Stratford simplifies Equation (17) by replacing the quantity

cn + , j&+1) ln + 2+
cn _ ,)tb-a >

by 10.7 x (2.0)4 l(n-a) which is within 1 per cent of the former quantity
when 6 Q n s 8.

This results in.

1

= le06 B (10-6R)10 . . ..(18)

The quantity 'II' is the flat plate (zero pressure gradient) comparison
profile at the point x = xs where suffix a denotes separation

xa '
the relevant Reynolds number being R, = --$ D Stratford found from
experimental data that a good approximation is
 - 38 -

The parameter B was found from an experiment by Stratford, In this experi-

ment the flow was maintained just at the separation condition throughout
the pressure rise and it was found that B was independent of Cp and has
the value

B = 0.66

However, in this experiment the value of dap immediately prior to separa-

dxa
tion had its greatest possible negative value and B will vary somewhat with
d2p
dxa '
To determine the effect of $$ Stratford applied the criterion to
four experiments in which separation of the turbulent boundary layer was
observed and found that, using a value of B = 0.66, the pressure rise to
separation was always too low, A clcse examination of the results showed
that the discrepancy in B increased with an increase in d2p varying_ from
da dx2 2
0 per cent when 3 was maximum negative to 20 per cent when dp was
dxa
large and positive.

In view of the insufficient data Stratford suggests that a crude

modification that would halve the error would be to take

B = 0.66 when 2 < 0

.o..(lY)
da
B = 0.73 when ---$a 0
1. _
Combining Equations (18) and (IY), using a value for n of 6, the criterion
for separation is, at Reynolds numbers of the order of IO6

r ?;
cp x dCp
dx = 0,3s (IO-* do
.( P

or in non-dimensional terms
I
0.39 (10m6 Re ~o~>'o ..0.(20)

when $$ 3 0 and Cp C Lc l the coefficient 0.39 is replaced by 0,35 when

7'
dap (0.
dx"
The pressure coefficient for incompressible flow is given by
 - 39 -

It will be recalled that this criterion was developed for pressure distri-
butions having an initial region of' constant pressure folloxed by a sharp
pressure rise, the distance x being measured from a point where the tur-
bulent boundary layer would have zero momentum thickness.

If transition occurs in the re ion of constant pressure then the

value of x to be used in Equation (20 has to be

2 = (X-q+q 0 0 0 0 (21)

where the distances 2 and j; are the distances from the point of zero
momentum thickness (pseudo origin) and the actual leading edge respect-
ively. The value of ?t is determined by the condition that the boundary
layer thickness for a fully turbulent boundary layer at &-, is equal to
that at Zt for the laminar boundary layer.

This results in

Rt = ,,,.(22)

where
8,=
For pressure distributions having an initial region of favourable pressure
gradient the distribution has to be converted to an equivalent one having
an initial region of constant pressure with a mainstream velocity equal to
the value at the transition point or the point of maximum velocity nhich-
ever is later.
The gronth of a turbulent boundary layer is given by, in non-
dimensional terms .- I
1+b
c St 1

J
zt
7' dz + constant I> ..&3)

The parameters a, b etc, according to various workers vary a little but

representative I7
values are a = a, b =-r;, c = 4, A = 0.016, The constant
is evaluated from the momentum thickness of the laminar boundary at
transition and is given, in non-dimensional terms, by

where
constant C =

% A
et=r- sv5
&r
Pt

0.470
vtal
l+b

2
w&4)

c-(25)
ii6 Rc o
 - 40 -

From Equations (2.2)) (23)) (24) and (25) we get .

. . ..(26)

where suffix-o now refers to conditions at the position of peak velocity

or at transition whichever is later and from (21)

x = (ii - X0) + I, . . ..(27)

Equation (26) can be re-arranged to give

This equation differs from Stratford's slightly in which the exponents 4

and -6 are 3 and $ respectively. The reason for this is that.the para-
meters m, n etc. in Equation (23) have been chosen as the mean values
according to various workers, which Stratford agrees is a better approxi-
mation.

Maskell method

Maskell' 8 made use of the Lud\;eig-Tillman skin friction law for

predicting the position of separation of the turbulent boundary layer.
This involves calculating the distribution of the shape parameter H and the
momentum thickness 0, the equations for which have been made more general
than before by making them fit flat plate data very closely and'by the use
cf some limited data for favourable gradients.

The equation for the variation of momentum thickness \?as derived

from the momentum equation, in a manner similar to Buri, and making use rf
the Ludweig-Tillman law.

The momentum equation may be written

0 0 0 ‘ (29)
 - J+l -

The Ludweig-Tillman law gives

Cf = G(H)Rin . . ..(30)

where
G(H) = aemmH

Combining Equations (29) and (30) we get

Rtf. g + (H + 2) 7eR: zdV = G(H).

Re-arranging ae get

where
r = eRe” dV
-yf-Y&

Xaskcll found that using experimental data the right hand side of
Equation (31) may be represented by a linear function of I', reducing the
equation to the form

& @Ii;;) = A - El’ o-(32)

It can be seen that Equations (31) and (32) can only agree exactly for
r = 0 ioeo, constant pressure, if 11is constant in plate flow. Ludweig
and Tillman found that the shape parameter H nas a function of Q and so
Equation (32) is necessarily in error for constant pressure.

To overcome this Maskell makes the substitution

I’ =
OR: dV
-I

V dx

into the momentum Equation (29). This results in .

Maskell determined the value of q such that when I' = 0, g = constant and
using experimental data found that the right hand side of Equation (33)
could be represented by

d@R;>
e - fI' where e = 0.01173, f = 4.2
dx =

Or
6R: dV
-I$ (OR:) + f v z = e v,here q = 0.2155

The value of f = 4*2 implies that H is constant and equal to 1.635 in

Equation (33). This is a linear equation of the first order for eg
whose integral is

fJRz Vf = e Vf dx + constant

In non-dimensional terms, the momentum thickness is given by

1
r 0.01173
ii

’ fj402~ + constant
7 -I_
I.
1
2155

e’ = ,45- 0.2155
. . ..(j4-)
V Re /
zt _
I i

where the constant is evaluated from the momentum thickness of the laminar
boundary layer at transition,
The approach used in finding an equation for the shape parameter H
was that of selecting the probable parameters affecting the variation of H.
Experimental results \iere then used to confirm that this choice of para-
meters was reasonable and to find an equation connecting them, The main
reason for adoptin, v this approach was because the available data was best
suited to it.

For constant pressure and favourable pressure gradient, the vari-

ation of the shape parameter is given by
 - 4.3 --

For unfavourable pressure gradient iaeo, pressure rises in the direction

of flow, it is given by a step-by-step solution of

,ieacia ,ooaes 0.268 &y

8 V Re - = @(l?*,H)
. dx:

where

@(r”,H) = O(O,H) + r(H)l?* 'nhen.I'* > Pi ...a(36a)

= s(H) + t(H)I'* when I?* c I?1 .oc.(36b)

and
- r(H) = 0.32 - 0.3H

s(H) = 0915 (1.2 - H)

t(H) = 0.15 (1 - 2H)

= sH - 9(O,H) = s(H) - @(O,H) 8

r4 . . ..(36c)
1 r(H) - t(H) 0.17

The function Q(O,H) is given by

Q(O,H) = 10-0'678H (0.01485 - 0,01399H) for H < I.4

..ob(37a)

O(O,H) = 0.0796 - O,O54H, for H > 1.6 ..0*(37b)

For the range 1,4 < H < 1.6 Q(O,H) is defined numerically to give a smooth '
transition from Equations (37a) to (37b) and the values are given below.
 -4-G-

@,H) ’ H
-.
1.4 -0,000533 ; 1.50 -0.00232 i

I 042 -0,000645 i 1.52 -0~00302 1

~44 -0.00086 ; 1.54 -0.00381 ::

1.46 -0.00120 i lo56 -0.00470 j

I .l+3 -0,00170 1.58 -0.00571 ’

1.60 -0eOO68 i
.. . :: ,. ,“. - .,’ .- .-:.- -.

The solution for the shape parameter H proceeds from the value of H
at transition, Maskell proposed a tentative procedure for predicting the
value of Ht' Briefly the procedure is

(i> H = f(RC)
for Ret > 2500 and for all pressure gradients.

(ii) H = f(Q)
for all Qt and zero and favourable pressure gradients.

(iii) H is defined by an approximate envelope.for Qt < 2500 and

unfavourable pressure gradient.

In (i) and (ii) H is given by Equation (35).

The local coefficient of skin friction Cf is given by the Ludweig

and Tillman law, in non-dimensional terms

= 0,246 e-105S1H (5 v Re)-ooas8

cf

= 0,246 jO-0’676H (8 v Re)-00a68 ,...(38)

The position of separation is determined by the condition Cf = 0. Since

the Ludweig-Tillman law cannot in fact give Cf = 0 the computation of Cf
is terminated after a rapid decrease in Cf has started, the position of
separation being determined by linear extrapolation from the steepest nega-
tive gradient of the curve of Cf against surface length 2.
Spencc method

Spence19 increased the usefulness of methods like those of

Truckenbrodt and Maskell by developing a method whereby the shape para-
meter H may be more rapidly calculated.

For the variation of momentum thickness Spence made use of

Equation (31) in Maskell's method. The assumption made was that in
 - 45 -

determining 0 there is no advantage in allowing for the dependence of Cf

on the shape parame ter and so the one-fifth power law for skin friction
was assumed, iOeo,
1
Cf = 0.0176 R; .eo.(39>

Combining Equations (31) and (39) we- get

where

The effect of H on the term H + -,2.2

D2 is small and taking H = constant =
1.5 then

& (OR!) = oeo106 + .!+I'

This equation can be integrated to give

X
1 .
0R;V" = 0.0106 V4 dx + constant

In non-dimensional terms, the momentum thickness is given by

s6
2
6 = & + constant *...(40>
fF-
I
where the constant is evaluated from the momentum thickness for the lami-
nar boundary layer at the transition point,

Using the energy and momentum integral equations Spence shows,

assuming

(i) power law for the velocity profile

 - 46 -

(ii) relationship for the shear stress distribution in the turbu-

lent boundary layer

(iii) the local skin friction coefficient is

Cf = G(H)Rin

that

where
r = --- 0Ri
dV
V dx

To ensure that good results are given for a flat plate with zero pressure
gradient Spence determined the form-of $(H) assuming the one-fifth power
law for skin friction and Cole's relationship for the shape parameter.
Using these assumptions and the momentum equation for steady motion then
it can be shown that

t’(H) = -0.00307 (H - I)" . . ..(42)

To enable the calculation of H to be rapid the form of the functipn Q(H)

was chosen such that Equation (4.1) could be integrated directly to give
the shape parameter H, For this purpose a quadratic was chosen

@CH)= 9.524 (H - 1.21) (H - 1) .' .-(43)

For the case when 0$ is large iaeo, the boundary layer is thick, the
right hand side of (41) is dominated by the first term. Assuming $(H) is
small compared to @(H)I' then

dH
-fg@(H)
dx=

thus

Combining Equations (43) and (44) then Spence shows that for Ho = ? .4 the
function @e(H)is in reasonable agreement with the functions used by
Maskell.
 - 47 -

Substituting Equations (L&Z) and (43) into Equation (41) and inte-
grating we get

Va 4.762 - &
\ constant - 0.00307
sq
X

i)
BR06

where the constant is evaluated from the shape parameter for the turbulent
boundary layer at transition,

In non-dimensional terms the shape parameter is given by

E
-6
v
Hz I + 40762 -1 .00307 -
6 1
va -s
0 Re'

where

Spence points out that the value of H at separation seems always to be

between 2 and 3 and suggests that the range 2.4 to 2.6 be taken.

D 76913/l/125875 K3 IO/66 R
 FIG. I.

TYPE A TYPE 0

SURFACE DISTANCE z

t
,iNG EDGE TRAILING EDGE
STAGNATION \
POINT

(CL) SUCTION SURFACE VELOCITY DISTRIBUTIONS

TURBULENT BOUNDARY
LAYER

TRANSITION DEVICE AT
LEADING EDGE SUCH \
THAT Ret500

eq

0 SURFACE DISTANCE 5 1-o

(b) GROWTH OF BOUNDARY LAYER OVER SUCTION SURFACE

FLOW MODELS
 FIG.2.

3000

REYNOLDS NUMBER R, = 2 X 10’

VELOCITY GRADIENT p= -0.5

1
/

I 0
MASKELL /
r/ 0 0

/ -(
/

/
0
/
/ 0

0
o-2 04 00 6 04 100
SURFACE DISTANCE %

MOMENTUM THICKNESS. TYPE A FLOW MODEL

 FIG.10
SHAPE PARAMETER H

-1 t.
\
\
\

-1

t,

---I TRUCKENBRODT
MASKELL
I.,-. . .. . ... SPENCE

O’OO!

LL LARGE 7
-O*Ol(

FUNCTIONS f,(W) f,0=0 f&-I,Re) IN %WE

PARAMETER EQUATION
 FIG.15

5,

\
\
\
4-
\

4
3-

2-

l-

FULLY TURBULENT BOUNDARY LAYER 500 AT LEADIN EDGE

O- - - - . - - --
0 cl*2 0’4 0’6 O*B
POSITION OF START OF ADVERSE VELOCITY GRADIENT x0

CURVES ARE LOCI OF POINT A IN ii v. ZDIAGRAM

ENVELOPES OF TYPE 6 VELOCITY DISTRIBUTION

FOR SEPARATION AT TRAILING EDGE
 FIG.16

1.2

1-l FULLY TURBULENT BOUNDARY LAYER \

Re ~500 AT LEADING EDGE

0’4 0-b U’U 1.3

POSITION OF START OF ADVERSE VELOCITY GRADIENT f,

CURVES ARE LOCI 0~ POINT A IN V V. Z DIAGRA~v~

ENVELOPES OF TYPE C VELOCITY DISTRIBUTION

FOR SEPARATION AT TRAILING EDGE
A.R.C. C.P. No. 868 A.&C. C.P. No. 068
namlb1965 -, 19%
Smith, ‘). J. l.. Smith, D. J. L.

- BOUNDARY LAYERTIECIRYAND ITS TUABULENP FWNDARYLAYER THWRYAliD ITS

APPLICATIONTo BLADEPrOFILE DESIGN APF’LICATIONTo BLUE PfDFfLE. DESIGN

Five methods oi predict lng the Incompressible, Lwo-dimenslonal Five methods of predlctlng the lncompresslble, Lwo-dlmenslonal
turbulerlt bomdary layer have been applled Lo floe conditions consir:ered turbulent boundary lE&yer have been applied to Ilow conditions considered
Lo occur over the suction surface of turbo machine blades and the measure LO occur over the suction surface of turbo machlne blades and the measure
of agteement between the separation criteria and boundary layer charac- of agreement between the separation criteria and bouudary layer charac-
ter1stiw assessed. The methods considered were those due Lo hurl, Leristirs assessed. The methods conslciered were those due to Purl,
TmckeubrodL, Stratford, tleskell and Spence. Truckenbrodt, Stratford, Maskell end Spence.

All of the criteria could be brought lnto tolerable agreement All of the crlterla could be brcught lnto tolerable agreement
prwldefl that a value or -G 04 was used tor EUri’S criteria and that pmvlded that a value or -0.01. was used for Burlis criteria, end that
ror Quckenbrodt snd Spencets methods the posiLi3n ol separatlm was for Truckenbrodt and Spence’s methods the positlan or separation was
detwlned by the condltlm that local skin Ir ctlon coelf.lcienL is zero. determined by the condition that local skin rrictlon coefllclent is zero.
It waq addltlonally necessary in the methods of tfsskell, Truckenbrcdt IL was additionally necessary in the methods oi Maskell, Tmckenbrodt
rwFRI - .--- --.- -- -- _- .-es-- .- .-!?E@--
A.R.C. C.P. No. 868
twch, 1965
Gmlth, D. J. L.

WRBJIENT BCUNDARY
LAY!ZRTHEDRYAND ITS
APPLICATIONTo BL+ADE
PRJFILS DESIGN

Five methods ol predicting the lncompresslble, Lwo-dimensloaal

turbulent boundary layer have been applled Lo flow conditions considered
LO occur over the suet ion surrace Or LUI%Omachine blades aud the measure
of agreement between the separation criteria and bouudary layer CharaC-
Leristics assessed. ‘lhe methods considered were those due to Purl,
Ttuckehbrodt, Stratlord, llaskell and SpeKe.

All or the crlLeCta could be brought into tolerable agreement

provided Lhat a value or -O-0:1) was used for BurlQ crlterla, and that
ror Tnxkenbrodt and Spence~s methods the posit ion or separation WA88
determined by the condltlon that local skln ir!CtlOn COerfiCimt is SerOm
It was addltlonally necessary In the methods of tlaskell, Txuckenbx’odt
~- - OVER/
and Spence for the calculation of the shape parameter to be started wlth a and Spence ror the calculation of the shape parameter to be started with a
value or l*I.. value 0r 1’4.

All of the crlterla except Spencecs were sensltlve to Reynolds All of the crlterla except Spencefs were senslt lve to Reynolds
number and showed that an Increase In Reynolds number delays separation. number and showed that an Increase in Reynolds number delays separation.

Stratfordts method ViBs extremely e-sy to apply, was the Stratrord~s method was extremely easy to apply, was the
simplest of the floe and predicted the lovlest pressure rlse to SeparatlOn. slmplest or the rive and predicted the lowest pressure rise to separation.

To ass 1st In the design ot blade prof lles, envelopes of To asslst In the design of blade prorlles, envelopes of
suction surface Velocity dlstrlbutlon have been constructed to give . suction surface velocity dlstrlbutlon have been constructed to give .
separation at the tralllng edge; these are considered to be separation at the trailing edge; these are considered to be
conservatively based. consmat lvely based.

and Spence ior the calculation of the shape parameter to be started with a
value of l-4.

All of the crlterla except Spencefs were sensitive to Reynolds

number and showed that an Increase In Reynolds number delays aeperatlon.

Stratiordfs method yas extremely easy to apply, was the

slmplest or the rive and predicted the lowest pressure rise to separation.

To ssslst in the design of blade proriles, envelopes of

suction surface velocity dlstrlbutfon have been COI’IStNCted to give
separation at the tralllng edge; these are cons ldered to be
conservatively based.
 C.P. No. 868

0 Crown copyright 1966

Printed and published by

HER MAJESTY’S STATIONERY OFFICE
To be purchased from
49 High Holborn, London w.c.1
423 Oxford Street, London w.1
13~ Castle Street, Edinburgh 2
109 St. Mary Street, Cardiff
Brazennose Street, Manchester 2
50 Fairfax Street, Bristol 1
35 Smallbrook, Ringway, Birmingham 5
80 Chichester Street, Belfast 1
or through any bookseller

Printed in England

C.P. No. 86f

S.O. Code No. 23-9016-6