r LEVEAGARD-AR- 175

AGARD ADVISORY REPORT No. 175

Propulsion and Energetics Panel

Working Group 12
on
Through Flow Calculations in
Axial Turbomachines

[DRISBIBTION STALVNA

IApprovod fai public releaa;__

C-3

DISTRIBUTION AND AVAILABILITY

ON BACK COVER
.82 01 04 004
 3 AGARD-AR-175

NORTH ATLANTIC TREATY ORGANIZATION

ADVISORY GROUP FOR AEROSPACE RESEARCH AND DEVELOPMENT

(ORGANISATION DU TRAITE DE L'ATLANTIQUE NORD)

AGARD Advisory Report No.! 75

PROPULSION AND ENERGETICS PANEL

WORKING GROUP 12

on

THROUGH FLOW CALCULATIONS IN AXIAL TURBOMACHINES

Edited by
Ch.Hirsch
Vrije Universiteit Brussel
Dienst Stromingsmechanica , "
Pleinlaan 2
1050 Brussel, Belgium
and
J.D.Denton
Science Research Council
Whittle Laboratory
University Engineering Department
Madinfley Road
Cambridge CB3 0EL, UK

ii

This Advisory Report was prepared at the request of the Propulsion and Energetics Panel of AGARD.
f ",I'
 THE MISSION OF AGARD

The mission of AGARD is to bring together the leading personalities of the NATO nations in the fields of science
and technology relating to aerospace for the following purposes:

- Exchanging of scientific and technical information;

f ,
- Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence
posture;
- Improving the co-operation among member nations in aerospace reseca ch and development;

-Providing scientific and technical advice and assistance to the North Atlantic Military Committee in the field
of aerospace research and development;
-Rendering scientific and technical assistance, as requested, to other NATO bodies and to member nations in
connection with research and development problems in the aerospace field;

Providing assistance to member nations for the purpose of increasing their scientific and technical potential;

-Recommending effective ways for the member nations to use their research and development capabilities for
the common benefit of the NATO community.

The highest authority within AGARD is the National Delegates Board consisting of officially appointed senior
representatives from each member nation. The mission of AGARD is carried out through the Panels which are
composed of experts appointed by the National Delegates, the Consultant and Exchange Programme and the Aerospace
Applications Studies Programme. The results of AGARD work are reported to the member nations and the NATO
Authorities through the AGARD series of publications of which this is one.

Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.

Parts 11, 111 and IV of this publication have been reproduced

directly from material supplied by AGARD or the authors.
b Part I has been set by Technical Editing and Reproduction Ltd.

Published October 198 1

Copyright 0 AGARD 1981
All Rights Reserved

ISBN 92-835-1400-9

Printed by Technical Editing and Reproduction Ltd

flarford House, 7-9 Charlotte St, London, WIP JHD
 CONTENTS

Page

GENERAL INTRODUCTION

PART I - AXIAL TURBINE PERFORMANCE PREDICTION

Notation for Part 1 3
1.1 Introduction 5
1.2 Review of Performance Prediction Methods 13
1.3 Influence of Correlation and Computational Methods on the Prediction of
Overall Efficiency 29
1.4 Influence of Correlations and Computational Procedures on Flow Field Predictions 33
1.5 Discussion and Conclusions 57
APPENDICES
A.I.1 Cambridge Turbine. Geometry, Test Data and Sample Calculation 59
A.I.2 The Two Stage Aero Engine Turbine 69
A.1.3 Ansaldo Steam Turbine. Geometry, Test Data and Sample Calculation 85
A.I.4 Hannover Turbine. Geometry, Test Data and Sample Calculation 99

PART II - AXIAL COMPRESSOR PERFORMANCE PREDICTIONS

11.1 Introduction III
11.2 Review of Loss and Deviation Prediction Methods i15
11.2.1 Survey on Diffusion Factors and Profile Losses 116
11.2.2 Survey of Models for Shock and Shock/Boundary Layer Interaction
Loss Prediction 127
11.2.3 End Wall Boundary Layer Calculation Methods 137
11.2.4 Correlation for Secondary Flows and Clearance Effects 151
11.2.5 Effects of Reynolds Number and Turbulence Level on Axial Cascade Performance 164
11.2.6 Survey on the Effect of Blade Surface Roughness on Compressor Performance 171
11.2.7 Part Span Damper Loss Prediction for Transonic Axial Fan Rotors 181
11.2.8 Deviation/Turning Angle Correlations 184
!I.•.9 Axial Compressor Stall and Surge 212
11.2.10 Summary of Answers to the Questionnaire 214
11.3 Presentation of Test Cases 219
APPENDICES
A.11.3.1 Single-Stage Transonic Compresor and Equivalent lane Cascade 221
A.11.3.2 BBC/Sulzer. 4-Stage Transonic Compressor 229
11.4 Results of Calculations 245
11.4.1 The Through Flow Calculations 247
11.4.2 Evaluation of Profile Loss Predictions Based on Diffusion Factors 269
11.5 Conclusions 281

PART II1 - REVIEW OF CALCULATION METHODS

111. 1 Axial-Flow Turbomachine Trough-Flow Calculation Methods 285
111.2 Blade-to-Blade Computations and Boundary Layer Corrections in Axial Compressors
and Turbines 307

PART IV - GENERAL CONCLUSIONS AND RECOMMENDATIONS 329

WORKING GROUP MEMBERSHIP LIST 331

LIST OF CONTRIBUTORS ACC'-' 3-Of- FOE' 333

-N Tt[S (E••''Ei
LIST OF CONTRIBUTIONS DTIC Ti i-I 335
J aI
J. t

_D1
ByIm,,/,

Av -, : : ' S
,"A . " 'il [o or

Ill
 GENERAL INTRODUCTION

The specialist meeting on "Through Flow Calculations in Axial Turbomachinery" organized

by the Propulsion and Energetics Panel in 1977 showed that, although the basic calcula-
tion methods could be considered as satisfactory, large difference appeared in the perfor-
mance predicted for some test cases by different authors (AGARD CP 195).

This was observerd even with results emanating Crom authors using the same mathematical
calculation method and obviously the central problem in through flow prediction is con-
nected to the various loss and deviation correlations used.
Following the conclusions reached after this meeting, the Propulsion and Energetics Pa-
nel of AGARD recommended the creation of a Working Group on "Through Flow Calculations
in Turbomachinery".

The terms of reference were defined as follows

1. A critical review was to be made of blade element and end-wall correlations available
in the literature or from sources not subject to military or commercial restrictions.
Emphasis was to be put on off design performance prediction and on transonic range of
operation.
This critical review was to be backed by comparisorabetween duct and through flow cal-
culationsusing various correlation combinations and selected well-documentes experi-
mental data on single and if possible, multistage reference machines (compressor and
turbine).
This phase was to be concluded by a precise inventory of the 3hortcomings of the exis-
: 4 ng information and recommandationson the most suitable correlation system.
2. Determination of the trends of the effect of variables such as streamtube thickness
and streamline change of radius in rotors, chordwise losses and turning variation and
other parameters detected through the study under 1, through advanced blade-to-blade
calculations. Tentative incorporation of additional correlationsin design systemsand
check on reference machines.
This phaae was to be concluded by the assessment of the validity of the "computer ex-
periment" versus test for the improvement of machine performance prediction and hope-
fully lead to improved correlations.
The Working Group activity was to be limited to steady state operation of machines.
The activities started in January 1978, under the chairmanship of Prof. J. Chauvin who
initiated this Working Group and kept on contributing to and stimulating the activities
even after having, for professional reasons, left the Propulsion Panel and hence, the
chairmanship of the Working Group.
The structure of the present report reflects the history of the activities and its orga-
nisation. From the start, two separate Subgroups were formed to analyze the turbine and
the compressor fields with their own specific problems, but both Subgroups considered si-
milar strategies in order to fullfill the objectives.

The objectives were soon recognized to represent an enormous task due to the large number
of existing co-relations, tle need to define their limitations and degree of validity,
separate the various effects at design and at off-design conditions and finally try to
synthetize and to bring improvements.

If obviously all the initial objectives were not met, it is felt that a considerable
amount of work has been accomplished. This is due essentially to the voluntary contribu-
tions from many people, not only from the members of the Working Group, representing all
sectors in the field of turbomachinery, industry, research organisations and Universities
with representations from nearly all KATO countries, but also from organizations which
were not represented in the Working Group, and from individuals. More specifically, we
acknowledge the efforts of all the organizations who answered the questionnaire of the
Compressor Subgroup, the Brown Boveri Company, which proposed a four stage compressor
test case, Dr. Groschup and Prof. Bammert of Hannover University, Germany for a single
stage turbine data, the Ansaldo Company (Italy) for providing a multistage turbine test
case, Dr. W.B. Roberts who produced two short reviews of his work on Re-number effects
and partspan damper losses and others who made calculations which .they submitted to the
Working Group.

The first task of the two Subgroups were to collect representative and reliable test ca-
ses of single and multistage machines. This appeared as a difficult task, due to the
lack of publicly available and well documented multistage data.

Nevertheless, each Subgroup succeeded in collecting and presenting in this report several
test cases : Three single stage machines (two turbines and one compressor), two 2-stage
machines (one turbine ans one compressor), a four stage compressor and a six stage low
pressure steam turbine. These data are fully donumented in this report when they are not
available elsewhere in the open literature. Also, because of the importance of blade-to-
blade calculations to sustain the correlations approach, some detailed compressor blade-
to-blade data are included in this report.

•- FI•O IMM FAZEg Ma-N0T naw~

V

-- aa
The second task was to review and attempt to summarize the existing correlation.. If
thin task could be handled more easily in the turbine field, the large number of compo-
nents entering a compressor correlation required a stronger division of the work.

Five main turbine correlations for turning and loss predictions based on loss component
analysis, including secondary and clearance losses were discussed. The complete and com-
prehensive correlations which are reviewed and considered in the turbine calculations are
those developed by the following authors a Ainley and Matthieson (with correction by
Dunham and Came), Traupel, Balje and Binsley, Dejc and Trojanovsky and Craig and Cox?
the latter two providing only the design point conditions. Of course predictions for
the outlet angle are also discussed and reviewed. None of these correlations appeared
to have a sufficiently strong and adequate physical basis in the light of our present
knowledge. This is especially true for high Mach number effects on turning and losses,
shock-boundary layer interactions, and the secondary flow phenomena.

Within the compressor Subgroup, an initial (unpublished) general review of existing cor-
relations was established by Ch. Hirsch as a starting working document. it appeared qui-
te clearly that if many correlations have common points, such as the concept of a dilta-
sion factor to characterise the loading level of a compressor blading, impcrtant diffe-
rences exist when more detailed effects have to be estimated. This is a consequence of
the severe conditions with rospect to the adverse pressuri gradient as well as the com-
plexity of the flow behaviour in a compressor and of our present limited knowledge of it.

In an attempt to answer some of the many questions and problems raised by this prelimina-
ry review, it was considered necessary to have some more detailed surveys of various to-
pics and also some more information about the way the compressor correlations are used in
the practice of the design work.

Therefore, detailed surveys were produced on the following topics : diffusion factor pre-
dictions and profile losses; shock and shock-boundary layer interaction loss prediction;
end-wall boundary layer calculation methods; correlations for secondary flows and clea-
rance effects; Reynolds number effects and influence of blade surface roughness; part
span damper losses; deviation angle correlations. A note was also prepared on stall and
surge. As the reader will notice some of these surveys are extensive reviews of the sub-
ject, while others reflect more the authors' experience. This is the case for the topics
on which little information is existing or available in the literature and the Working
Group was fortunate enough to benefit from the participation or collaboration of some
well known experts in the corresponding fields and willing to present and summarize their
knowledge and experience.
On the other hand, a questionnaire, reproduced in section 11.2.10 was distributed to va-
rious industries and organisations involved in design work in order to gain some informa-
tion on how some of the questions raised by the reviews are answered by the people desig-
ning the actual compressors. The answers from 14 i.dustrial and non-industrial organisa-
tions from six different countries were very stimulating and instructive and their con-
tribution is gratefully acknowledged.

These answers are summarized in 11.2.10 by J. Chauvin, and the shortcomings noted by the
designers should provide guidelines for future work and directions for improvements.

Both Subgroups attempted a series of through flow calculations with different correlations
on the various tcet cases. The calculations within the turbine subgroup were quite sys-
tematic. Different outlet anjle prediction methods as well as various loss distribution
schemes were compared with the data. As a general conclusion it can be said that detai-
led flow predictions are generally poor mainly due to the lack of prediction methods for
secondary flow deviations, although overall performance is fairly well predicted.
Similar conclusions were reached in the compressor calculations, although the secondary
flow effects are less important. Detailed radial distributions of loss and deviation
are, due to the lack of sufficiantly valid physical inputs in the correlations, poorly
predicted. Some important conclusions arose however, such as the need to distribute so-
condary losses and deviations along the span in turbines and the importance of introdu-
cing end-wall boundary layer calculations in the through 7low in order to handle in a co-
herent way the secondary and tip clearance losses.

Due to the importance of the through-flow and blade-to-blade calculation techniques in

the performance prediction method, two additional reviews have been included in the pre-
sent report on these two topics, as part III.

The organization of this report reflects the organization of the activities. The first
part contains the turbine group report while the second part is devoted to the compressor
subgroups' activities. Both parts contain their own conclusions and these are summarized
by some general considerations and recommendations in the last part of this report.

The key to the improvement of the loss and turntng correlations lies clearly in the simul-
taneous development of the physical understanding of the phenomena through basic experi-
ments and the use of two and three-dimensional viscous calculations.

It is the Working Groups' opinion that from a coherent use of fundamental experiments with
the systematic use and analysis of viscous flow calculations a more rigorous basis would
be given to the necessary correlations.

Vi
 PART I

AXIAL TURBINE PERFORMANCE PREDICTION

J
 3

NOTATION FOR PART I

a Throat width
b Back bone length
c Slade chord
d Blade channel inlet width
Blade rear suction surface curvature
Blade pitch (gap)
h Blade height, static enthalpy
H Total enthalpy
i Incidence angle
M Mach number
th Mass flow rate
n Rotational speed (RPM)
p StatiL pressure
PO Stagnation pressure
R Radius
Re Reynolds number
s Entropy
T Static temperature
T, Stagnation temperature
te Trailing edge thickness
U Rotation velocity
v, V Absolute velocity
W, W Relative velocity
Y Stagnation pressure loss coefficient
z Ainley's blade loading parameter
ot Absolute flow angle (from tangential)
Relative flow angle (from tangential)
, I3D Blade angles
7, Stagger angle, specific heat ratio
5 Deviation angle, tip clearance
Boundary layer displacement thickneqs
e Meridional pitch angle (tan' Vr/Vx)
p Fluid density
17 Turbine efficiency
0 Blade camber angle, boundary layer momentum thickness
Ainley's secondary loss parameter
4) Turbint flow coefficient, Vx/U
2
'p Stage loading coefficient 2AH/U
Stage loading coefficient 2Ahis/Ul
Kinetic energy loss coefficient, (VI - V2)/V'

Subscripts

a At throat
is Isentropic
m meridional
M Mid blade height
opt Optimal
s Static
NmO1Dm PAas sLANIl-T ulm
4

Total
tt Total to total (efficiency)
ts Total to static (efficiency)
x Axial
r Radial
S0 Circumferential
0oStator inlet
Stator exit, rotor inlet
2 Rotor exit

The term 'through flow calculation' as used througitout this report refers to an axisymmetric computation of the
flow through a turbine annulus, wherein the flow may be described by a series of concentric stream surfaces across which
no mass or momentum is exchanged. Flow parameters are circumferentially invariant (or represent circumnerential
averages if you prefer).

.4/
 Chapter 1.1

INTRODUCTION

1.1 OBJECTIVES OF THROUGHFLOW CALCULATIONS

Throughflow calculation programs are probably the most important tool of the turbine aerodynamic designer. At
the initial design stage a one-dimensional mean-I'ne calculation might be used to obtain estimates of blade height and so
to lay out a first approximation to the annulus line. Such mean-line calculations usually include estimates of blade loss
and deviation, so that predictions of turbine performance can be obtained, but these must be based only on the blade
geometry at mid-height so high accuracy cannot be expected. Although spanwise variations in flow are small for very
high radius ratio turbines these variations become significant at radius ratios below about 0.9. It is well known that
most turbine blades are remarkably tolerant to off-design incidences (compared to compressor blades), but even so
optimum performance, particularly at off-design conditions, cannot be expected unless the blades are matched to the
spanwise variation in flow. The main objective of a througlflow calculation is, therefore, to provide a prediction of
this spanwise variation so that suitable blade profiles can be selected to cope with the variations in inlet angle, turning,
Mach number; etc.
Throughflow programs may be either of the analysis or design type. In the former the blade geometry is specified
and solutions are sought for the resulting flow pattern. The design type of program is less commonly used and is provided
with the required enthalpy variations (and hence indirectly with swirl velocity) and obtains solutions for the meridional
velocity and hence flow angles. In either type of calculation accurate solutions cannot be expected unless the entropy
gradients resulting from viscous effects are reasonably well predicted. Similarly the solution obtained from an analysis
program is very dependent upon the values of blade exit flow angle specified. In a design program the flow angle is
calculated but it is not possible to lay out blade sections to produce this angle without first estimating the deviation.
A further problem associated with design programs is that the viscous effects have to be estimated at the stage before
blade angles become available.
Once loss and deviation models have been incorporated into a program it can be used to predict overall machine
performance, i.e. the pressure ratio ýmass flow a,•d efficiency : mass flow characteristics at off-design conditions. The
inclusion of spanwise variations in flow should eaable better estimates of these characteristics to be obtained than are
possible from mean-line calculations. However, the accuracy of the results is entirely dependent on the methods used
to predict loss and deviation and in particular their variation with incidence.

Experience suggests that the weakest links in this design procedure are the empirical correlations used to provide
estimates of loss and deviation. It was this realisation which led to the formation of the Turbine Sub-Group.

1.2 REVIEW OF THROUGHFLOW CALCULATION METHODS

It is not the purpose of this report to discuss in detail methods of performing the basic in iscid calculation on which
all throughflow solutions are based. These are well covered in the papers presented in AGARD CP 195 and in more recent
publications by Denton (1978), Doria et al, (1979), Hirsch (1978) and Spurt (1980). It should, however, be mentioned
that the theoretical basis of all such methods rests on the concept of performing the calculation on a mean hub to tip
stream surface, as originally advocated by Wu (195 1). Within a blade row the flow properties on the mean stream surface
are not the same as those of the passage averaged flow, as they are when the f -w is circumferentially uniform far upstream
and far downstream of the blade row. Many methods (so called duct flow methods) only place calculating stations at the
leading edge and trailing edge of blade rows and so do not need to distinguish between the flow on the mean stream
surface and the passage average flow, Methods which do include stations within the blade rows also usually ignore this
distinction. There is no evidence that this approximation has a significant effect on the resulting solution. In fact the
solution outside the blades seems to be remarkably insensitive to the form of the mean stream surface within the blades.
It is, however, highly dependent on the prescribed blade exit flow angles. The rnti, information required from the
solution within the blades is the relative thickness of the blade to blade stream surface. Since the whole concept of
axisymmetric blade to blade stream surfaces within blade rows is a major approximation some error in the prediction of
their thickness is probably acceptable.
The main problem encountered when developing through flow calculations for turbines, as opposed to compressors,
arises from the need to be able to calculate the flow through stages with high-pressure ratio and in particular with regions
of transonic relative flow. The latter is much more easily handled by streamline curvature methods than by stream

.. ,"-j
 function methods although severe difficulties arise even for the former type of method. Time-marching methods are
much better suited to calculating transonic flow but are not yet highly developed fur use in throughflow calculations
(see Spunf (1980)). Problems with calculating transonic flow are currently much more severe in steam turbines than in
gas turbines.
With the streamrli ne curvature method difficulties occur whenever the flow through any blade row approaches the
choking condition (see Novak & Hearsay (1976)). Denton (1978) shows how these problems can be overcome by
calculating a supersonic deviation at the trailing edge whenever the local exit Mach number exceeds unity but his model
does not permit calculating stations to be placed between the throat and trailing edge. This method enables calculations
to be continued to the point whiere the relative exit flow is supersonic over almost the whole blade height but as this
' condition is approached further problems occur due to the extreme sensitivity of the solution to the mass flow rate
specified. Most throughflow methods work with prescribed inlet flow stagnation conditions and mass flow rate but at
choking it is not possible to specify these arbitrarily. For fixed inlet stagnation pressure and temperature there isonly
one mass flow rate for which realistic choked solutions can be obtained. It is then necessary to prescribe the turbine
pressure ratio and to obtain the mass flow rate as part of the solution, This approach is also necessary for multi-stage
high-pressure ratio (but unchoked) turbines because of the steepness (large dP2 /ddh) of their pressure ratio: mass flow
characteristic. Denton (1978) also shows how it is possible to change the structure of throughflow calculations to work
in terms of a prescribed pressure ratio and a similar approach is used by Doria et al. (1979). This prescribed pressure
ratio approach is strongly recommended to anyone developing a program which might be used for high-pressure ratio

Apart from these problems arising from transonic flow and high-stage pressure ratios there are thought to be no
difficulties peculiar to turbine applications of the basic throughfiow methods.

1.3 PHYSICS OF LOSS AND DEVIATION

Methods of estimating loss and deviation will be reviewed in the next chapter but before studying these it isworth
first reviewing the physical processes involved so as to obtain a feel for the degree of understanding available. When
reading Chapter 2 it should be noted how little of this Physics is embodied in the correlations currently available.

Entropy increases in turbines occur due to viscous effects, shock waves and flow mixing (heat transfer is usually
considered negligible). Viscous effects are invariably assumed to be confined to the blade and annulus boundary layeirs
and dissipation inthe mainstream is assumed negligible. The resulting loss on the blade surfaces and in the wake is
referr ed to as profile loss whilst that on the annulus walls isusually (and misleadingly) called secondary loss.

Boundary-layer growth on the blade surfaces is influenced by Reynolds number, surface roughness, surface curvature,
turbulence level and, above all, by surface pressure distribution. In 2D flow it is believed that the effects of all these are
understood and boundary-.layer growth can be computed with fair accuracy (of the order ± 10% on trailing-edge momentum
thickness). However, a correlation method of estimating the loss is unable to take into account many of these factor,
(in particular it cannot know the surface pressure distribution) and a great deal of accuracy is lost. Denton (1973).
compared several correlations with a large body of cascade data and found that the best of them had a standard deviation
of the order 30%/ in the ratio of predicted to measured loss coefficient, In a turbine the curface boundary layers are
- subject to 3D, centrifugal and coriolis effects and to unsteadiness in addition to those found in cascade tests and the
predictions of correlations can be expected to be significantly worse. In particular it is not really known whether or not
it ispossible to maintain a laminar boundary layer on a blade in a real turbine environment.
Tho additional loss due to finite trailing-edge thickness is believed to be comparatively small in most cases but
cunrent methods of esti~mating it are very unsatisfactory (see Herbert (1972)). In particular estimates of the base pressure
acting at the trailing edge vary wildly. This is a particularly important parameter for blades with supersonic outflow

I when trailing-edge losses become much more significant.

In duct flow annulus wall boundary layers can be predicted with similar accuracy to blade surface boundary layers.
However, within blade rows very complex seconciary flows are set up by the influence of the circumferential pressure
gradient on the low momentum boundary-layer fluid. In most turbines it seems that the annulus boundary layers enter-
ing the blade rows are swept completely off the end walls and end up in a passage vortex displaced typically half a blade
chord from the end walls. As a result mainstream fluid isbrought into contact with the wall to form a new annulus
boundary layer. Initially this new boundary layer is very thin and highly skewed and hence is a source of high loss.
Details of this complex process are described for cascade flow by Langston et al. (1976) and Marchal & Sieverding (1977).
In a turbine the process is further complicated by skewing of the inlet boundary layer. The resultant loss per unit surface
area is several times greater than on the blade surfaces but the complete mechanism of loss generation is not yet fully
understood.

The secondary loss is dependent on all the factors influencing profile loss and in addition on the inlet boundary-
layer thickness and profile. When expressed as a loss coefficient averaged over the whole span the coefficient will clearly
vary inversely as the blade height.
 7

Methods of correlating this secondary loss were reviewed by Dunham (1970) who found a disturbing lack of agree-
ment between them with predictions differing by up to an order of magnitude, 'I here is also considerable evidence that
the magnitude of the loss may be different in turbines from that in cascades (see Dunham & Came (1970)).

Tip leakage flows are the other main source of entropy in turbines, The mechanism of loss generation differs
considerably for shrouded and unshrouded blades. For shrouded blades it is necessary to predict the leakage flow rate
as well as the loss because this flow does no work on the blades. Detailed measurements of the flow over shrouded
blade tips have been made by Denton & Johnson (1976) who show that the leakage flow behaves very much like that
through an orifice. The flow contracts to a jet just downstream of the tip seal and the flow rate can be closely estimated
from the ratio of leakage area to blade throat area multiplied by a contraction coefficient. The loss can be estimated
from the mixing loss of this leakage jet. It was found that the leakage flow suffered little change in angular momentum.

The flow over an unshrouded blade tip is much more complex. Fluid will flow over the tip from pressure surface
to suction surface forming a leakage jet whose subsequent mixing produces loss in a manner similar to that of a shrouded
blade. However, the pressure difference driving this leakage jet is not readily predictable in this case because it depends
on the blade surface pressure distribution which in turn is influenced by the leakage. An important point is that this
pressure difference does not decrease for low-reaction blading as does that over a shrouded tip seal. In addition to the
loss it produces the clearance flow will reduce the blade loading towards the tip thereby causing underturning, but no
loss, of the mainstream flow. This underturning results in reduced work and hence increased meridional velocities of
the flow leaving the tip region. A complete correlation of the effects of leakage over unshrouded blade tips should
include predictions of the underturning as well as of the loss,

Profile loss, secondary loss and tip leakage loss are the major sourcýo of entropy in turbines, Their relative magnitude
depends on the details of the design but for many machines all three are similar. At low aspect ratios (less than 1)
secondary loss tends to dominate whilst at very high aspect ratios profile loss may be the major component.

The exit flow angle from turbine blades is usually expressed as a deviation from the angle given by sinf-(orening/
pitch) (note: In this report all angles are taken to be measured from the circumferential direction). This is beca'.ase the
well known sine rule provides a very good approximation to the exit angle in 2D flow. The rule may be derive",
theoretically by applying the continuity equation between the blade throat and the downstream flow at the condition
when the blade is choked with an exit Mach number of unity. If losses downstream of the throat are neglected and if
the specific mass flow is assumed constant, at the sonic value, across the tI'roat the relationship

sin a=, a/g

results immediately. The same method can be used to predict the supersonic deviation which occurs when the exit Mach
number exceeds unity.

In practice there will be some degree of non-uniformity in the flow across the throat and this will reduce the choking
mass flow and hence increase the turning at a downstream Mach number of unity. Similarly the inevitable increase in loss
between throat and downstream will reduce the turning and produce a positive deviation.

Similar considerations (i.e. continuity equation) show that in incompressible flow the deviation increases (under-
turning) with the ratio of suction surface peak velocity to downstream velocity. Hence blades with a high suction surface
diffusion will tend to have a high deviation at low Mach numbers whilst blades with a "flat topped" suction surface
velocity distribution should show little deviation over the whole subsonic range.

These considerations show that the deviation is dependent upon the blade profile and is not solely a function of
the opening:pitch ratio and Mach number. It is also affected by radius change, rotation and stream surface 'hickness.
These factors are included in Traupel's method (1973) of prediction exit angles but are neglected by most other methods.

The 2D deviations discussed above are usually of the order of a few degrees. Much larger deviations can occur as a
result of the secondary flows near the annulus walls. It will be shown in Chapter 4 that the laiter oxe often the main
source of discrepancy between calculated and measured flows in turbines. These secondary deviations result from the
streamwise vorticity generated by the turning of the endwall boundary layers. A passage vortex is formed witb its centre
just off the end wall and its sense of rotation such that fluid very close to the wall is overturned wt.ilst fluid just outside
the inlet boundary layer is underturned.
Methods of predicting this secondary flow have been the subject of a grmat deal of theoretical research, however,
despite drastic approximations the methods are still too complex to include directly in throughiflow calculations. A
severe limitation is that they require a knowledge of the thickness and profile of the annulus boundary layers entering
the blade rows and in practice this is seldom available, Since secondary flow is inherently non-axisymmetric it cannot
be included directly in throughflow calculations. The best hope of including some allowance for it is via correlations
of the magnitude and spanwise distribution of the deviation. The only known correlation of this tvpe is by Bardon et
al. (1975).
1.4 INCORPORATION OF LOSS AND DEVIATION RULES INTO THROUGHFLOW CALCULATIONS

As far as is known there are no loss or deviation prediction methods which have been developed specifically for use
in throughflow calculations. All the overall performance prediction methods reviewed in Chapter 2 (e.g. Ainley and
Mathieson, Craig and Cox, Traupel) were developed as mean line methods and were not originally intended to be used
to predict spanwise variations of loss or deviation. Despite this limitation such methods are not universally used to
predict the loss and deviation on each streamk.ae of a throughflow calculation and this extrapolation beyond their
original intended application must be the cause of some of the errors found when using throughflow calculations for
overall performance prediction.

Loss correlations in general provide a value of loss coefficient, t or Y, whilst the equations used for throughflow
calculations require a value of entropy at each grid node. If there is no loss entropy is conserved along streamlines whilst
loss generation will cause it to increase along streamlines from upstream to downstream grid points. Given the loss
coefficient Z based on relative velocity at the downstream point the entropy increase is easily obtained from

As = AW2 /T.
Note that it is more convenient to use the loss coefficient t which is based on the actual relative velocity than the
alternative definition which is based on the isentropic exit velocity. The use of a stagnation pressure loss coefficient,
"1, is even less convenient as the stagnation pressure is not directly available to the calculation.

It would be realistic to include profile and secondary losses between any two calculating stations which lie within
a blade row. In order to do this the distribution of loss along the axial chord of the blade would have to be specified.
I
In practice this distribution is not known and the loss is usually only added at the trailing-edge station. it is, however,
important that the entropy increase is included in the radial equilibrium and continuity equations L, the trailing-edge
station, otherwise the effect of the loss on the blade exit velocity is not correctly modelled.

It is arguable that a more realistic way of including profile and secondary losses would be to introduce a blockage
factor wi.hin and at the i.railing edge of blade rows to simulate the displacement thickness of the boundary layers. An
entropy increase would then only occur downstream of the trailing edge to simulate mixing of the boundary layers
with the mainstream. As far as is known this model has not been used.
It is clear from the above that a value of loss coefficient must be available to the calculation at each grid point
along the trailing-edge station. In the simplest approach this coefficient may be estimated outside the program and
read in as data. It is, however, more usual to build correlations into the calculation so that the coefficient can be
obtained from the current values of incidence, Reynolds number; etc.

In addition to its effect on the entropy gradient it is shown by Horlock (1911) that the loss enters into the
throughflow equations via a frictional force, Tds/dm, acting along the streamlines. Only the component of this force
acting along the lines along which the momentum equation is being solved (i.e. the quasi-orthogonal lines) enters into
the equations and if these lines are truly nearly orthogonal to the streamlines this component will be very small. Since
the frictional force is anyhow small and not accurately calculable it is not usually included in the equations.

The treatment of tip leakage flow and loss is rather more complicated than that of the other loss components.
For shrouded blades the leakage flow may be assumed to do no work and so must be subtracted from the blade flow
for all calculatng stations between (and including) the leading and trailing edge. The leakage flow suffers an entropy
increase but may be assumed to do no work and to undergo no change in angular momentum (rV6 ), it must be mixed
with the blade flow downstream of the trailir.g edge in such a way that these quantities are globally conserved.
Although in reality this mixing is confined to the tip region and affects only a small part of the main flow it is probably
more usual to assume uniform mixing over the whole span. The latter model completely washes out the tip leakage jets
which are often found in practice.
For unshrouded blades it is much more difficult to formulate a realistic model because the leakage flow and blade
flow arc not readily distinguishable. It is probably preferable not to try to distinguish between them but to increase
the entropy and reduce the turning of the tip streamline to simulate loss and tip unloading respectively. Because it is
difficult to predict how much of the span will be affected by the leakage it is again more usual to distribute the loss
and deviation uniformly over the whole span.

As with tip leakage loss, it is clearly more realistic to concentrate secondary loss near to the end walls so that if
only affects part of the span (of order half a blade chord in length at each wall). Such concentrations of loss near the
endwalls can give reasonable results on a single-stage machine as shown in Chapter 4. However, on a multistage turbine
the accumulation of entropy near the walls can lead to very steep entropy gradients which may cause the solution to
predict reverse flow and hence possibly to break down ccmpietely. In practice such loss accumulations are prevented
by turbulent mixing of endwall and mainstream flow. This mixing is easily simulated in throughflow calculations by
allowing transfer of enthalpy, entropy and angular momentum between streamlines. For example the entropy on
streamline J at the Ith calculating station can be obtained from
 9

Su,= (I - O)S1..1, + ~j(.

001-u
1 1 + Si...ij+i) + as

where a is a mixing factor dependent on the rate of diffusion. Instead of from the usual

SuF S..Ij+ 4S.

The most realistic value of the mixing factor at is not known but values of about 0.5 cure any problems of entropy
buildup ai,.d have little effect on single-stage calculations,
It was stated in Section I1.2 that the shape of the mean stream surface within blade rows has little effect on the
calculated results outside blade rows. Hence it is only really worth including the deviation in the flow angle at the
trailing-edge station. This deviation must be predicted using the local value of opening:pitch ratio and current values
of incidence, stream surface divergence, radius change, etc. available to the program. Correct allowance for the latter
two effects is very difficult to obtain in correlation form and it is common to us-! 2D data, in the absence of better
information, even in highly flared steam turbine stages.

Inclusion of secondary deviation near the end walls is important if full details of the velocity profiles are to be
predicted (see Chapter 4). However, prediction of these deviations demands a knowledge of the annulus boundary-layer
thickness which is seldcm. available. As a result secondary deviation is not usually included separately in the calculation.

Calculations of annulus wall boundary-layer growth and blockage are common in compressor calculations but are
not usually included in turbine throughiflow methods, the reason being that in a turbine an annulus boundary layer does
not really exist within the blade rows. As described in 1.3 any annulus boundary layer at inlet is swept off the end wall
and a new, highly skewed, layer starts within the blade passage. When large regions of duct flow need to be calculated
the blockage due to annulus boundary-layer displacement thickness may become significant. In such cases the thickness
may be calculated using the methods reviewed by Hirsch and de Ruyck in the compressor sub-group's report.

( 1.5 COMPAR~iSON OF CALCULATIONS WITH TESTf DATA

The results from throughiflow calculations must be compared with as much test data as possible in order to establish
confidence in their predictions. The comparison may be made on the basis of overall performance (i.e. mass flow:
pressure ratio, efficient : pressure ratio characteristics) or may include comparisons of measured and predicted velocity
profiles. The latter type of comparison requires much more detailed test data and so is less common than the former.

The predicted efficiency is dependent mainly on the accuracy of the loss correlations and as will be clear from
Section 1.3 high accuracy cannot be expected. Nevertheless many claims to high accuracy art, made for efficiency
predictions, the reason being that it is common practice to scale loss predictions based on cascade data in order to
obtain the best possible agreement with turbine tests. For example Dunham (1970) developed a correlation for
secondary loss in cascades but when used by Dunham & Came (1970) in a turbine performance prediction method the
loss level was considerably increased. If all the many loss components are scaled in this way it is possible to get good
agreement with the turbine data used to determine the scaling factors. For example Dunham & Came claim an accuracy

i of ±2% on efficiency whilst Craig & Cox claim ± 11/4%. Clearly such comparisons of a prediction method with the test
data on which it is based are not really valid and the only valid comparisons are with data not embodied in the
correlations. Few such comparisons have been published but some are included in Chapter 3 of this report.
The mass flow : pressure ratio characteristic is mainly influenced by the deviation rule used. At low Mach numbers
the flow rate obtained at a given pressure ratio will be almost directly proportional to blade exit area, i.e. to sine cii
Hence, at a,1 200, at 10 error in deviation will cause about 5%error in mass flow. For multistage high-pressure ratio
machines changes in pressure drop across the first stage are greatly amplified by the downstream stages so the mass flow
obtained at a prescribed pressure ratio is determined mainly by the throat areas of the first stage. In particular, if the
first nozzle is choked, the mass flow is directly proportional to its throat area for fixed inlet conditions. Hence it is
important that the program is able to distinguish between throat area and blade exit area once the choking point isI
reached.
Compirisons of predicted velocity profiles, blade losses and exit angles are often limited by the accuracy of the
test data. It must be remembered that the flow in a turbine is unsteady and non-axisymmetric (see Hunter (1979))
whilst probe measurements are usually of the time average flow at one circumferential position. A useful check on the
accuracy of traverse data can often be obtained by comparing the integrated mass flow with that measured elsewhere
in the system. It is the author's experience that agreement to better than ±5% is hard to obtain, this error being theI
combined result of non-axisymmetric flow and effects of unsteadiness and shear flow on the probe calibration.
It is particularly difficult to measure blade losses accurately, except for those in the first nozzle row, and losses
must usually be inferred from the local stage temperature drop. This leaves the detailed breakdown of loss betweenI
nozzle and rotor indeterminate.
10

A further source of discrepancy arises if measurements and calculations are not compared at exactly the same
point. Traverses must inevitably be made some distance behind a trailing edge and may be compared with predictions
at the trailing edge. The relative flow angle in particular can change very rapidly between the trailing edge and a
traverse line just downstream of it due to changes in stream surface thickness. If this is disregarded a completely false
picture of the deviation can be obtained.

1.6 OBJECTIVES OF THE TURBINE SUB-GROUP

The objectives of the turbine sub-group can now be explained in the light of the review of its field of activity which
has been presented in the earlier sections of this Chapter.

It was realised early in the work that there were insufficient reliable test cases for which data were publicly available,
to support any attempt to produce new correlations. It was therefore decided to concentrate on reviewing existing
methods of predicting loss and deviation and on applying them to those test cases which were availalt.le. The limitations
of the methods could then be assessed and areas where improvements were prcessary could be pointed out. !t was not
considered that the turbine sub-group would have the time or the data available to implement improvements themselves.

In view of the extreme shortage of turbine test data it was felt that an important contribution would be to fully
document the test cases used by the turbine sub-group and this is done in the Appendices to this report.

REFERENCES

1. Denton, J.D. Throughflow Calculationsfor Transonic Axial Flow Turbines. ASME, J. Eng. Power,
April 1978.

2. Doria, G. Throughflow Calculationof Large Steam Turbines. Conference on Steam Turbines for
Troillo, M. Nuclear Power Stations, Plzen, Czechoslovakia, May 1979.

3. Hirsch, Ch. An Integrated Quasi-3D Finite Element Programfor Turbomachinery Flows. ASME
Warzee, G. Paper 78-GT-56, 1978.

4. Spurr, A. Prediction of 3D Transonic Flow in Turbomachines Using a Combined Throughflow and

Blade-to-Blade Time Marching Method. I. Mech. E. J. Heat and Fluid Flow, December
1980.

5. Wu, C.H. A General Throughflow Theory of Flow with Subsonic or Supersonic Velocity in Turbo-
machines of Arbitrary Hub and Casing Shapes. NACN TN 2302, 1951.

6. Novak, R.A. A Nearly 3D IntrabladeComputing System for Turbomachinery. Parts I and II. ASME
Hearsay, R.M. Papers 76-FE-19 & 20, 1976.

7. Denton. J.D. A Survey and Comparison of Methods of Predictingthe Profile Loss of Turbine Blade.
I. Mech. E. Paper C76/73, 1973.

8. Herbert, M.V. Some Notes on Loss Coefficients Relating to Turbine Blades with Finite Trailing-Edge
Thickness. UK Aero Research Council paper 34755, 1972.

9. Langston, L.S. Three-DimensionalFlow within a Turbine Cascade Passage. ASME Paper 76-GT-50,
Nice, M.L. 1976.
Hooper, R.M.

10. Marchal, Ph. Secondary Flow within Turbomachinery Bladings. AGARD CP 214,1977.
Sieverding, C.H.

11. Dunham, J. A Review of Cascade Data and Secondary Losses in Turbines. J. Mech. Eng. Sci., January
1970.

12. Dunham, J. Improvements to the Ainley-Mathieson Method of Turbine Performance Prediction.

Came, P.M. ASME, J. Eng. Pow., July 1970.

13. Denton, J.D. An ExperimentalStudy of the Tip Leakage Flow Around Shrouded Turbine Blades.
Johnson, C.G. CEGB Report R/M/N848, 1976.
 "1 1

14. Traupel, W. Prediction of Flow Outlet Angle in Blade Rows with Conical Stream Surfaces. ASME
Paper 73-GT-32, 1973.

15. Bardon, M.F. Secondary Flow Effects on Gas Exit Angles in Rectilinear Cascades. ASME, J. Eng. Pow.,
Moffatt, W.C. January 1975.
Randall, J.L.

16. Horlock, J.H. On Entropy Production in Adiabatic Flow in Turbomachines. ASME Paper 71-FE-3, 1971.

17. Hunter, I.H. Endwall Boundary-Layer Flows and Losses in Axial Turbomachines. Ph.D. Thesis,
Cambridge University, 1979.

LI
 Chapter 1.2
REVIEW OF PERFORMANCE PREDICTION METHODS
.
2.1 INTRODUCTION

Turbine performance prediction methods can basically be split into two major categories: The first category groups
together the so-called overall stage methods. The use of i!"ch methods does not in general require any details of the
blading and the complex flow patierns in the turbine are deliberately ignored. Turbines are treated more or less as
black boxes. These methods are in general derived from overall performance measurements of a large number of
turbines with similar characteristics. Their use is justified in the initial design phase for the selection of the turbine
design pat ameters.

The optimization of a turbine for a given duty requires, however, a deep understanding of the flow and only very
refined ptrformance prediction methods, which take into account details of the biading and of the meridional flow
channel can be of real help. This is attempted by the second category of perfarmance estimation methods which evaluate
the total losses as the sum of a great number of individual loss components each of which is influenced by a large number
of geometric and aerodynamic parameters. The problem consists in defining clearly the important parameters and in
making certain that the influence of each parameter cav be studied separately. Such systematic variations of the para-
meters can in general only be done on simplified models. There are, however, limitations to the degree of simplification
beyond which model tests become meaningless. A conistant effort has therefore to be made to simulate the real flow
conditions as closely as possible. The magnitude of the differences remaining between the real turbine flow and the
model flow determines whether the model test results can be applied to the turbine without any corrections or whether
the model test results indicate only the correct trends whilst the absolute loss level needs significant correction factors.
The need to adjust one or more less components in order to match the total losses to measured turbine efficiencies
presents a great problem in cases where one wishes to replace one loss component correlation by another beter one in
a given performance prediction method. Such a change is in fact only meaningful if the overall losses can again be
calibrated against a large number of reference turbine cases. Such test results only exist, however, in large turbine
manufacturing companies and are not accessible to the individual researcher and designer. This situation is dramatically
demonstrated by the efforts of the present turbine sub-group. Only three fully described turbine test cases could be
found in the literature.

2.2 OVERALL STAGE PERFORMANCE METHODS

2.2.1 Design Point Performance

As mentioned before overall stage methods are in general developed for families of similar turbines. Hence their
use is limited to turbines being designed with similar characteristics e.g. same range of degree of reaction, same aspect
ratio, same family of blade profiles. Latimer' has given a number of such correlations which have been used in the past
by Rolls Royce. The most widely known turbine efficiency correlation of Rolls Royce is the correlation of Smith".

The Smith correlation is based on data of some 70 HP and LP cold flow model turbine stages. All stages have the
following features in common:4
- axial velocity ratio of unity,
- design point incidence equal to zero,
- degree of reaction: for the HP-stages down to 20%, for LP-stages up to 60%. The degree of reaction is always
chosen to be sufficiently high to avoid losses associated with recompression at the blade roots.

The turbine efficiencies are presented as a function of the stage loading iý and the flow coefficient *, taken at
the mean radiuL. IF- 4"-..iticiencies were corrected to zero clearance. The Re-number effect on the turbine
efficiency was found to be negligible in the range 10-1 < Re < 3 - 10-. Smith claims an accuracy of ±t2% for his
correlation.

Latimner' produced the following design point correlation which ailows a very rapid estimation of the effects of
such parameters as: trailing-edge thickness, aspect ratio and blade clearance.

Note: The angle definitions used In this review are always those of the original papers.
 142
10-T= (WO),'0+
(h/0), ~ +(~+12 [2 I
- 8.5X + 100OBk/h + 6.8
where:
h/0 olade height to throat ratio,
--

te/O trailing-edge thickness to throat ratio,

?, Ainley's secondary loss parameter (derived from experimental data),
k -clearance,

B -clearance constant,
subscripts s and r indicate stator or rotor.

Soderberg's correlation' also seems to be based on high-speed turbine data. All losses, except the clearance losses
are calculated from the simple relation.

Q5 +/4 t(')(0.975 + 0.075 b/h) - 1I

where:
RH - hydraulic mean diameter at throat section (= 2 -h -g - cos ct2/(g. co
COa + h),
b/h - axial chord to height ratio,
- depends on the overall turning and the maximum blade thickness.

It should be noted that the above relation is only valid for optimum pitch to chord ratio and zero incidence angle.
The tip clearance effect can be accounted for by simply multiplying the blading efficiency by the ratio of the blade
area to the total annulus area (including the tip clearance).

The oversimplified relation ~'=f (flow turning) illustrates the lack of a physically sound concept for the loss
evaluation. It is unrealistic to assume that the degree of reaction for a given flow turning angle does not matter.

Presentation of Soderberg's correlation in the form of efficiency curves on the ~&-~diagram shows very similar
general tendencies to the Smith diagram. For the case of RH = 101 and b/h =3 it was found that Soderberg's
efficiencies were low by 0.5 to 2 points compared with Smith's efficiencies. This difference might at least partially be
explained by the progress in turbine design between the date of Soderberg's correlation ( 1949) and Smith's correlation
(1965).

2.2.2 Off-Design Performance

Based on the experience that turbines tended to sh-nw a unique variation of efficiency with stage loading Latimer'
attempted to produce a simplified model of turbinie off design behaviour. The best consistency was found when using a
logarithmic presentation of the efficiency-stage load relation.

As pointed out by Latimner himself care has to be taken when using his off-design curve because the tuirbine
performance curves depend strongly on whether the turbines are designed for zero incidence or for the incidence
resulting in lowest absolute losses. At zero incidence design the peak efficiency will occur at design load while for
optimum incidence angle designs the turbine efficiency will peak at part load. This difference has to be taken into
account when using Latimer's curve.

2.3 LOSS COMPONENT ANALYSIS METHODS

Authors who use this approach distinguish in general between two categories of loss:
(a) losses used to evaluate the blading efficiency: profile losses, secondary loss and annular losses which are
calculated separately for the stator and rotor blades;
(b) losses which are introduced only between the stages: clearance losses, disc windage losses, losses due to
partial admission lacing wire losses, balance hole losses and guide gland clearance losses:
these losses are taken together as a deficiency term Ai? when calculating the overall turbine efficiency as
77 ' ?ibnding -,&I?.

Although the various loss components are evaluated individually, their final value is adjusted in order to fit vhe maximum
possible number of turbine efficiency data points available to the authors. The difficulty lies not only in the quantitive
adjustment but in the correct selection of the main parameters influencing the prediction.
 is

The number of complete comprehensive loss correlations for estimating axir.l turbine performance is rather limited
as shown by the following list:
- Ainley and Mathieson 4 (with correction by Dunham and Came'),
- Traupel',
- BaiJ6 and Binsley7 ,
- Dejc and Trojanovskis,
- Craig and Cox'.

Ainley-Mathieson, Triupel and Craig-Cox provide means of predicting botlh design and off-design performances
while BaljU-Binsley and Djc-Trojanovski provide only the design point calculation.

Since all methods require a knowledge of the outlet flow angle let us discuss first the various methods of predicting
this most important parameter.

2.3.1 Outlet Flow Angle

There are basically 3 approaches to the determination of the outlet flow angle:
(1) Single airfoil approach using the flow deviation angle B. This approach is widely used for compressor cascades,
The correlations are based on experimental data.

(2) Channel flow approach calculating aý, basically via the throat to pitch ratio. The original empirical relation
ai = f(a/g) has later been substantiated through theoretical derivations by Traupel and other authors.
(3) Calculation cf outlet flow angle using a potential flow theory with a suitable closure condition at the trailing
edge.

Deviation Angle
Carter and Hughes suggest that the unstalled low-speed deviation for turbine cascades may be calculated by:

6= m'-"V/

with m - f (stagger angle and camberline shape),

0 camber angle,
g/c - pitch to chord ratio.

Dunavant and Erwin" provide a limited amount of data of 6 = f(0, ot) for some NACA turbine blades tested in
the range from impulse conditions to axiai inlet conditions.

Mukherjee" 2 seems to be the most recent author providing information on the deviation angle. His experimental
results are presented in the form:

3 = (610 +A)m-AA

where 6 = 6 for blade with 10% thickness,

m = f (blade thickness),
A f (blade camber).

Sine-rule
The sine-rule was in use a long time before a well-founded theoretical explanation could be given for the particular
suitability of this rule in the transonic range.

Ainley 4 proposes the following empirical expression for the Mach -lumber range M2 < 0.5:

c44g/e

where a* = f(a/g) is a linear variation valid for blades with a straight suction surface while the term 4g/e (e = radius
of rear suction side curvature) presents the influence of the rear suction side curvature. At M, 1 Ainley considers
the sine-rule to be valid. For the range 0.5 < M32 1.0, Ainley proposes either a linear variation of %h between the
angle calculated at M2 = 0.5 and at M2 = 1.0 or a smooth curve with an inflection point at M2 = 0.75. Ainley also
suggests that the effect of a positive incidence angle on the outlet angle can be expressed by a linear relationship with a
decrease of a 2 by 2' between zero incidence and positive stall incidence.
16

Bammert and Sonnenscheln 13 carried out low-speed tests on 4-blade profile with evenly thickened and thinned
blade contours. On the basis of these experimental results they arrived at a modified sine rule taking into account the
effect of the trailing-edge thickness.
sin o, = K (+ K2.g

where K, 0.8 and K,= a + 0.8 t (+K..

9 2g)
Traupell has derived a simple equation for outlet prediction Including the Mach number effect on aX2 by applying
the continuity and momentum equations to the space between the throat and far downstream plane. With the two
assumptions that
(a) the resultant pressure force in the tangential direction is negligible, and
(b) the flow is isentropic
he arrives at the following expression:
tan a2 a . ha . 1 I I l+-M
7+ M 1I-- .2 \22 l 7-1
Cosa
g h 2 cos aa 2 \ rcos•a/]
where ha, h 2 = blade height in throat and downstream,
aa, 012 = throat angle (direction normal to throat) and downstream angle.

The equation takes the effect of the suction side curvature into account via the throat angle. It has been verified that
the assumption of a negligible pressure force in the tangential direction is reasonahle for blades with thin trailing edges
and little curvature on the rear suction surface.

In the latest edition of his book "Thermische Turbomaschinen"' 4 Traupel shows a simple way of evaluating the
effect of stream tube divergence, radius change and rotational speed on the outlet flow angle.

PotentialFlow Theory
The successful use of potential flow theory for outlet flow angle prediction depends entirely on the judicious
choice of the closure condition at the trailing edge. At VKI practically all subsonic blade to blade calculations are made
with a compressible form of the Martensen method's. It seems that the Wilkinson condition (i.e. equal velocities on
opposite points on the blade contour near the trailing edge) at present provides the best closure condition. This closure
condition has, however, some shortcomings. It is the experience of K.Sieverding that if the location of the points in
which the Wilkinson condition is imposed is kept constant then the outlet flow angle (referred to tangential direction)
increases slightly with Mach number instead of decreasing as would be indicated by Traupel's equation. Cascade data
confirm the tendency given by Traupel's equation. Furthermore cascade tests at VKI have shown that the point of
equal velocity on pressure and suction side depends on the trailing-edge wedge angle and/or the rear suction side
curvature. The isentropic outlet flow angle can of course be corrected for the effects of boundary-layer thickness and
base pressure.
It seems that in modern turbomachinery practice turbine designers give preference to some sort of modified sine

rule or to outlet flow angle calculations from blade to blade potential flow calculation methods. Deviation angle rules
are nowadays practically out of use. The prediction capability of both modified sine-rule methods and blade to blade
potential flow methods is of the same order of accuracy. From comparison with experimental data on some 25 cascades
it appears that one can expect, at best, an accuracy of ±10. Tests at transonic outlet Mach numbers confirm that the
simple sine rule is very well suited to predict the outlet flow angle for well shaped conventional blades. It is also our
experience that outlet flow angles at transonic outlet Mach numbers calculated with time marching methods agree very
closely with the simple sine rule. It should be noted that the above related experiences refer strictly to two-dimensional
flows.

The turbine sub-group carried out a limited number of calculations for the prediction of the outlet flow angle of a
gas turbine rotor blade (a, = 600, arcsin a/g = 180). The following methods were used: singularity method, finite element
method and three different versions of time marching methods. Both the singularý"y method and the finite element
method predicted an increase of a, with increasing Mach number. The curves difter, however, by as much as 2.50.
The time marching methods predicted, in agreement with the experi ments, a decrease of a2 up to sonic outlet flow
conditions and an increase of a% for M2 > 1.0. The results of the time marching methods lie between those of the
singularity method and the finite element method. Compared with the experimental data their angles are too low by
0.5 to 20.

2.3.2 Profile Losses

In contrast to compressor loss correlations there exist very few turbine loss correlations which do not refer to a
specific blade family designed with one particular design concept. The reason lies in the fact that with some practice
 17

anyone can draw up, not optimal, but reasonable turbine blade profiles. The need for a systematic blade design
optimization is therefore less obvious. Nevertheless, we can distinguish 4 typical steps in turbine blade design practice:

(a) profiles with contours composed only from circular arcs and straight lines,
(b) NACA blqde series and British blade protile series,
Wc) blades with continuous blade curvature and small deceleration rates,
(d) hodograph methods and prescribed velocity distributions.
At present loss correlations distinguish, at the best, only between "old" profiles and "modern" profiles.

The table below indicates the main parameters considered explicitly by the correlations in Section 3 for the
evaluation of profile losses:

Correlation oi a g/c Profile te M2 Re Off-design

Ainley-Mathieson X X X t/c X X X X
Travpel X X X X X X X
BabJd-Binsley X X X X
Dejc-Trojanovski X X X
Craig-Cox X X g/b CR X X X X

Ainley-Mathieson (with corrections due to Dunham & Came) rely on cascade tests as well as on losses derived from
overall tests on a variety of turbine stages.
oil: limited to a range extending from nozzle to impulse blades.
Profile: blades belonging to group a and b (above) with limits of blade thickness 0.15 4 0.25 (outside
of these limits take limit values). c
M2: originally limited to M2 < 1.0. Extension to M2 > 1 by Dunham ind Came"t,

te: correction factors for trailing-edge thicknesses ( * 0.02) are based on theoretical work.

Re: based on chord length and outlet velocity.

Ainley and Mathieson give Re-corrections for both profile losses and overall stage efficiency. For
Re > 1.0 x 10 the correction is the same for both.
Off-design: the losses are evaluated as a function of the ratio i/is where the stalling incidence i4 is a function of
11,ai and g/c. Note: Zero incidence is the incidence giving lowest losses for a given blade.

Traupel'4 : Traupel evaluates the basic profile losses for zero trailing-edge thickness, tpa, on the basis of an
analytical expression for the dissipated energy using a dissipation coefficient of Cd = 0.03. The solution involves a
reasonable choice of the mean blade velocity. The basic profile losses are given as a fiuction of the inlet flow angle
(zero incidence) and outlet flow angle.
te: the evaluation of the trailing-edge losses includes both the losses due to the finite trailing-edge thickness
(Carnot-shock) and due to the difference between base pressure and the pressure ;n the trailing-edge
plane between the blades. The base pressure is based on the work of Hbrner"s.
Profile: the correlation does not refer to a particular blade family. In general Traupel's profiles should be
representative of group C above.
M2 : based on data from various sources Traupel concludes that the Mach number influence depends strongly
on the cascade geometry. For M > I Traupel recommends the calculation of shock losses on the basis
of the supersonic deflection model.
Re: based on chord length and outlet velocity; Re correction in the range I 0W < Re < 10". (For a sandgrain
roughness of k./c = 2 x 10-1 there will be no Re-influence for Re > 3 x 10s.)
g/c: the cascades are supposed to work at optimum pitch to chord ratios. The values of (g/c)opt are given
by Traupel as a function of inlet and outlet angle (theoretical derivation with adaptation to experimental
data).
Off-design: 'p,off = 'p,o + z(V) -
V2
where r',o presents the losses the blade would have if it was designed for the off-design angle oil.
Hence, Po is not a fixed value, it changes with ai. The factor z is a function of (ct, - at*) with
It as design inlet angle. Traupel recognises that there is a strong influence of the blade and cascade
geometry on z but is unable to find well-defined trends. He therefore suggests a bandwidth with an
uncertainty factor of approximately 2 for z.

hhL>L
•.,
- _..,.._,_,_- .... .

Note: Traupel points out that cascades, which are designed with zero-incidence at a given design inlet flow angle d,
might indeed have their optimum at a*. However, for inlet flow angies a* < 600 other cascades can be designed which
have their optimum profile losses at angles a, > a* but which at 4 still perform better than a cascade designed with
zero incidence at ot4, Example: for al* = 30 lowest absolute losses will be obtained with a cascade designed with zero
incidence at a, = 400.

Balij & Binsley7: The profile losses are calculated by assuming a linear variation of the velocity with camber-line
length from the leading edge to the trailing edge. The momentum thickness along the blade surfaces is calculated using
Truckenbrodt's equation for turbulent boundary layers. It turns out that the referred momentum thickness 0/i
(0- camberline length) depends only on the acceleration ratio sin Cq/sin a, and a constant k which is needed to
match the theoretical prdictions to available experimenta) data. The losses (referred to the isentropic downstream
velocity) are presented us functions of the inlet and outlet blade angle.
Te: it might be assumed that Baljd's tinal loss diagram is calculated for zero trailing-edge thickness. The
equation for the losses includes, however, an allowance for non-zero trailing-edge thickness. The base
pressure effect is not included: it could nevertheless easily be added (see Traupel),
Profile: no reference is made to particular blade families. However, use is made of particular functions, relating
both the stagger angle and the cumberline length to chord ratio to the inlet and outlet flow angle.
M2 : there are no indications with respect to thc Mach number level and the possible Mach number influence
on 'p. However, Baljt points out in another paper1 that the losses will decrease with increasing Mach
number since the acceleration ratio will increase. The fact that r -j•6and Binsley use the ratio
sin o,/sin a' for the acceleration rate rather than the velocity ratio suggests that the profile losses are
given for incompressible flow.
Re: the effect of Reynolds number is based on data by Gersten". The Reynolds number (based on outlet
velocity and camber!,ne length) affects the losses up to Re = 10', Since no inforn.ation on the effelt
of surface rot.ghness is given, care has to be taken when using the curve at high Re-numbers.
g/c: the loss diagram is given for optimum pitch to chord ratios (based on Zweifel coefficient 0.9).
0.z
No provision is made to calculate the losses at pitch chord ratios away from the optimum value.
Off-design: no provision is made for off-design calculations.

Deic & Trojanovski': This loss correl.ition is based on experimental data gathered at various Russian institutes.
The basic friction losses NRo are given as functions of inlet and outlet flow angles for optimum pitch to chord ratios at
an outlet Mach number M2 = 0.8 and zero trailing-edge thickness. The optimum pitch to chord ratios, are also given
for M 2 0.8 and te = 0.
%M2 due to the fact that the (g/c)opt is a function of M2 , the effect of a Mach number variation on the
Lsses can be associated with a (g/c) effect resulting from the difference between (g/c)optM,=O.8
aid (?/c)opi,M,2 *0.8.
g/c: the effect ot g/c O (g/c)opt is calculated from the equation
R --- gol+ AR)
where ANR is given as function of

Alt -gopt and (g/c)opt

tRopt
(g/c)Ii was calculated in the preceding etep as a function of the Mach number.
te : the profile losses including the trailing-edge losses are calculated from

kI
rp = r'S+ rte=o+k )

with rte=0- 0.08 and k = fNM

2 and Re).

Re: based on chord length and isentropic outlet velocity. The basic friction losses are representative of the
Ro-range 0.4 x 106 < Re < 1.2 x 106.
Profile: thee investigations refer to blades presented in the Russian profile atlas". The blades belong to the
gro-aps c mentioned above.
Off-design: no information is given.

It should be noted that Dejc and Trojanovski are the only authors to make a distinction between profile losses in
straight cascades and in turbines. The correction factor depends on the axial distance between stator and rotor.
 19

Craig & Cox' present a correlation based on experimental data from straight cascades. The main parineters are:
inlet and outlet flow angles, expressed in terms of a lift parameter, the ratio of pitch to backbone length g/b and the
contraction ratio of the internal cascade passage. These parameters result in a basic profile loss term for zero trailing-
edge thickness which is subsequently modified by correction terms for the influence of te, Re, M, etc.
g/c: the authors are the only ones who use the ratio of the pitch with respect to the backbone length u/b
rather than the chord length g/c. The use of the backbone length has the advantage of being closer
related to the boundary-layer development, however, it is more difficult to handle.
Profile: the effect of the profile shape is to a certain extent indirectly implied by the passage contraction ratio.
The correlation also includes the effect of the rear suction side curvature at high outlet Mach numbers,
M 2 > 0.8.
te: the trailing-edge losses are evaluated theoretically in a similar manner to Traupel.
M, : shock losses are evaluated for M2 > 1.0.
Re: based on throat and outlet velocity. The Re number effect is predicted for the range 10" < Re < 101.
Off-design: Craig & Cox evaluate the positive and negative stall incidence as well as the optimum incidence. The
correction factor for off-design is calculated as a function of the incidence parameter: (i - imin)/
S(istal - gmin)

Denton10 compared four of the preceding loss correlations (the method of Dejc used by Denton is different from
the method presented in this paper) plus his own method"' (which is based on theoretical calculations of a large number
of cascades) with experimental data from some 80 cascades. The comparison was made for near optimum (g/c)-values,
zero-incidence, high Re-numbers and low Mach numbers, It appears from this study that the method of Bali6 & Binsley,
followed very closely by Denton's method, shows the best agreement with experimental data. In spite of this favourable
result for Baljý & Binsley there seem to remain some inconsistencies in this correlation since the losses for nozzle blades
with 20* outlet angle are predicted to be higher than for a blade with the same outlet angle but 45* in!et flow angle.
VKI data confirm that the Craig & Cox method systematically overpredicts two-dimensional profile losses. An

important question is whether two-dimensional profile losses can be applied to turbines without any correction. It
could well be that Craig & Cox have assessed the general level of profile losses for the benefit of a better accuracy of
the final overall performance estimation for which the authors claim an accuracy of - ±1.25% based on the analysis of
more thian fifty turbines.

It seems that any progress in the reliability of profile loss correlations must be along Denton's approach i.e.
establishment of correlations on a theoretical basis for certain blade families where all blades have similar characteristics.
This implies in particular efforts to predict more accurately the boundary-layer transition point and transition length
and the effects of blade roughness and turbulence on both and finally to develop methods to predict locally and fully
separated boundary layers. As a back-up for the theoretical methods some well-defined and well documented special
purpose tests are needed. Such an experimental program could only be carried out with the collaboration of several
research institutes.

2.3.3 Secondary Losses

The number of correlations available bears no relation to the number of comprehensive data sets published in the
literature. There seem to be only two organizations which have provided any systematic data in this domain i.e.
members of the group of Schlichting and members of the University of Dresden (in particular Wolf). Apart from these
data there exists a number of published data of very limited value due to the fact that they are concerned in general
only with one particular aspect of the problem and investigated only a small number of cascade configurations, making
an eventual extrapolation very dangerous.

Most of the data were obtained on straight cascades. The effect of radial migration of losses in real stages is
acknowledged and demonstrated in some cases (i.e., Ref.22) but no systematic investigations are made. The secondary
loss correlations naturally suffer greatly from this lack of understanding of the real flow phenomenon: all correlations
provide an additional loss coefficient to be added to the two-dimensional or mid-span loss for non-twisted blades and
usually for unit axial velocity ratio. Information on the spanwise loss distribution is practically non-existent in the
literature.

It is in particular discouraging and trustrating that the most recent investigations into the effect of inlet boundary-
layer thickness cannot usually be applied to turbine stages because of the difficulty of defining an equivalent endwall
boundary-layer profile. Present investigations in this domain are therefore only justified to the extent in which they
serve as guideline for a better understanding of the physical phenomena, This effort needs to be extended both
experimentally and theoretically to flows in annulas cascades and finally to turbine stages where only full three-
dimensional treatment can provide real progress.

There exist several review papers (e.g. References 13 and 24 which discuss the numerous secondary flow correlations
in detail. As in the case of profile loss we shall only consider those correlations which are integral parts of more or less
20

complete turbine performance prediction methods. Table 2.1 presents the main parameters used in the correlations.
I le background of the various correlations as well as their limitations and peculiarities will be briefly discussed.

Ainley & Alathlesn (corrected by Dunham & Cames): The basic equation for secondary losses on nozzle blades
was mainly derived from test data on bent metal sheet blades. The secondary losses in rotor blades were derived from
performance measurements on turbines with conventional blading (i.e. belonging to groups a and b on page 17) using
the correlation already established for profile losses in stator and rotor, for secondary losses in the nozzle and clearance
loss in the rotor. Ainley showed that the secondary losses for both nozzle and rotor could be expressed by the equation

YS= X.z
2
C cos0 2
where z g- o is known as Ainley's loading parameter.

TABLE 2.1

Secondary Loss Correlations

orlain
Correlation Turning
Trangle Velocity
Angle or
ratio Aspect
ratio /c
g/C ro InleB. L.tEenino
Profile M2 Re second. reg. Off-design

Ainley-Mathieson X A id/od X
(Dunham-Came)
Traupel X V g/h X negl. ind X X
BaU• A h/c X X
Dejc-Trojanovski X A h/c X X X X
Craig-Cox X V h/b ind limited

Note: authors using loading parameters rather than explicitly turning angle and velocity (or angle) ratio are also listed
in columns I and 2.

Aspect ratio: X is a function of the inlet and outlet flow area of the blade passage and the inner to outer diameter
ratio of the blade row. For two-dimensional cascades the relation becomes

1 2/ '
2 cos
(Note: angles with respect to axial direction.)

McDonald' shows that X correlates the loss data much better than the original Ainley parameter if expressed as a
function of cos 23
132/cos P3,. This means that the original Ainley method does not properly account for the aspect ratio
effect. Dunham modifies Ainley's equation as follows:

where f(6*/c) presents the inlet endwall boundary-layer effect, Using Wolf's data2" plus data from other less systematic
investigations, Dunham concludes that the secondary losses increase constantly with 6*/h following the relation
f(6*/h) = 0.0055 + 0.078VS/-7.

This is contrary to the conclusion of Wolf who finds that the loss very rapidly reaches an asymptotic value.

In a later paper by Dunham and Came' it is proposed to replace f(e*/h, by a single constant f(6*/h) = 0.0334.
This modification appears to improve the prediction for low-aspect ratio turbines but it appears that secondary losses
for LP turbines are overestimated.

Re: Dunham & Came suggest that secondary losses are affected by Reynolds number in the same way as
profile losses: i.e. in proportion to
(Re ý-9.2
\2 x 101)

Off-desia: the use of the flow angle 013,in the loading parameter z instead of the blade angle P', implies a-priori
that situations in which 01 differs from 91, can be handled. However, this can scarcely be looked
upon as a method of calculating off-design conditions. Incidence angle effects not only influence the
overall load but also change the chordwise load distribution.
 21

Trmupel: the structure of Traupel's secondary loss correlation is derived frnm. Woins presentation of secondary
loss data obtakued on straight cascades. The quantitative values are, however, obtained by comparison
with unpublished results on turbine stages. Traupel's secondary losses are twice as high as those of
Wolf. The secondary losses are expressed as:

r =-12-F!. with F
rPo h
the factor P indicates that the secondary losses should be increased in the same ratio as that by
which the protfie losses exceed the basic profile losses r...
V1/V2 : the use of the ratio Vj/V2 instead of cos aI/cos a. implies that the correlation does account for
compressibility effects. Using the ratio VI/V 2 without limitation of V2 must certainly lead to an
excessive compressibility effect (due to the rapid change of V1/V,) in the high subsonic range at nearly
constant P32 /0 1,
g/h: in contrast to other authors Traupel uses g/h instead of c/h or similar expressions.
inlet boundary layer: Traupel refers to the results of Wolf indicating that the influence of a variation of the inlet
boundary-layer thickness is limited to a zone 5*/h < 0.03 which according to Traupel is of no interest
in real turbines.
extension of secondary flow zone: the critical height to pitch ratio at which the secondary zones grow together is
given as

The values of 'p should be limited to the subsonic range. Traupel also proposes the calculation of a total blade
loss coefficient in cases where the blade height is below the critical height using the equation

s+ (0.02 0.035) ( /h c/g

The secondary losses at off-design conditions are calculated in the same way as the profile losses at off-design, i.e. by

V2/

BaljO & Bnsley: the authors believe that Wolf's data"s are the most comprehensive set of experimental data available
at present. The authors present Wolf's data in the form of a minimum referred endwall loss coefficient for zero initial
endwall boundary-layer thickness as a function of inlet and outlet blade angle.

inlet boundary-layer thickness: the authors use Wolf's curve

s,6*=0 (

with a limiting value of k = 3 for 5*/h = 0.03.

In contrast to Traupel, who argues that the range of 5*/h < 0.03 is of no interest for turbines, Balje
and Binsley try to take advantage of the variation of r, with 6*/h and suggest applying different
coefficients k for nozzle and rotor.
Re: the authors rely on Wolf whose data do not show evidence of a Reynolds number influence.
extension of secondary flow region: again based on the work of Wolf, Baljb suggests evaluating the critical blade
height for which the secondary flow zones start to grow together through the equation.

_c
= AP
-+!!!A)0.022 in /3
P' . 5* .
h (6*/h) • (h/g) h
(Note: angles with respect to tangential direction.)

Delc & Trojanovski, present a correlation based on Russian experimental data as in the case of profile losses

Aoea[I
(c/h)Rem
w+B
a
[Pi +al ctio 2
ctg P)' gc c08 0 2

(Note: angles with respect to tangential direction.)

LJ-
22

The coefficients A, KI, in and B depend on whether one is dealing with impulse or reaction cascades with laminar
or turbulent boundary layers.

Craig & Cox 9 : the correlation i3 based on experimental data from straight cascaJes. The main variables are the lift
factor already used for the evaluation of the profile losses and the square of the ratio of inlet to outlet velocity:

Xs = Xs,b x NR X Nhr1
with Xsb - basic secondary loss factor,
NR - influence of Re,
Nh/b - influence of aspect ratio.

The authors draw attention to the fact that a distinction must be made between shrouded and unshrouded blade
rows. The correlation proposed here refers to shrouded blades.
(VI /V2 ) 2 : the authors maintain that the correlation holds approximately over a range of incidences (not specified
"bythe authors) provided that the correct velocities are used in evaluating the basic loss coefficient,
h/b: for high values of h/b the losses vary linearly with h/b. For small aspect ratios the tip and root loss
concentrations tend to merge together and the rate of increase is less than would be expected from a
simple inverse law.
Re: the Re number effect is supposed to be the same as for the profile losses.

A comparison of the preceding secondary flow correlations plus five other correlations was carried out by Chauvin"4
for both nozzle and impulse blades with the following conditions: h/c = 1, optimum g/c, low M, and Re = 2 x l0s.
The result is discouraging. The losses of the five correlations reported herein vary not only by a factor of up to five but
their slopes and shapes are completely different. Similar levels of disagreement were found by Dunham2 3 .

Considering the ar,,ount of research work which has been carried out in previous years and, keeping in inind how
few reliable, systenmatic and well documented data are available one must conclude that much of the work his contributed
very little to progress in this field.

Isolated research projects are of little use. What is needed for a better understanding of the generation of losses,
and as back-up for theoretical investigations, are systematic investigations with very well-defined test conditions on a
lari~e number of carefully selected bt ide profiles (possibly carried out in collaboration by several research groups)
coupled with a limited number of very detailed investigations of the whole three-dimensional flow field in stationary
and moving blade rows. The second type of tests are so time and moniey consuming that their number is severely limited.
Hence a general consensus between the various research groups should be reached as to the choice of the particular
configurations. Furthermore better collaboration is needed between theoreticians and experimentalists for analysis of
the test data, It often happens that valuable information gets lost because the. "'perimentalist cannot take full advantage
of the mass of recorded data which, if presented differently, could be helpful tv lhe theoretician in the development of
a new approach to the understanding of complex three-dimensional boundary h,:',%s.

2.3.4 Annulus Losses

This group of losses is dealt with separately only by Traupel and Craig and Cox.

Traupel: considers in particular the losses occurring in stages with shrouded blades, where the flow has to traverse
axial spaces with free shear layers at the ends of the blading. The main influence parameters are:
- the outlet angle from stator or rotor,
- the ratio of the axial spacing to the blade height,
- the overlapping between two successive blade rows (i.e. step changes in annulus height).
(A positive overlapping might result in an "ejector effect" which causes additional losses.)

Craigand Cox: base their annulus loss factor on the sum of 3 individual loss components:

(a) annulus loss factor due to controlled or uncontrolled expansion in the annulus area between two blade rows
(between stator and rotor in general small),
(b) cavity loss factor (i.e. cavities for internal water separation),
(c) cavity loss due to sudden expansion (eg. positive overlapping between two blade rows).

.4inley: considers the annulus losses as part of the secondary losses which of course excludes a detailed analysis
of effects such as overlapping and wall cavities.

"9,.i"
 23

2.3.5 Clearance Flows

The treatment of clearance flows must include both the evaluation of the clearance losses and the effect of the
clearance flow on the mass flow. Due to a lack of data on straight and annular cascades the clea,ance losses are in
general evaluated from turbine tests. This source of information is in general, however, only available to turbine
manufacturers. It appears that the only systematic investigation in straight and annular cascAdes was done by Hubert 2".
His data are used by both Traupel and BaU6 for their clearance loss correlations. Dejc and Trojanovski argue that a
direct application of Hubert's results to turbine stages is not possible because of the influence of rotation which 1.
particularly strong in the case of small clearances.

Detailed measurements of the tip leakage flow rates and losses in both a cascade and in a model turbine are also
reported by Denton and Johnson"7 .

Ainley & Mathieson (Dunham & Came): the authors present the clearance loss as a function of the same load
parameter as the secondary loss. The only difference is that the aspect ratio parameter is replaced by the clearance to
blade height ratio:
SYcl B- _ c -a

z
The difference between unshrouded and shrouded blading is taken into account by the factor B (B = 0.5 for
unshrouded, B = 0.25 for shrouded blades).

An obvious criticism is that the totally different flow mechanism in the two cases is not reflected in this approach.

Dunham & Came have introduced a modification based on Hubert's data which suggest that the usual linear
dependence of losses on tip clearance should be replaced by a power law. The final expression is:
yc C ( 8r, ''
ye, = B"_ - ,] •Z

where B = 0.47 for unshrouded blades and

B = 0.37 for shrouded blades.

In case of multiple seals, Br is to be replaced by

42
6; = 6, x (number of seals)-0 . .

The effect of the clearance on the mean outlet flow angle is evaluated by assuming that the mass flow through the
clearance is undeflected:
I32,cl =tan-' [( - X 1-3, ta
(c)s-F-I 3-X (') (::s 3:)tan l]
hCos
cs0

Traupel': this author takes completely different approaches for unshrouded and shrouded blades. For unshrouded
blades he proposes the equation

uVx 8r,-0.00 2 C

for the rotor and a similar expression for the stator

where K = f () and

G = f (inlet angle and outlet angle).

The structure of the functions K and G are based on Hubert's experimental data. The quantitative values have
been adjusted by experiments on turbine stages.

The effect on the mass flow is expressed by the mass flow coefficient k:

h= k'p V,.A
24

with k =B+C 6r 2 -H
(())DH\ C

for the rotor and a similpr expression for the stator

B = f (endwall displacement thickness) = 0.98 - 1.0,

C = 1.5 to 2.5,
H = f (inlet and outlet angle).

The approach taken by Traupel for shrouded blades follows the evaluation of flows through seals. Hence, the
clearance mass flow is calculated as:
Ii1cI = Ac I )
z P1
with p -- mass flow coefficient,
Acl- clearance area,
z - number of seals,
PI, p2 - upstream and downstream pressure taken at the blade extremity near the seal.

Having determined rhcI the flow coefficient k in the equation

rh = k-p-Vx'A
is calculated as

k =B+
mibi - rhi.

where ihbi = total mass flow through blade row,

B = 0.98 - 1.0 as in the case of unshrouded blades.

A graph presenting ¢ = f (. - ."is used for calculation of the clearance losses.

Baiet' bases his expression for clearance losses on the data by Hubert on straight cascades:

ch
hch = f(6u, A)

For the angle deviation due to the clearance flow BaIJ6 refers again to Hubert, who found that Af3cj varied linearly
with the relative clearance and that all experimental points fall on one single straight curve when relating Acl to an
appropriate expression for the blade lc,:iding:

a AP,-
OCI g 6,4 MO, c
6
where ,u,e = Ag3 2 ,e -Ago3 1 ,
mean vector angle for zero ..,earance flow,
3
2,e -. outlet flow angle for zero clearance flow,

02,c0 outlet flow angle in presence of tip clearance flow,

-
h
A90l = - (02,cl - 02,e) referred outlt angle.

Dejc & Trojanovskil base their correlation on data from actual tests performed by various Russian organizations.
For unshrouded blades the deterioration of the stage efficiency is mainly attributed to a significant distortion of the
meridional streamlines and a redistribution of the mass flow along the radius:

a Au mcu +Rm
 25

where RT, Rm - degree of reaction at tip and mean radius,

7r - pressure ratio across blading,
Al - annulus area between stator and rotor,
77u - stage blading efficiency (including profile, secondary and annular losses).
a 0.45 - 0.65.
For shrouded blades
where equ kr -rDT Sequ RT l+Rm~•)

where 5equ= 4-• 1'5"z'8

6a - axial clearance,
6r - radial clearance,
z - number of seals,

kr = A(2 +
(angles with respect to tangential direction),
A=I for blades with constant profile,
= 1.2 for blades with variable profile.

The authors also give a second equation which is easier to use:

h • sin ot,
tcI = kr h, Ine"?u Xcl

where h, and at refer to the stator exit,

- stage blading efficiency (including profile, secondary and annular losses),
× - f (degree of reaction at mid-span and mean diameter to blade height ratio).

Craig & Cox: For unshrouded blades the authors consider the data of Ainley as fairly representative "provided that
axial velocity renmains approximately constant across the blade row and provided that the relative velocities are well below
the sonic value". For shrouded blades they calculate an efficiency debit

A = FcI N-1 - (zero clearance)

Abl
where A,, - clearance annulus area,
Ab/ - blade annulus area,
= f(.hW2 W2)
Fc, (h' 2W I/tip'

Ah- amount of overlap.

The approach is similar to the one expressed by the second equation for 'cl given by Dejc and rrojanovski.

2.3.6 Miscellaneous Losses

2.3.6.1 Disc Windage Losses

The disc friction moment for a disc rotating in a housing is given by
2
M = 2.p'u "D 3'Cf

where Cf is an empirical friction coefficient.

BaliJ, Dejc & Trojanovski as well as Cox refer to the work of Daily and Nece 29 for the evaluation of the coefficient
Cf. Ainley refers to much older work and Traupel refers to Schlichting.
 26

2.3.6.2 PartialAdmission Loss

Ainley: no indication is given.

Traupel: this author refers to experimental work by Suter and Traupel3 . They distinguish between two loss
components:
(a) ventilation loss,
(b) filling and emptying loss at the ends of the admission sectors

I--e 0.30 • z b

ventilation sector end

loss losses

where C -coefficient dependent on whether non-admitted part of rotor turns in open or closed space and is further
dependent on mean diameter to blade height ratio,
e - partial admission,
z -number of sectors,
b -axial
S-VX/u,
width of rotor, 9,
- Ahis/(ul/2) (stage loading coefficient).

The tip clearance loss does not need special treatment except that for shrouded blades where the partial admission
coefficient e has to be introduced.

Balij & Binsley1 : consider the following loss components:

(a) filling and emptying losses,

(b) scavenging losses,
(c) blade pumping losses,
(d) leakage losses.

The authors refer to work of Stennming for loss component (a) and (b), to Mann and Marston" 2 for loss (c) and to
Linhardt and Silvern 33 for loss (d).

Dejc & TrojanovskiP: Consider only ventilation losses and sector end losses (as in Traupel's method). Referring to
Suter and Traupel, they state that their equation in general leads to too low an estimate of the loss.

Dejc & Trojanovski propose for the case of the uncovered rotor with one single admission segment and e = 0.3 to
0.9:
S
•=0.35 03Xf
03(
-•s--, II- _X•--V!
(D -sina1 e}y t
b
U

where Xf = •2 (Ahstage + V~)

D= D/Do and D. = I m.

In the case of several admission sectors the additional losses are evaluated as

I. 0.25 b 2 X
hA_ b(Z - I )

where b - axial width,

h2 - rotor blade height at outlet,
At - annular area at rotor inlet.

Craig & Cox 9 refer to the work of Suter & Traupel.

[ vI; -
 27

2.4 RADIAL LOSS AND ANGLE DISTRIBUTION

The only suggestion with respect to the radial distribution of loss is made by Traupel:

i?(r) = p -An

outside
of Al

where i? includes only profile and secondary losses. Al presents the extensioti of the secondary flow region:

Al = (3-*5),g y'p

The A?? values have to be chosen such that the integration over r correctly represents the mean value of profile and
secondary losses.

The only known method proposed for radial dictribution of the deviation angle is that of 1Rardon et al.s who base
their correlation on the results of secondary flow theory. Use of the correlation rquires a' wledge of the annulus
boundary-layer thicknesses at entry to the blades. Further discussion of means of distributi ,.' he secondary deviation
is included in Chapter 4.

REFERENCES

I. Latimer, R.J. Axial Turbine Performance Prediction. VKI, LS 1978-2.

2. Smith, S.F. A Simple Correlationof Turbine Efficiency. Journal of Royal Aeronautical Science,
Vol.69, p.467, 1965.

3. Soderberg, C.R. Unpublished notes. G-.sturbine Laboratory, MIT (1949) (see "Axial Flow Turbines",
Horlock, London Butt.rworths 1966).

4. Ainley, D.G. An Examination of the Flow and PressureLosses in Blade Rows of Axial Flow Turbines.
Mathieson, G.C.R. ARC R & M, 1951.

5. Dunham, J. Improvements to the Ainley-Mathieson Method of Turbine PerformancePrediction.

Came, P.M. Trans. ASME(A), Vol.92, No.3, p.252, July 1970.

6. Traupel, W. Thermische Turbomaschinen, Bd. 1, Springer Verlag, 1966.

7. Balj6, O.E. Axial Turbine Performance Evaluation. PartA: Loss-Geometry Relationship. ASME
Binsley, R.L. Transactions, Series A, J. Engrg for Power, Vol.90, October 1968.

8. Dejc & Unter~chungund Berechnung axialer Turbinenstufen. VEB Verlag Technik, Berlin,
Trojanovski 1973.

9. Craig, H.R.M. PerformanceEstimation of Axial Flow Turbines. Proc. Inst. Mech. Engrs, Vol.185,
Cox, H.J.A. 32/71, 1970-71.

10. Carter, A.D.S. A Theoretical Investigation Into the Effect of ProfileShape on the Performance of
Hughes, H.P. Airfoils in Cascades. ARC R&M 2384, 1950.

i1. Dunavant, J.C. Investigation of a Related Series of Turbine Blade Profiles In Cascades. NACA TN 3802,
Erwin, J.R. 1956.

12. Triebnigg, H. Ueber Verluste In Axialturbinenstufen und die Mdgllchkeiten einer optimalen Auslegung.
Mukherjee, D.F. VDI Zeitschrift, Reihe 6, Nr. 7, 1966.
13. Bammert, K. Der ELinfluss von verdickten und verdOnnten Turblnenschaufelnauf die Glttereigenschaften.
Sonnenschein, H. Archiv fur Eisenhuttenwesen, Heft 4, 1967.

14. Traupel, W. Thermische Turbomaschinen, Bd. I, Springer Verlag, 1978.

15. Van den Braembussche, Calculationof Compressible Subsonic Flow in Cascades with Varying Blade Height.
R.A. Journal of Eng. for Power.
28

16. Hoerner, S.F. Base Drag and Thick Trailing Edges. J. Aeronautical Sol., Vol. 17, pp.622-628, 1950.

17. BaUd, O.E. Axial Cascade Technology and Application to Flow Path Designs. PartI: Axial Cascade
Technology. J. of Engrg for Power, Vol.90, October 1968.

18. Gersten, K. The Inftuence of Reynolds Number on the Flow Losses in a Plane Cascade (in German).
Proc. of the Braunschweig Scientific Society, Vol.1 1, 1959.

19. Dejc, M.E. Atlas of Axial Turbine Characteristics. Mashinostroenie Publishing House, Moscow,
Filippov, G.A. 1965.
Lazarev, L.Ya.

20. Denton, J.D. A Survey and Comparison of Methods for Predictingthe Profile Loss of Turbine Blades.
Proc. Inst. Mech. Engrgs, Warwick, Paper C76/73, 1973.

21. Denton, J.D. A Method of Predictingthe Profile Loss of Turbine Cascades. Aeronaut. Res. Council,
33222 Turbo. 176, 1971 (unpublished).

22. Allen, H. ExperimentalInvestigation of Losses in Annular Cascade of Turbine Nozzle Blades of

Kofskey, M.G. Free Vortex Design. NACA TN 2871, 1953.
Chamness, R.E.

23. Dunham, J. A Review of Cascade Data on Secondary Losses in Turbines. J. Mech. Engrg Science,
Vol.12, No.1, p.48, 1970.

24. Chauvin, J. Turbine Cascade Endwall Losses. A review. VKI LS 72, January 1972.

25. Wolf, H. Die Randverluste in geraden Schaufelgittern. Wissenschafd. Z. der TH Dresden, 10,
Heft 2, 1961.

26. Hubert, G. Untersuchunguber die Sekundirverluste in axialen Turbomaschinen. VDI Forschungsheft

496.

27. Denton, J.D. An ExperimentalStudy of the Tip Leakage Flow Around Shrouded Turbine Blades.
Johnson, C.J. CEGB Report R/M/N848, 1976.

28. Utz, C. Experimentelle Untersuchungder Stromungsverluste in einer mehrstufigen Axialturb*ne.

Mittl. Inst. Tern. Turbomasch., ETH Zurich, Nr. 19, 1972.

29. Daily, J.W. Chamber Dimensions Effects on Induced Flow and FrictionalResistance of Enclosed
Nece, R.E. Rotating Discs. ASME Transct., Series D, J. Basic Engrg, Vol.82, No.1, 1960.

30. Suter, P. Untersuchungenaiber den Ventilations-verlust von Turbinenradern. Mittl. Inst. Therm.
Traupel, W. Turbomasch., ETH Zurich, Nr. 4, 1959.

31. Stenning, A.A. Design of Turbines for Hlgh-Energy-Fuel-Low-Po#erOutput Applications. MIT, Dynamic
Analysis and Control Lab., Report 79, 1953.

32. Mann, R.W. Friction Drag on Bladed Discs in Housings as a Function of Reynolds Number, Axial and
Marston, C.A. Radial Clearanceand Blade Aspect Ratio and Solidity. ASME Transact., Series D, J. Basic
Engrg, Vol.83, No.4, 1961.

33. Linhardt, H.D. Analysis of PartialAdmission Axial Impulse Turbines. J. American Rocket Soc., March
Silvern, D.H. 1961.

34. Bardon, M.F. Secondary Flow Effects on Gas Exit Angles in RectilinearCascades. ASME J. Eng. Power
Moffat, W.C. 97A, 1975.
Randall, J.L.
 29

Chapter 1.3

INFLUENCE OF CORRELATION AND COMPUTATIONAL METHODS ON THE

PREDICTION OF OVERALL EFFICIENCY

The process of designing a gas turbine or a steam turbine begins with an evaluation of the influence of component
design parameters on the overall operating cost. For both the steam turbine and the gas turbine, energy costs are rapidly
increasing and component efficiency is therefore a primary design objective. For the aircraft gas turbine the weight of
the component influences the fuel consumption and is therefore also an important energy cost consideration. The
prediction or turbine efficiency is a critical part of the design optimization process for both steam turbines and gas
turbines. ~
Turbine efficiency predictions, together with information which influence the weight and cost, are obtained from
flow field calculations which define the thermodynamic properties and velocity triangles throughout the turbine. These2
computational methods may be full span thr-ough flow calculations which predict the fluid properties from the hub to
the tip between each blade row or they may be mean-line calculations. In either case they are dependent upon loss and
deviation models for their effectiveness in the efficiency optimization stage of the design process. The loss and deviation
correlations whicl' are in common use by steam turbine and gas turbine manufacturers are frequently developed internally
and are maintained:!- proprietary information. The everall efficiency predictions which will be discussed are therefore
limited to results obtained using the correlations built into the methods which are described in the Appendices. For this
reason the quality of the results described in this chapter may not be a valid indication of the capability of today's exist-
ing turbine effiency prediction methods. In addition to being limited to published correlations, this study has also been
constrained by a very limited number of available test cases for which complete data are available. The test case
geometries and operating conditions are described in detail in the Appendices together with test data and sample
calculations. The influence of these correlations and computational methods on the prediction of overall efficiency is
presented in the following sections.

3.1 THE PREDICTION OF DESIGN POINT EFFICIENCY

3.1.1 The Influence of the Computational Method on Efficiency Prediction

It is generally accepted that the quality of any turbine efficiency prediction system is primarily dependent upon
the loss correlations which are used in support of the basic aerodynamic model. As a means of evaluating the importance
of the other factors which influence the efficiency prediction, four individuals using different computational methods
* have each predicted the efficiency of the Cambridge Turbine using the Craig & Cox loss correlations.

Cambridge Turbine - Craig and Cox Correlations

Computational Method Predicted 1 Pred. 7?- Test 17
Streamline curv. (by Denton) 91.47 -2.33
Streamline curv. (by Macchi) 91.9 -1.8
NISRE (by Sieverding) 92.2 -1.6
Finite Element (by Uger) 90.4 -3.4

The Cambridge Turbine has a test efficiency (total-total) of approximately 93.8% although this is subject to a good
deal of uncertainty, possibly ±3%. All of these efficiency predictions are substantially below the test result. This may
have been caused by extensive regions of laminar flow in the Cambridge Turbine which are not accounted for in the loss
corr elations. A significant finding in this comparison is the 1.8%variation in the efficiency predictions with the same
loss correlation. This is thought to be a result of different interpretations of the published correlation rather than the
effect of the same correlation on different calculation methods. In support of this interpretation is the finding that the
same loss subroutine gives very similar predictions of overall efficiency when applied in a mean-line analysis and in a full
throughflow calculation.

3.1.2 'Me Influence of the Loss Correlation on the Efficiency Prediction

Because of the possible significance of other factors on the efficiency predictions, the influence of the loss
correlation can be best evaluated when comparison predictions are made consistently using a single computational
method. This has been done for the Cambridge Turbine by Denton using the streamline curvature method with four
different loss correlations.
30

Cambridge Turbine - Streamline CurvatureMethod (Denton)

Correlationsby Predictedti Pred. i? - Test 17
Dunham & Came 89.87 -3.93
Craig & Cox 91!.47 -2.33
Traupel 91.4 -2.4

Denton 94.3 +0.5

It is apparent from the Cambridge Turbine comparisons that only the correlations used by Denton respond properly
to the low-bpeed, high-efficiency turbine test case. The Craig & Cox ar d Traupel correlations result in almost the same
efficiency prediction for this test case.

Comparisons of efficiency predictions with the Craig & Cox and Traupel loss correlations have also been made by
Sieverding using the NISRE method for the Hannover Turbine test case. The Hannover Turbine has a total-total
efficiency of 90.5 at its design point.

Hannover Turbine - NISRE Method (Sieverding)

Correlationsby Predicted i? Pred. t - Test 71
Craig & Cox 92.2 +1.7
Traupel 93.9 +3.4

For the Cambridge Turbine these *wo loss correlations produced consistent results but for the Hannover turbine
the efficiency predictions vary by 1.7%. Both predictions are significantly higher than the test recult. It is clear from
these comparisons that the selection of turbine loss correlations can have an important influence on both the level of
the efficiency prediction and on the prediction of the efficiency variation between turbines.

3.1.3 The Efficiency Prediction Model

It has been established in the previous sections that both the correlations and a consistent computational method
are important aspects of a turbine efficiency prediction model. Unfortunately there are not enough calculations or
turbine test cases available to give a fair evaluation of most of the correlations and computer methods.

However, it should be recognized that the most important quality of an efficiency prediction model is not its
ability to predict the absolute level of turbine efficiency. It is more important that the model be able to predict the
efficiency variation between turbines. This is the characteristic of the prediction model which is used in the design
optimization process. The streamline curvature model with the Denton Correlation was used to predict the efficiency
of the four test case turbines and two additional turbines with the following results.

Streamline Curvature Method (Denton) with Denton Correlation

Turbine Predicted 27 Test i? Pred. 77- Test q7
Cambridge 94.3 93.8 +0.5
Gas Turbine (2 stage) 92.0 88.4 +3.6
Hannover 91.4 90.5 +0.9
Ansaldo 89.7 90.0 -0.3
NGTE (Ref. 1) 91.5 92.2 -0.7
NASA (Ref.2) 91.5 93.0 -1.5

In four of the six cases the efficiency prediction is within one percent of the test level. It should be remembered
that variations in test procedures and experimental error can contribute to an apparent prediction error. The two cases
where the efficiency is not well predicted represent turbines with unusually high and unusually low efficiency levels.

3.2 THE PREDICTION OF OFF-DESIGN EFFICIENCY

The capability of predicting the off-design performance of a gas turbine is used primarily in defining the engine
characteristics over its operating range. In addition the examination of off-design prediction capability is useful in
evaluating the quality of the efficiency prediction system. Five different operating points are available for the
Hannover Turbine. The off-design performance has been predicted (a) using the streamline curvature method (Denton)
with the Denton correlations and (b) using the NISRE Method (Sieverding) with both the Craig & Cox and Traupel
correlations. As seen in Figure 3.1 the streamline curvature method with the Denton correlation predicts the peak
efficiency level within 1%but is less successful in predicting the shape of the efficiency characteristic. The efficiency
prediction is too high at low loading and too low at high loading. The NISRE calculation with both the Craig & Cox
and the Traupel correlations predicts the shape of the efficiency characteristic reasonably well over the moderate and
low loading range of the data. The predicted efficiency fall off at high loading is too steep with the Traupel correlation
 31

and much too steep with the Craig & Cox correlation, At peak efficiency the level predicted with the Caig & Cox
correlation is 1.7% higher than the test efficiency and the level predicted with the Traupel correlation is 3.4% higher
than the data. In general Denton's correlation is more successful in predicting the level while the others provide a more
accurate shape for the efficiency characteristic.

REFERENCES

I. Smith, D.J.L. Investigations on a Single Stage Turbine of Conservative Design. ARC R&M 3541, 1967.
Johnson, I.H.

2. Kofskey, M.G. Design and Cold Air Investigation of a Turbine for a Small Low Cost Turbofan Engine.
Nusbaum, W.J. NASA TN D-6967, 1972.

90. Oo

'a
U.-
80. /Test //

u / Denton Correlation
V4
V4. NISRE Muthod
S70."' by Sieverding
Craig & Cox Correlation

NISRE Method
by Sieverding
Traupal Correlation

60 . - ; .. .. . ,- ,, -
.3 .5 .7 .9

Loading ,.~
Fig.3. 1 Off-design efficiency prediction

"i.,.
 * -

Chapter 1.4

INFLUENCE OF CORRELATIONS AND COMPUTATIONAL PROCEDURES

ON FLOW FIELD PREDICTIONS

4.1 THE RADIAL EQUILIBRIUM EQUATION

In order to understand the influence of the empirical input on the results of throughflow calculations it is necessary
to examine the relative magnitudes of the terms in the radial equilibrium equation. This equation is effectively the
momentum equation applied along quasi-orthogonal lines which are roughly perpendicular to the streamlines through the
machine. For the simplest case of a radial line it becomes:
(I d2H - T Ls 42 V92 + Cose + V_ • sine+ F,. 4.1
dr dr dr 2r 2 dr rc dm

The last two terms, arising from the convective acceleration along streamlines and from the radial component of )
blade force respectively, are usually small and will be neglected.

Within or at the trailing edge of a blade row V0 is completely determined by Vi and by the geometry of the
prescribed stream surface and we may write

V9 = S1 r--V.coseo 4.2

where a is the angle imposed by the stream surface and would equal the blade angle in 2D flow. Equation 4.1 (with the
last 2 terms neglected) can then be re-written as

d ( a) = (H - rVe)--T Ls-
dV cot 2 a + 2
Vcosec' cotea+ -
,Vm
.M2 cose. 4.3
dr 2 o- dr dr r rc

It is clear by comparing 4.1 and 4.3 that the effect of the enthalpy and entropy gradients, and of the streamline
curvature term, on the gradient of meridional velocity is considerably reduced when ai is small, as it is at the trailing
edge of a turbine blade. Hence at exit from a blade row the velocity distribution is determined mainly by the variation
of exit angle, a, along the span and is only slightly influenced by loss, streamline curvature or inlet stagnation enthalpy
and entropy gradients. This may be illustrated by considering the order of magnitude of the terms on the RHS of 4.3 for
the case of a fixed blade row with no enthalpy or entropy gradients (i.e. a uniform flow) at entry; these are:

dd (H ---rVe) = 0 (1, = O)

Tdsd
dr, - •Vrn cosec•a/Ar
Tm- cot3 - VM2cot2 a/Ar
r
V2
2SIVM cot=i = 0 (&1= 0)
LM coset VMc/o r
rc

where Ar is the annulus height. Remembering that t is likely to be of order 0. 1 and at of order 200 it is seen that the
3rd term is almost an order of magntiude greater than the other terms. Hence it can be concluded that the velocity
profile at exit from a turbine blade is very largely determined by the prescribed distribution of exit flow angle and is
relatively insensitive to the estimates of loss and streamline curvature. Since the exit angle includes the deviation it is
clear that the method used to estimate this will have an important influence on the solution. Exactly the same argument
can be applied at the trailing edge of a moving blade row.

Within a duct region or at the leading edge of a blade row the flow is not confined to a prescribed stream surface
and equation 4.1 must be used. If we consider only changes in velocity profile between the closest upstream trailing
edge and the duct station in question, enthalpy, entropy and angular momentum will be conserved along streamlines
and changes in their gradients will only occur through changes in streamline radius. If the shift of the mid streamline
is of order 6 then the changes produced in these gradients will be of order 2 6/h and will be of opposite signs in the
34

inner and outer parts of the annulus. It may also be shown from actuator disc theory (e.g. Horlock (1978)) that the
change in streamline curvature between a trailing edge and downstream is of order 12l 6/h and that this is of the same
sign in both halves of' the annulus. Hence changes in velocity gradient between a trailing edge and downstream occur
mainly as a result of changes in streamline curvature and the velocity profiles calculated in duct regions are much more
dependent on the estimate of streamline curvature than are those at the trailing edge.

It is important to note that the estimate of streamline curvature at a trailing edge has a significant effect on the
downstream velocity profiles even though it has little effect at the trailing edge itself. This is easily seen by applying
equation 4.3 to a 2D flow with all terms on the RHS equal zero except the last (streamline curvature) term which
a constant, K. If ot is also taken as constant we get, at a trailing edge,
Vm = 2 sin 2Ca(A+ Kr) 4.4

where A is a constant of integration to be determined by the continuity equation. Hence, at the trailing edge we have

VO I Vm2cot Ic = 2 cos' ct(A + Kr). 4.5I

Now applying 4.1 downstream of the trailing edge at a point where the streamline curvature is zero we get

2 4.6
-(1V V -Cos' a (LA + 3K)
dr r

Since r has been assumed very large (i.e. 2D flow) we see that the radial gradient of m2 which was 2Ksinl a
at the trailing edge has become -6K(cos 2 o at the downstream station; i.e. the effect of the curvature at the trailing
edge is multiplied by - 3cot 2 c at the downstream station. This is likely to represent an order of magnitude
amplification.

These results obtained by simple consideration of the radial equilibrium equation will be illustrated by practical
examples in the subsequent sections of this chapter.

4.2 INFLUENCE OF STREAMLINE SLOPE AND CURVATURE

A complete investigation of the effects of streamline slope and curvature on the solution would involve using
different methods of calculating these terms in a program which was othecrwise unchanged and then applying it to the '
same test data. Because. of the effort involved in inserting different streamline slope and curvature routines into already
complex programs this has not been attempted. However, Novak and Hearsay (1976) report that they found little
difference between solutions obtained using spline fits and quadratic fits to obtain slope and curvature and that the
quadratic method was more stable. This result is in agreement with the findings of Wilkinson (1970) who investigated
the accuracy and stability of a large number of curve fits on a sinusoidal streamline. He concluded that polynomial fits,
using the same curve to obtain both slope and curvature, were the most accurate and stable whilst the use of splines and
of 'double differentiation' were undesirable.

A simple method of investigating the influence of the streamline slope and curvature on the solution, without the
need for any re-programming, is to vary the number of calculating stations used between the limits of a very coarse and a
very fine grid. This was done on a simple test case consisting of the single fixed blade row shown in Figure 4. 1. The
casing was made up of circular arcs and straight lines so that the discontinuities in curvature at the leading and trailing
edges and at mid chord would provide a severe test of the method. The inlet flow was taken to be uniform and axial
and the stream surface angle (tan- Vt/Vx) varied linearly with axial distance from zero at inlet to 650 at the trailin 6
edge. The blades were taken to have no thickness and, initially, zero loss.

Solutions were obtained using a quadratic curve fit (i.e. 3 points) to obtain both slope and curvature, initially withI
a coarse grid having no stations internal to the blade row and then with progressively finer grids having 1, 3 and 5 internal
stations in addition to those at the leading and trailing edges. Although there is no analytical solution for this case the
latter, fine grid, solution is expected to be almost exact.

Figure 4.2 illustrates the results at leading edge, trailing edge and far downstream (at x =0.4). At the leading edge,
Figure 4.2a, the velocity profile would be uniform were it not for the effects of streamline curvature. As expected
the use ot no internal calculating stations leads to an underestimate of the streamline curvature and hence velocity
gradients. Using one central internal point causes the parabola to overestimate the curvature at the leading edge but the
maximim error in Vx is less than 3%. This is roughly halved by using 3 internal points. The surprisingly accurate results
obtained with coarse grids are due to the fact that the curvature only varies rapidly, on the casing. At mid span changes
in streamline slope are much more gradual and reasonable results can be obtained with few points.

At the blade trailing edge the results, Figure 4.2b, confirm the conclusions of the last section that the solution is
very insensitive to the value of streamline curvature as varied by the number of internal calculating points. The steep
velocity gradient is almost entirely due to the radial equilibrium terms in equation 4.3.
 35

The changes between the trailing euge and far downs ream shown in Figure 4.2c are exactly as predicted at the end
or section 4.1. The effect of curvature is to slightly steel.en the velocity gradient at the trailing edge. When the curvature
is relaxed downstream the gradient is reduced to considerably below the value it would have had If there had been no
curvature at the trailing edge. The overshoot is greatest at the tip where the curvature was largest.

It should be emphasised that this is a severe test case with higher curvatures and pitch angles than are likely to occur
in most gas turbines, In most cases the use of a coarse grid with no internal calculating stations is likely to lead to even
less error than was found here.

4,3 INFLUENCE OF BLADE EXIT ANGLE

The effect of the blade exit angle on the solution was investigated using the same simple test case described in theI
previous section. All results described in this section used 5 internal calculating points with the flow angle varied only
at the trailing edge station.

F I Figure 4.3 shows the result of increasing the turning by 20 (i.e. - 20 deviation) over the whole span. The mass flowI
was held fixed (at 150 Kg/s) and so the 7.5% reduction in blade exit area required a higher exit Mach number and
consequently lower back pressure. The main effect on the solution is the increase in velocity level resulting from the
lower density but there is no significant change in velocity gradient either at the trailing edge or far downstream. This
result is predictable from equation 4.3 which shows that changes in the gradient of exit angle are more significant than
changes which are uniform across the span.

Figure 4.3 also shows the result of a parabolic variation in exit angle with 2* of overturning at root and tip. As
expected the effects on the velocity gradient are much greater than previously but because the change of exit area is less
there is much less change in velocity level. The result confirms that small changes in the spanwiie gradient of exit angle
have a large effect upon the predicted velocity profiles.
To illustrate the sort of discrepancies between measured and predicted exit angles which can occur in practice
Figures 4.4 and 4.5 show comparisons for the Hannover turbine and Figure 4.6 for the nozzles of the Cambridge turbine.
At mid span most of the deviation rules appear to be accurate to about ± 20 but Traupel's method appears to consistently
overestimate the deviation. The large differences between measurement and prediction which occur near the end walls
(particularly at the rotor root) are typical of the effects of secondary flow. They result in large discrepancies between
calculated and measured velocity profiles as illustrated by Figures Al1.2 and A4.3,

It is clear that accurate predictions of turbine velocity profiles cannot be obtained unless some attempt is made
t.n model the secondary deviation. The results of such attempts are described in section 4.5. The errors of a few degrees
in deviation at mid-span have comparatively little effect on the predicted velocity profiles but are very significant as
regards the prediction of the mass flow pressure ratio characteristic.

4.4 EFFECT OF LOSS ESTIMATES

- The effect of loss on the flow field predictions will again be illustrated using the simple test case of Figure 4. 1.
Figure 4.7 compares solutions with no loss and with a constant value of =0. 1 over the whole span. The increased
loss causes an increased pressure drop at constant mass flow and hence an increase in velocity level. However, it can be
seen that there is also a reduction in the velocity gradien~t at the trailing edge. This arises because, although the loss
coefficient does not vary across the span, the entropy produced (AS = I V2 t/T) varies as the square of the relative
velocity being greatest at the hub where the velocity is higher. The resulting negative entropy gradient can be seen from
Equation 4.1 to produce a more positive velocity graidient, Physically the frictional forces acting along the streamlines
are greatest at the hub where they produce a greater reduction in velocity.

The same explanation applies to the results with a parabolic variation in loss coefficient which are also shown in
Figure 4.7. The loss coefficient has been taken as zero at mid span increasing parabolically to 0.2 at the end walls. The
result is a significant reduction in velocity near the walls but no change in velocity gradient at mid span. This is a result
of a comparatively large variation in t. In many situations the spanwise variation of loss coefficient and the consequent
effect on the velocity distributions will be less than shown in Figure 4.7, In particular, for high aspect ratio blading, the
low levels of loss (of order a few percent) over most of the span will cause most of the effect to be concentrated near the
end walls. In this region the effects of loss gradients may well be dwarfed by those of angle variation. This is well
illustrated by Figure 4.9 which shows the result of distributing the loss in the rotor of the Hannover turbine in the
manner shown in Figure 4.8. It is clear that only a very small part of the disturbance near the end walls is due to loss
gradients. Similar results were obtained for the Cambridge turbine.

The spanwise extent of the region of increased loss near the end walls depends greatly on the inlet boundary layer
thickness, being typically about twice the overall thickness at entry to the blade row. However, there are no correlations
available to predict its extent and the boundary layer thic~kness ir not usually known. Hence any prediction must be
based on some aspect of blade geometry which is known. It may be argued that the diameter of the secondary vortex
36

is limited to the dimension of the blade opening. Hence, in the absence of any alternative, it is suggested that this should
be used as a measure of the extent of the region of increased loss. It must be emphasised that this suggestion is not based
on any experimental evidence but is presented as a hypothesis to be tested against whatever evidence becomes available.

4.5 DISTRIBUTION OF SECONDARY DEVIATION

It was concluded in section 4.1 that '.ne spanwise variation of blade exit angle was the most important factor
influencing the predicted velocity profiles. In view of the large and rapidly varying deviation angles caused by secondary
flows it is therefore important to try to include some measure of these in the throughflow calculation,
Secondary flows in turbine blades can only accurately be predicted by fully 3D methods and some results from such
a method are presented in the next section. The classical approach to the calculation of secondary flows is based on
analytical theory and involves assumptions (such as small disturbances and no distortion of stream surfaces) which are
not at all valid in turbine blades, Nevertheless these theories appear to give reasonable predictions of the spanwise
variation in pitchwise average exit angle. Unfortunately such methods are still too complex to include directly in
throughflow calculations, even if th. necessary data in the form of detailed inlet boundary layer profiles were available.
The only possibility therefore remains the use of correlations to predict the magnitude and extent of secondary flow,

Such a correlation has been proposed by Bardon et al. (1975) based on the application of secondary flow theory to
a range of cascades. Although its authors only compare its predictions with a small amount of cascade data, this correla-
tion remains the only method available for estimating and distributing the secondary deviation.
)
The deviation is distributed over two inlet boundary layer thicknesses in the simple manner shown in Figure 4.10.
The magnitude of the maximum overturning and underturning are the same and are obtained from the equation

Aa/e = 0.068 + 0.03 logj 0 P 4.7

where e is the blade turning and the parameter P is given by

P =(g' +gIh/2)82 .8
The parameter g' is a corrected pitch given by

g COS 2 49
=Cosa2,I g.

where -f is the flow angle measured from the normal to the stagger line.

The only parameter required which is not readily available to a throughflow calculation is the inlet boundary layer
thickness 5. As described in the last section it is reasonable to assume that the diameter of the secondary vortex is
limited to the blade opening and hence that the maximum underturning occurs at this distance from the end wall. On
this basis, it is suggested that in the absence of any other information the inlet boundary layer thickness 6 should be
set equal to the blade opening a.

It should be noted that the correlation breaks down at very low values of 6 when very high deviations are predicted
and also at very large values of 6/g' when the deviatiorn can become negative. In practice the secondary flows occupy a
significant spanwise distance, and the deviations remain finite, even when the inlet boundary layer thickness is zero. They
should tend to zero for very closely pitched blades. It is also not realistic to expect the blade height to affect the
deviation when the secondary flow regions do not interact as in the case of high aspect ratio blades.

Despite these limitations the correlation of Bardon et al remains the only one available and so was built into the
throughflow calculation method of Denton (1978) which is described in Appendix 1. At the same time the secondary
loss correlation was replaced by that of Morris and Hoare (1975) and the secondary loss was distributed linearly over two
inlet boundary layer thicknesses at each end wall. In order to obtain meaningful results from this distribution of loss
and deviation it is clearly necessary to have several streamlines within tv., boundary layer thicknesses of the end wall,
For normal calculations adequate accuracy can be obtained with only 9 streamlines but for the calculations reported in
this section 21 streamlines were used. This decrease of streamline spacing caused no instability problems when running
the program.

The method was first applied to the single nozzle blade (Figure 4.1) used as a test case earlier in this chapter.
Figure 4.11 shows the predictions with two different estimates of inlet boundary layer thickness. With 6 = 0.1 h the
predictions look realistic for a nozzle with thin inlet boundary layers. With 6 a the secondary flow region at the tip
appears unreasonably large in extent for a nozzle with this aspect ratio.
 37

The Cambridge turbine provides an ideal test case because the inlet boundary layer profiles are known accurately
and detailed measurements of the secondary flow were obtained. Figure 4.12 shows very good agreement of measured
and predicted exit angles from the nozzles arnd this is reflected in improved predictions of the axial velocity profile as
shown In Figure 4.13. The agreement is less good behind the rotor, as shown in the same Figures, but is still a significant
improvement on the prediction without concentration of loss and deviation at the end walls (Figures Al1.2 and Al1.4).

The annular cascade tested by Goldman and McLallin (1975) would also provide a good test caw, for this method
but the reference was not discovered until the calculations had been completed.
The inlet boundary layer thickness is not so well defined for the Hannover turbine but from the inlet velocity
profiles it was estimated to be 15 mm at the hub and ICmm at the tip; these values were used for both blade rows&
In this case the correl,'tinn appeared to considerably overestimate the deviation at nozzle exit, Figure 4.14,A
particularly at the hub where the exit angle is low (close to tangential). Figure 4.14 also shows the exit angles
predicted by Sieverding using a simple form of classical secondary flow theory.

The deviation is again overestimated at the hub but the underturning is in better agreement than the predictions
of Bardon's correlation. The exit angle from the rotor blade varies in a complex manner and as shown by Figure 4.14
the correlation fails to predict its details adequately. The increased deviation measured very close to the end walls is
an unusual feature, not included in the prediction, which was also found to a lesser extent behind the rotor of the
Cambridge turbine. (Figure Al .).

The inlet boundary layer thickness to the 2 stage gas turbine is not known and so the assumption that 6 = a was
applied to this case. The test results show that the flow in this machine is considerably influenced by secondary flows
particularly at exit from the rotor blades. Figures 4.15 and 4.16 compare the measured and predicted relative flow angles -

at exit from each blade row. It is clear that putting 6 = a considerably overestimates the extent of the secondary flow
region for the nozzles but underestimates it for the rotors. However, the magnitudes of the deviations are, if anything,
overestimated for the nozzles and reducing the estimate of boundary layer thickness would cause even larger deviations to
be predicted.
It is also noteworthy that at the tip of the LP rotor the secondary deviation predicted appears to be of the wrong
( sign (i.e. underturning at the wall). This was traced to a negative deviation being predicted by Bardon's correlations as
a result of eauation 4.9 producing a very low value of corrected pitch. The use of the corrected pitch is clearly suspect
for low camber tip sections where there is little difference between the stagger angle and flow exit angle (i.e. 72 in
equation 4.9 becomes small).

These studies of the effects of distributing the deviation have concentrated on comparisons of relative blade exit
angles. However, the trends of the velocity distributions can be inferred from these using the observation that the
magnitude of the relative blade exit velocity is not greatly changed by the deviation. Hence a region of under turning
will correspond to a region of high axial velocity and vice-versa.

In summary the results of distributing the secondary deviation are disappointing. Although encouraging results
were obtained in the Cambridge turbine neither the magnitude nor the extent of the secondary deviation were well
predicted for the other two machines. It seems likely that the secondary flows are greatly influenced by blade profile
and so cannot be at all well predicted by correlation methods.

4.6 FULLY THREE DIMENSIONAL CALCULATIONS

If the conclusions of the previous section are accepted it seems likely that the full co)mplexity of turbine secondary
flows can only be predicted by fully 3D methods. Such methods are now becoming available but are not yet widely
used because of the cost of the long computer runs needed. Viscous effects may also be included (e.g. Dodge (1976),
Moore &Moore (1979)) but these greatly hicrease the complexity and cost. The only 3D method available to the Turbine
Sub-Group was the inviscid method of Denton (1975) the latest form of which is described by Denton & Singh (1979).
This method has the ability to calculate shear flows although the amount of detail obtainable is limited by the number of
grid points which can be used. It was used on the Cambridge turbine and on the 2 stage gas turbine test cass.&. In order
to obtain sufficient detail in the secondary flow region separate calculations were performed for the inner and outer
halves of each blade row, the mid stream surface, as obtained from a throughflow calculation, being input as a solid
boundary with no shear layer on it.

Results for the nozzles of the Cambridge turbine are shown in Figure 4.17 and are very encouraging despite the factI
that only 3 streamnwise grid surfaces could be placed in the secondary flow region. Both the measured velocity profile
and flow angles after the nozzles were input to the calculation for the rotor blade. The results shown in Figure 4. 18 are
not as good as those after the nozzles but are still better than were obtained from throughflow calculations. Clearly the
complex 'Rngle distribution near the rotor root could not possibly be resolved with so few grid points.

3D calculations on the 2 stage gas turbine were reported by Denton & Singhi (1979). These were for the whole HP
stage (in one calculation) and did not include any shear flows. However, it is noteworthy that good agreement was
 r * -. . ...... .. ". . ".• - "- -.-.. w=° "- - --- " = • ( .

38

obtained for the relative exit Mach number distribution from each blade row despite the fact that the calculation was
inviscid and made use of no loss or deviation correlations, only geometric data being input.

The calculations with shear flow were confined to the HP turbine and each blade row was treated in two parts as
previously described. The thickness and shape of the inlet boundary layers are not known for this machine and so were
guessed, the thickness being guessed at 20% of blade height on each end wall. As shown in Figure 4.19 this was clearly
an overestimate for the nozzle blades, the calculated secondary flows are much too extensive and the amount of
underturning is too small. However. it was not thought to be worth attempting calculations with thinner boundary
layers because of the limitations on tn.- number of grid points previously mentioned.

The predictions at rotor exit show some signs of the large region of overturning near the hub but it is thought
likely that this is largely due to viscous effects (i.e. a flow separation) within the blade row. However, the blade surface
pressure distributions, which are obtained as an integral part of the 3D calculation, showed no cause to expect such a
separation. The spanwise variations of relative Mach number obtained from the 3D calculations were in excellent
agreement with the experimental data when the blade exit pressure (which is input to the calculation) was adjusted to
give the correct value on the hub.

On the basis of these limited comparisons it seems that fully 3D calculations show promise of being able to predict
secondary flows in turbines but that fine grids and hence expensive computer runs will be needed to obtain accurate
results with thin boundary layers. The calculations described each took about 8 minutes CPU time on an IBM 370-165.

ACKNOWLEDGEMENT

The 3D calculations described were carried out by Dr S.Usui of Nippon Kokan K.K., Technical Research Centre,
whilst he was studying at the Whittle Laboratory, Cambridge.

REFERENCES

Horlock, J.H. Actuator Disk Theory. McGraw Hill, 1978.

Novak, R.A. A Nearly Three-DimensionalIntrablade Computing System for Turbomachinery, Parts I and II.
Hearsay, R.M. ASME Papers 76-FE-19 and 76-FE-20, 1976.

Wilkinson, D.H. Stability, Convergence and Accuracy of Two-Dimensional Streamline CurvatureMethods using
Quasi-Orthogonals,I. Mech. E. Thermodynamics and Fluid Dynamics Convention, Paper 35,
1970.

Bardon, M.F. Secondary Flow Effects on Gas Exit Angles In RectilinearCascades. ASME J. Eng. Power,
Moffatt, W.C. Jan. 1975.
Randall, J.L.

Denton, J.D. Throughflow Calculationsfor Axial Flow Turbines. ASME J. Eng. Power, April, 1978.

Morris, A.W.H. Secondary Loss Measurements in a Cascade of Turbine Blades with Meridional Wall Profiling.
Hoare, R.G. ASME J. Eng. Power April, 1978.
Goldman, L.J. Cold Air Annular Cascade Investigation of Aerodynamic Performanceof Core Engine Cooled
McLallin, K.L. Turbine Vanes. NASA TM-X 3224, 1975.

Dodge, P.R. Numerical Method for 2D and 3D Viscous Flow. AIAA paper 76-425, 1976.

Moore, J. A CalculationProcedurefor 3D Viscous, Compressible Duct Flow, Pts. I and I1. ASME, J.
Moore, J.G. Fluid Eng., Vol.101, Dec. 1979.

Denton, J.D. A Time Marching Method for 2 and 3 DimensionalBlade to Blade Flows. ARC R. & M. 3775,
1975.

Denton, J.D. Time MarchingMethods for Turbomachinery Flow Calculation. V.K.I. Lecture Series 7, 1979.
Singh, U.K.

# * 'I ,
 39

mm

zr

enn

Loo

ItI
c..
 40

120
FIG 4.2o LEADING EDGE i
0
5 INTERNAL POINTS
110 -- x-- NO INTERNAL POINTS/

100p - o- -x- - __
M/S

80 I I I I
0 0.2 0.4 0.6 0.8 1.0
y/h

90 \
Vx
m/s
80 -- ,

[ 70 -
60- FIG 4.2b TRAILING EDGE

5 INTERNAL POINTS
-- x--- NO INTERNAL POINTS

500 I I 6
500 0.2 0-4 y/h 0'6 0.8 1.0

Fig.4.2 Ea'ect of number of internal calculating stations

--- --- - -
-- 41

Vx\ X
rn/s ,

80-

70 FIG 4.2c

FAR DOWNSTREAM -

5 INTERNAL POINTS
60- NO INTERNAL POINTS-9':

50 I I
0 0.2 04 0.6 0.8 1.0
y/h

Fig.4.2 (concluded)

100- -

o•X

Vx N
m/s

80 " -

TRAILING EDGE
70 -__ NO DEVIATION
- - 20 DEVIATION
-- 0o-- PARABOLIC DEVIATION
60 20 AT ENDS 00 AT MID SPAN
S. I I %
0 0"2 0"4 0-6 0./ 1-0

Fig.4.3 Effect of deviation

42

ANGLE CORRELATION SYMBOL

AINLEY - -o
TRAUPEL
BAMMERT
SIN- RULE

30°

loww

100 I00,

/ ,...... LA.,-,

0 0-2 0-4 y/h 0,6 0-8 1.0

Fig.4.4 Hannover turbine nozzle ex~it angles

 43

ANGLE CORRELATION SYMBOL

AINLEY
TRAUPEL
BAMMERT
SIN RULE ---- o

300

A %

200- \%-

loaf
I'* E PE MI ýE

0 0.2 0.4 y/h 0.6 0.8 1.0

t ~Fill.4.5 Hannover turbine rotor exit angles,

 + 00e
340

00001+

3Ole

if I'4
012-ee

401 000/h1
2i.. C2big ubn nzl xtage
 45

90-

m/s
80 _

70-- TRAILING EDGE

NO LOSS X%
60-- 60 ---- CONSTANT LOSS =0.1
-- 0- PARABOLIC LOSS VARIATION
0-2 AT ENDS.= 0.0 AT MID SPAN
! so I I I I
0 0"2 04 y/h 06 0.8 1.0

Fig.4.7 Effect of loss

46

LOSS DISTRIBUTION -SYM-BO-L

LOSS CORREL
CRAIG & COX EA
PARABOLIC
LINEAR
TRAUPELPARABOLIC

0-4

0,3

O.2 1%

a MEASUREMENTS
0 .11
0~1-%

0.2 0-4 y/h 06tB1

0
distribjution at rotor exit
Fig,4.8 Hannover turbine loss
 47

LOSS CORR. ANGLE CORR. LOSS DISTRIBUTION SYMBOLS

CRAIG &COX ANGLE LINEAR o
SF""PARABOLIC

Vax 22

1.0 /
a I

*MEASUREMENýTS

II
08I

I I Sg
0~6 I UI

t !t
V

04-

0-2-

02--

00 I I yh I I
0 0o2 0.4 y/h 0.6 08 1.0

Fig.4.9 Hannover turbine effect of loss distribution at rotor exit

L . i
-k" ... ... .. . . ..
 48

DEVIATION

AOL

I 26

Iy

Fig.4.10 Bardon's distribution of deviation

BO, n --
100
90 -

Fy

•._c
70----- TR ITING UTDED LOS \N P 10%

DEVIATION, 6H-= 6T - 0.1 x h •

DISTRIBUTED LOSS AND
50 ---- DEVIATION, 6 = OPENING

0 0-2 0-4 Y/, 0.6 0.8 0.

Fig.4.11 Test nozzle with distributed loss and deviation
 49

38, MEASUREMENT NOZZLE

--- BARDON EXIT
CORRELATION
MEASUREMENT ROTOR
34 BAROON EXIT
CORRELATION

300 I

260 104

22 0,-i

0 05 y/h 1.0

Fig.4,12 Cambridge turbine exit angle predictions

 Vx

II NOZZLE EXIT

0.5 1.
05 y/h 1"0

1"0 __0 " ,•0_••- - -

0-0 -0-0--0--0--0 i.,0• o%
r7=6 F.C EASUREaENT
PREDICTION USING
1'5 BARDON CORRECTIONS

Vx
MX, II

1"0
•.-00-- o-o-0o -o-
,i• ROTOR EXIT

0. 0-5 y/h 1-0

Fig,4.13 Cambridge turbine axial velocities

 1000

TEST DATA
300 in SECONDARY FLOW THEORY
(SIEVERDING)
------ BAROON'S CORRELATION

0 0X5y/ 1

* I Fig.4.14 Hannover turbine relative flow angles

52

300

100,

TEST DATA
300 - STANDARD PREDICTION
-x- BARDON'S CORRELATION

200- - =

100 H.P NOZZLE

0 0-5 y/h 1.0

Fig.4.15 2 tg a ozl xtage

 53

F300cpCr-nOmmO - mpO. O

20 0 - L.R ROTOR

100

300-

M
3o . - X ,,,• -s ., .. ....

--- o--- STANDARD PREDICTION

-x- BARDON'S CORRELATION
100
H.P. ROTOR

0 I
0 0.5 y/h 1.0
Fig.4.16 2 stage gas turbine rotor exit angles
54

z
SLU
<I

xx

Ei

-X--X-0

ts

xL

Ih II C)
 55
5

K ~z

VLLI LZ
(
roL t-

x
Ax 0.)

Ix x X X 0

x0

Ln-
C)0

xI1 Iso
0
X
56

30" 11.2 ROTOR

X3O

100-

M2iH.4 NOZ2ZtaE gstrieHP tg xtage

- ..
f.-0-0-.0--- -0.0-01,
 57

Chapter 1.5

DISCUSSION AND CONCLUSiONS

S.1 OVERALL PERSPECTIVE

The most striking finding of the Turbine Sub-Group was the lack of good, publicly available, test data on turbines.
There is certainly a large amount of proprietary data held by turbine manufacturers. Although unable to quote directly
from such data, the conclusions of the Turbine Sub-Group draw heavily upon its members' experience with data other
than that published in this report.
The Turbine Sub-Group agrees with the view current at the start of this work that the accuracy of througliflow
calculations is mainly limited by the empirical correlations used to predict loss and deviation. This view is supported
by the experience that when measured losses and angles are input to a calculation the measured flow field can be
recreated almost exactly.
The state-of-the-art of turbine throughflow calculations must be considered separately as regards overall perform-
ance and detailed flow field prediction.

The situation as regards overall performance is reasonably satisfactory. It seems that several methods are available
which are able to predict turbine efficiency to better than ±2%. Even remembering that this represents an error of the
order ±20% in loss prediction the result is considered surprisingly good. In fact when the complexity of the physical
situation is considered, together with how little we understand of the effects of factors such as unsteady flow, stream
tube divergence, centrifugal forces, boundary-layer skew, and many others, on the loss produztion Processes the
accuracy obtainable seems scarcely credible. There can be no doubt that it is achieved only by ad-hoc adjustments of
the many loss parameters so as to obtain agreement with as large a body of test data as is available. It would be highly
instructive to compare all methods against a large collection of publicly available test data which had not been used in
their development. Unfortunately such a body of data does not exist at present.

The situation regarding flow field predictions is less satisfactory. The calculated velocity profiLes are deternlined
mainly by the prescribed relative flow angles at exit from the blade rows. These angles are predictable to better than
±2' in 2D flow but in a turbine they are influenced by stream tube divergence, coriolis forces and, above all, by
secondary flows. The secondary deviations near to the end walls dominate the flow at aspect ratios of less than about
2.0 and at present they are not accurately predictable. The problem of predicting them within a throughiflow calculation
appears formidable since annulus boundary-layer growth would also need to be calculated. In the near future fully 3D
calculations offer the prospect of improved calculation of secondary flows. However, if these are ever to replace
throughflow calculations they will have to include vi3cous effects and to calculate through several blade rows and such
developments must be regarded as long term.

In summary it may be said that throughflow calculations for turbines have almost reached the li1mit of their possible
development. Future improvements in the prediction of loss and deviation are likely to be small until such time as fully
3D viscous methods become available.

5.2 CONCLUSIONS
I. There is an acute shortage of publicly available test data on turbines, particularly on high-speed machines.
2. The physics of loss generation in turbines, as opposed to in cascades, has not been widely studied and is poorly
understood.
3. Reasonably good predictions of turbine overall efficiency can be obtained at the design point. Agreement is less
good off-design.
4. In order to obtain good predictions of the flow field within turbines it is necessary to distribute both the loss and
the deviation along the span. However, variations in deviation are likely to have a much more significant effect
than are those in loss.
5. Current throughiflow methods can only give good predictions of the flow field for high aspect ratio machines.
Low aspect ratio machines are dominated by the effects of secondary flow which are not yet predictable.
6. Fully 3D methods appear promising for calculating secondary flows but they are only likely to replace conventional
throughflow calculations in the very long term.
 59

Appendix ALI.
CAMBRIDGE TURBINE - GEOMETRY, TEST DATA AND
SAMPLE CALCULATION

The Cambridge turbine is a large, low-speed research facility operated at the Whittle Laboratory of Cambridge
University Engineering Department. Full details of the turbine design and of the measurement made on it are given by
Hunter (1979) and the Turbine Sub-Group are greatly indebted to Dr Hunter for his help and cooperation in making
these results available.
Th ubn safe otxdsg ihhbtprto07 n ihicmr.,il lw ec nteasneo

viscous and secondary flow effects the axial velocity profiles would be everywhere uniform and constant. At design
conditions there should be no absolute swirl at exit from the rotor and the experimental flow rate was set to satisfy
this condition at mid-height. The turbine was tested with various values of rotor-stator spacing but only results for
in axial turbine design practice.

Details of the machine and blading are given in Tables ALI., Al .2 and Al .3. All geometrical details are based on
design rather than on measured data.

Flow measurements downstream of the nozzles were made with a 5 hole spherical head probe which was traversed
circumnferentially over two blade pitches. Great detail of the complex secondary flow regions near the end walls was
obtained. Flow visualisation on the nozzles showed them to have largely laminar boundary layers which partly accounts
for their remarkably low loss coefficient, about 1.4% at mid-span.

The flow behind the rotor was studied both with the five hole probe and with hot wire probes. The latter were
used in conjunction with a computer based data acquisition system to obtain ensemble averaged flow patterns. These
were obtained at different circumferential positions of the probe so as to study the variation of rotor exit flow pattern
with rotor position relative to the stator wakes. The resulting picture of the rotor exit flow was very complex making
comparison with throughflow predictions difficult. The results shown in Figures Al .2 and Al1.3 were obtained from
the averaged hot wire signals at what was thought to be a mean circumferential position. However, it should be
emphasized that circumferential variations in the average signal were significant and no attempt was made to obtain a
true circumferential average.

The stator loss coefficients were easily obtained from the 5 hole probe measurements but the rotor loss was not
measured directly. Instead, the overall efficiency and rotor loss were estimated by applying the Euler equation along
streamlines computed from the measured velocities. This process is subject to accumulation of experimental errors
and the final estimate of total to total efficiency (93.8%) is thought to be only accurate to t3%. Similarly the accuracy
of the rotor loss coefficient does not justify comparison with predictions although the large spanwise variations found
in it are probably significant. ti ahn mhsz h opeiyo h lwi
In general the detailed experimental results available from ti ahn mhsz h opeiyo h lwi
even the simplest of turbine geometries and in particular show that the flow behind a rotor is highly unsteady and
non-axisymmetric in an absolute frame.
The throughflow method used to illustrate the results for this test case is that of Denton (1978). This is a stream-
line curvature method with stations included inside the blade rows. The profile loss is obtained fr~m the method of
BaIJ6 and Binsley multiplied by a constant factor of 1.135 (see Denton (1973)), secondary loss from Dunham and
Came's (1970) prediction factored by 0.375 and tip leakage loss and flow from the method of Denton and Johnson
(1976). The factors multiplying the loss components have been adjusted to obtain good agreement with a large number
of measured turbine efficiencies but were not changed from previously optimised values for any of the calculations
reported here.

The method takes the blade exit flow angles as cos'l (opening/pitch) for all subsonic exit flows and adds a
supersonic deviation based on continuity for supersonic exit flows.
Profile loss and exit angle are evaluated separately on each streamline and 9 streamlines were used for the present
calculations. Secondary loss is evaluated separately at root and tip and then assumed to vary linearly with streamline
number (i.e. with mass flow). Tip leakage loss and flow are evaluated at the tip and then added uniformly to each
streamline.

FMJMNG P~Aa AA-ziOT IM

60

The results shown in Figures Al.1-Al .3 were obtained using the measured mass flow (20.1 kg/s) in the calculations.
However, with this flow the calculation predicted about 120 of swirl behind the rotor whilst the experimental flow was
set to have no exit swirl. In order to predict zero exit swirl the mass flow used had to be reduced by 10% to 18 kg/s.
It is significant that the mass flow rate which Hunter obtained from traverse measurements behind the rotor was also
8% higher than the measured inlet mass flow. These two results (computational and experimental) suggest that there
may be an error of about + 10% in the measured inlet mass flow at which the tests were performed.

As might be expected the computed velocit, ,rofiles for this free vortex turbine (Fig.Al.2) are almost uniform
and show none of the measured effects of secondary flow. The computed total-total efficiency was 94.3% which is in
good agreement with the measured value of 93.8%; however, considering the degree of uncertainty in the latter this
agreement is probably fortuitous.
It is clear from this comparison that better agreement between measured and computed velocity profiles can only
be obtained if some means of predicting the secondary flows and losses is included in the calculation.

REFERENCES

References to this Appendix are the same as in Part I. )

TABLE AI. 1

Cambridge Low-Speed Turbine

Geometrical and Flow Data

Stator Rotor

Number of blades 36 51
Blade chord (mm) 152.4 114.3
Axial chord root (mm) 106.3 106.1
mid 108.7 81.6
tip " 111.5 62.7
Hub diameter (mm) 533.4 533.4
Tip diameter (mm) 762.0 762.0
Tip clearance (mm) 0.0 0.63

Inlet Boundary Layer (natural) Hub Tip

Overall thic,§.ness (mm) 21.3 22.1

Displacement thickness (mm) 1.90 2.87

Rotational Speed 525 rpm

Inlet Stagnation Pressure 101.3 kN/m 2
Inlet Stagnation Temp,-rature 293 K
Air Flow Rate 20.1 Kg/s

JI
 61

rr

00 cro m mom e r4

-0-

ChQ~ 0 00

C4oC

m fn

oo
___ - a\%n0__M

*~~~ ' 00' 0~-

CDr.~~*~ 0

0 '. ~4 r
00i 1^~t
A. 4

1.0 N w v 0 OO It.f

0 0 0 0 00 ~"0
ot-l Nq t
66
62

o\ Cr c N 0

-.- 00~00 00 0 0

0 0 00 004kn00 0m%
enk n tn k n %

* Rqq

QO r- \0\ k^ 0n

I..,

k-t-NJO
N kn \0t w O
'- 6rSo-
[ 63

at -
ccIv

'U4
r~64
CAMBRIDGE TURBINE
-0 Measured
- x- Computed

Vi

0.6

0.2L

CAMBRIDGE TURBINE

-o- Measured
-x- Computed

VX,2
VX.2m

0.4 0.6 0.6 Y/h 1.0

0 0.2

Fig.AI.2 Axial velocity profiles after rotor and stator

 65

38 CAMBRIDGE TURBINE

34

-- o-- Measured
-x- Computed

30

26

018
22-

10

10 I i 1
0 0.2 0.4 0.6 0.6 Y1 h 1.0

Fig.A1.3 Stator exit angles

-. *1
66

36 CAMBRIDGE IURBINE

34
----- Measured
-x- Computed

30

P2

22

18

14 0 0.2 0A 0. 0.8 Yh 1.0

Fig,AI.4 Rotor exit an~es

t~
 67

0.2B

0.26 o MEASURED
x COMPUTED

0-22-
0.18,
0-20

• 0.16
APO
1/2 pV 014

0.10

0-08
0

004- 1 ! 1

0-02 V

0 0.2 o04 0"6 y/h 0"8 TO

Fig.Al.S Stator loss coefficients

'I
68

o MEASURED

"X COMPUTED

0 o2
-
aP0
2 I/
pw2
2d
0.1

00
\o•0

-0.1
0
0 .2 0.4+ 0 .6 y h N* 0~
0 81.0

0I ly/ h
Fig.AI .6 Rotor loss coefficients 08
 69

Appendix AI.2

THE TWO STAGE AERO ENGINE TURBINE

This turbine was tested by Rolls-Royce Ltd., and was used as a test case for the 1976 AGARD meeting on through-
flow calculations. The geometric details and test results ait published in AGARD CP 195 and are reproduced here,
Tables A2.1 A2.5, with no additional information being inciuded. Drawings of the blade sections are available in
AGARD CP 195.
Few details of the turbine and test procedure are available. The turbine was a model of the H.P. and I.P. stages of
an aircraft gas turbine with the two rotors running at different speeds. It was 'cold flow' tested using air at 420 K and
2.95 bar. Both hot wire and pressure probe instrumentation was used with radial and circumferential traverses after
each blade row. Only the circumferentially averaged flow properties are available to the Turbine Sub-Group and the
degree of confidence in the results cannot be established without more details of the experimental techniques. However,
it is considered likely that the measured loss coefficients of all but the first blade row are subjected to a great deal of
uncertainty as a result of the non-uniform total pressure field at inlet to these blades.

The design is unusual because of the very low aspect ratio of the I.P. nozzle row and the measurements indicate very
high losses for this row which are consistent with a flow separation within it. There are also signs of a possible flow
separation at the root of the first rotor. Because of this the overall efficiency is expected to be lower than might be
obtained for a well designed machine with this geometry.

Apart from the difficulties of estimating the losses correctly this machine provides a severe test of any throughflow
calculation procedure because both blade rows of the tt.P. stage are very nearly choked (some calculations predicted the
nozzle to be choked). Hence the pressure ratio:mass flow characteristic is very steep and it is necessary to work to a
prescribed pressure ratio rather than a prescribed mass flow.

The throughflow program used to provide sample results was developed by E.Macchi and is based on Vavra's (1960)
method of solution for axisymmetric and steady flow in axial turhomachines: it solves the complete radial equilibrium
equation accounting for enthalpy and entropy gradients and streamline curvature effects - in a specified number of
stations ahead of and after the blade rows. An orthogonal curvilinear system of coordinates, having the meridional co-
ordinates coincident with the generatrices of the flow stream surfaces, is used in tile solution.

It makes use of iterative methods for the numerical solution of the radial equilibrium and continuity equation at
each considered station. An iterative procedure is also applied for the location of the streamlines over the whole machine.
Streamlines curvature and slope are numerically derived from the radial positions of three adjacent stations. Entropy
gradients are assumed to occur only between stations ahead and after a blade row, and are computed from blade loss
coefficients. The Craig & Cox loss correlation is used: the profile losses are computed along each streamline, as a
function of the local blade geometry and flow condition; tile secondary losses are evaluated at mid radius and uniformly
distributed. No other losses are considered.

In the continuity equation, leakage flow rate is accounted for by means of the following relation:

diCL = ACL PT (P cos e

KCL VRTT
where:
ACL = clearance area
KCL = coefficient, assumed equal to 2 in the present calculations
PT. TT = total conditions ahead of the leakage region
(1) = flow coefficient, given by:

Flow angles in the stations after the blade rows are computed by the following relation, derived by applying the
continuity equation between the blade throat section and the station after the blade where the equilibrium equation is
solved:
 70

a h sine
gtsin' Ie lk_".

where
a = blade throat
g = spacing
ha = stream tube height at throat
h = stream tube height at considered station
H/K = boundary layer form factor (assumed 0.9 in calculations)
= loss coefficient at throat
= overall loss coefficient
e
C, = elevation angle at throat
e = elevation angle at considered station
X = I for unchoked throat
X = •er/c for choked throat, with the flow coefficient to be calculated for isentropic flow.
In the calculations, it was assumed V*= tPROFILE + itse •

Comparisons of measured and computed flow distributions are shown in Figures A2.1 to A2.12. For the H.P. stage
the relative Mach numbers at blade exit agree well but the very large secondary deviations at rotor exit (Fig.A2.5)
completely distort the measured axial Mach number profiles. (Fig.A2.4). The mean level of loss is reasonably well
predicted for both H.P. blade rows but the concentration of loss near the walls is not included in the calculation.

The relative exit Mach numbers are again reasonably well predicted for the I.P. stage although the computed stage
reaction seems to be higher than measured (relative Mach number low for nozzle high for rotor). The levels of loss are
underpredicted for both l.P. blade rows, particularly the rotor, and the complex variations in blade exit angle (Figs A2.8,
A2.1 I) are not reproduced by the calculation.

The overall efficiency of the machine is well predicted 89.1% (c.f 88.4% or 89.1 measured) but this agreement may
be misleading because significant tip leakage losses were included in the calculation whilst they are expected to be
negligible in the turbine (rubbing seals), If no tip leakage loss had been included in the calculation the low losses
predicted for the I.P. turbine would have caused the overall efficiency to be overpredicted.

The comparisons on this machine show that relative blade exit velocities profiles can be predicted much better than
can axial or meridional velocity profiles. The latter are greatly distorted by the effects of secondary flow which are
possibly greater than usual in this machine as a result of flow separations in the H.P. rotor and I.P. nozzle. For the same
reason the efficiency of the turbine might be expected to be rather lower than predicted by standard correlations.

REFERENCES

Vavra, M.H. Aerothermnodynarnicsand Flow in Turbomachines, pp.439-470, Wiley, 1960.

LAW,
 71

TWO STAGE AERO ENGINE GAS TURBINE TEST CONDITIONS

Mass flow = 11.5 Kg/s

Inlet total pressure = 2.94 bar
Inlet total temperature = 423 K
Rotational speed = HP turbine 8250 rpm
IP turbine 4693 rpm
Overall total:total pressure ratio 5.8359
Inlet total pressure distribution = not available
Inlet total temperature distribution = not available
Inlet velocity distribution (at station z = -68.6)

Radius Axial Mach Number Angle

0.25943 0.1019 90
0.26055 0.1524 90
0,26168 0.1502 90
0.26280 0.1468 90
0.26505 0. 1429 90

0.26955 0.1393 90
0.28080 0.1446 90
0.29205 0,1561 90
0.29655 0.1566 90
0.29880 0.1545 90
0.29993 0.1529 90
0,30105 0.1496 90
0.30218 0.1434 90

Radial clearance = Effectively zero (shrouded blades with rubbing seal)

Number of blades = first stator 54
first rotor 102
second stator 36
second rotor 148

Measured Efficiencies
continuity.
(1) Total-total based on brake work and total pressures from wall statics and
Overall: 88.4%
H.P.: 85.7%
I.P.: 89.6%

(2) Total-total based on total temperatures and total pressures from traverse data
Overall: 89.1%
H.P.: 89.8%
I.P.: 89.5%

I!
I]
Ii
72

~ 666600

o 00 0 n r00

ooDo000 o

wI
CDONc

S, 000

U4

q q 0

60 ) 00 00 00

0000 00 000

CA C

En 4\J0

0 0 0 0
 73

S0000l n00000t

a 0

0 0

t-

I ~M

:- C!n
IM
IWO 0~ 0

0 kn
QOO Q4O
t-00 0000000 %

O00ent in\ot-00 '

ON
74

* n w mt- NS1 n

tn en N0O

____ ;___________S __________

r-__________________
So________
o,_________
to "Io; .t- -t
:ý% ocqqqq

000iCD0 0 0000

*m
%A e

ga a~c~Z

0 0 CD C 0
C4I

00

00000 W, q 04 a o
d.o6t-doo t-
N a\

I
1-1 kn t 00 0\A

0
,~~-
1
qqqqqq
D cCD 0C 0

00
ip 0c 0"o-r4 o00 z v
 75

%0oo ON 0 0 t o

e4 el N N -- venI00
It

- ri e- e-~ r4 e-l4

00wl0 \0000r-00

t-.000 0\ mu00 00 t
-bO k %O\~ ,n'.0 kl -jI

o n \"0 r 0
0a 0 C'4.0M
00.0020oc0.\0t-0 0
20000500000 r-w 0C

m ITI I n )\

C ~ 0
00 It
~ 0 0D

t-t 000000000O 0 0 00

0 ~00~ ~
00 I 0.0.0 0,0

en0 Lný l 0O
76

C-0- .- D00
rlC Ci Il- "! C

0000Ar

a~~ CD CD "V

0ý to N
%0 N
m N- a

C=. C ;C 0 Cý a

00 r c 00 1 1 H 1

co. qt
~D- M - M < M-UIQ w

0 rC) 0 c 00M

= 0 -- - - -~

r- - co- l 0~ C4
0Cý4 I0~I

Q~j a 0
C'u*

00 r 00

C- C;7

=-) CD, -0 IM

%0 -i -(D1

0 C 0
0in
 1.4

1.2

LaJ
CALCULATED

X EXPERIMENTAL
u .6

.4 M l

0 2Q 40. 6L 80. 00.

PERCENT SPAN

Fig.A2,1 Comparison between experimental and calculated Mach numbers at H.P. stator exit

S •-- •'•,. • .... - _________-___.......

.. ___ .. ... . .____._.. __...__.... . ___
.. ____,_-______-_....__._._____._,_..,_,
... .-
78

40L

CALCULATED
30. EXPERIMENTAL
S~o'1

20.

10._

)
0 1I i I
0 20. 40. 60. 80. 100.
PERCENT SPAN

Fig.A2.2 Comparison between experimental and calculated absolute exit flow angles at H.P. stator exit

.4

.3 CALCULATED
EXPERIMENTAL

Z.2
w.
LL

wJ
0

-.1 I I I :
20. 40. 60. 0.1
PERCENT SPAN

Fig.A2.3 Comparison between experimental and calculated loss coefficients at H.P. stator exit
 79

1.2

0::

.13- CALCULATED
z[ ____EXPERIMENTAL

CID -I-

2~

0 20. 4. 60. 80. 100.

PERCENT SPAN

Fig.A2.4 Comparison between experimental and calculated Mach numbers at H-.P. rotor exit

40.

- - ----
-- A-- CALCULATED

EXPERIMENTAL
30-

10.1

0. 20. 40. 60. s0. 100.I

Fig.A2.5 Comparison between experimental and calculated relative exit flow angles at H.P. rotor exit
80

.3---------------- --------------------------- CALCULATED

EXPERIMENTAL

z.,2-

2 - - - -

0 20. 40. 60. 80. 10

PERCENT SPAN

Fjg.A2.6 Comparison between experimental and calculated loss coefficients at H.P. rotor exit

1.0

_EXPERIMENTAL

z
=M

0. 2a 40. 60. 80. 100.

PERCENT SPAN

Fig. A2.7 Comparison between experimental and calculated Mach numbers at L.P. stator exit
 40L

30.

S~20.

-,CALCULATED
• 10 EXPERIMENTAL

[ _ _ I I
0 20. 40. 60. 80. 100.
PERCENT SPAN

Fig.A2.8 _
Comparison between experimental and calculated absolute exit flow angles at L.P. stator exit

.3

-- CALCULATED
- EXPERIMENTAL

z
W
I U. .1

'0
.0

0 20. 40. 60. 80. 100.

PERCENT SPAN

Fig.A2.9 Comparison between experimental and calculated loss coefficients at L.P. stator exit
82

W
---- CALCULATED
EXPERIMENTAL

I2MV 2 r

0. 20 40 60. 80. 100.

PERCE NT SPAN

Fig. A2.I10 Comparison between experimental and calculated Mach niumbers at L.P. rotor exit

£40.

3Q

lol- EXPERIMENTAL

0 20. 40. 60. 80. 100.

PERCENT SPAN

Fig.A2. 11 Comparison between experimental and calculated relative exit flow angles at L.P. rotor exit
 83

.3

CALCULATED
EXPERIMENTAL

.2 _

Wu

0. 20. 40. 60, 80. 100.

PERCENT SPAN

Fig.A2.12 Comparison between experimental and calculated loss coefficients at L.P, rotor exit

SI ,

b v-
 8S

Appendix AI.3
ANSALDO STEAM TURBINE - GEOMETRY, TEST DATA AND
SAMPLE CALCULATION

As a result of the difficulty encountered by the Turbine Sub-Group in finding suitable test cases Dr Zunino offered
to make available data obtained by his company (Ansaldo) on a large low pressure steam turbine. Although the main
interest of AGARD isin gas turbines these steam turbine data were considered an equally valid test of loss and deviation
correlations and so the offer was gratefully accepted by the Turbine Sub-Group, several of whom had the ability to
calculate steam as well as gas flows.

The turbine geometry is fairly typical of large low pressure 3000 RPM machines, having 6 stages with a very high
casing flare to accommodate the large increase of volume flow. This results in very high blade speeds and pressure ratios
for the last stage. The Ansaldo machine has a last stage pressure ratio of over 6:1 which, coupled with a low hub:tip
radius ratio, results in very high relative Mach numbers (-' 1.7) at nozzle root and rotor tip.

The layout of the turbine is shown in Figure A3.1I and full geometric details of stages 3-6 are given in Tables
A3. 1-A3.9 (all dimensions in mm). Details of steam conditions, bleed flow rates and blade clearances are given in
Table A3.l10.

A complete set of all measurements necessary to obtain a heat balance for the turbine were performed in parallel
with the traverse measurements. These consisted of 42 pressure, 67 temperature and 6 mass flow measuring devices.
The overall electrical output was also monitored and the LP cylinder output was measured by means of a shaft torque-
meter and telemetry. These measurements enabled the LP cylinder efficiency to be evaluated separately from that of
the whole turbine.
The probe used for aerodynamic measurements was a 5 hole spherical head probe with a thermocouple incorporated
into its stem. This probe was calibrated in air up to a Mach number of 0.9. Traverses were made with the turbine at full
power and after allowing sufficient time for conditions to stabilise.

At each traverse point the probe was nulled in yaw and a complete set of readings taken at least 5 times before
moving on to the next point. As a result the measurements at each traverse point required from 2-10 minutes and a
complete radial traverse from 6-8 hours. The use of an average o1 5 or more sets of readings at each position should
ensure that random errors are negligible. Further details of the experimental technique are given by Accornero et al. (1980).

On the whole the measurements seem to have been very carefully performed and, although measurements in steam
are inevitably less accurate than in air, they should be considered reliable.

The throughflow calculation used to provide sample results in that described by Doria et al. (1979). This is a
streamline curvature type of method developed specifically to calculate the flow in low pressure steam turbines. It
includes allowances for: profile and secondary losses; lacing wire and windage loss; blade tip, gland and balance hole
leakage losses; flow extraction (bleed) for feed heaters and spanwise distribution of secondary deviation. The program
works with a prescribed mass flow rate and exit steam conditions (pressure and enthalpy) and the inlet steam conditions
are allowed to vary during the calculation.
The loss and deviation rules used for the calculation are not generally availabie. The losses are based on cascadeI
data for the nozzles and on Craig and Cox for the rotors. Blade exit angles are obtained from a modified sine rule with
an additional deviation based on continuity applied for supersonic exit flow. The method of Bardon et al. (Chapter 4)
was used to f~stimate and distribute secondary flow deviation.
Comparisons of measured and computed rkesults are shown in Figures A3.2-A3.13. On the whole the agreement
isremarkably good and similar results were obtained from other calculation methods on this machine.

The large distortions of the velocity profiles caused by secondary flow, which are present in the other test cases,
are not evident in this machine.

This is possibly because the high turbulence levels which must be present in a multi-stage turbine mix out the
secondary flows both within the blade rows and between trailing edge and traverse plane. The good agreement of
measured and calculated velocity profiles after the last stage is particularly gratifying since such a high pressure ratio,
low hub tip/ratio stages provides the most difficult test of a throughflow program. I.

pBNO~bp pAGEc BLN- OT

 86

REFERENCES

1. Doria, G. Throughflow Calculation of Large Steam Turbine. Proc. 6th Conf. on Large Steam
Troilo, M. Turbines, Plzen, Czechoslovakia, May 1979.

2. Accornero, A. Flow in a 320 MW Low PressureSection. Theoreticaland Experimental Evaluation.

et al. VKI Lecture Series, Steam Turbines for Large Output, April 1980.

72

'I
 87

oo E t-\D

200 W)r

enr4V

.5 v-t--0 0

4 00 0 0 04

-tn

4.

\,o Iix'

0\ 00 O

oo 00 1 00

C4. 00 00. 0.
88

ANSALDO TURBINE

TABLE A3.2
Geometry of the Third Stage

STATOR

Radiust Thoa Chr icE. Backbone Inlet

Rais angle Thot Cod Pth thickness length width

840.75 89.61 10.058 90.484 44.02 1.5 105.25 35.09

860.75 89.10 10.763 90.484 45.07 1.5 105.25 36.04
900.75 88.42 11.989 90.484 47.163 1.5 105.25 37.97
940.75 88.07 13.043 90.484 49.257 1.5 105.25 39.95
970.75 87.99 13.724 90.484 50.828 1.5 105.25 41,47

TABLE A3.3
Geometry of the Third Stage

ROTOR

Inlet T.E. Backbone Inlet

Radius age Throat Chord Pitch tikes lnt it

838.2 28.66 12.59 57.37 32.11 0.59 79.24 13.41

898.2 33.42 13.11 56.93 34.41 0.69 78.42 16.65
934.72 37.65 13.32 56.97 35.81 0.66 77.72 19,31
988.2 44.09 13.39 57.17 37.86 0.67 76.03 22.09
1018.2 48.61 13.36 57.47 39.01 0.66 74.89 24.05

TABLE A3.4
Geometry of the Fourth Stage

STA TOR

Rais Inlet Thra.hric E. Backbone Inlet

Radus angle Thot Cod Pth thickness length width

840.75 88.38 14,117 99.309 66.032 0.8 111.374 56.99

900.75 86.99 16,766 105.937 70.745 0.8 118.820 60.71
980.75 85.78 19.727 115.736 77.028 0.8 129.831 65.71
1040.75 85.49 21.301 122.927 '4 0.8 137.910 69.62

1068.75 85.6,; 21.698 126.170 4~3.940 0.8 141.554 71.54

 89

ANSALDO TURBINE continued

TABLE A3.
Geometry of the Fourth Stage

ROTOR

Rais Inlet hr Thra ic E. Backbone Inlet

Radus angle Thot Cod Pth thickness length width
838.2 27.22 17.20 81.28 45.01 1.64 113.23 19.44
918.51 38.18 17.94 77.99 49.33 1.32 104.83 26.19
958.21 40.75 18.31 76.54 51.45 1.57 101.87 28.71
998.85 41.74 18.63 75.50 53.64 1.53 100.115 31.71
1079.17 69.92 17.90 79.10 57.95 1.57 97.44 44.21

TABLE A3.6
Geometry of the Fifth Stage

STATOR

Radius Inlet hr Thra ic .E. Backbone Inlet

angle Thot cod Pth thickness length width
84075 87.69 14.857 99.31 66.032 0.8 111.374 56.83
940.75 85.99 18.693 110.87 73.886 0.8 124.365 63.11
1040.75 84.92 22.07 122.86 81.740 0.8 137.831 69.43
1140.75 84.79 24.387 134.49 89.594 0.8 150.90 76.06
1230.75 85.28 25.519 145.42 96.663 0.8 163.178 82.22

TABLE A3.7
Geometry of the Fifth Stage

ROTOR

Rais Inlet Thot Cod PthT. E. Backbone Inlet

Radus angle Thot Cod Pth thickness length width
838.2 28.09 19.11 86.52 45.8 1.49 118.35 21.02
914.4 34.35 20.32 81.51 49.96 1.47 109.85 26.41
1022.78 48.99 21.91 75.97 55.88 1.46 95.87 37.90

1133.55 76.06 22.25 77.14 61.93 1.48 86.72 53.17

1268.93 136.04 21.84 86.28 69.33 1.48 86.88 67.92
90

ANSALDO TURB[NE continued

TABLE A3.8
Geometry of the Sixth Stage

STATOR

Inlet 7'E. Backbone Inlet

Radius age Throat Chord Pitch tikes lnt it

850.72 91.286 17.82 183.48 89.087 2.0 213.77 71.44

950.72 89.89 22.15 207.19 99.56 2.0 241.48 79.16
1050.72 88.82 26.41 230.06 110.03 2.0 268.2 86.97
1150.72 88.19 30.49 242.77 120.5 2.0 283.04 95.77
1250.72 88.02 34.16 249,54 130.97 2.0 290.96 105.28
1350.72 88.07 37.20 262.25 141.45 2.0 305.80 114.35
1450.72 88.29 39.65 277.49 151.92 2.0 323.62 123.31

TABLE A3.9
Geometry of the Sixth Stage

ROTOR

Radius Inlet
age T.E. Backbone Inlet
Throat Chord Pitch tikes lnt it

838.2 33.39 28.67 194.80 56.03 2.02 238.59 28.67

965.84 41.67 33.54 172.23 64.56 3.04 205.33 41.25
1093.47 56.99 36.55 151.67 73.09 2.18 176.80 56.65
1427.48 128.37 37.25 132.10 95.42 1.52 133.73 94.61
1604.01 150.70 32.76 132.96 107.21 1.01 133.09 106.14
 91

TABLE A3.10

1. Inlet Steam Conditions to the Last 4 Stages

Static Pressure = ,. 176 bar
Static Enthalpy = 2895 kJ/kg

2. Inlet Mass Flow to these 4 Stages 99.826 Kg/s

3. Exhaust Pressure 0.0671 bar

4. Tip Clearance and Number of Seals for Each Blade Row

Stage Tip Clearance Actual Number of
N. (mm) Seals in Operation
1 1.8 4
2 1.6 3
3 1.8 3
4 2.0 5
5 2.3 7
6 7.3 19

5. Efficiency of Last 4 Stages - 0.890%

6. Bleed Flow Rates

Bleed I1"ow
N. (Kg/s)
1 3,582 Between first and second stage
2 7.578 Between second and third stage
3 8.13 Between fourth and fifth stage

7. Exhaust Geometry Downstream of D'D" (Fig.A3.I)

Point D' D" h' E" F' F"
Z (mm) 2004.0 1946.7 2081.55 1971.95 2217.3 1998.2
R (mam) 838.2 1697.25 838.2 1702.25 874.45 1708.5

tL
 I/In

I o o

U)

090

zo

itI
 93

ig

I'L

_ __ __ _
__ _ _ - i

0o

WM

:
40
q e

______ ________________ II

C4 co
94

ccS

ILI

ti

---- C

I 0 _ _i~

0_ __h

z1

0 _ 0
C4aqO
 95

U
U)
'a.
a
I-
.. L
URu

2U

ii U
i
2
('4 W

a
(I'U
U)
IL
o
I-

w
U ____

a
I-
I Cl
___ V

2
I
cII. 'U-

£
44
 __
_ ___ ___
SM
0I-
'a
I
U
UI I
U
-J 8 ____
4
(AR
z *

Ii
4
xz

- £ so

______ ______ _____ 0

_____

so
2
SM

___
- ____ 0 0I-'a
____

IL
I-o

SM
.1 -
2
- ______
U
____-
0 .0
I -
S
z
U
U

U C4
x

j
 97

co

IL

SM x

UU

o Mo

C4 Id

I.0,
r~d OD c~I
98

o ji l'0

_ _
_ __
_ _
_ _

-J

C44u

-n

_
.3
U

C40

faa

zi
z •

U, C4
 99

Appendix AI.4
HANNOVER-TURBINE - GEOMETRY, TEST DATA AND
SAMPLE CALCULATION

This turbine was investigated by G.Groschup af the In:,'itut tbr Str6mungsmachinen of the Technical University of
Hannover, Germany. The tests were part of the doc toral th.:is of Groschup in 1977 (Ref. I ). We are very grateful to
Dr Groschup for his assistance in compiling the full data of this turbine. Our thanks go also to Professor Bammert,
former head of the same Institute, for permission to publish these data in the present report.

TEST FACILITY

The turbine facility is an open loop configuration with exhaust to atmosphere. The air is supplied by three parallel
screw compressors with a maximum volume flow of I1 m3 /s at maximum pressure of 4 bar. The compressed air passes
through a cooler which maintains a constant inlet turbine temperature. A turbulence grid placed at I m upstream of the
turbine flange produces a turbulence level of about 2% just ahead of the turbine stage. The power of the turbine is
absorbed by an eddy-current-brake.

TURBINE

The facility was basically laid out for a four-stage turbine which was designed to have the same blade sections in all
stages at a given radius. The blading is of the free vortex type and the nominal mass flow of the turbine is th = 7.8 kg/s
and the nominal rotational speed is 7500 RPM.

The set up allows each stage of the four-stage turbine to be tested separately. The tests referred to in this paper are
concerned only with stage No.4 which was tested in the single-stage mode. Details of the meridional flow path are shown
in Figure A4. I. The turbine has a constant hub diameter of DH = 270 mm. The outer casing is cylindrical except across
the blading where it is conical. The rotor has extremely small radial clearances: S = 0.24 mm at nominal RPM. The
stator is shrouded and the leakage flow across the stator hub is minimized ': multiple seals.

BLADE AND CASCADE GEOMETRY

The blade sections are stacked radially along their centre of gravity as shown in Figure A4.2, the same figure also
defines all other relevant blade parameters. The values of these paiameters on 9 equally spaced "streamlines" are given
in Table I.

I MEASUREMENT TECHNIQUE

Overall Performance
The mass flow is measured with a standard nozzle (DIN 1952) in the inlet duct (0 500 mm) far upstream of the
turbine flange. The mass flow is determined with a precision of 0.2%. The torque is measured with the eddy-current-
brake with a precision of 0.5% of the full scale of the balance (range: 0 -* 60 bar or 0 - 120 bar).

Survey of Turbine Stage Flow

The measurement stations are indicate" in Figure I and are located as follows:
Plane 0" 20 mm upstream of guide vane.
Plane 1: Between stator and rotor. The axial distance between the blade rows is about 0.6 to 0.8 x axial
chord length of the stator. The distance between probe tip and stator trailing edge varies from
6 mm at the tip to 17 mm at the hub.
Plane 2: Close to rLtor trailing edge. The distance between probe tip and rotor trailing edge varies from
15 mm at the tip to 6 mm at hub.
Plane 3: 290 mm downstxeam of plane 2.

AM-%
100

Reference I quotes only data taken in planes 0, 1 and 2. All the data presented are taken with a combined wedge probe.
This probe measures total and static pressure, flow angle and total temperature. The meridional flow angle was not
measured but checks were made to ensure that the probe wns not sensitive to meridional flow angles of the order of the
ones existing in the turbine. The survey of the flow behind the stator was made by slowly rotating the stator.

The problem of measuring behind the rotor in an unsteady flow field is discussed by Groschup in Reference I.
Groschup comes to the conclusion that the error on the total pressure measurement is of the order of 0.25%.

TESTS RESULTS
Traverse and overall performance measurements were made at 5 different operating conditions. The inlet flow
conditions for these are given in Table 2 as traverses of inlet total pressure and temperature together with the ratios of
actual to design speeds and mass flow rates. Test point 4 is taken as being closest to design conditions.

The variation uf efficiency with blade loading parameter is given in Chapter 3 in comparison with computed results.
Traverse results are given for test point 4 only in Figures A4.3 to A4.7. Traverse results at the other 4 test points are
given in Groschup's thesis.

COMPARISONS WITH SAMPLE CALCULATION

The throughflow method used to provide sample results for this test case is that of Denton which has already been
described in Appendix 1. The comparisons with test results at test point 4 are shown in Figures A4.3 to A4.7. This
machine is of free vortex design and since the Mach numbers are low all axial velocity profiles should be uniform in the
absence of viscous effects, streamline curvature and secondary flows,

Figure A4.3 shows that the measured velocity profile after the stator is very uniform except near the end walls.
The calculated profile shows a positive slope near the tip due to the effects of streamline curvature and a negative one
near the root due to overestimation of the turning. The velocity profile after the rotor is highly distorted by the effects
of secondary flow whilst the computed profile is very uniform.

The large spanwise variation of exit angle from the stators, Figure A4.4, is reasonably well predicted apart from
the secondary flow effects which are confined to small areas near the end walls. However, these effects dominate the
flow behind the rotor, Figure A4.5, where the large and complex secondary flows at the root extend to mid-span.

The average level of loss is well predicted for both blade rows but once again the complex distribution of secondary
loss is missed by the calculation. The predicted efficiency of 91.4% (total to total) is in reasonable agreement with the
measured value of 90,5%. Further comparisons at off-design conditions are given in Chapter 3.

It is instructive to compare the results from this turbine with those from the Cambridge turbine. Both are free
vortex designs with 4imilar values of aspect ratio. However, the lower flow coefficient of the Hannover turbine results
in greater turning in each blade row and consequently much stronger secondary flows at the rotor root. These in turn
are the probable cause of the lower efficiency. Whilst calculation methods do reasonably well in predicting the
efficiency of both machines the large discrepancies between measured and computed velocity profiles after a single stage
of simple design are discouraging as regards the use of throughflow methods to predict velocity profiles in multi-stage
turbines.

REFERENCES I
I. Groschup, G. Str6mungstechnlsche Unterscuhung einer Axialturbinenstufe im Vergleich zum Verhalten
der ebenen Gitter threr Beschaufelung. Doktor Ingenieur Dissertation Technical University
of Hannover, 1977.
 I01

0go 00 o oo00 CO00 0000 N%'

N a,~ 0 0 fn 0 n %n0

W)t-Q M % 00 0L 0
000 0 QV

01 ~ f lt~0 0 Cf

00 m00

w oo ooo0

0000 000 000000000el

C'
0' D

In li -nC

IQ 0vUO 0n t- 0.t- --

aIm mI m mm0m0m en en

tnI
'102

0
00kn0 N
0Q

0; ;

oe4

1\00IS
\o \n 0I0

00

- - -- -4
--

0 -0
 103

1*0

I jI
C4C

Ln i I

I Un

ZCZ
-04

t~4.

4.-

00

4.44
 los

ccC

11

0W I
106

ulC

za

ccA

I.-

44 I0
MLL
 107

HANNOVER TURBINE

0,40

ROTOR
0.30
--0-- Measured

CR --x- Computed
0,20

Nt
0.10

II I I I

0 0.2 0.4 0.6 0.8 YIh 1.0

Fig.A4.7 Rotor loss cooficients

! ..
 109

PART 11

AXIAL COMPRESSOR PERFORMANCE PREDICUIONS

l•!•l FAGSlBj]
LANlnor It

Jut
 II.* . Introduction

in the field of axial compressors a very large number of correlations and correlation for-
mulas can be found in the literature and in the design groups of engine and industrial
turbomachinery manufacturers.

Although many, if not all, of the correlations have common points, as for instance the
use of a diffusion factor in estimating the minimum profile losses, large differencte ap-
pear when it comes to. estimating the effects of the various geometrical or physical vari-
ables on overall losses and deviations. This is obviously a consequence of the strong
complexity of the flow structure and of our present poor understanding of detailed flow
mechanism in an axial compressor.
Bearing this in mind, it was considered necessary to summarize the state of the art in
the field of the existing methods in an attempt to put forward the main influences and
the most sensitive parameters, as well as to proviAe some trendsor general concensus, at
least with respect to some components of the complex system of correlation formula's.
Therefore, a double action was initiated by cite Compressor Subgroup :detailed surveys on
some particular topics were written by various members of the subgroup, in particular on
diffusion factors and profile losses, deviation angle correlations, models for shock los-
ses and shock-boundary layer interaction, part-span damper losses, Reynolds number and
roughness effects, secondary flow and clearance effects on losses and turning, and end-j
wall boundary layer calculation methods. Moreover a review on blade-to-blade calculations
and a discussion of the through flow calculation methods were: produced.

On the other hand it was felt important to have some information or feedback from indus-
try and research organizations on the various topics and questions raised by the review
of the existing correlations and a questionnaire was prepared and distributed in that
sense.

Roughly 15 answers and comments were received from six different countries covering a re-
presentative cross-section of the aeronautical and non-aeronautical industrial and re-
search organisations. This allowed the Working Group to draw some guidelines and the
answers provided also a reasonably clear picture of the deficiencies and shortcomings of
the present-day situation as they are felt in industry. Also, indications about some im-

[( portant loopholes and directions where improvement is needed came forward.

In
order to clear some of the questions raised by the review of the present state of the
art, test cases were selected for calculations with various correlations and with inclu-
sion of end-wall boundary layer effects. Three compressor test cases, a single stage, a
two stage and a four stage compressor were selected as well as two compressor cascades for
additional blade-to-blade calculations in order to provide a criterion for distinguishing
among the various expressions for the diffusion factor and its relation to wake momentum
thickness found in the literature.

Several organisations contributed to the calculations on a voluntary basis and their con-
tributicon arid effort is gratefully acknowledged. Without their active support, the work
and activity of the Working Group would have lost much of its significance.

For obvious practical reasons of time and cost limitations, the amount of extensive and
systematic comparative calculations had to be limited. Therefore the analysis of the in-
fluence of a syst-'matic variation of certain parameters on the overall performance pre-
diction and (or) the internal flow distribution could not be extended as far as necesary.
However, very useful trends came forward from the calculation results.

This report summarizes and presents the various contributions, comments, results and con-
clusions obtained by the Compressor Sub-group. The first chapter contains the reviews of
the state of the art as written by the participants in the Working Group as well as a
summary and discussion of the answers to the questionnaire. The test cases are presented
and partly documented in the second chapter. Chapter three discusses the results of the

Finally, the last chapter contains the overall conclusions which have been reached after
several discussions between the participants and puts also forward trends and recommends-
tions for future research and improvements.

I
The objective in all correlations is the determination of deviation angles and losses
from a reduced set of parameters related to the blade row geometry and to the flow proper-
ties. This set of parameters should be as small as possible but sufficiently representa-
tive in order to describe all of the main effects and influences. It Is obvious that this
always will require a certain amount of empiricism but the actual trend and the ways for
improvements certainly lie in a better understanding of the basic flow phenomena, through
clean and fundamental experimental data as well as through systematic 2-Dl or 3-fl blade-to-
blade calculations.

Nearly all correlations define firstly basic two-dimensional low speed design (or refe-
rence) incidences, deviations and losses. Corrections are then introduced for Mach num-
ber effects, Reynolds number and non two-dimensional effects such as axial velocity den-
sity ratio and secondary flows. Based on the design values, off-design contributions to
the design values of losses and deviations are then estimated in function of incidence.

1111<711PAW......ýN-
 112f
Three different families of parameters enter as variables in all correlations : the cas-
cade parameters, the blade parameters and the aerodynamic variables.

The cascade factors are the blade spacing, the blade height and chord, the stagger angle
and the tip clearance. They are generally expressed in non-dimensional form through so-
lidity, aspect ratio (defined as height over chord ratio) and a clearance parameter
(clearance to chord or to height ratio).

The main blade parameters are, for a given family of profiles, the blade inlet and outlet
angles or the camber angle, the value and position of maximum thickness, the leading edge
and trailing edge radii and the blade roughness height. The two most influencing varia-
bles are the thickness over chord ratio and the camber. The trailing edge radius is not
often considered explicitly as a parameter, although it influences the blade profile los-
ses due to the wake mixing process and some indicaticne do exist that its influence has
perhaps been underestimated. The leading edge radius affects directly the leading edge
bluntness losses at supersonic incidence Mach numbers. The blade roughness, expressed
as a roughness Reynolds number has a non-negligible effect on losses above a certain cri-
tical value, where they become independant of Reynolds number.

The aerodynamic variables entering correlations are inlet Mach number, Reynolds number
generally based on chord and inlet velocity, axial velocity density ratio and a parameter
or a set of parameters measuring blade loading.

The measure of blade loading is a central point in every system of correlations and since
Lieblein's work it is a general practice to use the diffusion factor as the most represen-
tative loading parameter.
Actually, the diffusion factor is a measure of the ratio of suction surface peak velocity
to trailing edge velocity and many different chordwise distributions of loading could be
obtained on different blade profiles with the same value of diffusion factor. Hence, a
certain "regularity" in the velocity distributions is implicit in the use of this veloci-
ty ratio to represent the loading, and therefore its use at off-design can be questioned.
However, at design the correlations between D-factor and trailing-edge momentum thickness
are very strong. The discussion of the various formulas available in the literature and
connecting D-factor with inlet and outlet flow conditions are discussed in the review pa-
per by R. Dunker and H. Weyer.
These relations contain empirical constants which have been correlated through the rela-
tion between D-factor and trailing edge momentum thickness on one hand and the relation
between momentum thickness and loss coefficients on the other hand. For instance, Swan's
and Fottner's correlations are based on eq. (8); Montearrat's correlation is based on
eq. (9) while Koch and Smith's correlation is based on a wake mixing calculation.
Therefore, extrems caution should be given to use a consistent set of correlations, in
particular since the various relations between momentum thickness and loss coefficl.ant
have very different behaviours with regard to the effects of some variables like Mach
numbers, Reynolds number and axial velocity ratio. Also special attention should be given
to avoid duplication of influences of parameters.

The losses are generally separated into three contributions : Profile losses, including
shock losses, secondary losses including clearance effects and part-span damper losses.

The profile losses are expressed as design losses and off-design corrections with various
expressions for the influence of Mach number, axial velocity ratio, Reynolds number, and
so on. At high Mach numbers, shock losses, influenced by shoc;. boundary layer interaction
effects, become very important. Due to the complexity of these phenomena and their extre-
me sensitivity to detailed blade geometry, the available correlations are based on very
simplified assumptions and it is considered that this is one of the most important topics
deserving extensive further research. This opinion is also fully substantiated by the
answers to the questionnaire and most of the existing correlations are reviewed in L.
Fottner's contribution on shock losses and shock boundary layer interactions.
Mach. number influence on design losses is very difficult to summarize in a consistent way,
since the compressibility effects are introduced in the various correlations at very dif-
ferent levels. Koch and Smith's approach is perhaps one of the most consistent in that
sense. Predictions of choking and critical Mach numbers still require much resear--h ef-
fort in order to achieve a satisfactory state.

Reynolds numbers effects are considered to be important and a contribution by W. Roberts

for the low Reynolds number range is included in the present report as well as a contri-
bution by A. SchAffler and L. Fottner on the effects of blade surface roughness.

With regard to off-design profile losses, empirical relatiorw can be found relating the
loss increase to the variations of D-factor from its design value as in Swan's correla-
tion. However, a more reliable approach, very empirical in nature, is to connect the
off-design losses to the incidence variations refered to the choking or stalling incidence
range. This approach is advocated by Jansen and Moffat as well as by Novak (see AGARD
LS83) . The difficulties are displaced towards the prediction of the stalling and choking
incidences and the available expressions are indeed very rough and uncertain particularly
in the range of high Mach numbers (transonic and supersonic) and especially for the bla-
ding used in modern designs.

Finally, two controversial contributions on part-span damper losses by W. Roberts and by

AIn
 113

R. Novak (summarizing Koch and Smith approach) are included in the present report.

With regard to secondary losses, two ways can be followed. The more classical way in
which additional losses are calculated separately and added to the profile losses, being
distributed or not along the blade span. A large number of formulas can be found in the
literature and most of them are discussed in J. Chauvin and J. Salvage's contribution on
secondery flow and clearance effects. Another approach consists in defining an overall
efficiency drop due to the end-wall boundary layers and to calculated their evolution in
a coherent way following Mellor and Wood's original approach. This approach is discus-
sed in De Ruyck and Hirsch's contribution on End Wall Boundary Layer Calculations. It
appears, from the results of the test cases performed within the Working Group activities,
as well as from other sources that this is a right way to follow and it can be considered
as one of the recommendations of the Working Group.

The various problems, connected with the turning prediction are summarized in G. Serovy's
paper. Most important is the correct assessment of the interaction between the various
parameters, Mach number, axial velocity ratio and geometrical factors on design and off-
design deviation rules.

A note on axial compressor stall and surge by J. Chauvin has been included for complete-
ness in the present review.

It had been considered from the start, that the analysis of the basic computational as-
pects of through-flow and blade-to-blade flows was outside the scope of the Working Group.
However, since the interaction between the computational method and the correlations can
have a large influence on the produced results it was decided to include a summary of
both the through-flow methods and of the blade-to-blade methods. A contribution by G.
Serovy on the first topic and by G. Meauze and R. Van de Braembussche on the second topic
complete therefore the reviev of existing methods.

As already mentioned, three test cases representative of modern design techniques were
selected for through-flow calculations. They are described partly in the second chapter
and more extensive data and results can be obtained from PEP at AGARD headquarters. It
undoubtly appears that most of the available data have limite3 accuracy and that very well
controlled experiments, especially on multistage machines, are urgently needed.

The calculations performed by several participants to the Working Group, although limited
in scope, gave some interesting information. Notwithstanding the fact that the results
oi a through-flow calculation are depending not only on the correlation but also on the
computational aspects of the method, some trends could be defined.

The through-flow results as well as blade-to-blade results indicate that the systematic
use of numerical calculations is able to provide information which come closer to experi-
mental data than the methods of more empirical character.

It appears that the time has come when a drastic improvement of the methods for performan-
ce predictions, most of them originating in the 60's or even earlier, can be obtained
with a more extended and carefully defined use of computational tools. This should allow
a deeper understanding of the phenomenon and a reduction in the amount of empiricism by
an equivalent increase in the physical insight of the complex flow behaviour in modern
turbomachines.
 115

11.2. Review of Loss and Deviation Precition Mehtods

Introduction to the "Review Papers" and "Answers to the questionnaire"

The first review paper of Hirsch, %ritten at an early stage of the Working Group activi-
ties, raised several problems on the state of the art and indicated tentatively area.
where short-comings were evident. Two actions were set forth in parallel a a questionnai-
re on the state of the art was sent to industries and research organizations, and Working
Group Members were asked to produce review papers on specific items, i.e. a
Diffusion factor
Shock losses and shock boundary layer interaction
Secondary flow and clearance effects (correlations)
End-wall boundary layers
Deviation/TurnmnV angle correlations
Blade-to-Blade calculation methods
to which were added, as a result of the questionnaires and of the WG Work, two papers
* Reynolds number effect
* Part span damper losses (2 contributors)

This part of the final WG report contains the review papers, a summary and comments on the
answer to the questionnaire, and conclusions derived from the information contained in
this section.

It was also deemed necessary, although this was not included in the original task, to in-
clude a short discussion on surge line prediction.

Pni
 116

11.2.1 Survey on Diffusion Factors and Profile Losses

List of Symbols

Aa Annulus Area
C Chord Length
D Diffusion Factor
Deq Equivalent Diffusion Factor
i Incidence Angle
H2 Boundary Layer Shape Factor
M Mach Number
r Radius
t Blade Thickness
U Rotor Speed
V Relative Velocity to Blade Element

I)Relative Flow Angle to Blade Element

A Difference in Magnitude
a Solidity
Se Boundary Layer Momentum Thickness at Blade Trailing Edge
Profile Loss Coefficient

Subscripts

c Choke Condition
ink Incompressible Flow
kom Compressible Flow
max Maximum Value
m Meridional Component

3 Surge Condition
e Circumferential Component
1 Blade Inlet (LE)
2 Blade Outlet (TE)

Superscripts

Optimum Flow Condition

Corrected Values (Jansen/Mofeat)

Introduction

This section summarizes the various attempts published in the open literature to
correlate compressor profile loss with blade loading. The correlation techniques des-
cribed here are already briefly reviewed by Ch. Hirsch in (1).

The following subjects are discussed in some detail:

- diffusion factor and its development since Lieblein (2,3,4)

- profile loss correlation of:

W.C. Swan (5), N.T. Monsarrat (6), W. Jansen, W.C. Moffat (7), L. Fottner (8,9),
P. Strinning (10), L.H. Smith (11).

This calculation procedures used for compressor design and analysis are based more
or less on empirical correlations for flow losses and flow turning. For two-dimensional
incompressible cascade flow the profile loss is related to the boundary layer parameters
at blade trailings edge as follows: 23 .

The boundary parameters themselves depend primarily on the blade loading,

involving the maximum blade suction surface velocity as well as the cascade inlet and
outlet flow conditions.

L-A_ _
 117

Diffusion Factor
In 1953 Lieblein (2) made his first attempt, based on Buri's separation criterion,
to establish a relevant loading parameter that appeared to describe satisfactorily the
low speed relation betweer blade loading and losses at optimum flow condition (a. Fig.1).
Lieblein defined the so-called diffusion factor D as a worthwile loading parameter:

max 1 V91
D Vax - V2 D vth
- -- - 1(2)
V1 V1 2 GV 1

The model is based on low speed measurements of NACA 65-blades of 10 % thickness and
is thus limited to the prediction of blade element losses caused by surface friction,
boundary layer separation, and wake mixing.

Lieblein's later work (3,4) on blade losses led to a refined parameter called the
equivalent diffusion factor and expressed as the ratio of maximum suction surface velo-
city to outlet velocity: VVmax

Deq " (3)

This factor derived from Schlichting's boundary layer work offere a more general corre-
lation of blade wake momentum thickness i.e. profile loss with blade loading.

As far as cascade measurements are available all flow dta including maximum sur-
face velocity and wake flow distribution are known. As an example Fig.2 and 3 demonstrate
the wake momentum thickness as a function of D for minimum loss and off-design respec-
tively. Data for NACA 65 (C 1 oA1 o) 10 and C. 4 89rcular arc blades are shown.

For design purposes and off-design analysis the maximum surface velocity and the wake
parameters are generally unknown. Only the inlet and outlet flow data are prescribed
for design or otherwise evaluated for off-design analysis. Thus, Lieblein formulated the
following correlation between Deq and cascade inlet and outlet flow conditions:
[eq
- 1.12 + 0.61 (tam 5 - taA 4
e cos a 2

This equation was proved by the cascade tests mentioned before; it is therel'ore limited
to these types of cascades. Furthermore, it is only valid for the minimum loss situation
in incompressible flow.

Lieblein's method is still used for compressor design. There have been some success-
ful attempts to enlarge the range of applicability to high subsonic Mach numbers and ar-
bitrary blade shapes. Furthermore, some modifications were implemented concerning stream-
line shift and streamtube convergence in actual compressors (12). The details are des-
cribed in the succeeding chapter on profile loss correlations.

Profle Loss Correlations

The relations today used for predicting profile losses are simpifications of Eq.(1).
Approximately all correlations, valid for design as well as off-design, start with de-
fining the minimum (design) profile loss for a 2d-blade element in incompressible flow.
Then corrections for Mach and Reynolds number effects as well as for non-2-dimensionel
geometry are provided. The profile losses at off-design are estimated as a function of
incidence angle and blade loading variations based on the design values previously cal-
culated.

Swan's Correlation (5)

For design Swan first determines the equivalent diffusion factor with the well-
known inlet and outlet flow conditions. When using the following equation for Deqo
are incorporated:
streamline radial shift and blade thickness distribution

COS (,.12 + 0.61 (5

V m2 Cos 1 X

2
Cos rrrV
2 V2 22
where K rtan Im2 -2 tan LL) (2
L
1 1 Vmi m
Swan established correlations of trailing edge momentum thickness with diffusion
factor D, * (Eq.(5)) separately for rators and stators based on experimental compressor
element 8 ta. Blade height effects as treamline shift, boundary-layer centrifugation are
included as shown e.g. in Fig.4. The actual profile loss is estimated using the simpli-
fied equation:
118

-~ *2
() W 2 (6)

At off-design the analysis of the compressor element data showed that spanwise 1o-
cation of the element has little effect, however inlet relative Mach number does in-
fluence the loss-loading compilations. Fig.5 presents a typical curve fit (1/o - O/o0
versus D, - D 0) obtained for double-circular arc rotor blade elements demonstra-
ting the Indepfl~dency of blade height:

•.e2 3)
"= .827 14, - 2.692 K + 2.675 M (o 2
For DEQ > DEQO

3 2
e z
e
-- (2.80 KJ -
2
8.71 NM + 9.36 H I tDEQ" DEQ'
For DEQ c DEQ*

The off-design diffusion factor D

eq is defined as:
Vm 1 Cos 62 1.43 * O.6K
cr 1.02 + a(i - )143 01
•eq Vm2 con0 1

Co.s 2 r,2 Vm22 (1r 2),.

2 (7)
K [nt tan1 n7 $ _ .2( 1 i
It1IVml 1 rl

a 0.0117 for NACA 65-; a a 0.007 for C-Series-, DCA-blades

The off-design losses can be calculated by means of Eq.(6) incorporating the actual
flow angles. Some worthwile modifications of this off-design techniques are given by
Davis in (13).

Monsarrat Correlation (6)

For design purposes Monsarrat presents curves of wake momentum thickness versus
diffusion factor D (s. Eq.((2)) separately for rotors and stators based on experimental
compressor element data. The minimum losses occur at mid span. The losses increase when
approaching hub and tip as shown in Fig.6 and 7. Thus, tip clearance, secondary, and
end wall boundary layer losses are at least partly included.

Allowing for streamline radial shift and axial velocity changes the following ex-
pression for D is derived from Eq.(2):

D -1 V-+
2 r I 1 - r2 Ve2
V1 (r 1 + r) V1 Bia (8)

The design profile loss is defined as function of wake momentum thickness 9/c using
the following simplification of Eq.(1):

cos e2 (9)
There is no off-design loss correlation available.

Jansen/Moffat Correlation (7)

This correlation is based on cascade as well as on compressor blade element data.

The basic parameter used for design loss determination is the design diffusion factor D
as defined in Eq.(8). The relation between the design loss coefficient wO is determined
by Eq.(6) using the following equation for(9/c)} given in polynominal form:

(:) -(0.003 + 0.02375 D - 0.05 D 2 + 0.125 D 3) (10)

Mach number effects and incidence effects at off-design are accounted for using the
method of Wiggins (i4) which proceeds as follows:

1. If the inlet Mach number exceeds 0.7, the inlet flow angles for design, choke and
surge condition will be corrected by means of the following equations:
 119

•"l A s
3 = surge
A"1l0o 01 1.5 I c = choke

All" 3 1 A " z corrected values,

where AA8 10 M1 - 7.

2. The design loss coefficient W* will be corrected in a similar way if the inlet
Mach number exceeds its critical value. 'i'his correction is of the form:
of" z•G* (2 (M1 - Micr) + 1) cr = critical

where again the double prime denotes the corrected value.

3. For off-design operation a parabolic variation of loss with incidence is assumed

with choke or surge being achieved if the loss is twice its minimum value. The loss
coefficient therefore is given by the relation

*" (0.8333 52 + 0.1667 ++ 1.0) (12)

where S is defined as

S for 1
and Sic -
S- Si - *
for B1 >861
1

Fottner Correlation (8. 9)

For incompressible flow and design purpose Fottner extended the relation of wake momentum
thickness versus diffusion factor D* to %rbitrary blade thicknesses t. The following equa-
tions based on cascade data are proposed:

/co 0 1'
+ ir = 2
,~~~n.k C
cs--
• --
C os -- ( 13 )

(6.6 -t +0. 34) (0.0088 + 0.0107 D) 0.052 D2+ 0.116 D3 ) (14~)

V -V
SV~~max-V D aV 2 (15)

V max - 1.03 +(0.4 +

t )- I A V
-- + 0.7 -t (16)

Eq.(13) is again a simplification of Eq.(1). Fig.8 illustrates the comparison of re-

sults calculated with Eq.(15) for different blade thicknesses t to corresponding experi-
mental data.

In order to allow for higher inlet Mach numbers and to perform off-design analysis
Fottner uses the same procedure as described by Jansen-Moffat, however, proposing a
factor of 1.8 instead of 2 in Eq.(11) for Mach number correction.

Strinning Correlation (10)

This correlation technique valid only for design is based on cascade data and covers
the inlet Mach number range up to 1.4. At first the losses in incompressible flow are de- -
"termined using the equations of Fottner. For compressible flow the following diffusion
factor is defined:
 120

D a
Ska \ V1 / (17)

where
Vax 1 fVvmax 2
V1 kom N~
I - VI)n
k k-2
k - I
Sk +÷I

(The inlet Mach number ij related to sonic condition)

t V max\ t I A VO t
,a-1.03 a +(0.4 + - ) ,-0.7
V ink C V

At supersonic inlet flow the diffusion factor is calculated assuming an equivalent

subsonic flow ahead of the shock that leads to the same Mach r %ber at the position just
downstream of the shock.
normal shock equation:
The equivalent subsonic inlet velocity V
1
is determined by the
)
V
V1 " - I

The deceleratinr, from V assuming constant flow directlon changes the

tangential velocity componeAt:
( (A~
IV / ~ 0
.V1- V42 .I
(19)
V1

Eq.(19) is written with the actual flow velocity components, the term(&VO/V1 ý
is then used to calculate (V /VI) an dthe same procedure is followed to de-
termine the diffusion factor•. n

Dkom \V/kom Via

Smith Coro'elation (11)

This correlation technique only valid for design loss prediction is based on boundary-
layer calculations along blade surfaces with normalized roof top velocity distributions as
b shown In Fig.9. Compressible turbulent boundary-layer theory has been employed to relate
profile loss to suction surface diffusion ratio, streamtube contraction, Mach number.
Reynolds numbes, and specifi- geometric blade parameters.

These investigationr led to a more general definition of diffusion factor - including

compresaib'.ity effects, streamtube contraction, and :rbitrary thicknesschord ratios:

*q Wn 01 - r*)(1 \K- C0) 3 4 ' (21)

The various terms in Eq.(21) may be written:

rW V 9 1 - V 2
2

F V1I

- (rI + r 2 )

Cos ( .a l
Aal
 121

2
m*- - 1- H1 -•M (1"- A;" 1- tan 1 Or'os1 )

with: Ko 0.2$5

SK2 2 0.4458
K3 = 0.7688
K4 = 0.6024.

In (11) Koch and 3mith discuss in more detail the effects of Maoh number, Reynolds
number etc., however the general relation of losses, trailing edge momentum thickness,
and diffusion factor are not presented in detail.

Conclusion
World wide great emphasis is placed upon the development of new and sophisticated
stage and blale design procedures that finally will replace the simple loss prediction
techniques today in use. Momentarily however these new techniques are still difficult to
apply to real machine design because extensive iterative hub-to-tip, blade-to-blade, and
boundary-layer flow calculation is needed. Great and long term efforts are still neces-
sary to improve the physical basis and the handling of these programs. Thus in the near
future actual machine design and off-design performance prediction will utilize current
empirical loss correlations that however must be refined towards more general validity
and reliability without giving up their simplicity.
Parametric studies with blade-to-blade flow and boundary-layer calculation codes are
progressively attempted to refine the diffusion factor concept and loss correlation at
design as well as off-design compressor operation. The codes begin to demonstrate how the
flow parameters (incl. Mach- and Re-number) and the details of blade geometry interact
to produce performance characteristics, as was stated in a very encouraging paper on the
relation of measured losses to calculated blade-surface velocities which has been re-
cently published by T.F. Gelder et al. (15).

I.

1
122

List of References
(1) Hirsch, Ch., Survey of Deviation and Loss Correlations, ACARD-PEP Working
Group 12, Axial Compressor Performance Prediction.
(2) Lieblein, S., Diffusion Factor for Estimating Losses and Limiting Blade
et al., Loadings in Axial-Flow-Compressor Blade Elements,
NACA RM E53D01, 1953.
3) Lieblein, S. Loss and Stall Analysis of Compressor Cascades.
Trans. ASMP, Series D, J. of Basic Eng., Vol.81, 1959, p. 387.
4) Lieblein, S., Resume of Transonic-Compressor Research at NACA-Lewis Laboratory.
Johnson, I.; Trans.ASME, J. of Eng. for Power, 1961, p.219.
5) Swan, W.C. A Practical Method of Predicting Transonic Compressor Performance.
Trans.ASME, J. of Eng. for Power, July 1961, p. 322.
6) Monsarrat, N.T., Design Report Single Stage Evalkiation of Highly Loaded High-Mach-
et al., Number Compressor Stages, NASA V69-30869, 1969.
7) Jansen, W., The Off-Design Analysis of Axial-Flow Compressors, Trans.ASME,
Moffat, W.C. J.of Eng.ror Power, Oot.1967, p.453.
(8) Fottner, L. Kennfeldrechnung
72/008, 1972.
fUr Axialverdiohter, MTU, Teohnisoher Berioht
9
9) Fottner, L. Answer to Questionnaire on Compressor Loss and Deviation Angle
Correlations., AGARD-PEP, Working-Group 12, 1979.
(10) Strinning, P., Grundlagen des Auslegungsverfahrens und Vergleiah Messung/Reohnung,
Dunker, R. Forschungsberichte Verbrennungskraftmaschinen, Heft 235, 1977.
(11) Koch, C.C. Loss Sources and Magnitudes in Axial-Flow Compressors.
Smith, L.H. Trans. ASME, J. of Eng. for Power, July 1976, pp. 411.424.
(12) NASA Aerodynamic Design of Axial-Flow Compressors.
NASA-SP 36, 1965.

(13) Davis, W.R. Axial Flow Compressor Analysis using a Matrix Method,
Carleton Univ. Report ME/A 73/1, 1973.
(14) Wiggins, J.O. A Procedure for Determining the Off-Design Characteristics of
Multistage Axial Flow Compressors. MS dissertation, UnIv. of
Cincinnati, Cincinnati, Ohio, 1963.
(15) Gelder, T.F., Aerodynamic Performances of' Three Fan Stator Designs Operating
Schmidt, J.F., With Rotor Having Tip Speed of 337 Meters Per Second and Pressure
Esgar, G.M. Ratio of 1.511; NASA-TP 1614, 1980.
 123

MI

VI

.302

-.
40
I0 o ;) vN
A% CHAORD
S.0ROTOR TIP REGION9
U .0

II
SROTOR U

RIFFUBICJ
ORCTORR 0

200*
! Fig.1 Loss correlation with dif'tusion-f'actor tor compressor inlet stages.
.1- ANVSAO

.05 SOLIDITY,

Se.,04 o05

W.03 ¢ 1.0 6
5(A ojI01
SA ~125
v 1.5
-02- A 1,0 &
S1,0 )IOC.4
z00

0 1.2 1.4 1.6 1.8 2.0 2.2 2.4

EQUIVALENT DIFFUSION RATIO, Deq*

Fig.2 Correla'ions of wake momentum thickness with equivalent diffusion ratio at minimum
loss incidence angle.
 124

SLADE SOLIDITY
V .1O0-
u o I 0.5
4 a 1 .75
0 1,.0 0
; .'08- 1.25
Vw I 1.5
W 2 1.0
Z ' 3 1.0

Z .04i
a0-VV
z .02
0z

1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 St 3.2
EQUIVALENT DIFFUSION RATIO, D84

Fig.3 Correlatiorn of wake momentum thickness with equivalent diffusion ratic At inci-
dence angles larger than minimum loss incidence.

0.25

DOUBLE CIRCULAR ARC SUBSONIC

LOSS ESTIMATION CURVES (ROTOR)
TIP

STATISTICAL CURVE FIT

BASED ON PERCENTAGE
0.15 OF CHANNEL HEIMII'I

0.10

.05
0.0

HUB

1.0 1.20 I.40 .60 1.00 2.00

MIMIMUM LOSS EQUIVALENT DIFFUSION - DEQ*

Fig.4 Estimated variation of wake momentum thickness with minimum loss equivalent
diffusion.
 125

u.t i .I ! A

.10

.4
.76T

.-.20 0 .20 S•
.5U

DEQ- DEQ*

Fig.5 Estimated off-minimum loss variation of wake momentum thickness with equivalent
diffusion.

0•7
0,06 0-To
0,05 20 &
20 &700 .
- 00430- 50ý
'COS/3 2 Q0_o•Spo-aon _/_

2Q0.03-¶
0,02
0,01
0
"110 0,1 0,2 0,3 0.4 0,5 Q6 0.7
D-Factor

Fig.6 Rotor profile loss parameter vs diffusion factor D

% Span:
0,08 •-- - 20170.
QO8 5195,
10190,

0.07 30 00,
0
0.06_____________________

2a 0,04 ... .
003 - _ _ _ _

0.02

0 001
Q2 0,3 OA 0,5 0,6 0,7
D-Factor

Fig.7 Stator profile loss parameter vs diffusion factor D

126

0,07.
0,06- -- t Prof ilee
02H ACA 65 NACA4-ziff NTE C4 t=/,0,15
c 0.05 A c/,
0,05" - 0 06 0
010 0
U
U)1 )j 5 1 0,10

_o030,0
Re= 4 .105

E
0 0,02- "

00

0 0,1 Q2 0,3 0,4 0,5 0,6 Q,7 Q8O0,9

Oiffusion Factor 0

Fig.8 Blade trailing edge momentum thickness as function of diffusion factor D

1.5 '
1Vmao/VoTE:

>1>"- 1,2 -_,_ /--1.3

,0,
1 .0 - _ _ _ _ _

0.6 0 20 40 60 80 100

Fig.9 Blade surface velocity desitributions employed to determine Mach and Reynolds

number effects on blade profile loss using compressible boundary-layer theory.

 127

11. 2.2 Survey of Models for Shock and Shock/Boundary Layer Interaction Loss Prediction
List of Symbols
b axial chord length Indices
cP static pressure coefficient c Carnot loss
d blade thickness crit critical
F area D pressure surface
I chord length 9 circumferential
La Laval number, based on critical sonic A incompressible
velocity
max maximal
m mass flow min minimal
Ma Mach number, based on local sonic
velocity N leading edge
S shock position
P static pressure
r radius t stagnation condition
s entropy thr throat
t pitch 1 inlet
temperature 2 exit
- average value
w velocity
normal shock
w change of velocity
specific heat ratio A downstream of shock
ýN V supersonic expansion
W loss coefficient

1. Introduction
Increasing the inlet Mach number leads to a condition for which locally on the suction
surface the peak velocity is equal to sonic velocity. This inlet Mach number is called
the critical Mach number. Above this value a local supersonic field terminated by a
normal shock is built up (see Fig. la). For the so called choking Mach number this
shock limits the mass flow, because for this flow condition the throat velocity reaches
sonic conditions. For further increase of the inlet Mach number to supersonic inlet
conditions, there is a shock system (see Fig. lb) consisting of an oblique shock which
influences the region upstream of the cascade (for subsonic axial velocity) and a nor-
mal shock within the blade passage. It has been experienced that the oblique shock in
front of the cascade produces very low losses, whereas the normal shock within the
blade passage produces high losses for high shock Mach numbers (Fig. 2). The strong
increase of losses in transonic compressor cascades at high inlet Mach numbers is not
only attributed to the shock losses according to shock correlations, but primarily to
the shock/boundary layer interaction which leads to boundary layer separation due to
the strong pressure gradient of the shock.
These very complicated flow conditions have made it impossible up to now to accurately
calculate the shock losses with analytical methods. Therefore most of the published
methods are derived using very simple shock models and empirical shock loss correlations.
2. Empirical Shock Loss Correlations
The evaluation of cascade measurements shows that for low inlet Mach numbers (see Fic. 3)
the losses are independent of the inlet Mach number. Above the critical Mach number
the losses are increased due to local shocks and shock/boundary layer interaction. For
supersonic inlet flow conditions the gradient of the losses becomes more steep because
of the boundary layer separation downstream of the shock. The Mach number dependency
can be divided into three regions (see Fig. 4):
- in the region up to the critical Mach number the losses are independent of the
Mach number and are equal to the incompressible value (this case is treated
separately)
- the region between the critical Mach number and the sonic inlet condition with
supercritical inlet Mach numbers
- the region for supersonic inlet velocity

2.1 Critical Inlet Mach Number

The critical inlet Mach number is by definition the inlet Mach number for which the
peak velocity on the profile suction surface reaches the sonic condition. By applying
the gasdynamic functions for this flow condition, the following equation is obtained:
 128

LOicrit t - Cp(1 ) rain

and
La 1crit
M~lcrit
40.5 ( -1) Laicrit
The local static pressure coefficient

' ~Cl= P-P1.

is assumed to be constant in the region up to Maicrit,

this yielding:

{CpPmi=
(Cpi)min ( Cp)min ik = Wmax
I - (-!t- Ii

Applying one of the known modern pressure distribution calculation methods it is poss-
ible to calculate the peak velocity very accurately. But for fast loss estimation it
is convenient to use approximations. Ref. 1 contains, similar to Ref. 2 and 3, an
empirical correlation for NACA 65 profile series:
w
"x 103 + (0.dl+d/ )'t/IL Awg + 03 dl/ (4)
W ik W

A similar expression for the peak velocity, which must also be used for calculating
the diffusion factor is given in Ref. 4:

Wmax W 1 + 0.7688 d/I + 0.6024 t1 LW (s)

Wl Wthr Wi

2.2 Correlations for Supercritical Subsonic Inlet Condition

Ref. 1 contains a very rough empirical correlation taking into account the compressi-
bility effect on the profile losses as well as the losses due to local shocks:

l- tl- Wik [A (Ma -MOlcrit) +1] (6)

Pt1 - P,
for Ma1 > Malcrit.
The coefficient A is set to A - 2.0 in Ref. I but our evaluations of cascade mea-
surements suggest to set it A - 1.8. Practical application has shown sufficient agree-
ment with measurements, if the inlet Mach number does not exceed Ma - 1.2. An empi-
rical correlation very similar to equation (6) is given in Ref. 5. ior compressor cas-
cades with double circular arc profiles an evaluation of numerous cascade and stage
measurements gives:

W( {1lWlik[4 Ma
-CMQIcrit - 0.4)]'1
+ (7)

While these two correlations give corrections for the compressibility effects, Ref. 6
tries to calculate the shock losses by calculating the viscous pressure distribution
on the profile surfaces taking into account empirical corrections for the supersonic
field and the terminating compression shock. This shock is extended from the shock
position on the suction surface into the blade passage with a suction surface Mach
number immediately upstream of the shock corresponding to the supersonic expansion
on the profile decreasing to sonic condition at the sonic line. In the subsonic region
the Mach number is further decreased to the value on the ptessure surface.
The shock loss is calculated for a mean shock Mach number

Ma, (Mas +1.0) (8)

using the normal shock equation

LAgia
 129

W: L= [a - Las ] (9)

This loss has to be distributed in the blade passage between two blades to get a pitch
averaged total pressure at the exit of the cascade, yielding:

Pt - 1 M -1 ( ) (10)
t IMas -Mac Pt0
If the local Mach number on the pressure side is not known, it should be replaced
by the inlet Mach number.

2.3 Correlations for Supersonic Inlet Conditions

The first published correlation for shock loss prediction is gi .. in Ref. 7. The
method is based on a very simplified shock model, with an oblique shock in front of
the cascade with negligible losses, and a normal shock in the blade passage which is
assumed to be normal to the mean streamline. The Mach number immediately upstream
of the shock on the suction surface is calculated as a function of the supersonic
expansion from the inlet direction to the profile tangent at the shock position. The
Mach number immediately upstream of the shock at the leading edge of the adjacent
profile is equal to the inlet Mach number. In Ref. 7 the Mach number to be taken
for shock loss prediction is calculated as a mean value:

Ma = 2 (Ma• + Ma1 ) (11)

Ma is a function of the inlet Mach number and the supersonic expansion

Mas z f (MaiLLv) (12)

Ref. 7 contains some analytical expressions for this function applied to circular arc
profile parts.
If
and further details
the cascade of the
losses are blading axe known, i.e.
to be recalculated if the toblad'ng
(ccntrary designtask,
the design is available
for which
the blading is to be designed), Ref. 8 presents a method for detailed taking into
account the blade properties for calculations of the shock Mach number. The advantage
of this method is that it is also applicable to the off-design case.
Since the finite leading edge thickness leads to a detached shock ahead of the pro-
files, Ref. 9 proposes a further method for shock loss prediction on the basis of
Ref. 10 and 11. The shock model consists of a detached shock immediately upstream of the
profile and a normal shock in the blade passage. By use of the method of characteri-
stics (Ref. 12) the Mach number distribution upstream of the shock at each position
across the blade passage can be calculated. Thus, also the local shock loss is known.
The total loss of the compression shock system is calculated by a mass averaged in-
tegration across the blade passage.
Apart from the losses due to shocks, there are additional losses caused by the shock/
boundary layer interaction, which can be even higher because of possible boundary
layer separation induced by the shock. Ref. 13 contains a method for estimating shock
losses near the design point. The shock model consists of a X -shock in the entrance
region of the cascade causing boundary layer separation. In the region outside of the
boundary layer the suction surface velocity remains supersonic and is decelerated to
subsonic condition by a normal shock emanating from the pressure surface. Downstream
of the shock there is a mixing of the main stream with the separated boundary layer.
In Ref. 14 a method is developed taking into account a pseudo-shock-system in the se-
parated region downstream of the normal shock. Due to the separated region the flow
is reaccelerated to sonic conditions. The following deceleration to subsonic exit
conditions is accomplished by a Carnot deceleration, the losses of this mixing process
to be calculated using the laws of momentum and continuity:

T t, F ( 3

Pt2 F2 Fs
with

2
F2 + LasL(La 2 + 1) -La 2 (Las +1) 14)
Fs ~LQ2 -- L +La
( c
130

According to the justpentioned condition of sonic velocity due to reacceleration in

the separated region Laa has to be set to El a 1.0. It is clear that the boundary
layer does not separate for weak shocks, i.e.. low shock Mach numbers. Therefore a
correction has to be done for the Ca:^not loss coefficient of equation (13):

Pts
where ID - f (P/p) is a function of the static pressure rise of the normal shock.
Experience has shown that V/p > 1.35 leads to separation of the boundary layer down-
stream of the shock, i.e.

=1.0 for PP 1.35

Ref. 4 presents a shock loss model which determines the losses of the two shock parts
separately. For supersonic inlet conditions an additional supersonic expansion around
the leading edge has to be taken into account resulting in an additional loss. This
loss is determined In Ref. 4 in terms of an entropy increase:
AS
-=- in
rb
rN
sinp
[]
1.28 (Ma- 1) +0.96 (Moa ~1
J
(16) i
The entropy increase of the passage shock is calculated for an oblique shock, decele-
rating to sonic velocity.
Within the scope of MTU investigations on shock loss predictions methods, some of the
above described correlations have been applied to relevant cascade measurements, done
at DFVLR on specially MTU designed supersonic profiles for transonic compressors.
Fig. 5 shows the results of the correlations compared with the measurements. At first
an extrapolation of equation (6) had been done for high supersonic inlet Mach numbers,
only to show the limits of this correlation. As had been mentioned in chapter 2.2 this
correlation gives good agreement for Mach numbers lower than 1.2, but for higher Mach
numbers the correlation underestimates the losses. Therefore, for supersonic inlet
Mach numbers an additional normal shock loss had been calculpted for a shock Mach num-
ber equal to the inlet Mach number. This could be done because the investigated pro-
files had been designed with a wedge profile in the sup-'rsonic part of the profile in
order to reduce the shock Mach number. Quite good agre.:rent can be seen between cal-
culation and measurement. Finally in addition to the normal shock losses calculated
for the inlet Mach number, equation (13) has been applied to take into account the
shock boundary layer interaction downstream of the shock, also showing good agreement
with measurements.
Summing up, it had been shown that a very simple correlation, consisting of the normal
shock relation for the inlet Mach number and the compressibility correction term of
Ref. 1, gives good agreement with measurements for the investigated supersonic profile
types.
3. Analytical Methods
The losses due to shocks in a cascade with transonic flow can be calculated (with the
exception of shock/boundary layer interaction), if the flow field in the cascade is
known. This information comes from semi-empirical methods, or methods of characteri-
stics in the entrance region or in recent years from very sophisticated analyticil
methods taking into account the mixed subsonic/supersonic type of the transonic lascade
flow. Due to the effort necessary for application these methods are not considernd
within this survey.
4. Conclusions
The calculation of the shock losses has to take into account the losses due to normal
shocks as well as the additional losses caused by the shock/boundary layer interaction.
Due to the very complex flow behaviour, at present this is only possible by empirical
correlations for very simplified assumptions. Therefore, additional effort has to be
directed to the practical application of analytical methods and to better understanding
of the shock/boundary layer interaction.
Basically, transonic cascades have to be designed in such a way that the shocks are not
present or occur at minimum shock Mach number. It seems to be possible that for high
subsonic inlet Mach numbers the use of new blading design concepts (i.e. supercritical
profiles) can meet the first condition, whereas the second condition of minimal shock
Ma-h numbers has been successfully verified by the use of wedge type profiles.
 131

5. List of References
Ref. 1 Jansen, W. The Off-Design Analysis of Axial-Flow Compressors
Moffat, W.C. Trans. ASME, J. of Engng. for Power, Oct. 1967, 453-462
Ref. 2 Lieblein, S. Loss and Stall Analysis of Compressor Cascades
Trars. ASME, J. of Basis nngng., Sept. 1959, 387 - 400
Ref. 3 Members of the Aerodynamic Design c. Axial-Flow Compressors (Revised)
Compr. and Turbine NASA-SP 36, 1965
Ron. Div.
Ref. 4 Koch, C.C. Loss Sources and Magnitudes in Axial Flow Compressors
Smith, L.H.Jr. J. Eng for Power, July 1976, 411 - 424
Ref. 5 Dettmering, W. Machzahleinflun auf die Verdichtercharakteristik
Z. Fluqwiss. 19 (1971), 145 - 150
Vef. 6 Fottner, L. Ein halbempirisches Verfahren zur Bestimmung der rei-
bungabehafteten transsonischen Schaufelgitterstrbmung
mit Einschlufl von Uberschallfeldern und Verdichtungs-
st8Ben
Dissertation TH MUnchen (1970)
Ref. 7 Miller, G.R. Shock Losses in Transonic :ompressor Blade Rows
Lewis, G.W. Trans. ASME, J. of Engng. for Power, Jul. 1961,235 - 242
Hartmann, N.J.
Ref. 8 Balzer, R.L. A Method for Predicting Compressor Cascade Total
Pressure Losses, when the Inlet Relative Mach Number
is greater than Unity
ASME Paper No. 70-GT-57, 1970
Ref. ) Starken, H. Untersuchung der Strdmung in ebenen Oberschallver-
z~garungsgittern
DLR FB 71-99 (1971)
Ref. 10 Levine, P. Two-Dimensional Inflow Conditions for a Supersonic
Compressor with Curved Blades
J. of Appl. Mech. 24 (1957) No. 2, S. 165
Ref. 11 Mbckel, W.E. Experimental Investigation of Supersonic Flow with
Detached Shock Waves for Mach Numbers between 1.8
and 2.9
NACA RM ESODO5 (1950)
Ref. 12 Lichtfuss, H.J. Berechnung der ebenen Oberachallgitterstromung mit
Hilfe des analytischen Charakteristikenverfahrens
DLR FB 73-34 (1973)
Ref. 13 Gustafson, B.A. A Simple Method for Supersonic Compressor Cascading
Pekfnrmance Prediction Cascad
ASME-Paper No. 76-GT-64 (1976)
Ref. 14 Volkmann, H. Aerodynamische Entwicklung eines dreistufigen Trana-
Fottner, L. sonik-Frontgeblases
Scholz, N. ZfW 22 (1974) 4, 135 - 144
132

expansion

Fig.lbShok sytemforshpronckiltcodto
 133

Pt
,
1,0

P 0,92

1,2

0,1,4

0,68

- 1.5
0,64

0 4 8 12 16 20 24 28
AVo

Fig. 2: Shock losses as function of inlet Mach number and

supersonic expanFion
134

OFVLR measurement T
0,10
W 0,0 h--1 without trip wire
*-.-. with trip wire
0,06 -- NACA measurement

0,04-

0,02 , malcrit

P2_p._ I i
IZ

P2-PI 0,4
Pt' PI 0.2
0
0 0,2 0,4 0,6 0,8 1,0
Ma1
Fig. 3: Rasults of cascade measurements as a function of
inlet Mach number
 133

C0

ta..

C4C

M
r- 3 C__ Aa W

-3 w
c-_ _

U.1

- 141CL Go-

D 31

CE SO)L-

LC

CD ~~~ c~ D C 0 D
136

0,44
0 meosurement DFVLR

0,40 + (W Z- NREC

O Wm= ik +WI 0j÷+ WC

0,36 0 Woa WNREC + .___al _,

Profile: MTU 1]
0,32 MTU 2 13(0

MTU 1,1 o0

0,28 ARL e

0,24

0,20 _ _ _,,. O0

0,16 0

0,12 U 0 q 0•

0 0 ++
0,08 ..+
C3• +*+ +
0,04 * '%
*0 +
+ +
+
4+ +

0 ----

1,0 1.1 1,2 1,3 1,4 1,5 1,6 MlQ

1 1,7

Fig. 5: Comparison of measured and calculated losses for

"different loss models
 137

11.2.3. End-Wall Boundary Layer Calculation Methods

NOMINCLATURR

a velocity component Subscripts

Cf wall friction coefficient
CL lift coefficient m,y,t in streamline coordinates
F force defect x,ya in cartesian coordinates
f blade force e at the edge of the boundary layer
h annular height w at the wall
H total enthalpy
m mainstream coordinate
Smass flow Supscripts
p static pressure
Pt total pressure fluctuation term
r radius pitch averaged value
s pitch
t transverse coordinate
Tt c total
tip clearance
temperature
u mainstream velocity component
v radial velocity component
U rotational speed
C absolute velocity
w transverse velocity component
W relative velocity
x axial coordinate
y coordinat• normal to the wall
z tangeut..,I coordinate
0 flow angle
S blade angle
8 displacement thickness
c wall skewing angle
w efficiency
momentum thickness
p static density
T shear stress
* pressure coefficient
M pressure loss coefficient, rotor angular velocity
o solidity
INTRODUCTION

Two ways can be defined to take into account the secondary flow effects on the performan-
ce of an axial compressor. The more classical way consists in defining correlations for
the secondary losses as well as for the angle corrections and to distribute them in a
certain way along the span. These losses are added to the profile and tip clearance los-
ses and affect the overall efficiency. Although it is the most widely used method, cor-
relations for secondary losses are subject to a large uncertainty and a large variety of
formulas for the secondary loss coefficient can be found in the literature i 1.

Another way, which will be discussed in this note, is based on a totally different ap-
proach. The starting point lies in the consideration that all secondary flow effects
originate in the end-wall boundary layers along hvb and shroud walls of the machine. In-
deed, the existence of secondary flows is directly cornnected to the existence of an inlet
boundary layer along the end wall and tied to its three-dimensional character. Therefore
an estimation of the end-wall boundary layers and their development through the machine
will provide information about the secondary flow effects, and define the efficiency drop
due to the end-wall boundary layer losseei.

When using end-wall boundary layer calculation methods, the basic flow equations ara in-
tegrated in the pitchwise direction (pitch averaging) and over the boundary layer thick-
ness, simplifying considerably the amount of information which is to be analysed. While
detailed information is lost by integrating the equations, the method is able to take
into account various aspects of secondary flow phenomena in pitch-averaged form.
The important properties of the end-wall boundary layers and their influence on the over-
all efficiency are based on the following considerations. One can, generally, distin-
guish three sources of effects on the end-wall boundary layer
i) The deviation of the main flow between two adjacent blades
ii) The relative motion between blades and walls
"iii) The clearances
parameters at rotor tip and stator hub.
such as geometrical
details of the
These three factors will not necessarily act inblading,
the sameinitial
sense boundary
with the consequence that
layer thickness,
twisting of the blade, rotor-stator distance, ea. could play an important role too.

The basic idea behind the calculation of the efficiency drop due to end-wall boundary
layers lies in the fact that the blade sections in the boundary layer work under high in-
cidences and hence do not perform any useful work, while the energy input per unit mass
 138 $
flow is increased in theme regions. Hence, there is a lose of efficiency compared to a
situation without end-wall boundary laeros. An elaborated end-wall boundary layer cal-
culation procedure is able to deliver i
i) A blockage factor which includes secondary flow effects, and which also affects the
overall efficiency of the machine.
ii) Terms defined as "force defect" which represent the blade force variations inside
the boundary layers, and which on their turn affect the overall efficiency.
The importance of these force defects is fully recognised in
recent years and their
non-zero value is responsible for possible equilibrium stage situations in multi-
stage compressors where the boundary layer can reach an approximately constant
thickness after a few stages.
"iii) With the int'oduction of families of velocity profiles in the boundary layers, ra-
dial variatiwns of incidences and turning in the wall regions can be obtained. Also,
with an extra assumption on the static pressure variation inside the boundary layer,
full profiles of total pressures and hence of louses can be deduced (this would
form a "local" alternative to the global approach consisting of defining an overall
EWEL efficiency loss). It therefore appears that the end-wall boundary layer me-
thud, with some adequate information for the force defect terms, provides a cohe-
rent and general approach for the through flow, including the influence on the
overall efficiency.

SEVERAL PROPERTIES OF END-WALL BOUNDARY LAYERS

1. The three-dimensional character of the end-wall boundary layer

The turning of the flow through the blades gives rise to a transverse(pitchwise) pressu-
re gradient which holds the centrifugal force, arising from the turning, in equilibrium.
This pressure gradient is impressed on the wall boundary layer on account of the usual
boundary layer properties, but centrifugal force is reduced inside the wall region due
to velocity decrease. Hence, the dominating transverse pressure forces will act on the
boundary layer flow and deflect the streamlines more strongly than outside the wall
region. This simple fact can also be understood by considering that in order to main-
tain the same centrifugal force, and to satisfy the equilibrium equation

c2
at-rl
where t indicates transverse direction, the decrease in velocity has to be compensated
by a decrease in curvature radius r.

Since this transverse pressure gradient acts from pressure side to suction side, the
boundary layer streamlines will be deflected towards the suction surface. This implies
the existence of transverse velocity components in the end-wall boundary layer and,
through continuity, the existence of a secondary flow pattern (Fig. X1.2.3.1), which is
not confined exclusively to the boundary layer region. It should be noticed, however,
that the existence of a secondary flow in the outlet plane of a blade row can, at least
partly, be considered as an inviscid effect although one has to assume the existence of
a wall boundary layer in the inlet plane of the blading.
In the language of inviscid fluids, this means the existence of a non-zero vorticity com-
ponent in the transverse direction at inlet, which through turning round a spanwise di-
rection, will give rise to a longitudinal component of vorticity, much in the same way
as a gyroscopic effect. The intensity of the induced streamwise vorticity is proportio-
nal to the velocity gradient at inlet (the transverse vorticity of the inlet boundary la-
yer) and the turning of the main stream.
This effect leads to an increase of flow deflection in the boundary layer compared with
the deflection of the main stream, that is, an increase of circulation and hence of loa-
ding in the wall region. Also, as seen from Pig. 11.2.3.1, the outlet angle, measured
with respect to the axial direction, decreases in the wall region (Fig. 11.2.3.2). One
often speaks of overturning at the wall. It is clear then, supposing a two-dimensional
wall boundary layer at inlet of the caucade, or what is also called a collateral boun-
dary layer, that the boundary layer at outlet will no longer remain two-dimensional or
collateral. One has a three-dimensional boundary layer where the velocity vectors are
not all in the same plane. Projecting the velocities on the mainstream direction, one
has profiles of the form indicated in Fig. 11.2.3.3. An important parameter of these
three-dimensional profiles is the angle of the streamline at the wall, or the limiting
streamline, Ew defined by
1 -w ( 'w)t
e m t tad y-10
l r im lmi 2- -
y40 w (2)

where (T ) and ( ) are the wall shear stresses respectively in the transverse and the
main direction. T$emangle C is then also the angle of the resultant wall shear stress
with the mainstreaw direction, u being the meinstream and w the cross-flow velocity com-
ponent.
i "Various experimental and theoretical works have been carried out in order to make clear
the properties of a turbulent three dimensional boundary layer (3DTBL). Most of the

jWL-IJ
 139

efforts have been devoted to the analysis of the mean velocity profiles with the hope of
putting forward a general form in the spirit of the law of the wall and wake profiles for
two-dimensional turbulent boundary layers. This will not be discussed in this note.

2. The relative motion

The relative motion between a rotor and the shroud wall and between the stator and the
hub wall exerts a profound influence on the nature of the boundary layer entering the bla-
F de row. In the following, we will concentrate on the rotor tip region.

2.1. Wall bonaylyretrn h oo

Supposing the wall boundary layer to be collateral kefore entering the rotor, the law of
velocity composition C-i+U allows to verify that the incoming boundary layer will not re-
main collateral relative to the rotor.

Moreover, Fig. 11.2.3.4 shows that the absolute value of the entering velocity will notI
tend to zero when approaching the wall. On the cintrary, the magnitude of the relative
velocity will remain nearly constant. Fig. 11.2.3.5 shows an isometric view of the boun-
dary layer profile effectively entering the rotcý: as well as its resolution into the ex-
lyappears that one has a highly skewed three-dimensional boundary layer entering the
roowhere the cross-flow components are oriented in a direction opposite to the blade
velocity U, hence are directed from suction side to pressure side for a compressor.

Frmthe energetic point of view one can say that the boundary layer receives a kinetic
enrycontribution from the relative motion and enters the rotor in a reenergized situ-
ato.One can also notice that the flow enters with higi angles of attack in the boun-
daylayer region.

2.2. Wall boundary layer deviation in the rotor

The secondary flow originating in the passage between two successive blades will act on
the relative velocity profile entering the rotor. Whatever this profile may be, the ac-
tion of the transverse pressure gradient is always directed towards the suction side of
the blades, so that the transverse component at outlet will tend to be directed towards
the suction side, that is in the direction opposite to the entering cross-flow component
if the turning in the blades is sufficiently high.

In any case (Wu~ .will be very near to the direction of WyUt if the turning is

Thi inicaes
hatthedeflection will be such higher in the boundary layer than in the
external flow, and also (Fig. 11.2.3.6) that the absolute outlet angle aotwill be much
higher in the boundary layer.

Consequently, one can expect the following behaviour at rotor outlet

i) The absolute outlet angle aut~ increases when approaching the wall, and so does the
deflection through the blade.
ii) The tangential velocity variation AcZ is distinctly higher in the boundary layer
than in the external flow, with as consequence that the circulation for the whole
blade row r - 27rrAcz increases with decreasing distance from the wall.
iii) The input energy for unit mass (or the input power for unit mas3 flow as measured
by Euler's formula, E = UAc , has the same behaviour as r' and will also increase in
the boundary layer.

iv) Since the angles of attack in the boundary layer are large, they are generally not
well matched to the blade geometry and the blade sections in this region will work
under high incidences, leading to increased danger of local profile boundary layer
separaL-ion with the subsequent increase in local profile losses. Added to the end-
wall boundary layer losses, one can expect large total pressure drops or increased
energy losses in these regions and therefore the blade sections in the end-wall re-
gion will operated under unfavorable conditions.

2.3. Wall boundary layer leaving the rotor

AS shown in Fig. 11.2.3.7 through an isometric view of the outcoming wall boundary layer,
the composition of velocities when going over from the relative System to the absolute
System leads here again to an absolute velocity profile whose magnitude is nearly con-
stant and which does not go to zero at the wall. This is the situation inverse to the
incoming profiles, where the role of absolute and relative velocities are interchanged.
Fig. 11.2.3.7 also shows a representative sketch of the axial and tangential profiles
where the energizing effect of the relative motion on the wall boundary layer is seen
and compared to the profiles of a "normal" houndary layer. it should however be observed
here that these profile Shapes, in particular the non-zero velocity at the wall# will

soon be attenuated with axial distance due to the non-slip condition at the wk.ll, and the
skin-friction losses, indicating the importance of the axial clearance between rotor and
Clalthe boundary lyrflow is rergzdwhen leaving the rotor andthdipae
metthickness will be reduced, even if it was large at inlet. One does therefore not
necessarily have an incesinbudrlar thickness with axial dsac ftesm
amutas in ordinary boundary layers developing normally on a wall.
140

Summarizing, one can say that the regions near che wall recieve an important amount of
energy, since the flow is nearly tangential. This amount is larger than outside the
boundary layer since (Ac 3)b.1 is larg .r than (Aca)e. However this energy cannot be nor-
mally converted into useful work since the blades are operating under unfavorable condi-
tions. The total pressure increase - or the useful work - does not follow the input ener-
gy increase, and the greatest part of the latter will be dissipated into heat, while part
of it will also serve to reenergize the axial velocity profiles as shown in [ 2 .

One can also expect then that in a multistage machine after a few identical stages, an
equilibrium will be estaolished between the energy supply and the dissipation in the
boundary layer, so that the thickness of the wall boundary layer will remain nearly con-
stant instead of increasing. This prediction as well as the energy supply in the wall
region measured through the total temperature profiles according to
C Ac " c AT (3)
x z '?t
for a perfect gas, are well substantiated by experiments, as shown for instance by L.H.
Smith [3 1 (Fig. 11.2.3.8 an.ý 9).

2.4. Implications on the forces acting on the fluid

The energy exchange takes place through the action of the tangential forces Fz, following

dP dF
U "- Z (4)

where P is the shaft power, and di the mass flow of a streamtube of wilth dr at radius r
dz - PC xdS - Pc 2ffrdr (5)

The total power at the shaft is given by

P . UdF . rt Uf dS - 27rUf dr (6)
fr h Jr hrh

where the force f for unit surface dS - 2irrdr of the axisymmetric streamtube in intro-
duced. Outside t~e boundary layer, where the tangential shear stressen may be neglected,
one has
dF
• Zm "A~z••
d-•- cand f - Pxe Ac ze (7)

Considering this expression as valid also in the boundary layer - as a first approxima-
tion - one sees that fz will vary in the boundary layer as the product of two conflicting
variable quantities. The increase of Ac will compensate in a certain way the decrease
of aAial velocity. Thus it is not impossible that f is larger than the tangential force
outside the wall layer f in certain regions of the wall bourdary layer.
ze
It is however important to notice that the differences between f and f could be large
and consequently it is not certain that one could neglect this differenze with respect to
the shear stresses in the boundary layer. Indeed, these differences are to be kept in
the equations in order to arrive at a consistent picture.

3. Tip clearance effects

In the tip region of a rotor the wall relative motion is directed from suction side to
press,.re side of the next blade, for an observer moving with the rotor. Due to the wall
shear stresses, fluid will be swept along by this wall motion in a sense opposite to the
secondary flow deflection. Moreover, the pressure difference in the clearance region,
from pressure side to suction side of the same blade will also induce a leakage flow op-
posite to the secondary flow. This clearance effects will therefore act against the in-
fluence of the relative motion on the wall boundary layer region. Since these effects
increase when approaching the wall, one will observe a decrease of absolute outlet angle,
tangential velocity variation, circulation and hence input energy (Fig. 11.2.3.12) when
approaching the wall. On the other hand, the total pressure increase in the clearance re-
gion will be reduced again with respect to the total pressure variation one would have
without clearance, due to the leakage flow.

One has therefore a conflicting influence. On one hand, a small leakage flow will act
against the negative influence of the skewed wall boundary layer, improving the flow con-
ditions in this region. On the other hand clearance always reduces the useful work done
on the fluid in the leakage flow region. Besides, a large clearance will strongly dis-
turb the whole boundary layer flow but also will badly influence a large flow region out-
side the boundary layer. One understands then that an optimal clearance could exist, even
if any clearance decreases the overall efficiency of a machine. This is indeed substan-
tiated by cascade experiments [4 ] but it seems that optimum clearance would be too small
to be realized in a machine.

Up to now, nothing has been said about the corner flows. This is indeed a complicated
flow pattern where several concentrated vortices act in opposite senses, see e.g. [5 1.
Without clearance, one has a passage vortex, originating from the secondary flow which
localizes in the corner between wall and suction side, where low energy boundary layer
fluid accumulates and rolls up into a vortex, the passage vortex. With clearance, the
leakage flow also produces a localized vortex, the tip clearance vortex, in the suction
side corner, rolling up in opposite sense to the passage vortex and remaining next to it.
 141

Moreover the motion of the blade relative to the wall in the clearance region has a
"scraping effect" on the flaid which will accumulate in the pressure side corner of the
blade (for a compressor, the pressure side is the leading side) and also finally roll up
into a vortex with the same sense as the tip leakage vortex, the scraping vortex. The
optimum clearance in cascades is explained by the conflicti,g action of the vortices in
the suction side corner where, for small clearances, the ti, clearance vortex attenuates
the corner stall danger produced by the passage vortex.
In a real machine a given stage recieves in fact a time averaged or equivalently the
pitch averaged flow from the preceeding stages and so does also a standing measuring in-
strument behind a blade row. Thereforeit is not unreasonable to assume that little of
the corner vortices will be noticeable in the pitch averaged flows. This is confirmed
by the experiments of [2 1 , where the transverse velocity component, as seen by a stan-
ding instrument behind the rotov, obeys a simple distribution law, the so called Mager
profile law.

With regard to stall, one can expect that it will be conditioned by and arise in the re-
gions where the flow is in the most unfavorable situation, that is in the end-wall re-
gions. It is therefore possible to consider that stall will appear as soon as the end-
wall boundary layers become too important, which is indeed an established experimental
fact.

The quantitative influence of the clearance on the pitch averaged end-wall boundary la-
yers and the effect on the efficiency will have to be introduced through some semi-empi-
rical law connecting the force defect terms with the tip clearance.

REVIEW OF THE END-WALL BOUNDARY LAYER CALCULATION PROCEDURES

1. Basic equations

All the existing theories of wall boundary layer calculations in compressors involve sol-
ving the same basic boundary layer equations, mentioned here in their most general form.
Each theory consists practically of making assumptions about two parameters, the direc-
tion of the force defect and its magnitude.

Integrating the Navier-Stokes equations over pitch and boundary layer, the following ge-
neral equation is obtained [ 6 1 , all variables denoting pitch averaged values
d 2 -
d. -... xc 6 ~ *d t + - (8)
dx -xe 0+cxe 6x dx_ 8

2
where c 6e (C-C)cxdy (W-W)cdy (9)
xe eJ xa- 9
0 0

ce (Cxe-cx)dy (10)

F-F + F + F (
1 2 3

F 1 - Tt J (Ap-Ap)dy (12)

- " -x pe-p)dy (13)

2 x fo

F3 -P (C'c'-C'c')dy (14)
-x d3 x

or F - - (C -Cc')dy (15)
3 dx fj xc x-

The difference between (14) and (15) for F3 which contains the fluctuation terms is due
to different approximations when averaging over the pitch, (14) being obtained by first
integrating over the boundary layer thickness and (15) by first averaging the Navier-Sto-
kes equations. From F, F :nd it follows that the direction of the total force defi-
cit term is by no means obvious.

Equation (8) differs from standard integral momentum equations for three dimensional tur-
bulent boundary layers by the presence of F in its
play a fundamental role and represent the integratedr.h.sdifference
These different defect forces
in blade force over the
boundary layer with respect to thelade force outside the end-wall boundary layer, in-
cluding the "secondary stresses" C'c' where C',cl are the deviations of the velocity com-
ponents from axisymmetrl, and are due partly to secondary flows. In order to close the
system, additional assumptions are to be made with regard to the force defect terms F.
 142

2. Calculation of overall efficiency

According to section I and after solving the end-wall boundary layer equations, the cor-
rected overall efficiency of the machine can be obtained through the definition of two
"boundary layer" overall efficiencies which will be denoted as nu nd n2 .rI represents
the overall efficiency drop only due to the boundary layer pressure losses and is a mea-
sure of the loss in output energy due to the presence of the end-wall boundary layers,
while n2 represents efficiency variations due to blade force variations inside the end-
wall boundary layers and rotor shaft friction losses, and is a measure of the variations
in rotor work due to the presence of the end-wall boundary layers. If • denotes the
adiabatic efficiency of the machine which would be the total efficiency in absence of
end-wall boundary layers, then the total efficiency is given by
n- nI n2 (16)

Aftip

" hub (17)

Atip

hub

tip
" • •Jhub
and ad- (18)

jhub

The efficiencies and are defined byAiJ f H 5 dth

Aftu He ddi

) I Atpas(9a
•hub

tip
Ahub d

hub (19.b)
n2" tip R i
•hub d• 24

The head denotes values as they would be calculated in absence of end-wall boundary
layers.

Both numerators and denominators of nI and n2 can be related by the following equations.

Ahtip H' di - Athu

t is dii - 21r E Aprc 2xe (cxex+ Czeez 2C
"hub hub h,t
where the second term in the r-h-s is an approximation for the energy thickness
-2 ~ tip FeC Ce 2 CC 2 d ZP c0 C6) c 2(21.a)
x
eCx Jhub PCfe(-- - -)dr P(cO
xe x ze ,1 xe

and
ti tip 1 out* d-
xe aX am,
- prcr
'hub •ubhtfin
hubH d ii- 2w hE
Afhub w
CrUbladeF z - Irrwwall
wU T zwII - d

(21.b)
where the subscript w denotes wall values.

The total efficiency can be approximated in a one dimensional approach by [3 1

1 - E 6
hth
n h h(
F 22)
i z---
R
hit ze

It is to be noted that the displacement thickness 6 can strongly be influenced by an

axial force deficit F through equation (8). In general the axial defect force seems to
affect the total efficiency through the displacement thickness in a more important way
than the transverse component F .

.1z
 143

3. Review of different assumptions

Starting from the basic equations, written here for convenience in cartesian coordinates
+ * dcxe T F
d 2
_c
dx xe6x +c xe 6 x --
dx 0
• + _A
P (23)

d 2 * dc T F ((24)
e2 + C 6 - -- +
dTx Cxe xe x - dx p

or in streamline coordinates [77

d 2 *du e 2 (0 da Ft
dm e a x e 1 2 dm + -P(5

d 2 2 do Tp
d u e + u (6 +e -6 tg + Ft (26)
dm e2 e x 1 2 dm P P
L where 61 . T. Cosa (27)

22 T.1t cosa (28)

and Tt= m tgEw (29)

the following review can be made concerning the different theories.

- Railly and Howard [8 1 (1962)

Railly and Howard solved eq. (8) with F - 0. They considered an "unbounded" axisymmetric
!boundary layer and didn't consider the interaction blade - boundary layer represented by
F. The shear stress terms were evaluated using a shear stress coefficient related to
the .(eynolds number based on the boundary layer thickness. This method allows for theexis-
tence of a cross flow through eq. (26) which can be rewritten as, witf. F-0

d 2 ( dtga + 1t30
dx ue 0 2 Cxe
-2 dx pcosa
02 being a direct measure for the cross flow. The first term of the r.h.s. of eq. (30)
represents the pressure gradient which always tends to create an overturned flow, the
second term is the cross shear stress 4hich always acts against the cross flow. No ef-
ficiency corrections were introduced by Railly and Howard.

- Ccoke and Hall 19 1 (1967)

Cooke and Hall proposed principally the same equations as Railly and Howard, but written
in streamline coordinates and using four impulse moment thicknesses.

- Stratford (10 1 (1-67)

dtratford's main intention was to estimate the blockage factor due to the end-wall boun-
dary layers without considering detailed secondary flow effects. Stratford assumed the-
re is no variation of flow angle within the boundary layer, always giving a collateral
flow. In this way, only one of the momentum equations is to be solved, namely eq. (23),
the axial defect force F being zero. In this case the cross flow equation (26) reduces
to Coll
Fz S..
F . . c 2°11
2 16+6x)* 11
. C (31)
p pcosa xe dx
The transverse defect force denoted Z:asFI vanishes with the pressure gradient term in
eq. (26), eliminating the origin of the passage secondary flow. Physically, this means
a decrease in blade force inside the boundary layer, giving an increase in efficiency
through equation (22).

Thus the method implies the existence of a defect force directed in the transverse direc-
tion z and which should be taken into account when calculating the overall efficiency of
the machine. The validity of this simplified method is questionable, since the assump-
tion of collateral flow cannot be valid at inlet of a multi-stage compressor blade row,
where a strong inlet cross flow is present due to relative motion (see Fig. 11.2.3.5).
However it should be repeated that Stratford Ceveloped this method only to compute the
end-wall boundary layer blockage without introducing explicit efficiency correlations.
- L.H. Smith [3 1 (1969)
L.H. Smith presented an experimental analysis of the dependency of the end-wall boundary
layer on aerodynamic loading, blade-to-blade passage width and tip clearancef in the case
of a repeating stage configuration. He showed the experimental evidence of defect for-
ces and analysed the repercussion of the end-wall boundary layer parameters on the over-
all stage efficiency. Moreover, some simple correlations based on test results for tip
clearance effects and force defects are also given. L.H. Smith proposed to impose a
growth in blockage due to the tip clearance t C through
144

, ,tc.O
. + t C(32)
cmax
This correlation is illustrated in Fig. 11.2.3.11. Extrapolating this correlation to
the basic equations, it would imply an axial defect force component proportional to t /C.
On Fig. 11.2.3.12 an idea in given of experimental results for a non-dimensionalized C
tangential force defect defined as
F Z (33)
S ze

where v - (fand f oCc

P (34)
se 0
Tentatively a value of 0,65 was given for v/8 . It is to be noted that the force defi-
cit defined by L.H. Smith does not include all the force defect terms defined in eq.
(11).

- Horlock [11 1 (1569)

Horlock considered a three dimensional bounded boundary layer by introducing distributed
body forces representing the action of the blades. He used hypothetical axisymmetric
equations written in streamline coordinates and found extra terms - absent in Cooke and
Hall's equations - which are due to the presence of a non zero vorticity normal to the
wall outside the boundary layer, while Cooke and Hall's hypothetical flow was irrotatio-
nal. These equations are equivalent to the basic ones (8) with zero defect forces and
were applied by Horlock and Hoadley [12 1 . They allowed for variations of the flow an-
gle across the boundary layer and assumed Coles and Mager velocity profiles. They used
an averaged entrainment equation. The force defect terms were not considered.

M-ellor and Wood [13 ] (1970)

Mellor and Wood considered the bounded boundary layer equations and averaged them in the
cartesian coordinate system. All fluctuation terms, including blade profile shear stres-
ses were grouped in generalized, effective blade forces. The basic equations were inte-
grated over the blade length, connecting inlet and outlet thicknesses directly. They
assumed the defect force nearly perpendicular to the blades (introducing a small parame-
ter) and eliminated its magnitude by combining both momentum equations. A semi-empirical
correlation for the cross flow thickness 62 at outlet of a blade row was introduced based
on secondary flow considerations, this thickness being zero in absence of tip clearance.
With present notations, this correlation can be written as

(ue2 1 C1/2 sinAc

e 2 outt - KtcctueL sin (35)
In this case, the cross t-mentum equation (26) becomes, assuming a linear variation of
2
u 8
-U2 - .u2 out (36)
dx e 2 x-X out
implying a defect force in the t direction
Fcoll
t Ftol+ (s ot (37)
QcosO ocoss X-x out
eout ocosn

This concept allows for negative values of the axial defect forces, with as a consequence
the possibility of a quasi-equilibrium of the boundary layer after a few stages, through
the axial component of equation (8) or through eq. (23). In this way a complete repea-
ting stage configuration can be predicted.
It is to be noted that Mellor a Wood's method was the first to deliver a complete useful
method, including the main secondary flow effects and the corrected efficiencies discus-
sed in section 2.2. Although both hypothesis of direction of defect forces and cross
flow at outlet of a blade row have no direct confirmation, the method delivers good re-
sults [14 1 115
Marsh and Horlock [16 ] (1971)
Marsch and Horlock made a detailed analysis of the different ways to obtain the momentum
equations and presented a large part of the present review. They analysed the defect for-
ces in the case of a "many bladed" cascade, assuming the cross flow as practically inexis-
tent. They presented the defect force corresponding with a collateral flow as
F_Az. c2 0x+*1 dtg (38)
p xe ( x dx

•, Fx
F - -Fz tgo (39)
cell neato(3)
This corresponds to the value of Ft in equation (31).

The concept of using these forces does not take into account the effects of a cross flow
at inlet of a blade row, as is the case in a turbomachine. This difficulty can be solved
 145

by using Mellor a Wood's method, which is equivalent with Marsh a Horlock'u equations in
the came of zero cross flow at inlet and in absence of tip clearance (eq. 37), or by ad-
ding an extra defect force, as dune by Horlock a Perkings (17 1.

The always positive axial defect force predicts an increase in boundary layer thickness.
However, the corresponding efficiency drop is coMpensated by positive values of F in
eq. (22). Lindsay [181 proposed the use of F and zero axial defect force, 4 hich
principally corresponds to Stratford's method, giving principally lower values of blo-
ckage.

- Daneshyar (19 1 (1972)

Following Marsh A Horlock 1 16 1 , Daneshyar presented the complete derivation of equation
(8) in both cartesian and streamline coordinate systems where the different force defect
terms are discerned. He also proposed the use of force defects which partly compensate
the transverse pressure gradient terms in eq. (26) using an empirical factor giving
F - kFcoll (40)
z a

- Horlock atid Perkins [17 1 (1974)

Besides an interesting review, Horlock a Perkins presented a detailed calculation method
which is principally based on the calculation of the overall secondary losses. Tkey
separated completely the calculation of blockage, tangential defect force F and extra
secondary losses. In order to estimate the blockage they started from Stratiord's prin-
ciple by integrating only the axial momentum boundary layer equation (23), but they added
an axial defect force depending uniquely on the tip clearance.
Kft
tx ff (41)
fxe being the axial blade force component outside the boundary layer and K being a cor-
rection factor for high clearances.
t
Kf - 2t (42)
max
tmnx being the maximum thickness of the blade profile. This method corresponds princi-
pelly with the tip clearance correlation presented by L.H. Smith [3 1 when Kf-l.

Horlock a Perkins used an entrainment relation and a simple skin friction relation borro-
wed from Stratford's method. They did not solve the transverse momentum equation (24),
but only estimated the transverse force defect. This defect force is based on the "col-
lateral" force defect discussed earlier, but they added two extra correction terms
i) The effect of a skewed inlet flow is included. Assuming this inlet skew to be only
due to relative motion, the following force defect is necessary to remove the flow
in the collateral direction, L being the le~ngth over which the collateral flow is
restored F
E- -- L xee x )inlet (43)

ii) The tip clearance also leads to a tangential defect force qince the blade force vani-
shes over the clearance gap. The corresponding force defect was proposed as
F
E . Ktt f (44)
p tc e
where Kt is a correction factor proposed as 0,44.

The total tangential defect force used by Horlock & Perkins, taking all these effects in-
to account was F
f +Kt (45)
me
Besides the end-wall boundary layer losses, extra secondary losses were finally taken
into account in the estimation of the blade profile loss coefficients. Horlock A Per-
kins only considered the effect of the highly skewed inlet flows inside the boundary
layers, which induce incidence losses. This loss was taken proportional to the averaged
kinetic energy normal to the inlet reference flow, being a measure for the inlet skew,
giving K 22
U~c2 6
2 "E,- (46)
h h,tht u.4 ~inlet (6

They suggested a value K -0,2. For turbines, they suggested extra secondary loss corre-
lations for passage secondary flow and tip clearance losses.

Their cquations were all written in a meridional coordinate system, taking radii varia-
tions into account.

- Hirsch 1 7 J (1976)
Hirsch presented a method principally based on Mellor a Wood's method but with equations
written in streamline coordinates. He introduced Power and Mager laws for the velocity
profiles, which led to the use of passage averaged boundary layer equations written in
146

the streamline coordinate system. A modified version of the secondary flow equation (35)
of Nellor a Wood was used allowiiug for non-zero cross flows at outlet of the blade rows
by introducing an extra empirical parameter K2 .
1/2 (7
(u ae2out - (KItc-K 2 c)ua C/L signAcas (47)

- De Ruyck, Hirsch and Kool (6 1 (1978), [20 1 (1979)

- be Ruyck and Hirsch (21 1 (1980)

In reference [6 1 the authors presented equation (8) and introduced mainstream and cross-
flow boundary layer profile models in order to obtain a complete description and predic-
tion of the radial flow distribution including the end wall boundary layer regions.

An analysis is made of the different force defect assumptions and the following expres-
sions are finally proposed
Ft 8
p - -ku 2es 2 k N 1 (48)

for the transverse component of the force defect while it is shown theoretically that the
axial force defect component has to be proportional to the end wall losses, namely secon-
dary and clearance losses, leading to

FFx U2 h .a
uh
- D 2 q (49)

where g is the axial chord.

The drag coefficient C is given by

C - CDs + CDT

where Cs is due to the secondary losses and taken to be

C 0.01 CL (50)
D Lh
and CDT is due to the leakage losses and is taken to be
2
C K C -ha (51)
DT L h
with K-0.3.
In FJ- II.2...13 a comparison is given between different assumptions for the blade force
defes showing that assumption (48) predicts the correct behaviour for boundary layer
flc.-s with inlet skew.

RESULTS TO BE EXPECTED FROM A EWBL APPROACH

The introduction of an end-wall boundary layer approach in a performance prediction

method allows accounting for the secondary flow effects in a consistent way, as well as
the calculation of the end wall blockage.

The through-flow calculation coupled interactively with an and wall boundary layer calcm.-
lation method leads to a prediction of the efficiency drop due to the secondary flow ef.-
facts whertby the through flow calculation acts as a mainstream "potential" type flow for
the boundary 1- ir.

i-.terac' - the through flow and the wall layers is illustrated in Fig. 11.2.3.10
and it shou-z be noted that the losses in the through-flow calculation have to be restric-
ted, within this approach, to the profile and shock lossev, all secondary losses being
considered as part of the EWBL efficiency drop.

From results obtained in (20 1 and [21 1 it appears that most of the known effects are
correctly predi- .cU by the end wall boundary layer calculation. For instance, Figure
11.2.3.14 illu3 ,s the calculated maximum efficiency in function of tip clearance for
various aspect .os.

The influence of aspect ratio and tip clearance, Fig. 11.2.3.15 and 16 is seen also to be
well predicted as well as the influence of the axial gap between rotor and stator, Fig.
11.2.3.17. Moreover, the introduction of velocity profiles, as in the mithod of (20) and
(21) allows the prediction of complete profiles over the whole span. Figures 11.2.3.18
and 19 show velocity and angle variation in the main3tream and the wall regions compared
to experimental data for two different compressors. Thus an estimation of the stall li-
mit as an end wall boundary layer separation is obtained and figure 11.2.3.15 to 17 show
that this sst~mation is a valid approximation to the stall limit as function of mass flow
rate.
 147

CONCLUSIONS

The end-wall boundary layer approach to secondary losses is based on the integration of
the pitch-averaged three-dimensional boundary layers along the hub and tip walls, with
extra assumptions for the blade force defect terms. From there an efficiency drop is
calculated including secondary flow effects, an well as blockage.

The method allows in principle for the influence of inlet blockage, clearance, axial gap,
aspect ratio and other parameters to be included through their effect on the end-wall
boundary layers.
From the knowledge of the velocity profiles inside the boundary layers, the full radial
variations of incidence and turning can be predicted (Fig. 11.2.3.19) by coupling a ra-
dial equilibrium calculation to the EWBL calculation.

Similarly, with an assumption on the static pressure variation within the boundary layer
total pressure and hence local loss coefficients can be estimated in the boundary layer
regions as function of radius. This could lead to an alternative to the present approach,
by using the loss and turning predictions in the EWBL region to estimate local efficien-
cies instead of the global efficiency drop (see calculation of overall efficiency).

This alternative illustrates the generality of the EWBL approach, based on a consistent
set of equations, radial equilibri1im and boundary layer equations, and a limited set of
semi-empirical data for the force defect terms.

Since this approach is based on the integration cf two equations it was not possible
within this short review to discuss all the implications. However, more experimental da-
ta should allow to gain more insight in the validity of the various assumptions with re-
gard to the turbulent properties of the end-wall boundary layers as well as with regard
to the force defect hypothesis.

REFERENCES

[I J.W. SALVAGE, "Investigation of Secondary Flow Behaviour and End-Wall Boundary La-
yer Development Through Compressor Cascades", VKE, technical Note 107, 1974
(2 A. MAGER, J. MAHONEY, R.W. BUDINGER, "Discussion of Boundary Layer Characteristics
near the wall of an Axial Flow Compressor", NACA Report 1085, 1952
[3 L.H. SMITH, "Casing Boundary Layers in Multistage Axial Flow Compressors", Flow
Research on Blading, Brown Boveri Symposium, Elsevior, 1969

[4 J.H. HORLOCK, J.F. LOUIS, P.M.E. PARCEVAL, B. LAKSHMINARAYANA, "Wall Stall in Com-
pressor Cascades", Trans ASME, Ses D, 88, 637, 1966
5 A.G. HANSEN, H.Z. HERZIG, "Secondary Flows and Three-Dimensional Boundary Layer Ef-
fects", NASA SP36, Aerodynamic:; Design of Axial Flow Compressors, Chapt. XV, 1965
[6 1 J. DE RUYCK, Ch. HIRSCH, P. KOOL, "An Axial Compressor End-Wall Boundary Layer Cal-
culation Method", ASME Journal of Eng. for Power, Vol. 101, n02, 1979, 233-249
(7 1 Ch. HIRSCH, "End-Wall Boundary Layers in Axial Compressors", ASME paper i0 74-GT-72
(8 I J.W. RAILLY, J.H.G. HOWARD, "Velocity Profile Development in Axial Flow Compressors"
J. Mech. Eng. Sci. p. 166, Vol. 4, 1962
19 ] J.C. COOKE, M.G. HALL, "Boundary Layers in Three Dimensions", Ch 2 in Progress in
Aeronautical Sciences, Vol 2, Boundary Layer Problems, Pergamon, 1962
[10 ] B.S. STRATFORD, "The Use of Boundary Layer Techniques to Calculate the Blockage from
the Annulus Boundary Layers in Compressors", ASME paper n* 67-WA/GT-7
[11 1 J.H. HORLOCK, "Boundary Layer Problems in Axial Turbomachinas", Flow Research on
Blading, Brown Boveri Symposium, Elsevier 1969
[12 ] J.H. HORLOCK, D. HOADLEY, "Calculation of the Annulus Wall Boundary Layers in Axial
Flow Turbomachines", Aero Research Council n* 31, 955, 1970
(13 ] G.L.
MELLOR, G.M. WOOD, "An Axial Compressor End-Wall Boundary Layer Theory", Trans.
ASME, Series D, Vol 93, p 300, 1971
14 1 T.F. BALSA, G.L. MELLOR, "The Simulation of Axial Compressor Performance Using an
Annulus Wall Boundary Layer Theory", ASME J. Eng. Power 305-318, 1975
15 E.G.
K GRAHL, "Predi tion of Compressor Aerodynamic Performance for Multistage Compres-
sors", Dynalysis oa Princeton, Report n*29
116 ] H. MARSH, J.H. HORLOCK, "Wall Boundary Layers in Turbomachines", Joun. of Mech. Eng.
Vol. 14, 6, p 411, 1972
(17 ] J.H. HORLOCK, H.J. PERKINS, "Annulus Wall Boundary Layers in Turbomachinea", AGARD
AG 185, 1974
(18 W.L. LINDSAY, 'Tip Clearance Effects in the Growth of Annulus Wall Boundary Layers
in Turbomachines", Ph.D. Thesis, Cambridge Univ. 1974
(19 ] M. DANESHYAR, "Annulus Wall Boundary Layers in Turbomachines", Ph.D. Thesis Cambrid-
ge Univ. 1973

-J , .. _ _ . •.y.=•
W • - . . ",- I . • • •._NO •
 148

mainstream•
""rtilensverse proit
taper

F,,t

Figure 11.2.3.1 Secondary flow Nnd

boundary layer du.i to Figure 11.2.3.3 1 Velocity profiles
skewing )

0rPLECTION OUTLET ANGI.E

GIRCULATION

DISTANCE FRON WALL DISTANCE FRON WALL.

Figure 11.2.3.2 4

C4

i ¢1 -- -- boundary layer volOtily

outside boundary layer

Figure 11.2.3.4 z Skewing of relative profiles due

to relative motion

y
IFI

*rsgsv.ra. LW
5
opals

Figure 11.2.3.5 : Skewing of relative velocity Figure 11.2.3.6 s Wall boundary layer at
profile due to the relative outlet of rotor blade
motion between blades
 149

S1' U

eI T
//

II
(11118

e
onerglized ve tocity profites
MUDi100e'ltue T,-lm-)
t/
/I .... Ni"i~t
IIo
o-- .- .

Figure 11.2.3.7 ,igure 11.2.3.8 and 9 : Trensvere

Connection between absolute and 1
Figurmne btained froatvers
relative outlet velocity p-mles1o measurements obtained from a twelve
stag compressor, from 131

.04

C \

\ • 3
k-I
48)
2 eq(

Figure 11.2.3.10 4 ,•(36) Mallet

- F "- .
1Q.1 t tq

• , '• , _ ---

'nf -
axial coordinate (inch)

0 2-5 0 055o5 Figure 11.2.3.13 - Comparison of

40o5-& a7A
0-62-- .. different assumptions for the force
a XC 005o defects with experimental data from
... .
W"",,f e0r kh
I9 0 d l ot kdO Moore and Richardson, from [61

.0
" 3.22 St4 -V~ OI ItI CAt 1a -1- -
* 90

'U49

,I
_ _ _I __

.65 *.8

Iin
O .9...,
(.
, ,b. ,. . .i,5
O. _

h..
.04 .08 .12 .16

_iure _I ___. .11__ nd_12_ GE 4 stage compressor Fi g 11.2.3.14 : GE 4 stage c ompre ssor
eerenabonaylayer data, from [ 3] experimental an~d calculated data for
various aupect ratio and clearance [ 6]
S.
150 .9

I. h, 1,

.75 . 45 ,.3 .75 " , .

.7
, I cow ,0 ,i0 ,11,0 g

.405 .75
,0.8 2 AD.65 ,t1

.50 ,0.50 *O .50

F u5 .115 .,1 1
.55 650 .6 .7.5555 .0 .5 .7 S S 6

0claa0c,
ti fro [201
.0 .55 .60 ,65 .7 6 0 .0 .55 A.62i0 5 ,70 ,ti .55 .0 ,a5 .1 0 1

Geea Elcti Fou Stage Compresso

Figure 11.2.3.15 Low asGEct F9auro Moan asbect
FI.2.3.16 FicUro
1 1.2.3.17 : Two vmlues
ratio ratio of axial gap
General Electric Four Stage Compressor
Comparison between calculated and
measured characteristics, two valuedof

clearance, from (20] atip

.8 . 9612 .i. 2i i i i *

0. 4-

1. DE Hn . ,

• ,961
Bon- L -oJ--u

Mr0. .2 .4 Cx/Utlp .6 o. .2g.4fFz/(rU . er

PUZtlp V, 13 0w
--4 g• e 1s

Figure 1I.2.3,18 GE configuration 9 Figure 11.2.3.19 : TASK II copressor

axial velocity and tangential 0°/0 schedule
blade forces, solid line : calculated Compariso~n between calculated and
frcm (20 1 measured stator incidence and deviation
angles near design point, from (21 ]

20 3 J. DE ROYCK, Ch. HIRSCH, P. KeeL, "Investigation on Axial Compressor End-Wall

Boundary Layer Calculations", mnt. Joint Gas Turbine Congress, July 1979,
HAIFPA- ISRAEL
[ 21 1 J. OE RUYCK, Ch. HIRSCH, "Investigations on an Axial Compressor End-Wall Boundary
Layer Prediction Method", 25th Annual mnt. Gas Turbine Conference - N4ev Orleans,
March 1980, ASME Journal of Eng. for Power, Vol. 103, n°1, 1980, p. 20
 151

11.2.4. CORRELATION FOR SECONDARY FLOWS AND CLEARANCE EFFECTS

INTRODUCTION

As indicated by Hirsch (1) in his review paper made for WG 12, In the field of secondary flow effects, to
which belongs the influence of clearance, two approaches are possible when these effects are to be expLi-
citeLy treated .
- Additional Losses and modification of deviation can be calculated separately and added to the basic
profile effects, for inclusion in the procedure for generaLised ratfiaL equilibrium calculation, either as
an average correction on blade height, or as a function of radius., For the clearance case, attempts can
be made to introduce specific corrections, independent of the oneit due to secondary flows with zero
clearance.

- Calculation can be accomplished through a pseudo end wall boundary Layer approach, Leading either
to a global correction on efficiency, and/or to a detailed description of the flow near the end walls,
which is then matched to the "main fLow" in the meridionaL plane, as calculated by the generaLised radial
equilibrium equation, in which only the profile and shock Losses are introduced.

In the opinion of the author, this second approach is preferable as it is physically more sound, and Leads
to a coherent system for the prediction of bLocage, force defect and additional Losses due to secondary
flow and clearance, taking into account, aLbeit in a synthetic way, the effect of reLative motion between
successive blade rows.

However, to provide a basis as complete as possible, the WGhas decided to review some of the information
available in the open Literature on the explicit correlations for Losses and turning due to secondary
flows without clearance and to clearance effects, corresponding to the first approach. This is the sole
object of this paper, the boundary Layer approach being discussed elsewhere.

For the aspect of cascade secondary flow effects, without clearance the readers are referred to Salvage's
comprehensive paper pubLished byVKI(2) ConcLusions on their use will be presented at the end of this paper.

For the clearance aspect correlations for Losses and turning will be reviewed and compared for simple cases.
As most of the data has been obtained from cascade tests, it will then be attempted to ascertain the vali-
dity of the application of cascade date to axial compressors.
Finally, the possible ways of utilizing the date for compressor performance prediction will be briefly
discussed.
For both secondary flow and clearance effect, correlations specific to turbines were included, because
their formulation covers basic factors relevant also for compressors.

it is clear however that the experimental calibration which has been made for the turbine case cannot be
applied as such to compressors.

LIST OF REFERENCES FOR THE INTRODUCTION

1. HIRSCH, Ch. : "Axial compressor performance prediction survey of deviation and cross-correlation".
WG12 Working Paper, March 1978.

2. SALVAGE, J.W. : "A review of the current concept of cascades secondary flow effects".
VKI TN. 95, March 1974.

I. REVIEW OF EXISTING CORRELATIONS

A. Losses

There exists a Large number of correlations or of theoretical expressions for the clearance Loss
calculation. Most of them are derived from a straight cascade model (cascade with constant blade profile,
chord, stagger and pitch-chord ratio), some specifically for turbine application, others covering both
compressor and turbines. Practically all the formulae have been derived for incompressible flows, with
AVDR equal to unity and are meant to describe a pure clearance eftect, the secondary flow influence being
estimated separately and supposed independent of the clearance effect.

1. Parameters considered

Loadin% effects

The formulae can be classified as follows

- those which attribute the Losses to a drag induced by the tip vortex created either by a Lift reduction
in the clearance space or by the rolling upof the cLearance flow, considering only theeffect of clearance from
pressure to suction side and not from upstream to downstream. Theinduced drag is then proportionaL toCtm. Note
that inthe incompressibLeregime the Lift coefficient is Linkedto both the upstream-downstream pressure
difference (through the axial force) andto the pressure sidesuction sidepressure difference (through the tangen-
tiaL force) asshown bya reasoningsimiLar tothe oneused for establishing the diffusion factor formulae.
- those which attributethe Lossto thedissipation of the kineticenergy containedin theclearance flow.The
induced drag is then proportional to Cdm.
- those which consider the forces which create the clearance flow, i.e. the pressure difference between
suction and pressure side and/or upstream and downstream, and express a relation, between those pressure
differences and the clearance Loss.
 The Loading parameter used is that existing as mid-bLade height, either ideally (infinite height) or in
presence of the clearance effect.
f .Effect of blade height
It is usually accepted that the clearance flow ii independent of blade height, i.e. that the contribution
of the clearance Losses to the global Losses is inversely proportional to the blade height. This is
acceptable as Long as the end-wall effect does not modify extensively the character of the cascade flow.
Ref.(1) indicates that for the genmetr~es considered the Limit is reached for an aspect ratio of 1.3 for
compressors and I forturbine cascade. For a clearance of the order of 3 X chord Ref. (15) gives a Limit
of the order of 1 and considers the Loss as constant for Lower values.
Salvage (8) has shown, for the case of secondary flows (without clearance), that the Limit aspect ratio
is a function of the non-dimensional upstream boundary Layer displacement thickness based on blade height,
of the Loading, characterized by the turning, and of the aspect ratio, in presence of clearance, this
Limit should also be a function of the reLative clearance based on blade height.
The effect of the blade height is sometimes expressed in function of aspect ratio, which Leads to a non-
dimensional expression of the clearance based on chord.
Effect of clearance heig~ht
Physically,for constant blade chord and thickness,the crons soction offered to the clearance flow is
proportional to the effective cLearance height, taking into account,if needed, the presence of Labyrinth
seals, the chord being the other characteristic Length if the pressure-side/suction side pressure
difference is predominant. Tht blade thickness is the other characteristic dimension if the uptstream- J9
downstream pressure difference is the major driving agent.
It appears (1), (5), (6), (9), that for unshrouded compressor blades, there is an optimal clearance, of
the order of 1 to 5 % of the chord. This is explained physically by the fact that secondary and clearance
vortices are of opposite direction of rotation. For clearances higher that the optimum, a quasi Linear
increase of Losses with clearance is first notod, then, a progressive slope reduction, probably due to
the effective relative height effect. This evolution is represented in Fig. I (VKI measurement on a highLy
Loaded blade).
None of the formulae available in the Literature considers the existence of an optimum clearance. ALmost
all
When ofusing
themaspect
consider thatbased
the losses varyonelinearly
ratio on chord, obtains with
also aanon-dimensional cearanrrifbased
Linear Law in function on blade
clearance basedheight.
on
chord. The appearance of chord as a reference parameter for clearance appears to be justified only if
the upstream-downstream pressure difference is the dominating factor in the flow mecharism.
2. Selected formulae available in the Literature
GroupI - Induced drag approach
= L (1) Mohdaht M) (10)
(turbines)

CD2 = Ca 0 . (2) Vavra I (correction MehdahL) V

M Lm F cos02 (turbines) (11) 1

C 2 cos1Ba
Y 0.5 (in) (3) Ainley
cos 3Bm (turbines) (12)

2
c 0,78 C~*m 2 COS 02
Y 0.7 R (j/c) (---) (4) Dunham-Came
SCos3 D.C. (13)

(correction of (4) on the basis of part of the results of (1) for turbines).

Group II - Kinetic energy approach

SC =0.29 /h (5) Vavra 11V (11)

K : contraction factor 0.5 (valid for both turbine and compressor cascades)
X : characteristic factor of clearance resistance (= 0.8).
 153

"Mixed" formulae Lakshmlnarayana - L (5)

cos B2
W, M 0.7=O3
C0.Lm /h Ci -- + 7 QCL)3/2
L~m TTh
c (j/t)3/2 CO/t
(6)
3 3
cos 2B cos

W, = 0.7 C2 jlh cit 1 + 10 / t W)(6

The first term of the R.H.S. of (6) corresponds to an induced drag, the second to the kinetic energy Lost
by the spanwise motion of the blade boundary Layer (valid for turbine and compressors).

ALL the preceding formulae indicate a continuous increase of the Losses with clearance.height.

Group III - Formulae based on pressure difference

AW
s = Y = 0.0696 tanh (13 j/c) c W Cos (7)
Wa a h W
a osm

(formulae derived from the turbine cascade experimental data of Ref.(1)).

The pr e dB1 identifies wa with Y , which comes to multiply

2
the losses given in Ref.(1) by 1/cos B2 .

CA Wu
w = f(j/c) c tan (Bauermeister(1))
a m

(formulae derived for compressors).

The pressure difference is expressed as

u tan 1 P2 - PI
m P V

VA a
The function f(j/c) is not explicitely defined. A particular shape is given in Fig. 1

high Loss for zero clearance with a rapid decrease to a minimum corresponding to optimum clearance, then
linear increase folLoweo by a flatter part.

The available data do not allow a statistical evaluation. The optimal clearance, which, as said before,
is of the order of 1 to 5 X chord in cascade seems to be much lower in actual machines.

Cross checking on the compressor cascade data in (1) and in view of the precision of measurement, and of
the domain of practical interest (0 i/c •5-6%)one can propose the following Linear relation to be
V
introduced in (8)

f(j/c) = 0.44 j/c

disregarding, like in all the other formulae, the existence of the optimum clearance, and the possible
non linearity with clearance height.
An additional correlation, based on the concept of pseudo-end wall boundary Layer derived from Low speed
multistage compressor data is presented in (15), allowing estimation of efficiency reduction due to
secondary and clearance losses
S1 ( h* + a) / h
h (9)
1-1-(vh + vt)/

with • the efficiency without end wall effects, a* Mt)and v. the displacement and tangential force
1 I
defect of the pseudo-end wall boundary Layer at hub and tip. It is based partly on data, partly on
general designing experience.

A correlation is given for the sums of a " as function of a stage pressure rise coefficient, non
dimensionalized by its maximum value, and mid-span axial gap (Fig. 2, 3, 4). Fig. 2 gives expLiciteLy
the normalized clearance effect (clearance normalized by mean diameter staggered tangential spacing).

This correlation provides an interesting global evaluation for clearance effects.

( 1 *subscrit i - h or t
MI
 154

3. Comparison between the various formulae

The correlation formulae, except the last one, have been compared for two types of compressor cascade,
corresponding respectively to the tip condition of a first LP compressor rotor, and root condition for
a rotor or a stator of an MP or HP compressor.
Cascade : =65 to 72*5 ; 02 = 55 ; t/c = 1 ; h/c = 3
=45 to 5205 ; 2 = O° ; t/c a 0.5 ; h/c = 1.5
Comparison of the Losses for constant clearance (i/h = 0.01) is given in Fig. 5 and 6. The Lack of
coeherence of the data is evident.

Differences ofan order of magnitude appear. Lakshminarayana and Dunham-Came formulae give the highest
values. The empirical constants have been evaluated from machine data, essentially turbine for the latter and
compressor for the former. Bauermeister's formulae gives average values. It is based on cascade test.
Fig. 7 and 8 give the clearance effect at constant Load ($1 = 65* or 450), for Lakshminarayena,Came and
Bauermeister. The slope for the first one is very high. elt
Conclusions on the respective value of the formulae could only be obtained by systematic evaluation
with machine data. The dispersion of data can be attributed to the fact that the formulae do not really
'I, hdcover the real physical mechanism. The evaluation of the best formulae of this type of approach seems
hardly warranted.
B. Angles
It is shown also that the clearance effect modifies the local and average flow outlet angle. Ref. (5)
proposes an approximate method for the clearance induced angle correction, as function of blade height.
Clearance effects appear to be more important than secondary effects. The pitch average correction is
given by

2 = tan [.25 CL o (I- ) a + j yj

A 2 =0 y>a +j

with a the radius of the clearance vortex supposed to be of the forced type ; j the centre of the vortex ;
a and y must ve calculated. The procedure is rather long but it can provide a spanwise evolution of the
angle. More recently, in (17), in the scope of the development of pseudo end-wall boundary Layer method,
under the impetus of Prof. K. Papailiou, on the basis of Lakshminarayana's approach on the one hand, and
of Yokohama's on the other, a method was developed for describing the pitch average angle correction from
inlet to outlet of the blading.

Both these approaches could eventually be used to develop further simple correlations.
in Ref. (1) a pitch and spanwiseaveraged correction is given for turbine as
(2 W W cos2
(A ad)° 2 K j/h t/c -u- cos 8m
a Cos8
with K = 1.50 / % j/c
Fig. 9 and 10 show the evolution for the two cascades used in the Loss comparison. The correction for the
second one is small. The validity of a turbine correlation to compressor can be set in doubt.
II. VALIDITY OF THE CASCADE T'3iz
Cascade flows can differ from machine flows on the following accounts
- Reynolds number and turbulence Level
- often, Mach number
- radial pressure gradient
- relative motion blade-end wall
- inlet boundary layer including effect of change of coordinate system from one blade row
to the next.
I A relatively systematic study of the first
in Braunschweig a number of years ago, and
four effects was carried out by Prof. SchlichLing's group
is summarized mainly in Ref. 1 and 2. Tests have been
carried out on large aspect ratio cascade, thus avoiding direct contamination of the mid-span flow by
the end-wall effects. The end wall blocage was thus small.

1. Reynolds number effect (1)

The cascade Losses due to the clearance effect (as obtained by difference between secondary losses
with and without clearance) and the average angle correction are independent of Reynolds number
(1.5 x 101< Re S 8.5 10') even when the flow on the central part of the bLading is affected by a
laminar separation bubble. The Losses and turning in the real machine, where the level of turbulence
is usually much higher than that of the cascade tunnel should present all the more the same characte-
ristic.
.-
 155

Note, however, that Sierverding and Marchat, VKI, (private communications) have noted differences in the
extent of the Laminar separation bubble on the central part of Low aspect ratio turbine cascade blade,
with the Reynolds number. This would imply a variation of the extent of the zone of influence of
secondary flows with Re.
The Reynolds number effect is probably negligible for compressors. We will use this assumption, also
accepted in (15).
2. Mach number effects (2)

Tests have been carried on turbine cascades in the high subsonic domaine (0.5 S M S 0.8). In this range,
no effect of Mach number or clearance loss is apparent, if the Load variation due to compressibility is
taken into account, for a given cascade.

Griepentrog (private communication) reaches the same conclusion for the secondary flow effects in highly
cambered compressor cascades, operating with Local supersonic pockets.

In a recent paper (16) Sieverding and Wilputte showed that there was some effect of Mach number
for the turbine case, but which were not much outside of measurement accuracy if the inlet Losses were
taken into account.

At this stage we will consider that the Mach effect is negligible, if the blade Loading is correctly
taken into account, its change in function of M being due to density change, back-pressure in choked
regime or incidence in the unchoked one. Mach number effect is also neglected in (15).

3. Effect of radial pressure gradient (1)

Test on fixed annular cascades of turbines (high and Low camber) and compressor blades (NACA 8410 -
Stagger 30* - inlet angle 450, 206 turning) with no twist, but with variable pitch and variable hub-tip
ratio from 0.5 to 0.8, for one of the turbine cases, show the same global Losses than for the equivalent
straight cascade tests. The Loss and angle evolution in function of blade height can even be correctly
calculated from the straight cascade data, if the pitch variation and the component of radial velocity
due to the radial equilibrium are taken into account. The position of maximum Loss only is slightly
different.

For this type of conditions where the pressure difference between hub and casing is of the same order than
the pressure difference between blade inlet and outlet, the mechanism of the clearance effect does not
seem to be fundamentally altered.

Even for radial pressure gradients closer to the circumferential ones, the global effect in compressors
seems small, the radial equilibrium effect being opposite on the clearance flow on the pressure and suction
side (hindering or fostering). Secondary flows are more affected, and the same is true for centrifugation
effects.

Further information based on detailed tests in machine or systematic application of end-wall boundary Layer
calculations (2 and 3 D) would be needed to ascertain exactly the effect of radial pressure gradients and of
centrifugation. At this stage, we will accept that it does not seem to affect the clearance flow mechanism
fundamentally and that it could eventually be taken into account by an increase of the thickness of the
end-wall pseudo boundary Layer.

4. BLade - end wall relative motion

This effect has been studied in an approximated way in (1) either by rotating the part of the hub immediately
under the blade in a fixed annular cascade, or the whole hub, from the nose bullet to the cascade exit, for
compressor and turbine cascades. Clearance has been varied from 0 to 10 X and non dimensional hub rotational
velocity varied from (0. Up/Va5 2). The corresponding value of the velocity of rotation can be as high as
3 in actual machines. The scraping vortex can be expected to increase the clearance effect for decelerating
cascade. However, if, as indicated by the visualisation of (3), taken in a straighý cascade with a mobile
end wall, the Local phenomena are modified, the results of (1) show that for the compressor cascade the
effect on the globaL Loss and the average outlet angle are negligible (Fig.1). Figuie 2 is the corresponding
figure for turbine annular cascades. Although ( 1) does not give details on the clearance flow mechanism,
it can be concluded that a system of calculation of global clearance losses and mean angle correction based
on cascade data without rotation effect will yield results not too far from the truth when applied to
compressor blading.
multistage axial suggests thatof the
pumps modification effect of
Rains (4) onwhothestudied
rotation clearance flow isthelimited
in detail Localin Lift
to a flow
clearance increase, without the fundamental
flow mechanism.

Lakshminarayana (5) comparing his calulation results with the compressor data of (6) concludes that the
effect of the scraping vortex is to reduce the optimal clearance from 2 to 5 X chord to - 1,5 X.

Grieb (7) proposes, on the basis of experimental compressor data, to weight differently the clearance
effect of the rotor and stator, giving the Latter half the contribution of the former in the efficiency
reduction due to clearance. This difference could be attributed to the reLative motion effect and/or to
centrifugation, but might also be due to the energy exchange defect. The tests of (1) do not cover the
effect of change of reference system from one blade row to the next. This has been shown to modify
extensively the boundary Layer state at blade inlet and outlet, and could tend to increase the clearance
effect for the compressor case. Data is missing and only the pseudo-boundary Layer approach could provide
the needed information.

fac
 I
156

5. Inlet boundary Layer effect

Apart from the skewing effect due to change of coordinate system referred to above, a wealth of information
exists concerning the effect of boundary Layer initial thickness on secondary flow. However, practically
nothing is available with regard to the influence on clearance effects, except for the h.ib rotation effect
in (1), as discussed above. Rains (4) estimates indirectly that the initial boundary Layei effect is
negligible. Lakshminarayana (5) attributed part of the clearance Losses to the kinetic enervy lost through
the radial motion in the pressure and suction side boundary Layer, in the vicinity of the clearance. The
blade boundary Layer thickness is certainly modified by the end wall boundary Layer effect. Ref. (5)
indicates that this part of the Losses is proportional to boundary Layer thickness based on the pitch.

The empirical values of the coefficients which have to be introduced in (5) to fit experimental results
would Lead to boundary Layer sevent times as thick as the pitch, indicating that the physical description
given is probably not sound.

The main boundary Layer effect is most probably that of the change of reference system.

For a global approaximation of clearance influence, this effect could be neglected, assuming that it
bears mainly on the secondary flow effects (i.e. as existing without clearance). This seems coherent with
the compressor derived correlation of (15).

6. Conclusion

It can be concluded, guardedly, that correlations obtained from straight cascade data, will at Least
provide correct trends for the effect of clearance in compressors. As usual, with cascade data, the effect
of variation of blade circulation with span is not known. This will cause problems if spanwise distributed
correction for Losses and angle are to be considered.

For turbine it is well known that for secondary flow effect the Loss Level as obtained from cascade data
may differ of an order of magnitude from the machine Losses. This is Linked most probably to the effect of
relative motion when passing from nozzle to rotor or vice-versa. In compressors this effect trends to bring
back the end-waLL boundary layer to a more collateral state, and the difference migbt be somewhat Less.

III. POSSIBLE USE OF THE CORRELATIONS : Clearance effects.

Supposing that one of the formulae could be validated, the fact that they are derived for constant geometry
constant circulation bLading introduces further problems for their application

- if a global averaged correction is deemed sufficient, it could be applied to the mean diameter section
(or to the section nearest to the clearance). For the angle correction, data corresponding to blade
height averaging should be obtained, using the method presented in (5) and (17).

- if a Local correction is needed, a distribution Law must be found or assumed. An exercise similar to that
carried out by the turbine group, on the effect of the Loss distribution, could be carrieu, but this is
worthwhile only if the psejdo and wall boundary Layer approach is deemed unsatisfactory.

- as practically all formulae have to be more widely calibrated against machine data, it seems that a
formulation of the type of (5) which is partly calibrated or, for the sake of simplicity and some analogy
with the D factor approach, Bauermeister's formulae could be used for Losses.

We cannot recommend the continuation of the use of the correlation type of approach on its present physical
basis. The approach can only provide - at most - an indication of order of magnitude for a first rough
evaluation for instance in a 1-D calculation.

The pseudo end-wall boundary Layer should be further developed including the evaluation of its development
inside the blade passage. The concept is very sound physically and its application very easy, requiring
Little additional computer time when coupled with a through flow calculation method.

LIST OF REFERENCES

(1) Hubert G. andBauermeister K.J. : Probleme der Sekundlrstr6mung in axiaLen turbomachinec

Tail I. Untersuchungen Ober die SekundbrverLuste in axiaLen Turbomaschinen- Teil II.Uber den ElnfLuss
der Schaufelbelastung auf Verdichtergittern.VDI Forschungsheft 496 -Aufgab B - Band 19.
(2) HebbeL H.H. : Untersuchungen Ober den EinfLuss der Nachzal auf die SekundirverLuste eanes geraden
Turbinen SchaufeLgitters bei hohen UnterschaLLgeschwin - Z.F.W. Heft 9, Sept.1965, pp.324-333.
(3) Hansen A.G. and Herzig Z. : Secondary flows and three-dimensional boundary Layer effects.
Chap. XV of Aerodynamic Design of Axial FLow Compressors-NASA SP36, 1965.
(4) Rains A. : Tip clearance flows in axial flow compressors and pumps.
Report no.5 Hydrodynamics and Mechanical Engineering Laboratories. CaLif.Institute of Tech., June 1954.
(5) Lakshminarayana B. : Methods of predicting the tip clearance effects in axial flow turbomachinery.
ASME Journal of Basic Engineering, Sept. 1970, pp. 467-481.

(6) Horlock J.H. : Some recent research in turbomachinery.

Proc. Inst. of Mech. Eng. (London), Vol. 182, Part I. no.26, p.571.
 157

(7) Grieb H., Schaffer A., Laudenberg H. : EntwickLungsprobLeme an Axialverdichtern fOr KLeingasturbinen.
NTU-12-11, 1974.

(8) Salvage J.W. : Investigation of secondary flow behaviour and end wall boundary Layer development through
compressor cascades.
VKI, TN-107, june 1974.

(9) Raw H.G. : Turbomachinery tip clearance design considerations. VKI PR 72-300, 1972.

(10) NehdaL A. : The end Losses of turbine blades. Brown Boveri Review, VoL. 18, no.11, pp.356-361,
november 1941.
(11) Vavra N.H. : Aerothermodynamics and flows in turbomachinery.
Wiley and Sons.
(12) AinLey D.G. and Nathieson G.C. : A method of performance estimation for axial flow turbines.
British ARC and N. 2974, 1951.
(13) Dunham J. and Came P.M. : Improvements to the AinLey-Nathieson method of turbine performance prediction.
ASNE Paper GT-70-GT2.

(14) Balje E., BinsLey R.L. : Axial turbine performance evaluation. Part A. Loss geometry relationship.
ASME Journal of Engineering for Power, October 1968, pp. 341-347.

(15) Koch C.C., SMITH L.H. "Loss sources and magnitudes in axial-flow compressors"
ASNE Paper 75-WA- GT-6.

(16) Sieverding, C.H. and WiLputte, Ph. : Influence of Mach number and end wall cooling on secondary flows
in a straight nozzle cascade . V.K.I.

(17) Ohayon G. : Contribution h L'6tude des dcouLements secondaires dans Les compresseurs axiaux avec effets
de jeu radial. Th&se de 3Ame Cycle Universit6 CLaude Bernard, Laboratoire de MRcanique des FLuides,
EcoLe CentraLe de Lyon, July 1979.
(18) Stewart W.L., Whitney W.J., Wong R.Y. : A study of boundary Layer characteristics of turbomachine blade
rows and their relation to overall blade loss.
ASME Transact. Series D J. Basic Engineering , VoL. 82, 1960, pp. 588-592.

LIST OF SYMBOLS
A Area
AR Aspect ratio h/c
C Chord
C,
4-
Lift coefficient
CD Drag coefficient
h Blade height
j CLearance height
K, k Constants or proportionality factor
M Mach number
m Mass flow
N Rpm
P Pressure
Q VoLume flow
r Degree of reaction
R Radius
Re Reynolds number
t Pitch
T Absolute temperature
U Peripheric velocity
V Absolute velocity
W Relative velocity
Y AinLey Loss coefficient Y = PO1R- PO2R/ PO1R p P 2
a Absolute angle (from axial direction)
0 Relative angle (from axial direction)
Y Specific heat ratio or stagger angle (from axial direction)
n PoLytropic efficiency
STraupeL's Loss coefficient • = V~js - V / V~,5
SFlow coefficient

Io
0

p
Pressure coefficient AWu/U
Solidity c/t
Mass density
Wl Loss coefficient w, = PolR - Po2R/ PoIR - Pi
WA Loss coefficient WA = PolR - Po2R/ 1/2pVa 2
 158

Subscripts

A Axial

Upstream of blade row (rotor)

2 Downstream of rotor blade row (upstream of stator for a stage)

3
Dowstream of stator for a stage

m Average (based on mean velocity vector or at mean radius)

MS Mid Span

o Stagnation conditions

R ReLative

S Secondary

id Ideal

Averaged

LIST OF FIGURES

1. Losses in function of clearance for a compressor cascade

BLade 10C7-45-C50

2. Correlation of reference (15) for displacement thickness of end-wall boundary Layers

3. Correlation of reference (15):effect of axial gap between blade row and end-wall boundary Layer
displacement thickness
4. Correlation of reference (15):correlation of tangential force thickness

5. Comparison of the different correlations formulae for a rotor tip section. Effect of clearance in
function of Loading

6. Comparison of the different correlations formulae for a rotor or stator hub section

7. Effect of clearance in function of clearance height. Same cascade as in figure 8

8. Effect of clearance in function of clearance height. Same cascade as in figure 9

9. Angle correction due to clearance for the cascade of figure 8

10. Angle correction due to clearance for the cascade of figure 9

11. Effect of relative motion clearance end-wall (annular compressor ctscade)

12. Effect of relative motion clearance end-wall (annular turbine cascade).

List of Figures

Fig. la: Mixed flow field for supercritical inlet condition

Fig. lb: Shock system for supersonic inlet condition
Fig. 2: Shock losses as function of inlet Mach number and
supersonic expansion

Fig. 3: Results of cascade measurements as a function of

inlet Mach number
Fig. 4: Loss model including shocks and shock/boundary
layer interaction
Fig. 5: Comparison of measured and calculated losses for
different loss models
,lnlý
Ak .. ,
 19

10-c7-45C-50

12

10-

FIG. I Losses in function of clearance for a compressor cascade. Blade 10,C7-45-C50

Courtesy VKI.

'v&mOAS'

414" hubln
a IIMWI a ll i s|b
4LOO~ •R) 1.03l

FIG. 2 - Correlation of reference (15) for displacement thickness of end-eaLL boundary Layers.
160 j

Yt, ii w -t 1 14 13 I
Saa
a Qemaega,
. All & "r

ftg 6 E~fct go oxii gap botw..antad. mw .g'oft aneiwaS ba"Mya

FIG. 3 - CorreLation of reference (15) effect of axial gap between blade row and end-wall boundary-
Layer displacement thickness.

L I I
__. I

U0 11,-

121-.- As
°,r,.
P- ! Li

FIG. 4 - Correlation of reference (15) a: coreaio o aneniL c e thickness.

5 Coparsonof the different correlations formulae for a rotor tip section. Effect of clearance
FIG
of Loading. CASCADE I

A Laksh. 0,., .
Dun.-Came ,

A Med.

O Bauer.
V A.M. CASCADE II

o Vavra 1
+ Vavra 2 0.0

x BaLje 0-03 45 7-5 50 '52.5

0-02- - V A.M CASCDE 1

001

Bau2e5 P1
FIG. 6 -Comparison of the different correlations formulae for a rotor or stator hub section.
 161

8
(01 X BAUERMEISTER
%10 o LAKSM
6 _0 DUP4-AM -CAME
CASCADE I

0 0.01 0.02 0,03 0.04 0,05 J/h

0.03 0.06 0.09 J/c

FIG. 7 - Effect of clearance in function of clearance height. Same cascade as in figure 5.

24
O/o

20- .

V ~16 -

12

41.

0 0.01 0-02 0-03 0.04 0,05 0.06 jlh

0-015 0-03 0-045 0-06 0.075 0.09 i/c
FIG. 8 - Effect of cLearance in function of cLearance height. Same cascade as in figure 6.

CASCADE 1 4 __________

010

65 67.5 70 72-5
FIG. 9 -AngLe correction due to cLearance for the cascade of figure 7.

2Aj
162

Af32
CASCADE II 3t.0,

1.0 - -I _...--

0
45 47.5 50 5;65

FIG. 10 - AngLe correction due to cLearance for the cascade of figure 8.

0.10
WA

0.08

P2 26 0.0{6 _ _ !

00-.--
22 0.02-

g S-
0,02----- ___ -___

S14 0
0 0'18 08 1,2 1.6 Up 2

FIG. 11 - Effect of relative motion cLearance end-waLL (annuLar compressor cascade)

(from Ref.1).
 163

0.24 ..
WA -0. - - m-

0.20 ,,,

0.15

54- 0.12 , * .. 4

P2

SB- 0.089

62 - 0-04 --- ____ )~1 ____

o1o2 ..... 08 Up2

F G. 1tt

FIG. 12 - Effect of retative motion clearance end-walL (annuLar turbine cascade).

164
FI
11.2.5 EFFECTS OF REYNOLDS NUMBER AND
TURBULENCE LEVEL ON AXIAL CASCADE PERFORMANCE

A semiemp irical theory is developed which will predict the

behavLot of the shear layer across a laminar separation bubble.
The method is proposed for two-dimensional incompressible flow
and is applicable down to short bubble bursting. The method can
be used to predict the length of the laminar bubble, the burst-
ing Reynolds number, and the development of the shear layer
through the separated region. As such, it is a practical method
for calculating the profile losses of axial compressor and tur-
bine cascades in the presence of laminar separation bubbles. It
can also be used to predict the abrupt leading edge stall asso-
ciated with thin airfoil sections. The predictions made by the
method are compared with the available experimental data. The
agreement could be considered good.

NOMENCLATURE
AVR - axial velocity ratio S - distance along a surface
C - chord length s - the same as S
ChtT - inlet turbulence intensity,
Cd - dissipation coefficient,
2 -T(9uN dz 7
p e E T.F. or TF - Taylor's turbulence factor,
Cd(i) - mean dissipation coefficient Te. i/
Cf - friction coefficient, U = free stream velocity in
Tw//(1/2)pUes the direction of the main
CC - correlation coefficient, flow, U U + u"
u'I.-,2 U - time averaged component
V . /t Ue - local free stream velocity
D overall diffusion factor, u - velocity in the boundary
(1_U•+ AU layer or free shear layer
4.9 =01j parallel to Ue
9 pitch or blade spacing u. - randomly fluctuating com-
H .6/8* ponent of velocity
- /e* x - distance along an x axis
or chord line
i incidence
Ki a constant, 0.016 p probe separation V
y - distance normal to x
K& a constant, 0.0011 z - distance normal to s
L total length of a laminar a - angle of attack
separated bubble 1- cascade inlet flow angle
L 8x - di•t
macroscale of theCC inlet -cascade outlet flow angle
turbulence, Vp CC d(Xp) - stagger angle
11 - distance between separation 5* - displacement thickness
and transition in a laminar
separation bubble 8 deviation angle
it- distance between transition E - energy thickness
and reattaclnent in a - deflection angle oi-01
laminar separation bubble
Pe - total or stagnation e* M momentum thickness
pressure v - kinematic viscosity
q - dynamic head, I/2pUt AR W reattachment criterion
Rc - Reynolds number based on
chord p - density
"",a) - momentum thickness Reynolds ao solidity, l/(
(8*number at separation (c
R-(1) W Reynolds number based on 11 T - turbulent shear stress
Z - lose coefficient, Pnm-Pb(e) - blade camber angle
ql
 165

NOMENCLATURE (Continued)

Subscripts and Abbreviations

- bursting
M-M - maximum loss, minimum turning
m - mean value
R a reattachment
S n separation
SB - sub-bursting
T M transitior.
1 M upstream
2 0 downstream
6 circumferential
VKI von Karman Institute

I. Discussion of the Flow Phenomenon

This discussion was developed from a thorough review of the pertinent litera-
ture. It is later confirmed by the author's experimental results.
If, for the time being, we postpone consideration of the factors that define
whether a bubble is short or long, the conditions necessary for the formation of a
laminar separation bubble are:
1. an adverse pressure gradient of sufficient magnitude to cause laminar sep-
aration, and
2. flow conditions over the blade surface such that the boundary layer will
be laminar at the separation point.
Inherent in the second condition is that the blade surface be smooth, that the
free stream turbulence level shall be relatively low, and that the distance between
the stagnation and separation points be moderate (or more precisely, that the boun-
dary layer Reynolds number at the laminar separation point, Re , be less than that
required for transition). s
The essential features of a laminar separation bubble are illustrated in Fig. 1.
The laminar boundary layer, starting at a stagnation point upstream of the
separation point S, separates from the surface at S to reattach downstream at the
point R. Between the points S and R, the flow may be !i--ided into two main regions:
1. THE FREE SHEAR LAYER, contained between the outer odge S"T"R" of the viscous
region and the mean dividing streamline ST'R.
2. THE RE-CIRCULATION BUBBLE contained between the mean dividing streamline
and the blade surface STR.
These two regions may then be further subdivided into parts upstream and down-
stream of the transition point T. Upstream of T, the free shear layer is laminar
and is incapable of doing any significant diffusion, because weak viscous shear
stresses operate in this region. As shown in Fig. 1, the surface velocity is prac-
tically constant between separation and transition. This constant pressure "plateau"
is a general feature of the laminar part of the separated flow.

So far, we have been discussing laminar separation bubbles in general, without

making any distinction between short and long bubbles. Perhaps the most basic way
of making this distinction was proposed by Tani (1)*. He has suggested that the
difference between a short and long bubble lies in their effect u pon the overall
velocity and pressure distribution. The short separation bubble has only a slight
effect upon the pressure distribution. Outside of the short bubble region, the
pressure distribution is a close approximation to the inviscid distribution about
the profile, apart from a slight reduction in the magnitude of the suction peak
ahead of separation. On the other hand, a long bubble is one which interacts with
the exterior flow to such an extent that the pressure distribution is appreciably
modified from the inviscid model in a way that the velocity peak and circulation
are decreased. This difference can be seen in Fig. 4 where Caster's (2) Scries II
pressure distributions have been plotted.

W-u-ers in parentheses denote referece at the end of the paper.

. ...................................-...--
•~
166

We can characterize laminar separation bubble behavior in the following way

(for constant turbulence level):

When the overall blade chord Reynolds number gets low enough, the laminar boun-
dary layer reaches the separation point before transition is achieved. .fter the
laminar boundary layer separates, it forms a laminar free shear layer that eventually
undergoes transition to turbulence. The turbulent free shear layer is able to do
enough diffusion by entrainment of high energy free stream fluid to reattach to the
surface as a turbulent boundary layer. This short bubble is seen as a small pertur-
bation on the pressure distribution; its effect on the flow outside of the bubble
region is minimal.

As Rc continues to decrease, the laminar free shear layer grows in length (the
mechanism of this growth will be discussed in more detail in subsequent chapters).
This growth causes the turbulent free shear layer to do more diffusion to reattach
at a pressure near the inviscid pressure value (see Fig. 1). Finally, Rc becomes
so low and the laminar shear layer so long that the turbulent entrainment process
can no longer support the diffusion required for reattachment with a value close to
the inviscid pressure level. This is when the bubble starts to significantly affect
the flow outside the bubble region. The velocit eak and circulation decrease,
thereby reducing the pressure gradient over the ubble. This allows the turbulent
shear flow to reattach as a long bubble -- that is the short bubble has burst into a
long bubble.

As Rc is further lowered, the velocity peak and circulation is further decreased.

Finally, the bubble is so long that reattachment on the blade surface is no longer
9
possible. The flow is then completely separated, and there is little change to the
flow field around the profile with continued decrease of Reynolds number.
To summarize, within the incidence operating range of the cascade, there are
four flow regimes possible across a large Reynolds number range (disregarding turbu.-
lent separation due to off-design conditions):
1. Rc sufficiently high for transition to occur before separation;
2. short bubble region (before bursting);
3. long bubble region (after bursting);
4. complete separation.
These four flow regimes are shown schematically in Fig. 2.

II. A Modified Semiempiricql Theory For the Growth and Bursting of Laminar Separation
Bubbles

Equations and Discussion. The method proposed here is a mod.fication of Horton's (3)
semiempirical theory. Using Horton's theory, calculations of cascade performance were
compared with selected results of this author'* trsearch and that taken from the
literature. It was found that, in general, Horton's method gave a delayed prediction
of bursting and underestimated the losses. This was probably due to three factors:
1. Inadequate transition criterion;
2. possible error in the value of the reattachment criterion, AR; and
3. the low value of the mean dissipation coefficient, Cd(m).
This author has done a thorough analysis (4) to correct these faults. The re-
sulting modified theory is presented below:
1. THE LAMINAR FART .
(a) Transition Length

1-1 ,- . 2.5 X 104 Lot, (cothUTF) X 10))

where - c

(If TF is not available, then T(7') could be used in equation (1)

as a first approximation.)

(b) Growth
do*
eS T(0,

*S
• T* (2)
 167

2. THE TURBULENT PART OF THE BUBBLE

(a) Reattachment Criterion

8*" dU e- - "d ,
(R)
(A e rR
a)C

(A)theory - -0.0059. (3)

(b) Length Relations

-
'UR)-CI ,1 -
B(l2eR) 1 + TI, (4)
BB 1 -

-AR
B,
=

em

(c) Growth
Cdm R7(1-Ue)
-+ (5)
3
(UeR)' Cm (UeR) (l-UtT

T2 - l/e*,

Cd 0.035, (6)
dm

H =1.5.
Cm

A bar indicates normalization with respect to values at separation.

The equations in the foregoing follow from three approximations that are rea-
sonable for a short bubble:
1. the perturbation to the inviscid velocity distribution is negligible,
except over the length of the bubble itself.
2. the external velocity over the laminar part of the bubble is constant to
a good approximation.
3. the external velocity falls linearly between the transition and reattach-
ment points.
The modified method proposed in the foregoing has been successfully used to
predict the bursting Reynolds number and bubble lengths for isolated airfoils, as
well as for airfoils in cascade. Along with a reliable boundary layer method, it
was also used to predict the losses of two-dimensional axial compressor and turbine
cascades where separation bubbles were present (see Figures 3 and 4).
Note that Equation (1) denotes the effect of the turbulence on bubble behavior.
For
low the same Reynolds
turbulence levels. number, the flow regime could be quite different for high and
168 ,CE

_REFERENCE

1. Tani, I., "Low Speed Flows Involving Bubble Separations." Progress in Aernautical
Science, Pergamon Press, 1964.
2. Gaster, M..* "The Structure and Behavior of Laminar Separation Bubbles," NPL Aero
Report, 1181, ARC 28.226, 1966.
3. Horton, H. P., "A Semri-Emnpirical Theory for the Growch and Bursting of Laminar
Separation Bubbles." ARC, CP 1073, 1969.
4. Roberts, W. B., "The Effect of Reynolds Number and Laminar Separation on Axial
Cascade Performance," ASME Transactions, Journal of Engineering for Power, Vol. 97,
Series A, No. 2, April 1975.
5. LeFoll, J., "A Theory of the Representation of Boundary Layers on a Plane, Proc.
Seminar on Advanced Problems in Turbomachinery, von Kerman Institute, Oct. 1V53.

39mMTI TUI.DUUM

r ..
FIGREla

RR

WMOW=

FIGURE Ib.
The
Corbl (pndn
SuraeGreatsyre Disrtribtio

R. IN
 169

$- SERMAT'IO.
T- T1USNITION.
A- RZ0TTAmuTl.

1.'

1.2 TIUUTIhILL INVISGCD FRISSUNK DISTRIBUTION.

1.0

(2).
S "284 (4).•
i = 232 (.5).

0.6 /s R

0.4
R AT
I - 16 La.

X DISTANCE FROM LEADING EDGE, inches.

FIGURE ic.
Pressure Distributions in the Vicinity of Separation
Bubbles for the Series II Experiments of Gaster (2),
Bubbles 1 to 4 are Short, 5 is a Long Bubble.

FIGURE 2.
I Schematic of the Four
Flow Regimes Possible with
Varying Reynolds Number
.... . ........... 0(i0-iO) of Over
Surface the Suction
a Compressor
/ / "-'--•" ""Profile.

(Low Speed Flow, a<a Stall)

1. HIGH Re: No separation,

velocity distribution approx-
imates inviscid distribution.
2. MEDIUM Rc: Short separation
bubble, distribution approxi-
mately inviscid outside
bubble region.
3. LOW Rc: Long bubble after
bursting, distribution
signficantly affected.
4. LOWER Rc: Long bubble with
complete separation.

INc
170

01 I j

1.1 FIGURE 3.
Suction Velocity Distribution
for a Cascade of NACA 65-4(AlO)10
Profiles: 5/c - 1.0, 01-560,
.I a-6.6 0 , Mid-Spnn A.V.R. - 1.07,
Rc-2.55 x 10', T.F. a 0.006.
Oe /U.

1.05.
0 Experimental Mid-Span
Data From VKI.
Modified Experimental
1.0 Distribution Used with
LeFoll Method (Ref. 5)
to Predict Laminar
Separation Point.
FOR CumC=U U. U Bubble Region as predicted
0.gs
by the Modified Semi-
Empirical Theory.

0.2 0.4 0!6 0 X/c

NACA 65-4(AlO)l0: g/c - 1.2, B0

1 600, a - 3.P, T.F. * 0.006.

&
4: 1i ,' uuL-
DllhItCAL 1131.

2 1

FIGURE 4
Loss Coefficient, ;q, as a Function of Blade Chord Reynolds Number, Rc,
for an Axial Compressor Cascade:
Modified Semi-Empirical Theory versus Experiment.
M-M: Point of Maximum Loss (and Minimum Deflection).
 [ 17!
Compressor Performance
11.2.6 Survey on the Effect of Blade Surface Roughness on

List of Symbols

a m average distance between Indices

roughness peaks a axial component
Sf c m mid span chord length individual stage
k pLm roughness height, defin. Fig. 2 k technical roughness
ks ,Um sand type roughness height ks sand type roughness
lm mm roughness measuring length crit, 1 "lower" critical Reynolds
Ra * m arithmetic average roughness, number
identical to AA or CLA
p maximum permissible rough-
Re Reynolds number ness for hydrodynamically
w M/s air velocity relative to smooth boundary layer flow
cascade pol polytropic
x m length coordinate crit, u "upper" critical Reynolds
y m roughness height coordinate number
z number of stages 1 entry to cascade
d change of parameter ref reference condition
Sefficiency - arithmetic average
2
Sm/s kinematic viscosity

1. Introduction
Blade surface roughness affects blade profile drag when the surface roughness elements
protrude from the laminar or viscous sublayer of the turbulent boundary layer as known
from the classic Nikuradse (Ref. 1) experiments. As long as the roughness elements are
submerged within the viscous sublayer no further reduction of friction losses is achiev-
able below the so called "hydrodynamically smooth" value. Friction drag in that regime
depends exclusively on Reynolds number.
As soon as the roughness elements protrude from the viscous sublayer duu to a thinner
sublayer as the result of increased Reynolds number, the friction loss function turns
toward independence of Reynolds number within a small Reynolds number interval.
After that transition region friction drag is exclusively a function of the roughness
height itself (Ref. 2).
The reaction of compressors to excessive blade surface roughness manifests itself in
a distinct trend toward constant polytropic efficiency despite increasing Reynolds
number (see Fig. 1). The Reynolds number where this "kink" occurs is called "upper
critical Reynolds number" or "roughness boundary" and depends on the type of roughness
and its height as well as the ratio of flow velocity over kinematic viscosity w/lV
or in other words Reynolds number per unit length Re/c.
The steadily increasing pressure ratios and flow velocities in modern gas turbine com-
pressors increase the Reynolds number over chord length ratio in the back end of the
compression system to an extent that even with the best presently available manufac-
turing methods noticeable losses of potentially achievable efficiency gains must be
accepted. The latter applies prinarily to all missions with high inlet pressures, i.e.
low kinematic viscosity. The roughness problem occurs primarily in the higher Reynolds
number regime where turbulent attached flow predominates.
2. Literature Survey
A number of experimental results are available in the literature on tests with sand
type roughness, but little is published on technical roughness, especially for blades
produced by modern manufacturing techniques. Basic information on the aerodynamic
mechanisms of surface roughness can be taken from Ref. 3, 4, a very complete survey
of roughness effects on compressor flow is given in Ref. 5 which includes a useful list
of all relevant literature. New data from three high pressure compressors with blades
produced by present day manufacturing techniques as grinding, forging/etching and
electrochemical machining are desribed in Ref. 6. The latter will essentially be used
in the present paper.
3. Criterion for Hydrodynamically Rough Boundary Layer Flow. Roughness Types and
Definition of a Technical Roughness
3.1 Criterion for Hydrodynamically Rough Boundary Layer Flow
The classic Nikuradse (Ref. 1) experiments on sand type roughness show that the maximum
permissible roughness up to which hydrodynamically smooth boundary layer flow can be
expected is given by a simple relation, called critical roughness Reynolds number:

Reks,,,RekSP~
ks w -1100
 172

The above mentioned equation does mean that the maximum permissible characteristic
surface roughness height k for a certain roughnessa type depends exclusively on the
velocity to kinematic visc~sity ratio or Reynolds number per unit length wv-t - Re/c
and the average geometry of the roughness elements and not on a characteristic length
as blade chord or compressor size.
For other roughness types large differences can be found in the literature for the
critical roughness Reynolds number reaching from 50 - 260 (Ref. 2, 7).
3.2 Characteristics of Roughness Types Different from Sand Type Roughness
Investigations on rows of regularly spaced spheres, hemispheres, cones, sharp edged
metal strips etc. (see Ref. 2) show a large variation of their effect on drag relative
to sand type roughness. The distance to height ratio is of decisive importance, ge-
nerally giving the impression that drag reduces linearly with the distance between
peaks above a ratio of a/km-8, which means every peak acts as a single body and hence
the specific shape of each element matters. The technical surfaces produced by the
typical blade manufacturing process as grinding, forging and electrochemical machining
look like very flat hills in wide valleys with a slope hardly exceeding 10 degrees
and typical distances to height ratios of a/k - 50 - 200 for the largest, aerodynami-
cally effective peaks. It is a totally different type of roughness compared with sand
type roughness.
3.3 Measurement and Definition of a Technical Roughness Height
The characterization of a technical roughness is very difficult due to various aspects
as:
- individual shape of peaks
- density of essential roughness peaks per unit length
- spectrum of roughness height
Whereas sand type roughness described in text books is made up of sand grain of essen-
tially constant diameter and quite small average distance between pek, a technical
roughness contains a whole spectrum of peak heights and possibly also large scale
waviness. The multitude of shapes is the reason for the very limited data published on
technical roughnesses. Roughness is usually described in industry by an arithmetic
average value (Ra, AA, CLA) measured by a stylus which travels along a rough surface
picking off roughness heights at constant distances. This arithmetic mean value alone
is not too helpful in defining the hydrodynamic characteristic of the surface because
a large portion of the smaller roughness elements is fully submerged in the laminar
sublayer and therefore not at all felt by the turbulent flow, but are counted in the
arithmetic mean value.
A more characteristic but still reasonably simple roughness height k is defined
which describes the flow affecting peaks rather than the average.
Roughness k is arbitrarily defined as the difference between the arithmetic averages
of the 10 highest peaks and the 10 deepest grooves which exist per millimetre lenght,
i.e. in an average distance of 100 p.am. Measuring length is 5 mum in chord- and span-
wise direction on the 20 % chord position of the suction surface, the forward 50 % of
which are regarded as decisive for friction drag produced by surface roughness. In
case that chordwise and spanwise measurement reveal different roughness levels, the
higher value is chosen on the experimental evidence obtained in wind tunnel tests
F with a milled flat plate having considerable differences of roughness height in di-
rection of milling and normal to it. The tests described in Ref. 8, show that only a
10 degrees incidence relative to the low roughness direction makes the high roughness
fully effective.
Fig. 2 shows the definitions of the roughness parameters, where k yields essentially
similar data as a more sophisticated definition used by HUrlimann,Ref. 9,for estima-
ting roughness effects on steam turbine blades. A considerable number of blade surface
roughness measurements done on forged and electrochemically machined blades with a
roughness measurement device revealed a sensible correlation between k and the readi-
ly obtainable arithmetic average Ra (AA, CLA):
k - 8.9 Ra
The factor scattered from 7 - 12 with 80 % of the samples placed between 8 and 10. The
measurements are done with a stylus of 5,Mm tip radius, a cut off length (wave length)
of 0.075 mm and a stylus contact force of 4.9 N.
4. Critical Roughness Reynolds Number for Technical Roughness and Permissible
'Roughness Height
4.1 Critical Roughness Reynolds Number
Critical roughness Reynolds numbers for blade surfaces produced by typical manufacturing
methods were obtained from test results of three multistage high pressure compressors,
which met the roughness boundary during variable Reynolds number tests (Ref. 6). From
Figs. 3, 4 a mean Reynolds number was picked of f at the crossing point of the elonga-
tions of the increasing efficiency slopes in the hydrodynamically smooth or interme-
diate regime with the constant 11 line from the hydrodynamically rough regime.
This "kink point" was used to cal~fiate individual stage roughness Reynolds nmuber
from the measured blade roughnesses of the various stages. Roughness Reynolds number
is based on relative cascade inlet velocity, roughness height k as defined on Fig.2
 173
and kinematic viscosity calculated from static pressure and temperature ahead of the
respective cascade. A mean roughness Reynolds number was obtained by arithmetically
averaging the individual stage data omitting atypically high values which were due
to distinctly higher than average roughness levels. The analysis gave critical rough-
ness Reynolds n,,mbers of Rak • - 88 for forged and electrochemically machined blades
and Rek,p - 137 for ground lades
& as given on Tab. 1.

Tab. 1: Stage Roughness Reynolds Number at Roughness Boundary

Compressor Res. HPC HPC "A" HPC "B"

(5-stage) (6-stage) (6-stage)
Reynolds number at roughness 3.1 • 10 6.0 * 10 5.0 * 105
boundary

"L
wR stage
stage
1 111 83 69
2 119 91 74
stage 3 137 70 147+
stage 4 153 97 105
stage 5 165 173* 101
stage 6 - 101 166 +

Re - k 137 88 87
k,p V
production method grinding ECM ECM

+ omitted in averaging process

The significant difference between the ground blades and the forged or electrochemi-
cally machined blades may be connected with the definition of maximum roughness in
case of the ground blades (see 3.3)which showed a 2:1 ratio of spanwise to chordwise
roughness as opposed to forged and electrochemically machined blades which do have
similar roughness in both directions. If average values of spanwise and chordwiae
roughness were used, which would be equally sensible on the argument that over at
least 50 % of the blade height of a high hub/tip ratio compressor the flow is essen-
tially chordwise, i.e. in direction of the lower roughness with flow angles certainly
below 10 degrees relative to the minimum roughness direction and therefore unaffected
by the increased spanwise roughness, the roughness Reynolds number would drop to
Re = 103. This is very close to the critical roughness Reynolds number of sand
tyN&Proughness and that of forged or ECM manufactured blades.
The values of Re - 88 however tie in extremely well with a value of 90 proposed
by Koch and Smith'#n Ref. 10. Obviously more experimental evidence is needed for cases
with significantly different roughness heights in chordwise and spanwise direction.
The average critical roughness Reynolds numbers given on Tab. 1 can be used to locate
the Reynolds number where the individual stages enter the hydrodynamically rough boun-
dary layer flow regime. These points are encircled for instance on Figs. 3 and 4.
Limited experience shows that a distinct curl over of the efficiency/Re number function
has to be expected if only about 2/3 of the decisive cascades enter the rough boun-
dary layer flow regime. The reason for this behaviour is not really understood as one
would expect a steady curl over as more and more stages enter the hydrodynamically
rough Loundary layer flow regime.
The observed effect of the technical roughness on the compressor behaviour is consi-
derably more pronounced than could be expected from an extrapolation of data using
hemispheres or cones (see Ref. 2) as a simulation of technical roughness elements.
In fact the very similar critical roughness Reynolds numbers of sand type and technical
roughness in spite of their totally different appearance, especially with respect to
the distance/height ratio a/k, is surprising.
4.2 Permissible Blade Surface Roughness
The permissible blade roughness for hydrodynamically smooth flow is gjven on Tab. 2
for forged and electrochemically machined blades over a range of Reynolds numbers
per unit length using the critical roughness Reynolds number of Rek,p ' 88.
Tab. 2: Permissible Roughness Height Re k 88 for Hydrodynamically Smooth
Boundary Layer Flow Conditions kp
Re 8 W1 - -1
10-
-- -- 108 m 0.10 0.25 0.50 0.75 1.00 1.25 1.50
kp /m 8.8 3.5 1.75 1.2 0.9 .7 0.6

I - - _ _
 '74

Fig. 5 shows the permissible blade surface roughness height k against w1 /V ratio
for the roughness Reynolds numbers of 88 and 135 in Ref. 6-teits, showing the rapid
reduction of surface roughness necessary in the back stages to keep the flow hydro-
dynamically smooth. An indication of which stages are running into problems at the
various flight cases can be seen from the columns drawn at a surface roughness of
k ' 2,mm which is about the best quality presently achievable with forging and electro-
chemical machining. The necessary surface roughness reduces as pressure increases from
stage to stage.
engines operate in the hydrodynamically rough boundary layer flow regime at low level
flight conditions; for civil engines the back end of the compression system is also
limited in the potentially achievable efficiency gain due to increased Reynolds number
by excessive surface roughness. The basis for the columns drawn is a compression
system made up of present day average flow velocity according to Fig. 6.
5. Simple Method for Estimation of Roughness Effect on Efficiency
With the use of the obtained critical roughness Reynolds number the upper critical
Reynolds number can~ be obtained for each stage and hence for the compressor by averaging:

Recritu 1 e ,p -

Below this upper critical Reynolds number (hydrodynamically smooth region) polytropic
efficiency follows a power function:
-11YPO Re ~-
-1 Poi ref ýR rret I
Quite good agreement was found between the experimentally obtained exponent n and
the Wassel prediction (Ref. 11). The exponen~t had a value of n x 0.11 for some multi-

I
stage compressors and appeared to be valid also at part speed conditions.
If the exponent n is known for the hydrodynamically smooth regime frora easily obtain-
able subatmospheric tests or a good correlation the potential efficiency vs Reynolds
number relation can be drawn for hydrodynamically smooth flow. By making use of the
experience that the bend over occurs within a small Reynolds number band the average
upper critical Reynolds number Re can be marked on the Yt 0 -Re function as the
crit u..o
bend over point towards constant efficiency for a certain roughness chosen.
The potential efficiency losses can be picked off for various Reynolds niimbers between
the constant and the increasing efficiency line for hydrodynamically rough or smooth
flow, respectively. If the picture is handled parametrically for various roughness
heights it provides an easy and good insight into the potential efficiency losses due
to roughness of blades involved in high Reynolds number operating points.

6. Conclusions
Analysis of limited experimental data reveals that critical roughness Reynolds numbez
of compressor blades manufactured with typical present day methods as forging/etching
and electrochemical machining are around 90 and therefore very close to sand type
- roughness if roughness height is based on the largest peaks.
Modern high pressure ratio engines suffer from blade surface roughness in the back
stages of the compression system. Surface quality needed to keep the flow hydrodynami-.
cally smooth exceeds considerably the best present day quality achievable, thus li-
miting the potential efficiency gain at high Reynolds number flight conditions, i.e.
essentially low level flight conditions.

----
 175 *
7. References
Ref. 1 Nikuradse, J. Oesetzm~aigkeit der turbulenten Str~mung in glatten Rohren
Forschungsarbeit im Ingenieurwesen, H. 356, 1932
Ref. 2 Schlichting, H. Grenzachicht-Theorie
Verlag Braun, Karlsruhe, S. Auflag, Kapitel XX, XXI
Ref. 3 Bammert, K. Boundary Layers on Rough Compressor Blades
Milsch, R. ASME-Paper No. 72-GT-48, 1972
Ref. 4 Bammert, K. The Influence of the Blading Surface Roughness on the
Woelk, G.V. Aerodynamic Behaviour and Characteristics of an Axial
Compressor
ASME-Paper 79-GT-102, 1979
Ref. 5 Bammert, K. Untersuchung des Rauhigkeitseinflusses auf die Strdmungs-
Woelk, G.V. verluste im Axialverdichter
Sonderforschungsbereich 61, Bericht 18, Abschluabericht
zum Teilprojekt C5 1977, Techn. Universitht Hannover
Ref. 6 SchAffler, A. Experimental and Analytical Investicration of the Effects of
Reynolds Number and Blade Surface Roughness on Multistage
Axial Flow Compressor
ASME-Paper No. 79-GT-2, 1979
Ref. 7 Speidel, L. Einflun der Oberfl~chenrauhigkeit auf die Str8mungsverluste

Ref. 8 Scholz, N.
in ebenen Schaufelgittern
Forschung im Ingenieurwesen 20, S. 124 - 140 (1954
Aerodynamik der Schaufelgitter, Band I
9;
Verlag Braun, Karlsruhe, 1965
Ref. 9 HUrlimann, R. Zum EinfluO der Oberfldchenrauhheit, inabesondere der Fer-
tigungsgUte auf die Strdmungsverluse von Dampfturbinen-
schaufeln
VDI-Berichte Nr. 193, 1973
Ref. 10 Koch, C.C. Loss Sources and Magnitudes in Axial-Flow Compressors
Smith,L.H. Jr. Trans. of the ASME, J. of Engn% for Power, July 1976, 411-424
Ref. 11 Wassel, A.B. Reynolds Number Effects in Axial Compressors
ASME Journ. of Engng. f. Power, April 1968

N.

\.-- . . . - - ~ .
 176

Lam. Separation Bound. Roughness Boundary

Laminar rTurb. Attached Flow rTurb. Attached Flow
Separation * Hydrodyn. Smooth Hydr. Rough Surface
?,pol Surface

Laminar Sublayer

I I

Re
SRe crit, L< 12.05 Recrit, U f (2. i k)

Effect:

Reduced Turning Normal Turning Normal Turning

High Losses
-n
Steep • Increase 1- ýPl ReS" const.
• ~~pot =cnt

Steep Flow Flow Increases Flow Increases

Increase Slightly
Stall Line Stall Line Line
Affected Unaffected Unaffe c ted
Fig.I Effect of boundary layer conditions on compressor behaviour
 177

S---*.. -

F GD
"> inE CU
-0

0 W 0E Ei
oD

C.

x 0

00 E
a LL
6.i 0)

4n -r C

0)0

0 Go .E.o
> a

) U)
U)-
GD0 C *C C 0
E Z I-,

II

0 4))J
0~
178

Roughness Bound.---51 6 4 2 13
[oioj ~Rotor 's...-

HPC A~,.

Rogns Boundary
-2- -0!

-3"

HPC B0-- Rework.Blades)

---- 2
n 0 367 I
Re~ =88]-

Vl -
3

-34
Lamninar Separation
Boundary

10, 2 3 46 810 -6/

Rotor 1 Mid span Re -Nt'mber

Fig.3 Effect of Reynolds-Number and surface roughness on polytropic efficiency

of two 6-stage high pressure compressors
 179

o Original Rough Blades

0 Smoothed Blades Rer~f a1,1 106

pot Correlation Exponent: n= 0,115

-1 *--'---•

, L 0/oJ/
- 2 --- .. . . .• ....... 4~
.0......
TrasitonRegion
-3,---
Rotar 2 1

.. '-Roughness Boundary
Rek:31 V

-5-
10 2 3 4 6 8 106
Rotor 1 Midspan Re-Number

Fig.4 Effect of Reynolds-Number and surface roughness on polytropic efficiency

of a 5-stage high pressure compressor

Manufacturing
Difficulty

10 Easy- \ A

C 6
"C"
==-\
l
Medium
E C \
- \-

\
IHydodn.Rough
135(Grinding)
S4 SECM Eaoh A 1kmForgingECM
o. Addition
'2°,-L fP IP I .
S Add it io nal /////.•\\\\''

its _ Ma=
01 - ,0, ,
+15 Mono
"E Polishing
a.S1 -Necessary• 1H\drodn. Smoothl -

016 1 1'_
6 16/2 , 6 10- 15

per Unit Length "c - = 4 I -

Fig.5 Permissible surface roughness for hydrodynamically smooth

boundary layer flow
7' ~

fAssumption:
~180-

w, :385 rn/s :const,

__

Wi

IP- Stages IP-Stages HP- Stages

Re is.001
A 0I

1 2 .340 6 1 150 3

00I
LP S.0e P Sa e IA St ge
HP

1 2 3 45 6 8 10 1S20 10

r Stages zf
Fig.6 Reynolds-Nu'nber per unit length as a function of compressor
pressure ratio and flight condition
 181

11.2.7 Part Span Damper Loss Prediction For Transonic Axial Fan Rotors

There are only a limited number of references in the open literature on loss calculations for
transonic axial fan rotors. Obviously the problem is quite arduous and many of the results are
proprietary.

Evaluation of part 3pan damper (PSD) losses is among the more difficult of these studies; two
attempts have been docurmented in the open literature.

(i) Optimization of PSDs requires the knowledge of the minimum loss that can be achieved,
This is the Koch & Smith (Reference 1) approach. They consider the situation t.o be divisible Into two
parts.

First one could identify a profile drag contributed by the series of PSDs that is equivalent to
the drag of an annular shroud, not interrupted by the blades. This shroud is then treated as an infinite
uncambered airfoil. Considered in this fashion, the significant Mach number is the merldional corn-
ponent of the relative Mach number. For all relevant cases this is always subsonic. Thus, even in the
case of a trar,;,onic fan there is no PSD leading edge shock loss to be considered.

Second, one can defire an interference drag coetilcient, once again a function of the merid-
ional Mach number.

These two drag coefficiente are, of course, dependent also on the geometry of the PSD cross
section. In most cases an equivalent elliptical section tan be defined and 'he corresponding drag
coefficient can be estimated. The pressure 1s.is coefficient Will then be related to the entropy increment
that can be enpressed in the following way

(1) 4A_ JC.LmZ A,

R C 2 Aa

where CD is the drag coefficient, Mach number M is based on the vector mean of the upstream and
downst. earn velocities and on the average static temperature; A. is the PSD ring area, Aa is the
annulus area,

Examinition of experimental data shows that to a theoretical shroud drag of the order of 0. 3
thete corresponds an effective drag coefficient that can be derived from experimental pressure losses
using relation (1). This coefficient varies between 0.3 and 1.3.

In order to ,-wderstand the reason for such scattering the experinmental data should be exam-
ined in detail: total pressure, Wotai temperature and flow ar.le traverse downstream of a transonic
fan show that the efect of PSD is merely the increase of total temperature and a swing in absolute air
angle due to damper Aalez. These effects increase with angle of attack and may be attributed to flow
separation and interference effects with the blades.

From the 'heorctical approach a probable minimum value of PSD losses can be derived. The
theoretical expression (1) has to be amended by means oi an empirical coefficient of the order to 1. 8
and the expression

,1.8 CD I M22 A,
R 2•

gives the minimum entropy increment that can be expected from PSDs on a transonic fan.

(ii) Correlation of excperimental data was attempted by Roberts (References Z and 3). A
series of correlations has been frmulated awl propcsed to predict the design and off-design aerodynamic
performance of PSD on transonic rotors.

At the design point, the total pressure losses attrilutable to the damper were correlated with
the following parameters;

I- The sho0'"- loss coeffic.ent, w., for the blade elerntnt containing 'he damper, which is the
total pressure loss coei'cierit associated with a normal shock of strength h1M. (,A, is the relative Mach
number at the rotoe: inlet. ) Although it was shown by Koch and Smith that the flow relevant to'the shroud
is usually subsoni-; since it corresponds to the meridional Mach number, these authors did not take into
account the interruption of the blade shock surfaces by the P5SDs that g-nerate.i this loss;

2- A blade aerodynamic loading parameter, such as the camber ý divided by the solildity C"
at the damper spanwise locadion;

3-- The geometry of the damper itself, characterizod by the leading edge a -d the trailing edge
damper radius normalized by the mean span and damper chord respectiveiy (as a usual order of magni-
 182

tude .ae can take rle/h = 0.001 and rt /Cdamp = 0. 005.)

Correlation of eixperimental results obtained on more than twenty different fans leads to the
following relation for the maximum pressure loss coefficient at the design point

(3)
S---PAD,
*
M 5 0 0 h
(~fl - ~8
Cdamp 0

The excess pressure loss due to the PSD can then be approximated by means of a normalized
distribution, knowing that the spanwise region of influence can be taken as 12. 5 times the damper thick-
ness.

(4) x = 12.5 t

For off-design conditions, the maximum loss due to the damper and region of influence at
design point are calculated and these values are used as reference quantities for all off-design calcu-
lations. The point of intersection of a constant throttle or aria operating line through the design point
with a speed line delines the reference point where the PSD loss is minimum. From this point the
suction surface incidence at the damper location is determined from experimental data or off-design
analysis code. Then the damper maximum loss region of influence is calculated for any point orn the
speed line by taking the difference in suction surface incidence at that poin'tand the reference incidence
and using it in equations (5) and (6).

WPSDh PD,, M + ' iss)

(6) X/h (X/h)* + 0.06 Ž( iss)

where iss =iss - isse

(iii) To minimize damper loss, several things should be taken into consideration. Both
approaches agree on the following:

1- The drmper should be as thin as possible so as to minimize the area influenced;

2- The damper should be located as near to the hub as possible;

3- Work input should be minimized at the PSD location.

Correlation of experimental results shows further that damper trailing and leading edge thick-
ness should be as sharp as possible.

References

(I) Koch, C.C. and Smith, L.H., Loss Sources and Magnitudes in Axial Flow Compressors -
Journal of Engineering for Power, July 1976.

(2) Roberts, W. B., A Design Point Correlation for Losses Due to Part-Span Dampers on
Transonic Rotors, Journal of Engineering for Power, July 1979.

(3) Roberts, W.B., Crouse, J.E., and Sandercock, D.M., An Off-Design Correlation of
Part-Span Dampe- Losses Through Axial Fan Rotors, ASME Paper No. 79-GT-6.

,4
 183

Nomenclature

Aa Annulus Area

As Part-Span Damper Area

CD Drag Coefficient

Cdamp Part-Span Damper Chord

Siss Incidence Angle To Blade Suction Surface, Deg.

Mm Meridional Mach Number

M1 Mean Inlet Relative Mach Number

PSD Part-Span Damper

Specific Gas Constant

rle Leading Edge Part-Span Damper Radius

te Trailing Edge Part-Span Damper Radius "

t Part-Span Damper Maximum Thickness

X Part-Span Damper Spanwise Region Of Influence

6 Ratio Of Specific Heats

'L s Entropy Increment

Blade Camber Angle, Radians

Solidity, Ratio Of Blide Chord To Spacing

w Total Pressure Loss Coefficient (P 1 - Pz)/ql

Subscripts

PSD,M Maximum Additional Loss Due To Part-Span Damper

* Design Point Conditions

i-
 184

11.2.8. Deviation/Turning Angle Correlations

INTRODUCTION

In the past, and at present, compressor through-flow solutions have used substantial input from
experimental measurements made in actual fan and compressor flow passages or in flows simulating real
compressor conditions. For example, almost total reliance has been placed on experimentally supported
methods for estimation of fluid turning angles in individual blade rows. Compressor design is, as a
result, heavily dependent on the quality of the experiments and data correlations associated with these
methods.

The three principal currently-used prediction methods for flow turning angle were developed during
the period 1945 to 1960. One of these methods directly predicts fluid turning angles in terms of cascade
geometry and aerodynamic parameters. The others predict the exit flow deviation angle,.defined as the
angle between the average exit flow direction and the direction of a line tangent to the blade section
camber line at the trailing edge. There has been no substantial modernization of these deviation/turning
angle procedures since 1960, while during this time the methods have been widely applied to airfoil section
profiles and in aerodynamic regimes far outside the limits suggested by the original derivations and data
correlations.

BLADE-TO-BLADE FLOW FIELD COMPUTATION

Most compressor aerodynamic design/analysis systems include blade-to-blade flow field computation to
aid in optimizing blade section shapes and their orientation in the blade row, but it must be remembered
that the optimization is carried out in an iterative procedure in which the hub-to-tip distribution of
flow is used to set the upstream and downstream boundary conditions for the blade-to-blade solution. These
"boundary conditions include the upstream and downstream relative flow angles and stream tube areas. Even
\,nder these conditions a complete capability to compute the viscous and compressible blade-to-blade surface
\ow field in axial-flow compressor blade row geometries has not been demonstrated up to the present.

Many partial computation procedures do exist for blade-to-blade flows and these are the procedures
used optimization. Nearly all, for examples Refs. 1-5, begin computation by working with an inviscid
flow, id viscous influences are introduced in an iterative sequence if at all. References 1, 6 and 7
represen some specific attempts'\to predict exit flow angles and/or losses for compressor blade rows by
blade-to- lade flow field comuta on. These were constructive attempts but there is little evidence of
regular us of these or othermetho to replace experimental turning angle and loss correlations in the
iterative hu -to-tLip calculation of n overall design system. Increased utilization of high relative Mach
numbers combi ed with high loading leves has made the task of prediction even more difficult (Refs. 8 and
9). \

Since the cuent 'overall trend in bla&l-to-blade flow field computation appears to emphasize develop-
ment of ability tocontr'1 and optimize the fow patterns in the blade row rather than to predict the flow
deflections and losses produced, the continuation of both qualitative and quantitative improvement of
experimental correlations for turning angle and loss it essenLial, ý-t the possibility of improvement exists
only if a correct conceptual model of the blade-to•lbade flow is established and understood. The following
sections describe the most important turning angle correlation systems available to compressor designers in
the published literature.

-•EXISTING DEVIATION/TURNING ANGLE PREDICTION EQUATIONS

Three principal correlation equations are used as the bases for prediction of changes in relative flow
angle across compressor blade rows. The initial development of all of these correlations occurred before
1960, and there has been little substantive change in the base equations since that time. All of the
equations are semi-empirical in nature, with an analytical background supported by the results of linear
cascade experiments. As a consequence of the theoretical asd experimental flow field conditions associated
with all three deviation/turning angle correlations, certain real compressor blade row problems are not
resolved by the basic procedures:

1) Stream surface radial location, radius change and shape are not linear cascade variables. The
problem of effective blade section ;rofile definition does not occur.

2) Changes in radius do not affect blade spacing or effective chord length determination. Prob-
i-ms of cascade geometry definition do not occur.

3) Stream tube effective area changes are not prediction equation vLriables. Axial velocity-
density ratio (0) was assumed equal to 1.0 for analysis and experiments used to generate the
correlations, or it was considered not to be a strong influence on the results.

4) The prediction methods are directed toward the selection of suitable cascade geometry combina-
tions for design point operation in conservative, non-extreme compressor configuration.
There is insufficient theoretical or experimental support in terms of marginal loading levels
and setting angle-solidity combinations.

The conditions associated with the initial development of the three deviation/turning prediction
methods are considered sufficiently important to justify a separate descriptive subsection and an Appendix
devoted to each one. The Appendices also collect in one location material drawn from numerous sources
which are sometimes difficult to locate. The Appendices are organized so that the restrictive conditions
connected with each correlation are given in a standard format. Naturally in the application of the basic
 t8S

correlations to the design or analysis problems for real compressor blade rows, strategies have been devel-
oped to adapt the definitions of equation variables to compressor geometry and to extend the ranges of
geometric and aerodynamic variables outside the limits suggested by the base correlations. Each Appendix
contains some examples illustrating these strategies.

It should be remembered that the objective in the case of each deviation/turning angle method was to
permit estimation of the change in average relative flow direction across a cascade arrangement operating
in a two-dimensional flow at specified entrance flow conditions including a defined entrance flow direc-
tion or incidence angle. The base incidence angle was different for each correlation, was not intended to
represent a design incidence, and was indicative only of prevailing ideas concerning favorable aerodynamic
operation.

1. NATIONAL GAS TURBINE ESTABLISHMENT (NGTE) CORRELATION

The overall development of axial-flow compressor' esign methods in the United Kingdom during the period
1939 to 1950 included research leading to the NGTE dev tion angle prediction equation, commonly called
Carter's rule. Figure 1 shows the primary pattern of de lopment of the correlation by reference to sup-
porting analysis and experiments.

The NGTE correlation was based on trends in deviation a le predicted by analysis of two-dimensional
potential flow through cascades with sharp trailing-edge blade ections (to permit use of the Kutta-
Joukowski circulation condition). This analysis was supported ad the equation for deviation was adjusted
by the results of plane cascade experiments. The correlation was nitially proposed to give the deviation
angle corresponding to a nominal flow turning angle, equal to 0.8 tes the turning measured at cascade
stall (maximum turning). In the theoretical studies the base incide e was either arbitrarily defined
(Ref. 14, ic - 4 deg) or set at the value for maximum lift/drag ratio ef. 15). In Appendix A the iopt
values of Ref. 15 were considered to be the base incidence. This Appen x summarizes the background and
subsequent application of the NGTE deviation method.

In the recommended terminology of this report, the base equation of the GTE correlation is
a\

Reference 15 also included generalized performance curves to permit prediction o cascade turning

angles for non-optimum incidence.

2. NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS/NATIONAL AERONAUTICS AND SPACE ADMINISTRATI (NACA/NASA)
CORRELATION

The NACA/NASA procedure for estimation of deviation angles for axial-flow compressor cascades was
developed at the Lewis Flight Propulsion Laboratory of the NACA between 1952 and 1956 as a part of the
preparation of a summary of NACA compressor design technology.

The NACA/NASA correlation used previous theoretical work (Refs. 16 and 17) as well as the NGTE corre-
lation to suggest the form of the base equation, but specifically attempted to recognize and eliminate
deficiencies noted in the NGTE method:

1. The NGTE correlation predicts zero deviation angle for zero camber blade section profiles.
Both theoretical and experimental studies indicated a non-zero deviation angle for two-
dimensional flow across staggered cascades with zero camber and non-zero thickness at zero
incidence.

2. Both camber line shape and blade section thickness distribution were believed to significantly
affect deviation angle.

3. Use of a constant exponent on the solidity parameter in the NGTE correlation was considered
unsatisfactory.

In the recommended notation of this report the NACA/NASA correlation equation is (Ref. 23).

6 -- a60
ref

To define the correlation, available plane cascade experimental data were evaluated to isolate a
base of experiments which could be considered two-dimensional (0 - 1.0). This eliminated a substantial
number of sources, including most potential variants in blade section profile. The final correlation was
strongly influenced by NACA experiments (Refs. 18-20) carried out at low cascade inlet Mach numbers.
Unfortunately these experiments also were sequences of data points at constant inlet flow angle 81 with
the variation in incidence obtained by changing stagger angle. This created difficulty in defining a base
incidence angle related to the operation of real compressor cascade arrangements, where the stagger
remains constant as inlet flow angle changes (Ref. 27).

The predicted deviation angle corresponds in the NACA/NASA correlation to a reference minimum loss
incidence angle at low In~et Mach numbers. A procedure was derived for determining the effects of inci-
dence changes. The equation is
 186

5 =dre + (i - irf) re
refdief ref

ur it is limited to the relatively small range of incidence at low inlet Mach number in which the slope

(L'
'ref
is constant.

Figure 1 shows the development pattern of the NACA/NASA correlation, and a more complete summary of
its content is given in Appendix B.

3. USSR CORRELATION
This correlation is the product of extensive systematic plane cascade experiments performed in the
Soviet Union between 1951 and 1956. Two independent sets of data are involved (Ref. 24 and 25), reporting
results for cascade entrance Mach numbers up to 0.92 with Q in the range 1.0 to 1.15 (not controlled).
The correlations developed (Ref. 26) are almost entirely empirical in nature, -:t they cover a wide
range of cascade blade row geometry and substantial variations in blade section profile, unfortunately
only including two camber line shapes with very limited movement of the position of the maximum camber
location.

The principal prediction equation estimates the low-speed flow turning angle corresponding to two-
dimensional flow across a cascade operating at an optimum incidence angle. In this case the optimum
incidence is based on the minimum value in the loss-incidenco curve at a specified high-subsonic Mach
number level (Ref. 25). A relatively simple equation is suggested for the prediction of optimum incidence.
K
In the recomended notation of this report

A$oB)d/c
LAso A8)dc=0.4]~~o
E 4
KFo/
J_) +L
B]
/ 0.10] 00
K - (5 - 2 (90 - 2 - 1000 + 100 (5.5 - 26)

2
B 8 17) + 16

(As°)d/c 1 - 0.28 (d/c - 0.4) for - = 0.3 to 1.0

(AWo)d/c = 0.4 c

(Ao)t/c 1 - 1.6 (0.10 - t/c) for - = 0.0125 to 0.125

(ABo)t/c = 0.10 c

Generalized curves are given in ReZ. 25 for the influence of cascade entrance Mach number and incidence
variation above and below optimum on cascade fluid turning angle A 0o. Appendix C outlines the basis for
the USSR correlation. Little is known about specific applications.

DEVIATION ANGLE CORRELATION FOR ADVANCED FAN AND COMPRESSOR DESIGN/ANALYSIS

1. EXPERIMENTAL DEVIATION!TURNING ANGLE DATA

a. Plane Cascade Data

Figure 1 shows that a very limited part of the linear cascade data available in the literature has been
used in development of the principal existing deviation/fluid turning angle correlations. In the majority
of linear cascade experiments carried out before 1950, the results were not suitable for correlation
because of unsatisfactory control of aerodynamic variables in the test programs and because of inadequate
measurement and data reduction methods (Refs. 23 and 28). However, after 1950 and up to the present,
numerous linear cascade facilities have been uted for experiments with both subsonic and supersonic inlet
flow through compressor blade section profiles in cascade arrangement. Of the published results of experi-
ments conducted during the past 25 years, some but not all might be satisfactory for extension or develop-
ment of deviaticn/turning angle prediction methods. Certain preconditions should be considered in evalua-
tion of the data:

1) Is the blade section profile and cascade geometry adequately described for definition of
"geometric variables?
2) Have the independent aerodynamic variables been satisfactorily controlled, so that the
parameters influencing fluid turning angles can be considered numerically reliable?

3 Are the measurement and data reduction methods clearly described so that the basis of
 187

numerical values of both deviation/turning angle and total-pressure lobe parameters can be
understood?

4) Is there sufficient data available to assist in extension of the range of applicability of

existing correlation forms or to support the development of a new prediction method?

A large number of data sources were investigated during the present study and in the case of linear
cascade results several sets of published data were considered suitable in term of most of the above
criteria. These includes Refs. 29,30,31,20 and 32-47. In addItion, some unpublished data were obtained.
In general both the published and unpublished results met all requirements except for number 4 above.
While the data has been acquired in order to meet a limited research or development objective, it is not
systematically planned so as to allow the generation of a new prediction method.

b. Fan and Compressor Cascade Configuration Daza

Annular cascade results from flow passage surveys in rotor and stator rows of axial-flow fans and com-
pressors are not available in significant quantities, especially where the data quality is adequate for
possible correlation. A large number of reported and unpublished test data sets were considered as candi-
dates, with the vast majority of these sets from rotor and stage experiments carried out by the NASA and by
NASA contractors.

The principal question of data adequacy for deviation/turning angle and loss correlation exists as a I
result of the presence of part-span dampers or~shrouds in the rotor blade rows involved. While the damper
geometries do not reduce the value of the data for many purposes, the aerodynamic effects cover enough of
the flow passage downstream from the shrouded row to confuse the data for angle and lose correlation (Ref.
48). Probable exceptions exist in the case of Refs. 49 and 50-53, and possible additional sets may be
evaluated as satisfactory in the future (e.g. Refs. 54-59).

Some reservations must exist concerning the use of fan and compressor passage survey data measured with
conventional pressure and temperature probes. The data-averaging characteristics of many probe types remain
in question.

INFLUENCE ON DEVIATION/TURNING ANGLE OF AERODYNAMIC VARIABLES

NOT INCLUDED IN PRINCIPAL EXPERIMENTAL CORRELATIONS
This section discusses several aerodynamic quantities known to have an effect on cascade deviation/
turning angle, but not accounted for in the principal correlations of experimental data. These are quanti-
ties which in many typical axial-flow compressors will fall in a range of values outside the correlation
boundaries. As a result some correction to correlation predictions should be made in desigu or performance
prediction by through-flow methods.

Brief discussions of the principal variables follow, but the major relevant work is sum'arized in
Table I to conserve space. Two kinds of references are called out for each variable. The first group
consists of valid data sets indicating the quantitative order of the effect of each variable as a separate
influence on turning angle. The second group includes methods for correction of two-dimensional, optimum
incidence fluid turning angle/deviation predictions. It should be noted that many of the corrections
account for the isolated effect of a single parameter, while in general the variables are highly-interrelated
in the blade row of a compressor.

TABLE I. DATA BASE FOR EFFECT OF NON-CORRELATED CASCADE AERODYNAMIC VARIABLES ON TWO-DIMENSIONAL, OPTIMUM
INCIDENCE FLUID TURNING ANGLE/DEVIATION

AERODYNAMIC VARIALBE REFERENCE DATA SOURCE TURNING

" ~CORRECTION
METHOD

Inlet Mach Number, M, Andrews 21 X

Todd 76 X
Dunavant, et al. 36 X
Briggs 39 X
Howell 11, 12 X
Robbins, et al. 77 X
Swan 78 X
Jansen, Moffatt 79 X
Novak 67 X
Davis, Millar 60, 80 X

Reynolds Number, Re Howell 11, 12 X

Horlock, et al. 83 X
Hebbel 82 X.
Roberts 83 X
Hoheisel, Kiock 84 X
Davis, Millar 80 X
Free-Stream Turbulence Barsun 85 X
Schlichting, Das 86 X
Kiock 87, 88 X
Evans, B. 89 X
188

AERODYNAMIC VARIABLE REVERENCE DATA SOURCE TURNING

CORRECTION
METHOD

Free-Stream Turbulence (cont'd) Evans, R. 90, 91 X

Incidence Angle, I or i Howell 11, 12 X

C 6s Lieblein 23 X
Swan 78 X
Davis 60, 80 X
Novak 67 X
Axial Velocity-Density Ratio, S Jansen, Moffatt 79
Heilmann 40, 41 X
Masek, Norbury 44 X X
Ilyas, Norbury 45 X
Starken, at al. 42 X
Gustafson 92 X
Starke 93
Stark, Hoheisel 94 X
X X

Aerodynamic Loading Level Lieblein 23 X

Swan 78 X
Davis 60, 80
Heilmann 40, 41 X

Secondary Flows End-Wall Effects Lakshminarayana,

Tip Clearance Horlock 95, 96 X X
Bardon, Moffatt 97 X X
Smith 98 X

1. UPSTREAM RELATIVE MACH NUMBER

In axial-flow compressors the Mach number level of the upstream relative flow may be either subsonic
or supersonic. The influence of M on the cascade flow field has been extensively studied in linear,
annular and rotating blade cascade&, witi. the primary objective of blade section loss optimization. Only
a few data collections have focused on deviation/turning correlation.

In general, there is a minimal effect of relative upstream Mach number for M1 < M1 .t.l. This is
shown in Refs. 23 and 77, with experimental support from rotor, stator and linear cascade iara Above the
critical M 1 , correction equations are proposed in Refs. 78, 79 and 80, and correction curves derived prin-
cipally from subsonic compressor data are given in Refs. 27 and 77. For choked cascade and supersonic
upstream M, correlations are rare, with most reported designs showing no direct correction procedure for
M1 .

2. REYNOLDS NUMBER

Turning/deviation corrections for Reynolds number level have been studied only in limited-objective
experiments. Generally linear cascade correlations are reported as valid above a specified Re value, and
the principal conclusions related to effects of Re. on boundary layer transition, flow separation, and
cascade loss coefficient.

3. FREE-STREAM TUR3ULENCE PARAMETERS

Compressor design/analysis systems do not usually account for turbulence parameters as independent
variables. This is certainly due to the limited data reporting measured turbulence quantities in typical
compressor configurations for locations other than the compressor inlet face. As is the case for Reynolds
number, most linear cascade test documents have concentrated on the influence of turbulence parameters on
boundary layer development and cascade loss soefficient.

4. INCIDENCE ANGLE

For two-dimensional linear cascade data, the change of deviation angle with incidence near the optimum
or design incidence is very small unless the cascade aerodynamic loading is high. Near the optimum incidence,
the correlated values of Ref. 23 may be used. As a general off-design performance prediction aid, these
values are not acceptable.

5. AXIAL VELOCITY-DENSITY RATIO

Axial-velocity density ratio (AVDR), defined as

.P,
• ) trailing edge plane
xV
x)leading edge plane
 189

has for many year. been recognized aa an influential va-iable in linear cascade tests. However, lack of
control of this independent variable has reduced the validity of numerous linear cascade date sets
including some of those associated with the principal deviation/turning angle correlations..

Several of the experiment*l studies of AVDR influence have suggested that a correction equation of the
form

is valid, with the derivative term determined by cascade geom.etry.

The difficulty of generalization of AVDR effects is due to several problems:

1) The AVDR effect in through-flow computation, as well as in linear cascade tests is a measure of
the change in effective stream-tube area through the cascade row geometry. The rate of area
variation through the cascade has an influence on both lose and turning.
2) The AVDR effect in higher Mach number cases is naturally related to the shock wave pattern exist-
Ing in the blade row, and is therefore difficult to isolate for correlation.

r3) The AVDR effect Is completely interdependent and not separable from the ;eneral effect of cascade
or row aerodynamic loading level.

6. CASCADE DIFFUSION FACTOR LEVEL

Turning/deviation characteristics of linear cascades have been demonstrated dependent on diffusion

loading as measured by any of the recognized limit parameters. However, the level of diffusion has not
been directly incorporated into turning/deviation angle prediction equations because the optimum incidence
condition for all but a limited number of correlated cascade geometries was representative of low diffusion
parameter levels. Lieblein (Ref. 23) made the observation that there was an effect of D on deviation in
two-dimensional linear cascade results, and excluded from the MACA correlation all data for D > 0.62.

FLUID TURNING ANGLE ESTIMATION FOR THROUGH-FLOW CALCULATION METHODS

USING COMPUTATION STATIONS WITHIN BLADE-TO-BLADE PASSAGE

Existing and future through-flow computation for axial-flow compressors requires initial estimation of
the fraction of total cascade turning for each internal computation station. This estimation Is significant
in flow field determination. Most existing intrablade computation reported uses guessed approximations for
turning rate. Some useful optical measurements in compressor rotors that may aid in the improvement of
turning rate approximation have been generated in recent years (Refs. 58, 59, 99).

APPENDIX A

NATIONAL GAS TURBINE ESTABLISHMENT DEVIATION ANGLE CORRELATION1

Base Equation

at6 me (Ref. 15)

i -io (Ref. 15)

oPt

5 deviation angle, angle between cascade exit average fluid angle and line tangent to blade section
camber line at trailing edge, degrees

mfunction of blade section stagger angle and camber line shape, Figure A-3.

S blade section stagger angle, angle between chord line and axial direction, degrees

8 blade section camber angle, angle between lines tangent to section camber line at leading and
trailing edges, degrees
a blade spacing, tangential distance between equivalent points on adjacent blade sections

c chord length, length of straight line connecting points where camber line intersects leading and
trailing edges

i incidence angle, angle between cascade inlet average flow angle and line tangent to blade section
camber line at leading edge, degLees

ISymbols and notation defined in Appendix A correspond to original publication of correlation except for
sign convention on stagger angle. See also Figures A-1 and A-2.
190
S incidence angle predicted
to give maximum lift/drag
ratio for given cascade
l°rdegrees geometry,

aI blade section camber line angle at leading edge, measured from axial direction, degrees

Q1
2 blade section camber line angle at trailing edge, measured from axial direction, degrees

Q1 average fluid angle at cascade inlet, measured from axial direction, degrees

Q2 average fluid angle at cascade exit, measured from axial direction, degrees

E fluid turning angle, degrees

VI fluid velocity at cascade inlet

V2 fluid velocity at cascade exit

Base Airfoil Section Profiles

C.1 (Ref. 11)

C.2 (Ref. 11)

C.4 (Ref. 22)

Base Camber Line Shapes

Circular-arc (Ref. 11)

Farabolic, ! - 0.40 only (Ref. 11)

Base Cascade Geometry Limits

a J 0.50 circular-arc camber line

c 0.40 parabolic camber line

0.33 C.1 section

d 0.30 C.2 section
c 0.30 C.2 section
k0.0C.4seto

_ Ref. 15 based on analysis and data for t/c - 10%

c

ro.08t C.l section

LER(00.12t C.2 section
LO.12t C.4 section

O.02t C.1 section

TER( O.02t C.2 section
Lj.06t C.4 section

Sm
defined for ý from 0 to 60 deg

c_ Ref. 15 based on data for 1c from 0.5 to 1.5

a Applicable camber limited by section loading

Base Aerodynamic Variable KLage

M, analysis-incompressible flow
results used in Ref. 15 all were for low-speed plane cascade flow

Re results used in Ref. 15 refer to efiective Re of about 4 x 105 based on chord and exit velocity

Q Q - 1.0, two-dimensional flow assumed in analysis and experiments

Turbulence data not available

Design/Analysis Application Examples

A. Modified Base Equation and Incidence

The NGTE rule has been used to predict deviation angle for a range of incidence levels when cascade
diffusion loading is low. This is done on the assumption thbt deviation variation with ir-idence is not
 191

large when profile losses are low.

Various investigators including Refs. 60 and 61 have given equations for a based on Fig. A-3.

Application of the NOTE rule to design has frequently included a capability to add an arbitrary angle
correction (for examples, Refs. 62 and 63).

Ref. 49 applied a modified equation to design

6 --

with

me - 0.92 + 0.002 a

Ref. 63 suggests the substitution of an equation for mc of the form

incm
2
(0.219 + 0.0008916y + 0.00002708,y ) (221x-0052y+00097 2
9
The experimental support for this modified m is not clear, but the trends associated with its use are simi-
lar to those of other application modifications. It should be noted that the definition of the angle y
given in Ref. 63 is slightly different from that used in this report and in the NOTE correlation. This
difference is not considered material.

B. Extended Blade Section and Cascade Geometry Limits

Airfoil Section Profiles and Camber Line Shapes

The NOTE rule has been used for NACA 65-series profiles on circular-arc camber lines (Ref. 61) and for
double-circular arc and multiple-circular-arc (DCA and MCA) profiles (Refa. 49, 62, 63, 64). It has been
applied to a polynomial camber line with a polynomial thickness distribution (Ref. 65).

Ref. 63 includes an NOTE rule option for the sections defined in that report.

Cascade Geometry Limits

a Equations or curves for prediction of deviation for a/c values less than 0.4 and greater than
c 0.5 are given in Refs. 49, 62, 63, 64 and 65.

S mvalues extrapolated by equation to C > 60 deg

Blade-to-Blade Surface Radius Change

The NOTE rule has been used for radius change and 11- 1.0 cases by substituting an equivalent circula-
tion turning angle into an equation of the form (recomended notation)

1B_- i -2e

This approach is used in Refs. 61, 62, 63 and 64. See Appendix E.

C. Extended Aerodynamic Variable Range

M Rule used without M 1 correction for supersonic entrance flows in Refs. 49, 62, 63, 64 and 65.

Re Correction suggested in Refs. 10, 11 and 12.

Q See the method under raAius change above and in Appendix E. This approach used in Refs. 61, 62,
63 and 64.
 192

APPENDIX B

NACA/NASA DEVIATION ANGLE CORRELATIONI

Base Equation

6ref 60 +-b see Figures B-5, B-6, B-7 and Ref. 23

with

60 -(Kl)sh (Kl)t (60)10 see Figures B-3, B-4

at

i = iref (Ref. 23)

k. 6ref cascade exit average deviation angle measured from tangent to blade trailing edge camber line
r direction for cambered cascade, degrees

6 deviation angle measured from camber (chord) line for zero camber cascade with same fluid inlet
angle and solidity as the cambered cascade, degrees

(60)10 value of 60 for cascade with NACA 65-series blade section airfoil p:ofile and maximum section
thickness 10% of chord length, degrees
(Kd)sh dimensionless correction factor to (ý,.lo for effects of blade section profile not NACA 65-
series
(Ka)t dimensionless correction factor to (60)10 for effects of blade section maximum thickness not
10% of chord length

Tp blade section camber angle, with equivalent circular arc camber angle used for NACA A10 camber
line shape, degrees

o cascade solidity, ratio of blade section chord length to blade-to-blade spacing or pitch

r0 ,= rate of change of deviation angle with camber angle for cascade with solidity of 1.0
b correction exponent accounting for variable influence of solidity on 6- slope associated with
different fluid inlet angles

a blade spacing, tangential distance between equivalent points on adjacent blade sections

c chord length, length of straight line connecting points where camber line intersects leading
and trailing edges

i incidence angle, angle between cascade inlet average flow angle and line tangent to blade sec-
tion camber line at leading edge, degrees

ire* reference minimum-loss incidence angle, degrees (Ref. 23)

blade section camber line angle at leading edge, measured from axial direction, degrees
1

K2 blade section camber line angle at tladilng edge, measured from axial direction, degrees

B1 average fluid angle at cascade inlet, measured from axial direction, degrees

a2 average fluid angle at cascade exit, measured from axial direction, degrees

AB fluid turning angle, degrees

V1 fluid velocity at cascade inlet

V2 fluid velocity at cascade exit

Base Airfoil Section Profiles

NACA 65-010 blower blade section (on NACA A camber line only)
C-series airfoil section thickness distribuiona (on circular-arc camber line only)
Double circular are airfoil sections

iSymbols and notation defined in Appendix B correspond to original publication and correlation. See also
Figures B-1 and B-2.
 193

dass Camber Line Shapes

NACA A10 (a a 1.0)

Circular Are

Base Cascade Geometry Limits

a
{0.50
0.50 NACA A10
circular arc

0.30-0.33 C-series
0.40 65-010 blower blade section
C 0.50 double circular arc

(k 6 )t given in range 0 to 0.12

c

LER
10.08
0.00 6 7c 65-010 blower blade
to 0.12t C-series

R 0.0015c 65-010 blower blade

0E02 to 0.06t C-series

0(I1is correlation angle)

a (6 ) given in range a - 0.4 to 2.0

Applicable camber limited by section loading to D < 0.62 with 60 deg implied by 0 < C10 < 2.4
in Fig. B-5

Base Aerodynamic Variable Range

very small effect of on deviation at iref was predicted up to limiting M, where rapid
increase in loss occurs
Re above 2.5 x 10 based on chord length and cascade inlet velocity

n data correlated for Q Z 1.0 only

Turbulence data not available

Design/Analysis Application Examples

A. Modified Base Equation and Incidence

The NACA/NASA rule has been used for i • iref with the equation (Ref. 23)

= ref +(iref) di) ref

Values of K have been used in Figs. B-3 and B-7 to replace 81 (Ref. 66).

Application of the NACA/NASA rule to design has frequently included a capability to add an arbitrary
angle correction. Examples are in Rafe. 63, 66 and 67.

B. Extended Blade Section and Cascade Geometry Limits

Airfoil Section Profiles and Camber Line Shapes

The NACA/NASA rule has been applied to exponential, polynomial and arbitrary camber line shapes with
polynomial thickness distribution (Refs. 66, 68, 69 and 70).

Cascade Geometry Limits

a
a equation is given for deviation with - as a variable in Ref. 66
c c
Ref. 71 suggests caution in application when a is low and 81 > 60 deg

Blade-To-Blade Surface Radius Change

Equivalent circulation turning angle (see Appendix E) is used in NACA/NASA deviation option in Ref.
63.

-iL
194

Ref.72 suggests a solidity and thickness/chord ratio

correction for average blade-to-blade stream
surface slope in meridional projection.

C. Extended Aerodynamic Variable Range

M1 - correction suggested for above critical value in Refs. 60 and 67

-B Ref. 71 suggests caution in application when 81 > 60 deg

- see radius change above

- Ref. 23 suggests procedure for determining incidence correction

APPENDIX C
1
USSR FLUID TURNING ANGLE CORRELATION

Base Equation

-
0 (Amo )- 'x c
(Ai); Oc
0 10 1

K- (5 - 2t/b) v2 - 10 + 100 (5.5 - 2.6 t/b)

B - 8 2 17 +16

(&i )-x
0 X
c -1- 0.28 (c - 0.40) for xc 0.3 to 1.0

(Ao) - 0.4
c

and

(Aas)-c 1- 0.016 (10 - C) for c = 1.25 to 12.5%

(Ao)
(Ac*)-c 10%

At Base Incidence

1 Fr
i . ai -sin

Amo fluid turning angle for cascade operation at base incidence i° and low M,

V blade section stagger angle, angle between chord line and tangential direction, degrees

F blade section camber angle, angle between lines tangent to section camber line at leading and
trailing edges, degrees

t blade sp~cing, tangential distanc2 between equivalent points on adjacent blade section

b chord length, length of straight line connecting points where camber line intersections leading
and trailing edges

i angle of incidence, angle, between cascade inlet average flow angle and line tangent to blade
section camber line at leading edge, degrees

i optimum incidence angle, degrees

01I blade section camber line angle at leading edge, measured from tangential direction, degrees

m2 blade section camber line angle at trailing edge, measured from tangential direction, degrees

m1 average fluid angle at cascade inlet, measured from tangential direction, degrees

ISymbols and notation in Appendix C correspond to original publication. See also Figures C-1 and C-2.

4
 195

a2 average fluid angle at cascade exit, measured from axial direction, degrees

6 deviation angle, angle between cascade exit average fluid angle and line tangent to blade sec-
tion camber line at trailing edge, degrees

V1 fluid velocity at cascade inlet

V2 fluid velocity at cascade exit

Fr minimum flow passage width in blade-to-blade channel (throat) (Ref. 25)

Base Airfoil Section Profile

A-40 symmetric profile (Ref. 25)

(see xc / below)

Base Camber Line Shapes

H45 parabolic used for camber angle . 55 deg (Ref. 25)

K50 circular arc for all camber angles a 55 deg (Ref. 25)

Base Cascade Geometry Limits

xf $ 0.45 parabolic camber line

b ( 0.50 circular-arc camber line

0.30 A-30
0.40 A-40
xc 0.50 A-50 (Ref. 25 gives profile construction)
c 0.65 A-65
1.00 A-100

c 0.0125 co 0.125

LER 0.055 c

TER 0.05 c

V 20 to 110 deg

b 0.7 to 2.5 range of validity suggested in Ref. 26

S 5 to 85 deg

Base Aerodynamic Variable Range

H1 0.30 to 0.92
Generalized curves are given for predicting change in AO with H. in Ref. 25

Re 2 x 105 to 8 x 105 based on chord length and inlet velocity

fl 1.0 to 1.15

Turbulence not reported

Design/Analysis Application Examples

Not available

1r

iValues given are ranges reported in Refs. 24-26.

 196

APPENDIX D

COMPRESSOR BLADE SECTION PROFILES AND CAMBER LINE SHAPES

Table D-I is a listing of some blade section profile geometries which have been adequately described
and have been used in linear cascade experiments or in compressor blade row design. References are given
which contain information cn or instructions for layout or construction of each profile. In those cases
where reasonably general .-orrelation equations exist for aerodynamic performance, reference documents are
also listed.

APPENDIX E

i, DERIVATION OF EQUIVALENT CIRCULATION FLUID TURNING ANGLE

The equivalent circulation fluid turning angle concept was developed to rationalize the use of corre-
lations tnd parameters based on two-dimensional linear cascade experiments for the design and analysis of
axial-flow compressor blade rows. In this context it accounts for the Influence on effective cascade
diffusion loading of changes in stream surface radius and meridional or axial velocity across the blade
row.

An example velocity diagram for a rotating compressor cascade with r 2 # rI is shown in Figure E-1.
The relative fluid turning angle indicated is

AO= I 82

and the circulation for the row is

row = r 2 V ,2 - rVo
6 1

If the flow through the rotor occurred with no change in radius but with the same total circulation, the
equivalent exit tangential velocity would be

V,2e r V6,2

and

WU V U r2V
P,2e 2e V 2 "U -3,ve,2

The equivalent circulation velocity diagram is based un these exit tangential velocity components and

VM ,2e -V ml

The equivalent relative fluid angle at the rotor exit is

UI Ur-2
rr 62

B2e =tane Vm,

and

8
A =
-e R1 - 2e

r 2 Vm 2 + Uatan 1 - 2
rlVml 2 Vm l

For a stationary blade row

-1 r2V tB
A = tan r2V tanS 2 (E-2)

In both the rotor and stator cases the eqaations for Aae sh.?ti that boti. Vm, 2 < V, 1 and r 2 < r 1 cause anr
increased equivalent turning angle.
 ..... _ __

The equivalent circulation approach was suggested in 1957 by Lieblein (Ref. 76) as a means for extend-
ing the application of two-dimensional linear cascade diffusion loading parameters to compressor flo•t con-
ditions. It also is similar to a method for correction of l~near cascade turning angles for axial
velocity changes used by Erwin and Emery (Ref. 28). The idea was iurther developed by Klapproth (Ref.
S61) 72) to
for determine
diffusion equivalent and was angl(,s
parameters turning subsequently modifiedangle
for deviation by Seyler and Smith
prediction.
(Ref. 62) and Wright (Ref.
Equivalent circulation turning
angles have been used for selection of compressor cascade geometries based on cascade plane projection
(Ref. 62) and on the conical stream surface approximation (Ref. 63).

SYMBOLS AND NOTATION

The symbols and notation defined below are recommended for correlat. )n of deviation/flUid turning
angle data as described in the body of this report. Separate lists are given for existing correlation
systems in the Appendices.

AVR axial-velocity ratio across blade row, Vx2/VxI

a location of maximum camber point (see Fig. A-l)

a acoustic velocity

a
0J
acoustic velocity at local total temperature

b maximum camber (see Fig. A-1)

b exponent (see Appendix B)

c chord length (see Fig. A-1)

D cascade diffusion parameter

D casc&de equivalent diffusio, parameter

eq

Si
d location of maximum blade section profile thickness (Fig. A-1)
incidence angle, angle between cascade entrance relative flow direction and line tangent to
camber line at leading edge (see Fig. A-2)

i auction surface incidence angle, angle between cascade entrance flow direction and line
tangent to suction surface where leading edge shape fairs into thickness distribution

1 arc length os! camber line

M Mach number, V/a

mc parameter In NGTE deviation angle correlation (see Appendix B)

ma modified parameter in deviation angle correlation equation (aep Appendix B)

Rec Reynolds number beset on chord length, Re. - P]VIc/l

Sr radial coordinate
a blade spacing, distance in tangential or circumferential direction between corresponding
points on adjacent blades (see Fig. A-2)

t maximum thickness of blade oaction ptofile (see FI#. A-i)

U blade velocity

V fluid velocity

W fluid velocity measured relative to rotating blade row

X coordinate parallel tc chord line

x axial coordteiate direction

Y coordinate perpendl,!ular to chord liue

y tangential coordinate direction in linear cascade arrangement

z spenwise coordinate in linear cascade arrangement

Greek

, £luid angle of relative flow measur,2d from axial direction (see Fig. 1,-2)

y blade-chord angle or stagger angle, angle between blade section chord line and axial
direction (see Fig. B-2)

r f.* ~ .h ~ ~ ~ ~ j l
 198

"specific heat ratio

deviation angle, angle between cascade exit relative flow direction and line tangent to
camber line at trailing edge (see Fig. A-2)

e tangential or circumferential coordinate in compressor

K camber line angle, angle between line tangent to camber line and axial direction (see Fig.
B-2)

p fluid density

) dynamic molecular viscosity

G cascade solidity, c/s

* blade secti.on camber angle (see Fig. A-2)

Q axial velocity-density ratio, p 2 V /P V

2 , 2 1 X,l1
W angular velocity of rotor

SubrLrfpts

m meridional component
o zero-camber in NACA/NASA correlation; optimum in USSR correlation (see Appendix C)

ps pressure surface

T radial component

ref reference minimum-loss

rel measured relative to rotating blade row

as suction surface

x axial co-vonent

E circumferential component

1 cascade entrance

2 cascade exit

REFERENCES

1. Fottne., Leonhard. "Eln Verfahren zum Bestimmen der Verlutbehafteten Trsnnsoniqchen

Schaufelgitterstr1o,•uig bei vorgegebener Druckverteilung." Forscb. Ing.-Wes. 38(3):69-81. 1972.

2. Wilkinson, D. H. "Calculation of Blade-to-Blade Flow in a Turbomachine by Streamline Curvature."

ARC Rep. and Memo. 3704. 1970.

3. Miller, Max J. and Serovy, George K. "Deviation Angle Estimation for Axial-Flow Compressors Using
Inviscid Flow Solutions." J. Eng. Power. Trans. ASME, Series A. 97:163-172. 1975.

4. Delaney, Robert A. and Kavanagh. Patrick. "Transonic Flow Analysis in Axial-Flow Turbomachinery
Cascades by a Time-Dependent Method of Characteristics." J. Eng. Power. Trans. ASME, Series A.
, 98:356-364. 1976.

5. Sanger, Nelson, L. "Two-Dimensional Analytical and Experimental Performance Comparison for a Compres-
sor Stdtor Section with D-Factor cf 0.47." NASA Tf1D-7425. 1973.

6. Hansen, E. C., Serovy, G. K. snd Sockol, P. M. "Axial-Flow Compressor Turning Angle and Loss by
Inviscid-Viscous Interoction Blade-to-Blade Computation." J._Eng. Power. Tranr. ASME. 102:28-34. 1980.

7. Calvert, W. J. and Herbert, M. V. "An Inviscid-Viscous Interaction Method to Predict the Blade-to-Blade
Perfnrmence of Axial Compressors. Aeronautical Quarterly. 31:173-196. 1980.

SGostelow, J. P. "Review of Compressible Flow Theories for Airfoil Cascades." J. Eng. Power. Trans.
ASME, Series A. 95:281-292. 1973.

9. Lichtfuss, H. J. and Starken, H. "Supersonic Cascade Flow." In: Progress in Aerospace Science,
Vol. 15, pp. 37-149. New York Pergamon Press, Ltd. 1974.

10. Constant, F, "Perfotmanca of Cascades of Aerofoils." R.A.E. Note No. E.3696. A.R.C. 4155. 1939.
(Unpublishel)

11. Howell, A. R. "The Present Basis of Axial Flow Compressor Design. Part I. Cascade Theory and Per-
formance." Aeronautical Research Council. Rep. and Memo. No. 2095. 1942.

_ ii
 |. 199

Proc. of the IME. 153:441-452. 1945.

12. Howell, A. R. "Fluid Dynamics of Axial Compressors."

13. Howell, A. R. "Note on the Theory of Arbitrary .,erofoils in Cascade." Philos. Hag., Series 7,
39:913-927. 1948.

14. Carter, A. D. S. and Hughes, Hazel P. "A Theoretical Investigation into the Effect of Profile Shape
on the Performance of Aerofoils in Cascade." ARC Rep. and Memo. 2384. 1950.

15. Carter, A. D. S. "The Low Speed Performance of Related Aerofoils in Cascades." ARC CP 29. 1950.

16. Weinig, F. Die Str3mung um die Schaufeln von Turbomeachinen. Leipzig. J. A. Barth. 1935.

17. Schlichting, Hermann. "Problems and Results of Investigations on Cascade Flow." J. Aero. Sci. 21:
163-178. 1954.

18. Emery, James C., Herrig, L. Joseph, Erwin, John R. and Felix A. Richard. "Systematic Two-Dimensional
Cascade Tests of NACA 65-Series Compressor Blades at Low Speeds." NACA Report 1368. 1958.

19. Herrig, L. Joseph, Emery, James C. and Erwin, John R. "Effect of Section Thickness and Trailing-
Edge Radius on the Performance of NACA 65-Series Compressor Blades in Cascade at Low Speeds." NACA
RM L51 316. 1951.
20. Felix, A. Richard and Emery, James C. "A Comparison of Typical National Gas Turbine Establishment
and NACA Axial-Flow Compressor Blade Sections in Cascade at Low Speed." NACA TN 3937. 1957.

21. Andrews, S. J. "Tests Related to the Effect of Profile Shape and Camber Line on Compressor Cascade
Performance." ARC Rep. and Memo. 2743. 1955.
22. Howell, A. R. "A Note on the Compressor Base Aerofoils C.1, C.2, C.3, C.4, C.5, and Aerofoils Made
Up of Circular Arcs." Power Jets, Ltd. Memo M. 1011. September 1944.

23. Lieblein, Seymour. "Experimental Flow in Two-Dimensional Cascades." In: "Aderodynamic Design of
Axial Flow Compressors." Chapter VI, NASA SP-36. 1965.

24. Komarov, A. P. "Investigation of Flat Compressor Grids." FTD-MT-24-69-68. 26 April 1968. Transla-
tion of "Lopatochnyye Mashiny i Struynyye Apparaty." Sbornik Statey. Vypusk 2. Mascow.
Izdatel'stvo Mashinostroyeniye. 1967. pp. 67-110. AD 682829.
25. Bunomovich, A. I. and Svyatogorov, A. A. "Aerodynamic Characteristics of Foil Compressor Cascades at

High Subsonic Speed." FTD-MT-24-69-68. 26 April 1968. Translation of "Lopatochnyye Mashiny i

Struynyye Apparaty." Sbornik Statey. Vypusk 2. Moscow. Izdatel'stvo Mashinostroyeniye. 1967.
pp. 5-35. AD 682829.

26. Bunimovich, A. I. and Svyatogorov, A. A. "Generalization of Results of Investigation of Foil Compres-

sor Cascades at Subsonic Speed." FTD-MT-24-69-68. 26 April 1968. Translation of "Lopatochnyye
Mashiny i Struynyye Apparaty." Sbornik Statey. Vypusk 2. Moscow. Izdatel'stvo Mashinostroyeniye.
1967. pp. 36-66. AD 682829.

27. Miller, Max J. and Sk~nberg, Torbjorn. "Reference Incidence Angles in Constant Stagger Cascades."
Iowa State Jniversity. ISU-ERI-Ames-99985. June 1971.

28. Erwin, John K. and Emery, James C. "Effect of Tunnel Configuration and Testing Technique on Cascade
Performance." NACA Report 1016. 1950.

29. Mikolajczak, A. A., Morris, A. L. and Johnson, B. V. "Comparison of Performance of Supersonic

Blading in Cascade and in Compressor Rotors." J. Eng. Power. Trans. ASNE, Series A. 93:42-48.
1971.

30. Fleeter, Sanford, Holtman, Robert L., McClure, Robert B. and Sinnet, George T. "Experimental Inves-
tigation of a Supersonic Compressor Cascade." Aerospace Research Laboratories, ARL TR 75-0208. 1975.

31. Huffman, G. D. and Tramn, P. C. "Airfoil Design for High Tip Speed Compressors." J. Aircr. 11:682-
689. 1974.

32. Emery, James C. "Low-Speed Cascade Investigation of Thin Low-Camber NACA 65-Series Blade Sections at
High Inlet Angles." NACA RM L57E03. 1957.

33. Savage, Melvyn, Felix, A. Richard and Emery, James C. "High-Speed Cascade Tests of a Blade Section
Designed for Typical Hub Conditions of High-Flow Transonic Rotors." NACA RM L55F07. 1955.

34. Emery, James C. "Low-Speed Cascade Investigation of Loaded Leading-Edge Compressor Blades." NACA
TN 4178. 1957.

35. Erwin, John R., Savage, Melvyn and Emery, James C. "Two-Dimensional Low-Speed Cascade Investigation
of NACA Compressor Blade Sections Having a Systematic Variation in Mean-Line Loading." NACA TN 3817.
1956.

36. Dunavant, James C., Emery, James C., Walch, Howard C. and Westphal, Willard R. "High-Speed Cascade
65
Tests of the NACA 65-(12A1 0 ) 10 and NA'A -(12A218b) 10 Compressor Blade Sections." NACA RM
L55108. 1955.

L/
200

37. Emery, James C. and Dunavant, James C. "Two-Dimensional Cascade Tests of NACA 65-(CIOAIO) 10
Blade Sections at Typical Compressor Hub Conditions for Speeds Up to Choking." NACA RM L57H05.

38. Dunavant, James C. and Emery, James C. "Two-Dimensional Cascade Investigation at Mach Numbers Up to
1.0 of NACA 65-Series Blade Sections at Conditions Typical of Compressor Tips." NACA RN L58A02.
1958.

39. Briggs, William B. "Eliect of Mach Number on the Flow and Application of Compressibility Corrections
in a Two-Dimensional Subsonic-Transonic Compressor Cascade Having Varied Porous-Wall Suction at the
Blade Tips." NACA TN 2649. 1952.
40. Heilmann, W. "Experimentelle und Grenzschichttheoretische Untersuchungen an Ebenen Verzggerungagittern
bei Kompressibler Str6mung, inbesondere bei Xndeiung des Axialen Str3mdichteverhiltnisses und der
Zustromturbulenz." DLR FB 67-88. 1967.

41. Heilmann, W. "The Influence of Axial Velocity Density Ratio on Compressor Cascade Performance in
Compressible Flow." In: "Boundary Layer Effects in Turbomachlnery." AGARD-AG-164. 1972. Paper
1-13.

42. Starken, H. "Untersuchung der Strrmung in Ebenen Uberschallverz~gerungsgitlern." DLR FB 71-99.

1971.

43. Starken, H., Breugelmans,

Ratio in a High Turning Cascade."
F. A. E. and Schimming, P.
ASME Paper 75-GT-25.
"Investigation of the Axial Velocity Density
1975. 3
44. Masek, Z. and Norbury, J. F. "Low-Speed Performance of a Compressor Cascade Designed for Prescribed
Velocity Distribution and Tested with Variable Axial Velocity Ratio." In: "Heat and Fluid Flow in
Steam and Gas Turbine Plant." IME Conference Publication 3, pp. 224-236. 1973.

45. Ilyas, M. and Norbury, J. F. "Effect of Axial Velocity Variation on the Subsonic Flow Through a
Compressor Cascade." In: "Heat and Fluid Flow in Steaim and Gas Turbine Plant." IME Conference
Publication 3, pp. k76-287. 1973.

46. Ohounu, E. H. and Shaw, R. "Effect of Axial Velocity Variation on Deviation for Compressor Cascades."
In: "Heat and Fluid Flow in Steam and Gas Turbine Plant." IME Conference Publication 3, pp. 110-
114. 1973.

47. Stark, Udo and Starke, Jorg. "Theoretische und experimentelle Untersuchungen uber die quasi-
sweldimensionale inkompressible Stromung durch vorgegebene ebene Verdichtergitter," Forsch. Ing.
Wes. 40:172-186. 1974.

48. Sandercock, Donald M. Unpublished comnmunications concerned with NASA Lew~s Research Center fan and
compressor data evaluation, 1975-1977.

49. Monsarrat, N. T., Keenan, M. J. and Tramm, P. C. "Single-Stage Evaluation of Highly-Loaded High-Mach
Number Compressor Stages." Design Report. NASA CR-72562. 1969.

50. Sulam, D. H., Keenan, M. J. and Flynn, J. T. "Single-Stage Evaluation of Highly-Loaded High-Mach-
Number Compressor Stages. !I - Data and Performance Multiple-Circular-Arc Rotor." NASA CR-72694.
1970.

51. Keenan, M. J. and Monsarrat, N. T. "Experimental Evaluation of Transonic Stators." Preliminary

Analysis and Design ReporL. NASA CR-54620. 1967.

52. Keenan, M. J. and Bartok, J. A. "Experimental Evaluation of Transonic Stators." Final Report. NASA
CR-72298. 1969.

53. Harley, K. G. and Burdsall, E. A. "High-Loading Low-Speed Fan Study. Part 2: Data and Performance,
Unslotted Blades and Vanes." NASA CR-72667. 1969.

54. Lewis, G. W., Jr. and Tysl, E. R. "Overall and Blade-Element Performance of a 1.20-Pressure-Ratio
Fan Stage at Design Blade Setting Angle." NASA TM X-3101. 1914.

55. Osborne, W. M. and Steinke, R. J. "Performance of a 1.15 Pressure Ratio Axial-Flow Fan Stage with a
Blade Tip Solidity of 0.5." NASA TM X-3052. 1974.

56. Kovich, G. and Steinke, R. J. "Perforpance of a 1.15 Pressure Ratio AMial-Flow Fan Stage with a
Blade Tip Solidity of 0.65." NASA TM X-3341. 1976.

57. Moore, R. D. and Steinke, R. J. "Aerodynamic Performance of a 1.25 Pressure-Ratio Axial-Flow Fan
Stage." NASA TM X-3083. 1974.

58. Urasek, Donald C., Gorrell, William T. and Cunnan, Walter S. "Performance of Two-Stage Fan Having
Low-Aspect-Ratio, First-Stage Rotor Bleding.: NASA TP 1493 (AVRADCOM TR 79-49), 1979. (Test config-
uratlon for this AGARD study).

59. McDonald, P. W., Bolt, C. R., Dunker, R. J. and Weyer, H. B. "A Comparison Between Measured and
Computed Flow Fields in a Transonic Compressor Rotor." ASHE Paper 80-GT-7, 1980. (Test case config-
uration for this AGARD study).

60. Davis, W. R. "A Computer Program for the Analysis and Design of the Flow in TurbomachInery, Part B-
Loss and Deviation Correlations." Ottawa. Carleton University. Division of Aerothermodynamics.
 201 J
Report IM/A70-1. 1970.

61. Wright, t. C. "Blade Selection for a Modern Axial-Flow Compressor." In: NASA SP-304, Part 2, pp.
603-626. 1974.

62. Seyler, D. R., and Smith, L. H., Jr. "Single Stage Experimental Evaluation of High Mach Number
Compressor Rotor Blading. Part I - Design of Rotor Blading." NASA CR-54581. 1967.

63. Crouse, James E. "Computer Program for Definition of Transonic Axial-Flow Compressor Blade Rows."
NASA TN D-7345. 1974.

64. Ball, Calvin L., Janetzke, David C. and Reid, ,ontkie. "Performance of 1380-Foot-Per-Second-Tip-
Speed Axial-Flow Compressor Rotor with Blade Tip Solidity of 1.5." NASA TM X-2379. 1972.

65. Wennerstrom, A. J. and Hearsey, Richard M. "The Design of an Axial Compressor Stage for a Total
Pressure Ratio of 3 to 1." d.S. Air Force Systems Command. Aerospace Research Laboratories, ARL
71-0061. 1971.

66. Hearsay, Richard K. "A Revised Computer Program for Compressor Design. Volume I: Theory, Descrip-
tions and User's Instructions." U.S. Air Force Systems Command, ARL TR 75-0001, Volume I. 1975.

67. Novak, Richard A. "Flow Field and Performance Map Computation for Axial-Flow Compressors and
Turbines." In: "Modern Prediction Methiods for Turbomachine Performance," ACARD-LS-83. 1976.

68. Frost, G. R., Hearsey, R. X. and Wennerstrom, A. J. "A Computer Program for the Specification of
Axial Compressor Airfoils." Aerospace Research Laboratories. ARL 72-0171. 1972.
69. Frost, G. R., and Wennerstrom, A. J. "The Design of Axial Compressor Airfoils Using Arbitrary Camber
Lines." Aerospace Research Laboratories. ARL-73-0107. 1973.

70. Frost, G. R. "Modifications to ARL Computer Programs Used for Design of Axial Compressor Airfoils."
Aerospace Research Laboratories. ARL 74-0060. 1974.

71. Lieblein, Seymour. "Incidence and Deviation-Angle Correlations for Compresso-, Cascades." J. Basic
Eng., Trans. ASME, Series D. 82:575-587. 1960. (Discussions by L. H. Sm4 .ch, Jr. and G. Sovran
relate to B1, a limits).

72. Lieblein, Seymour. "Loss and Stall Analysis of Compressor Cascades." J. Basic Eng., Trans. ASME,
Series D., 81:387-397. 1959. (Discussion by J. F. Klapproth relates to D for annular cascades.)
eq
73. Hearsay, Richard M. and Wennerstrom, Arthur J. "Axial Compressor Airfoils for Supersonic Mach Num-
bers." Aerospace Research Laboratories, ARL 70-0046. 1970.

74. Stanitz, John D. "Approximate Design Method for High-Solidity Blade Elements in Compressors and
Turbines." NACA TA 2408. 1951.

75. Klapproth, John F., Jacklitch, John J., Jr. and Tysl, Edward R. "Design and Performance of a 1400-
Foot-Per-Second-Tip-Speed Supersonic Compressor Rotor." NACA RM E55A27. 1.955.

76. Todd, K. W. "Tests Related to the Effect of Profile Shape and Camber-Line on Compressor Cascade
Performance." Aero. Res. Council, Rep. and Memo. No. 2743. 1955. (NGTE Rep. R. 60, Oct. 1949).

77. Rubbins, William H.. Jackson, Robert J. and Lieblein. Seymour. "Blade-Element Flow in Annular
Cascades." In: "Aerodynamic Design of Axial Flow Compressors," Chapter VII, NASA SP-36, 1965.

78. Swan, W. C. "A Practical Engineering Solution of the Three-Dimensional Flow in Transonic Type Axial
Flow Compressors." WADC Tech. Rep. 58-57. 1958.

,1. Jansen, W. and Moffatt, W. C. "The Off-Design Analysis of Axial-Flow Compressors." j.,Eng. Power,
Trans. ASME. 89:453-462. 1967.

80. Davis, W. Roland and Millar. D. A. J. "Through Flow Calculations Based on Matrix Inversion: Loss
Prediction." In: "Through-Flow Calculations in Axial Turbomachinery." AGARD-CP-195. 1976.

81. Horlock, J. H., Shaw, R., Pollard, D. and Lewkowicz, A. D. "Reynolds Number Effects in Cascades and
Axial-Flow Compressors." J. Pn.ow__.e•,
Trans. ASME. 86:236-242. 1964.

82. Hebbel, H. "Uber den Einflusa der Machzahl und der Reynoldazahl auf die Aerodynamischen Beiwerte von
Verdichter-Schaufelgittern bei verscheidener Turbulenz der Str~mung." Forechung Ing.-Wesens, 33*1/.1-
150. 1967.
83. Roberts, William B. "The Effect of Reynolds Number and Laminar Separation on Axjial Cascade Perform-
ance." J. Eng. Power. Trans. ASME. 97:261-274. 1975.

and Kiock, Reinhard. "Zwanzig Jahre Hochgeschiwindigkeits-Gitterwindkanal des

84.
Inst:tuts Heinz
Hoheisel,
6r Aerodynamik der DFVLR in Braunschweig." Zeit. Flugwiss. Weltraumforech. 1:17-29. 1977.

85. Barsun, K. "Experimentelle Untersuchungen Uberden Einfluss der turbulenzgrades und eines
Turbulenzfadens auf die kompressible Unterschallstromung durch ebene Verdichtcegitter." DFL Rep.
68-34. Braunachweig. 1968.
202 _ _.2.6.19

86. Schlichting, H. and Dan, A. "On the Influence uf Turbulence Lavel on the Aerodynamic Losses of Axial
Turbomachiners." In: "Flow Research on Blading." Amsterdam. Elsevier. pp. 243-274. 1970.

87. Kiock, R. "Messung des Turbulenzgrae.es

dreistufigen Axialgeblase." DFL Bericht hinter
68/32. elnem ebenen Schaufelgitter sowi,
Braunschweig. 1968, in einem

88. Kiock, R. "Einfluss des Turbulensgrades auf die aerodynamischen Eigsneahaften von ebonen
Verzogerungsgittern." Forachung mi. Was. 39:17-28. 1973.

8. Evans, B. J. "Effects of Free-Stream Turbulence on Blade Performance in a Compressor Cascade." Ph.D.

Dissertation. Cambridge University. 1971.

90. Evans, R. L. "Turbulent Boundary Layers on Axial-Flow Compressor Blades." Ph.D. Dissertation.
Cambridgr. University. 1973.

91. Evans, R. L. "Turbulence and Unsteadiness Measurements Downstream of a Moving Blade Row." J..Ing.
Power, Trans. ASME. 97:131-139. 1975.

92. Gustafscn, B. A, "Some Observations from Low-Speed Cascade Tests Concerning Side Wall Boundary Layer
Suction." In: Secondary Flows in Turbomachinem. Paper 19 ACARD CP-214. 1977.

93. Starke, J. "The Effect of the Axial Velocity Density Ratio on the Aerodynamic Coefficients of Com-
pressor Cascades." ASME Paper 80-CT-194. 1980.

94. Stark, U. and hoheisel, H. "The Combined Effect of Axial Velocity Density Ratio and Aspect Ratio on
Compressor Cascade Performance." ASME Paper 80-GT-138. 1980.

95. Lakohminarayana, B. "Methods of Predicting the Tip Cleprance Effects in Axial-Flow Turbomachinery."
ASME Paper 69-WA/GT-26. 1969.

96. Lakshminarayana, B. and Horlock, J. H. "Effect of Shear Ilows on the Outlet Angle in Compressor
Cascades - Methods of Prediction and Correlation with Experiments." J. Basic Engra. Trans. ASME.
89:191-200. 1967.

97. Bardon, M. F., Moffatt, W. C. and Randall, J. L. "Secondary Flow Effects on Gas Exit Angles in
Rectilinear Cascades." 1. Eng. Power, Trans. ASME. 97:93-100. 1975.
98. Smith, Leroy H., Jr. "Casing Boundary Layers in Multistage Axial-Flow Compressors." In: "Flow
Research on Blading." Amsterdam. Elsevier. pp. 275-304. 1970.

99. Strazisar, A. J. and Powell, J. A. "Laser Anemometer Measurements in a Transonic Axial Flow Compres-
sor Rotor." In: "Measurement Methods in Rotat~ng Components of Turborachinery." ASME. pp. 165-176.
1980.

ii

'I
 203

WEINIG 1935 (16) CONSTANT 1939 (10)

HOWELL 1942 (11)

HOWELL 1948 (13) HOWELL 1945 (12)

CARTER-HUGHES 1946 (14)

CARTER 1949 (15)

HERRIG-EMERY-ERWiN 1951 (18)
HERRIG-EMERY-ERWIN 1951 (19) BUNIMOVICH-
FELIX-EMERY 1953 (20)
ANDREWS 1955 (21) SCHLICHTING KORAROV SVYATOGOROV
HOWELL 1944 (22) 1954 (17) 1956 (24) 1956 (25) )

LIEBLEIN 1956 (23) BUNIMOVICH-SVYATOGLROV 1967 (26)

NACA/NASA NGTE CORRELATION USSR CORRELATION

CORREMATION APPENDIX A APPENDIX C
APPENDIX B

Fig. l Pattern of development of deviation/fluid turning angle correlations.

Numbers in parentheses are reference numbers

MAXIMUM THICKNESS, t

a = LOCATION OF MAXIMUM CAMBER

b = COORDINATE MAXIMUM CAMBER
M,-
c = CHORD LENGTh
d - LOCATION OF MAXIMUM THICKNESS
LER - LEADING EDGE RADIUS
TER - TRAILING EDGE RADIUS
X1 - ANGLE BETWEEN LINE TANGENT TO BLADE SECTION CAMBER LINE AT LEADING EDGE AND
CHORU LINE, DEGREES
X2 r ANGLE BETWEEN LINE TANGENT TO BLADE SECTION CAMBER LINE AT TRAILING EDGE AND
CHORD LINE, DEGREES

Fig.A-1 NGTE blade section profile terminology (References IIand 15)

A
204

U
U

- U

ii

),.C Ui U LL

II DI

'I 1 %
aYJ
CC W
a.) Ix L j.

ziz

NKbb
 205

MAXIMUM THICKNESS, t

b ORDI TER

-a b, CAMBER LINE

S~d
C

a - LOCATION OF MAXIMUM CAMBER

b = COORDINATE OF MAXIMUM CAMBER
C = CHORD LENGTH
d = LOCATION OF MAXIMUM THICKNESS
LER = LEADING EDGE RADIUS
TFR = TRAIL ING EDGE RADIUS

Fig.B-I NACA/NASA blade section profile terminology

02
/2

• 1p
V
,

FLOW SUCTION AXIA

, 6 = B~~2- K2 B=)+i-
NAK

Fig. B-2 NAgA/NASA cascade terminolog"y

206

SOL IDITY
1.0 65-SERIES SECTIONS 2,0
K 1.1 C-SERIES SECTIONS
V Io.O.OUBLE-CIRCULAR-ARC
i4 -SECTIONS 1.6

Ce 1.4

1.0
S0.8 /Po

2)

10 20 30 40 50 60 70
FLUID INLET ANGLE, 01, deg

Fig.B-3 Zero-camber deviation angle at reference minimum-loss incidence angle deduced from
low-speed-cascade data for 10-percent-thick NACA 65-series blades

1.4 ____-____________________

1.2

1.0

,v08
:0.8

c---0 .6

0.

0.4 I

l
I' I I
0 0.02 0.04 0.06 0.08 0.10 0.12
MAXI1I4.J-THICKNESS RATIO, t/c

Fig.B-4 Maximum-thickness correction for zero-camber reference minimum-loss deviation angle

S. ...... . ... . .................- -

 32 60

28- 56

24 52-
52

L16- 44~

3~12 40~

8 36
Lu L

4 32

*.j.......
28
0 4 8 12 16 20 24
DESIGN (ISOLATED-AIRFOIL) LIFT COEFFICIENT, COEFFICIENT, C
10

Fig.B3-5 Equivalent camber angles for NACA A10 camber line as equivalent circular arc
(see Reference 18 for camber line construction with CIO0* 10)

0.

0 10 20 30 40 50 60 70
INLET-AIR ANGLE, 01, deg

Fig.B3-6 Factor m,0= Iin deviation-angle rule

208

1.0

0.9-

0.8

0.7

0.6-

0.5-
0 10 20 30 40 50 60 70
INLET-AIR ANGLE, 01, deg

Fig. B-7 Value of solidity exponent b in deviation-angle rule deduced from data for
65-series sections on NACA A10 camber line

MAXIMUM THICKNESS, c

ER
" "T
ZLER

xf f CAMBER LINE
b

Xf = LOCATION OF MAXIMUM CAMBER

f =
COORDINATE OF MAXIMUM CAMBER
b =
CHORD LENGTH
Xc =
LOCATION OF MAXIMUM THICKNESS
LER LEADING EDGE RADIUS
=
TER =
TRAILING EDGE RADIUS
X =
ANGLE BETWEEN LINE TANGENT TO BLADE SECTION CAMBER LINE AT LEADING EDGE AND
CHORD LINE, DEGREES
X2 = ANGLE BETWEEN LINE TANGENT TO BLADE SECTION CAMBER LINE AT TRAILING EDGE AND
CHORD LINE, DEGREES
1 = LENGTH OF CAMBER LINE
= MAXIMUM RELATIVE THICKNESS

Fig.C4l USSR blade section profile terminology (Ref.25)

 2

I I'
b

v t

/
S~~FLOW SSUCTION AXIAL

Ct SUFACEDIRECTION""
, SRESUR
t
S•URFACE
IF/ DIRESSION

A 2 1 FLOW aj SUCTON
AIAL

DIRCTIO
SURFAE ii

PRESSURE
 210

IC
CIL, 4 co

to4. 0'

~~41
4 to
4** .* .ft 4 441.

.- 4 4 -4 to..
.~~4
44 41* 1W 41 ý41 4
44. 41. 41.4'
144

"toJ'03 4 ~ .4 ,4 - .

44W
4
~~
4.414 ~ ~ .4 ~
1.44
44 ~ . 4' W w.

m 00
4.4 44 4'4 4..1

4" 'A4 '4 W. 1 . "4

0 00 0 A" $00 .4

-4 14

4414405
0~~~

4'.4

o~4.4
~~6 ý4 0'u u'
4'~"~' .4 ..ow '
'0 '
 ----. -. ....... .- .•_- -

211

022

W2 M,2, 2 •
U I **•
• v*u2
V ul

Ve,2

I Fig.E•-I Rotating Cascade velocity diagram components

i,.
212

11.2.9 AXIAL COMPRESSOR STALL AND SURGE

INTRODUCTION
ALthough the evaluation of the so-callod surge line of compressors was expliciteLy excluded from the
original task ifthe Working Group due to the fact that it would constitute a full task in itself, it
was requested at a Latar date that a short discussion on this subject be introduced in the review of
the state of the art,
First, the name of steLL Line or surfe Line is improper and confusing. Stall refers either to a
critical drop in performance or to some kind of steady or unsteady flow separation from walls or blade
surge is characterized by Large oscillations of mass flow rate. Neither are sufficient to define the so-
called "surge" or "stall Line", which in reality, is a Line defined by couples of pressure ratio-mass
flow values which Limit, for each speed Line, the region of safe operation of the compressor.
In the vicinity of this Line, the flow changes from a stable generally axisymmetric configuration
characterized by a large "sound flow region" and boundary Layer type flows near walls and blade surfaces,
to another stable configuration, with Large low flow or zero flow zone which can be axisymmetric or not,
steady or not. The change is made by passing through a transient (unstabLe) regime, characteristic of
the compressor design and mode of operation, and determined by the axisymmetric original, flow behaviour
near its Limit of existence.
Roughly, the second stable regimes can be described as follows when the rotational speed is increased
part-span or "small" rotating stall appearing in the front stages (without Large modification of the
overall characteristic), which, for smaller mass flow degenerate into "full span" or "Large" rotating
%talL, affecting in a Large way the characteristic of the stage concerned, and the overall compressor
characteristic ; increasing speed one can then meet a "wall stall" or annular zone of Low or zero flow
covering a non negligible part of the span near the end walls, flutter generated by separation on blades,
)
then again waLL or blade steady stall, Large rotating stall in the Last stages and again flutter (in the
transonic range). This wouLd be typical of an advanced highly Loaded jet engine type compressor. ALL the
type of stall or aeroeLastic effects are not necessary met for all compressors.
After that, there is a "system response" to change in flow nature, which depends on upstream and down-
stream volume, throttle characteristic. etc. One typical response is surge. The response can be affected
by inlet distortions, as well as the form and nature of the second stable regime. The Logical process to
determine the Limit line of "safe operation" should then be : evaluation of the unstable regime of the
nature and characteristic of the "second stable regime" and of system response. This type of approach
is now beginning to be applied.

Reference papers
An excellent paper, summarizing the present knowledge of the subject is about to be put-Lished by Greitzer(l).
Important progress is being made in the determination of instabiLity appearance and characterization.
Ferrand (2) for instance has developed on an original idea by Fabri a general method to treat effect of
distortion, the appearance oa stall or sur2e, and is extending it to flutter appearanc,. A most remarquabLe
series of studies have been Led jointly at MIT and at the WhittLe Laboratory by Greitzer, Cumpsty, Day et
al (see (3,4, 5) for the characterization -through reasoned correlations- of rotating stall, instabilities
and especially system response. Their efficient approach is simiLar (but much improved) to that of
Yershov (6) dating from 19601p whose work was unhappily unnoticed for a Long time. Additional significant
papers are those of Takata and Nagano (7) and Oo'ner (8).
State of the art
References (2) to (8) open the way to the logical approach outlined above. They cover one-dimensional and
quasi 3-b approaches, and work being done on 3-D approaches. The use of sound 1-D models has already aided
iiiimproved optimisation at the design stage, after having been either calibrated by a Limited number of
experiments, or validated by their use in parametric studies.
In general practice, however, the so-called "surge Line" is either predicted from experience and extrapo-
Lation of previous data, or by methods which do not consider the change in flow regime, but when calibrated
against experiments, gi'.,efor particular forms of design the desired limit. Such methods are discussed in
more detail in (1), which also contains an extensive bibliography.
One such method consists in identifying the surge Line through the use of a Limit Loading, expressed either
by a pressure rise, or more often, through a Limit diffusion factor or a conventional Loss increase. This
is done either on a one-dimensional basis (mean Line approach) or in the scope of a through flow calculation
method.It is impLiciteLy assumed that all problems arise from blade separation. The more real justification
is that such Limit Loadings are rough indicators of the stability Limit, which has been shown (see (2) for
instance) to strongly depend on the Loss gradient in function of incidence or Mach number. Additionally, the
criterion is Local and appLicabLe at most to a blade row characteristic. Thus additional criteria have to be
defined to see when the whole compressor characteristic is affected (by stage stacking techniquesthrough
flow calculation or otherwise). Such an approach cannot give rise to general correlations, but is quick,
extremely useful and accurate enough when applied to forms of design for which they have been calibrated.
The most satisfactory correlations are however proprietary.
Another method of the same type, but probably more general is to use, either with a mean Line or a through
flow calculation, a Mellor type method for pseudo-end wall boundary Layer:when reducing mass flow, a sudden
growth of the end-wall boundary Layer is noticed, and can be taken as a Limit for the stable regime (it is
assumed that the Limit of safe operation is due to annular stall) (Hirsch private communication).
 213

CONCLUSION
The so-called "surge line" evaluation is an extremely complex area, which deservte in itself full attention.
The intention here was only to define the problem.

In summary, the existing methods of prediction do not take the real physics of the flow into account, and
cannot explain or consider or identify all the parameters of importance. They are however quick and useful
for the range in which they have been calibrated, but are mostly not available in public Literature.

A more generalized use of the pseudo-end wall boundary layer approach seems interesting.
Satisfactory solution, will come however only from the Logical approach of instability detection, second
flow regime aod system response characterization, for which solid elements are now in hand but require
further development.
LIST OF REFERENCES
(1) Greitzer E.N. : Review-axial compressor stall phenomena.
J.Basic.Engr.,Trans. ASME ; 102 : 134 - 151.1980.
(2) Ferrand P. and Chauvin J. Theoretical study of flow instabilities and inlet distorsions in axial
compressors. ASNE Preprint 81-GT-211.
(3) Day I.J., Greitzer E.H. and Cumpsty N.A. : Prediction of compressor performance in rotating stall.
ASH J. Engr. Power Vol. 100 pp 1-14, Jan. 1978.
(4) Greitzer E.M. : Surge znd rotating stall in axial flow compressors.
Part I and 1I, ASME J. Engr. Power, Vol. 98, April 1976.
(5) Cumpsty N.A. and Greitzer E.M. : A simple model for compressor stall cell propagation.
G.T. and PDL report 148, MIT, Feb. 1980.
'6) Yershov V.N. : Unstable conditions of turbodynamic rotating stall.
Foreign Tech. Division Translation, Wright Patterson A.F.B., Aug. 1976.
(7) Takata H. and Nogano S. : Non Linear analysis of rotating stall.
ASNE Transactions 72-GT-3, 1972.
(8) Oiner N. : Rotating stall in axial flow compressors.
Von Karman Institute lecture series "Unsteady flow in turbomachinery" (1979).

6
214

II.?.1O SUMMARY OF ANSWERS TO THE QUESTIONNAIRE

1. INTRODUCTICN

To ascertain better the state of the art, a questionnaire bearing on particular points raised by the general
survey paper of Hirsch (1) was prepared and distributed in industry and research groups, through the
national representatives of the WG.18 questions, divided in 3 categories (compressor Loss correlations,
12 questions - l.ow turning correlations, 3 questions - Secondary and CLearance effects, 3 questions) were
asked. They are given in Appendix 1. 14 answers were received from six countries, covering a typical cross
section of aeronautical and non aeronautical industries, research establishments and university Laboratories.
(The List is given in Appendix I1). Additional comments were received from N.G.T.E. and Rolls-Royce Aero
Engine Div. (Bristol). Contributors are to be thanked most heartily, both for the time spent and for the
extremely useful information provided, which has proven to be useful not only in assessing more accurately
the state of the art but also in orienting the work on the WGand in shaping its conclusions.

A summary of the answers is presented below.

2. SUMMARY AND COMMENTS ON THE ANSWERS

The general impression conveyed by the answers is that the state of the art does not satisfy the users who
have to complement the information provided by non-universal corrections. The shortcomings are particuLarly
important when the correlations are not based on a sound understanding of the relevant physical phenomena.
The need for a quick assessment of performance at both design and off-design remains essential., and does not
ease the task of representing correctly the physics of the flow.

The general consensus appears to be that progress will be considerably helped by systematic use of numerical
experiments from which to develop more soundly based correlations (for instance, blade to blade calculations
for a better assessment of Losses and turning) or by the inclusion of fast, integral numerical method of
general application in 'he evaluation process. An example of the possibilities of the first type of approach
is given in the paper of Koch and Smith (2) , while the method derived from MeLLor's concept of pseudo-end
wall boundary Layer (3) typifies the second one.

The answers also show that the use of ttVrough flow calculation for prediction of off-design performance is
not fully generalized partly due to the Lack of information in suitable form on Loss and turning away from
the design point.

More details are given bel.ow.

A. Compressor Loss correlation

General remarks

The concept of diffusion factor and its correlation to Losses is universally used at design point,
introducing velocities and not angles, in its formulation. Its use for off-design prediction is not
generalized and often questioned. Three questions implicitely arise z which of the many forms of diffusion
factor (4) to use, which relation between D factor and momentum thickness applies at the TE, and relation
between the Latter and the Loss coefficient.

For the first question, as suggested by one of the contributors, the interest Lies in a form as general
and simple as possible, but which gives a correct measure of WMAx/WTE, i.e. represents correctly the
Level of diffusion.

Note that all the expressions of diffusi.' factor have a common deficiency for off-design purpose : they
neglect the pressure side effect which is important in the Low incidence range.

The answers are favourable to the type of approach followed in (2) which, however, still need additional
experimental cross-checks.

For the second question,asin(2), the effect of the flow and bLading characteristic (Mach, Reynolds,
incidence, etc.) should be better evaluated in terms of trailing edge momentum thickness. In actual
practice either the value of diffusion factor is changed, while maintaining its relation with the Losses
or the D factor definition is kept and the evolution of momentum thickness with it, changed. Sometimes,
both approaches are used simultaneously, and care should be taken to avoid accounting twice for the same
effect.

The third question: several formulae express the Loss coefficient in function of trailing edge thickness
outlet angle and trailing edge momentum thickness, Leading at high Mach number to differences in Loss
coefficient of the order of 100 %.

Data used for assessing the relation between D and losses, and thus for pe;frmance evaluation is for most
responses, cascade data (straight or annular) corrected or ,oot 6y compressor oata, and applied mostly on
conical surfaces. However, more and more the use of blade to b[lade calculation which car. take into account
real machine effeuts, such as streamtube area and radius changt is used and/or advocated, either to derive
correlations, as in (2) where it was used to calculate the diffusion factor expression, or to get directly
the information.

Two remarks are in order : first, using compressor-corrected data (Like a change of Loss curve in function
of radius) implies that all effects (area ratio, secondary flow and clearance effects, etc.) are
impLiciteLy incorporated and indistinguishable from each other. The validity of the correction is thus
Limited to similar designs, and in many cases is coherent only in a given through flow method structu'e
as such a method is frequently used to reduced the compressor data. Second, the blade to blade method
becomes cheaper and cheaper to use and more accurate, for all domains of operation. However, they are
 215

still generalky unable to daL,with very few exceptions,with the :ases of separated flows, be it incidence
or shock induced.
It seems possible to continue to use the diffusion factor concept in the transonic-supersonic range, and
also at off-design but only by modifying the Loss correlations.

Shortcomings

- The need for simple means of predicting both Low and high Reynolds (and roughness) effects was
emphasized. Hence, papers on the subject are included in the Wr report.

- The information on Maoh number effect is insufficient. There is a need for a simple "Loss critical Mach
number" evalustion, i.e. the inlet Mach number above which Loss grows rapidly due to shock presence,
but doubts are raised about obtaining a general single formulation. BLade to blade calculations are
advocated again here. Shock Losses models in use (1 or 20) and evaluation of interaction Losses are
deemed insufficient, and valid mostly at design and for particular bladings. Cost of experimental cascade
testing in time and money, for that range is deemed u.npracticaL. BLade to blade calcylation is advocated
but the problem of separated flow remains.
- Effect of trailing edge thickness appears to be important for high performance multistage compressors,
(blade row interferenceO and is not well understood.

- 30 effect and rotation effect must affect DF correlations but are not well ascertained.

B. Flow turning correlation

- There is a need for a Mach number correction to a Carter type formuLation, above the critical one, and
obviously for supersonic exit Mach number. Use should be made of the super-criticaL turbine experience.

- Except for very Low Reynolds number, the influence of the Latter is accepted to be small.

- The effect of axial velocity or axial velocity density ratio on turning is evaluated very dlfferentLy
from negligible to important, with an agreement of the fact that no good "out-house" correlations exist.
The effect, which is both of the "potential" and "viscous" type -the Latter at or near separation- couLd
be assessed by blade to blade calculation.

- Opinions on the need for correction in function of incidence in the non-separated region are again very
different from one response to another. For those who deem it necessary, the Linear (NASA-type) approxi-
mation is not sufficient. This point although it may be not of first importance needs cLaraficiation.

Some answers rightly indicate that the Larger source of difference from the prediction of angle by the
Carter-type correlation is to be sought in secondary :Low and clearance effects which are generally not
taken into account explicitely.

C. Secondary and clearance effects

BLockage evaluation is usually made by some sort of end-viLL boundary tayer calculation. Opinion on the
evaLustion of Loss and turning corrLtion due ' -j secondw;-y and clearance Loss tends to a combined use ot
experimental correction and pseudo-end wall boundary Layer ipproach, although the Latter is seen to be the
best way, once improved upon. Tip clearance effects are mostly taken into account by an efficiency
correction only.
three remarks are in order

- As pointed out in the answers, in the present state of the art, blockage and deviation corrections cannot
be chosen independently. They have the same effect on the accuracy of over&LL performance prediction by the
through flow approach. The good results are, in someway, achieved by a "coherent compensation of errors"
calibrated to get the right results. This indicates that the real physics of the phenomenon are not yet
included in the Loss and deviation subroutines, as shown often by the discrepancy between calculated and
measured radiaL flow distribution.

- The Mellor-type of approach has the potential of accounting for not only global, but Local secondary and
clearance flow effects. These developments, especially that of Hirsch as reported elsewhere in the WGreport
and at EcoLe Centrale de Lyon, constitute a major step in a correct flow modelling, allowing a proper
separation of effects.

- Their main shortcoming might be in the anglecorrection evaLuation, outside the pseudo-end wall boundary
layer, where the effect of clearance and secondary flow is still felt.

Lakshminarayana's approach (5) has been applied successfully by one of the industrial participant, and is
being further developed at EcoLe Centrale de Lyon.

"i~
216

APPENDIX I
List of questions
A. COMPRESSOR LOSS CORRELATION
____Insbsoic_______ eL below the criticaL Mach number the f'low Losses (blade friction Losses)
at desig are accepted to be a function o diffusion fator "U"'or e uivaLent diftusion factor "e u. DF".
ARFI~
idcaorSoudbe an~dd
to identity the 'LOW reiehrelses nces strongLy.

A _2 _______ g compressor btading do you use cumpressor data or cascade data or both ?
A.3.Arecascde ataworthwhile o ~ce design on cnaLsurface orisccaepoctn

A.4.
s th hat
v~ld;;,of S of L.H. Smith (s. Hirsch paper) compared to the
__________________ eoveumr generalizod fo arirr Lade shapeS and cascad

A.S. What are the effects of

- 3 0 ftLow phenomena
- streamline curvature
- system's rotation (influence of strcam~ine shift and boundarr'-Layer centrifugation)
on the
diffusion factor concept7
A.6. Influence of Re-number :It is recommended that a correlation for the critizal Re-number
indicatingj prsence of Laminar separation bubbles shouLd be deveLoped as well as one for' -aRe.W-mbar
n______ is n o the separatit-,i bubbLes. The continuous variation of Losses in tuncti-4n of Re-
number isconsidared to beof.econdar importance. Another bIG s planned to Look at this point in more
det~iL However, commert on your basic experience.
A.7. What is the vaLidit of diffusion factor (aLso egu.DF)i at off-design ?
Additional7
ine'raiW~~~~ empiricaL constants must Be obandfom axitn tests or from
ralc-u =aon-s. Thn -correlationsshould also be extended for small (5 to 10 9 chord)_separated zone.
A.&. Is trailing edge thickness accepted to be of seconder/ importance ?
A.9. Inthe transonic f'low regime additional tosses occur due to the appearance of shocks.
ran the profile ls rdict ion any_ Longer be based on diffusio factr?
A.10, What is the validlity of 2D shoc, models tot Loss prediction in compressor blade rows at
design and ol`f-dUý'ý
gn ?

A.11.
'spaxia ratio to bo considered for shock Loss prediction besides its
effect on tt- MahTF7uberhai qcCk ?
A.12. Is there an additional fraction of tosses to be taken into account for shock-boundary-layer
interaction and boundr-ae ?~orto

B. FLOW TURNING CORRELATION

8.1. Ourrent deovia-tion nrt'te prediction (based on soLidity ,thicknes,tcamber,s aggr angLe and
blade shape) cov ri prim rL;' mietr'icaL parameters and potential' flwefcs an insvai t design point.
ShouLd the methods be f~xtended to account for roughness, Mach and Re-number ?
8.2. InfLuence of axial veLocit density ratio and of viscosity (with and without separation) must
be in,Luded Can you give reco edt'os~on~~'
8.3. Is correction in function of incidence frthe non-separated region important ? This can
probably be answered by the people who make calculatios of test cases by comparnrsutobiedwh
and without this correction.

C. SECONDARY AND CLEARANCE EFFECTS

C.l. Secondary f'Low corrections ar ?,imotant with respect to blockage an& clearance effects.
Are there vati býoWg corltosAvAialfo all types of compressors mdC. multistage achnes?
PLease comment on MeLLor's methid of end w-allbo-u-ndaylyrca ltos

C.2. Should the Loss and turning correction be approached by correlation or by end-wall boundary
Layer computation or by Moth ?
C.3. How do you correct for the tip clearance effect 7
 217

APPENDIX I1

Organisation$ invoLved for the answers to the questionnaire

1. Nuovo Pignone, it.

2. NGTE U.K.

3. Rolls Royce, Bristol U.K.

4. Genova University It.

S. Centemeri it.

6. "TU Go.

7. T.H. Darmstadt Ge.

8. GHH Ge.

S9. BBC Ch.

10. Ecole Centrale de Lyon Fr.

11. NASA USA

12. Novack USA

13. Serovy USA

14. Detroit Diesel ALLison USA.

Comments were aLso received from NGTE and RoLts Royce, Great Britain.
218

CONCLUSION TO THE "REVIEW PAPERS" AND "ANSWERS TO QUESTIOrUNARE"

1. The concept of the basis for the correlation on tosses and turning dates from the period 1953-1958
as well as the basic corresponding experimental information. Application to quite different types of tasks
and of geometry has been made by extrapolation based on a very Limited number of additional data. As a
result the information is not always sufficiently based on a correct ktiowLedge of the physical processes.
The experimental data can also be questioned as all flow parameters were not always properly controlled.
Especially, data at off-design is insufficient.

2. The industry although using more and more sophisticated approaches - including fully' 3D calculations-
still has a strong need for a simple but physicaLLy-fnunded approach of the correlation type, and favors the
use of the diffusion factor concept and of the Carter's type of approach for deviation rules -wher; applicable.

3. It seems highly desirable, in the frame of the existing through flow approaches, to separate the
main flow and end wall (including clearance) contributions to Loss and deviation, the only sound alterna-
tive being a 3D approach, which is being developed but still too costly and has not enough good experimental
data to back it up. The practice of correcting blade element correlations globally for "all other effects"
Lacks of generality as it is concealing in a recipe, several phenomena whose contribution is very much
function of the compressor design.

4. For pure blade Loss, the concept of diffusion factor based on a boundary Layer approach is sound
in a wide range. Many forms of diffusion factor to eylst, as well as several correlation between Losses
and diffusion factor. A form has to be found which expresses properly the WMAX/W2 ratio , for the widest
possible range of flow conditions and bLading geometries. The relation of the ditfusion factor with the
momentum thickness should incorporate the relevant variables. An approach of the type followed by Koch
and Smith (2) which combines experiment and blade to blade calculations for its basis is promising,
9
although further checks (some of which are carried in the scope of the WGactik-Ities) are needed.

5. The nrincipLe and the bulk of the information on the deviation correlations date from 1945-1960.
Even disregarding the secondary and clearance influence, they need up-dating , especially in function of
Loading, Mach number (especially above critical) and incidence.

6. Information on the Re-number effect have been added. Robert's approach for the Low Re range and
Fottner and Schaffer's for the high Re number range are recommended.

7. Existing global models for shock and shock-boundary Layer interaction are insufficient, and
applicable only to a Limited number of geometries and flow conditions. Prediction methods for critical
and choking Mach number are not general enouS:i.

8. It is expected that to remove the shortcomings mentioned above, extensivo use of the blade to blade
calculations (including boundary Layer) backed by a few chosen validation experiments will be madn. The
main Limitation of those methods is their present inability to predict pqrformance in the separated regimes.
An assessment of the value of the best existing method in their use in numerical, parametric experiments
is needed, and is partly undertaken in the WGwork.

9. A Limited amount of information on part-span damper losses has been reviewed and presented. Two
approaches are given, the second model underestimating the Losses by quite a Large amount, the first one
is essentially empirical. Data on geometries is insufficient to draw conclusions.

10. Secondary flow ens clearance effects should be evaluated by a Mellor type of approach, whose most
developed versions have the potential to provide both global and Local information, when coupled with a
good through flow method. Angle correction outside the pseudo end-wall boundary Layer needs to be taken
into account. A method of the type described in (5) is suggested.

Experimental correlations in this area do not properly account for the real flow phenomena and their
improvement is not deems,! useful. If a quick, global assessment is needed, the simplest one should be used
(see conclusion of reLevaht paper).

LIST OF REFERENCES
(1) Hirsch, Ch. :Axial compressor performance prediction survey of deviation and Loss correlations.
WG 12 - Survey paper.

(2) Koch C.C. and Smith L.H. : Loss sources and magnitudes in axial flow compressors.
ASME paper 75-WA.GT. 6.

(3) Do Ruyck J., Hirsch Ch. and Kool P. : Investigation on axial coipressor end-wall boundary Layer calculation.
Preprint 1979 Int. Gas Turbine Congress and Exhibition. July 9-11, 1979. Technion, Israel.

(4) Dunker R. and Weyer H.B. : Survey of diffusion factors and profile Losses.
Through flow caLcuLatioi in axial turbomachines. WG12 Report Chapter 11.2.1.
(5) Lakshminarayana B. : Method of predicting the tip clearance effects in axial flow turbomachinery.
ASME Journal of Basic Engineering, Sept. 1970, pp 467-482.
 219

11.3 PRESENTATION OF TEST CASES

Three test cases, representative of actual designs, were submitted for the Working
Group; a single stage, a two stage and a four stage compressor.
The single stage compressor, developped and extensively tested at DFVLR was already
used as test case for the PEP-47th Meeting (1). Classical as well as laser measurements
inside the rotor are available. Appendix AII presented by H. Weyer, contains the
geometrical data as well as performance mapsand some traverse obtained with conventional
instrumentation. Appendix 3.1 gives also the coordinates of some blade sections at 45
span and the corresponding blade to blade velocity distribution at various inlet condi-
tions, in cascade configuration and in the actual rotor.
The second stage compressor, submitted by NASA Lewis is fully documented in the
literature (2), NASA TP 1493, August 1979, and is not introduced in this report at an
appendix. The report includes measured as well as calculated data. It is to be noted
that the measurements are performed with conventional instrumentation in the following
way. Total pressures and temperatures ar.d tangential flow angles are measured at stator
outlets along circumferential traverses at fixed radii. The stagnation pressure and tempe-
ratures are mass averaged while the flow angle is arithmetically averaged. Moreover, a
static pressure is measured in the middle of the passage.
All other variables at stator outlet and all variables rotor outlet are calculated
along the design streamlines and for the design values of the end-wall blockage factors.
In particular, the average total pressure at rotor outlet is taken equal to the value
measured at stator exit outside the wake.
This two-stage fan has a low aspect ratio blading particularly in the first rotor,
aspect ratio being equal to 1.56.
Several new design concepts are introduced in this rotor, such as elimination of
the part-span damper blade maximum thickness is moved rearward, allowance for inlet tip
boundary-layer by giving to the rotor a leading-edge end-wall bend. This resulted in
avalue
peak-efficincy of 0.846 at a pressure ratio of 2.47, even higher than the design
(2.4).

This test case is quite challenging with regard to performance prediction, espe-
cially at off-design as illustrated in the following chapter on the calculation results.
The four-stage compre3sor has been kindly presented by Brown Boveri & Cie./Sulzer
and is typical of an induetrial multistage configuration.
The geometrical data iire presented in appendix AII.3.2 and the data include full
performance maps as well as radial traverses or stagnation pressures and temperatures
at the compressur outlet, for 6 operating points along the nominal speed lines. No in-
formation is available with regard to the accuracy of the measurements.

It is to be noted, that due to space limitations, only a summary of the geometrical

data and of the results of the single stage compressor are published here. More exten-
sive information is available, on request, from the Propulsion and Energetics Panel at
AGARD headquarters.

List of References
(1) AGARD-Conference Proceedings, CP 195, "Through Flow Calculations in Turbomachines".

(2) D.C. URASEK, W.T. GORREL, W.S. CUNNAN; "Performance of Two-St.ge Fan having Low-
Aspect Ratio First-Stage Rotor Blading", NASA TP 1493, August 1979.
 221

Appendix AII.3.1 Single-Stage Transonic Compressor and Equivalent Plane Cascade

Introduction
The Wa 12 activities cover primarily turbomaohinery off-design performance predio-
tion, however include also blade-to-blade calculation on relevant test cases to get more
information on those basic flow phenomena that affect flow losses and turning in blade
rows.
One of the test cases selected io t•:e DFVLP single-stage transonio compressor that
has been tested in great detail with ,ie :,esu'tt well documented (1, 2, 3). Particularly
intrablade velocity data of the rotoI Are ava. .bl,, from extensive flow studies with
laser velocimetry. The data allow to compare blade-to-blade calculations to actual tran-
sonic compressor flow. Additionally to the votor tests the rotcor blade section at 45 %
span was investigated in the DFVLR transonic cascade windtunnel (4, 5).
Single. Stage Transonic Compressor
The transonic compressor without inlet guide vanes was designed for a total pressure
ratio of 1.51 at a mass flow rate of 17.3 kg/a and a tip speed of 425 m/s. The isen-
tropic efficiency was estimated to be about 80.5 %. The rotor inlet diameter of 400 mm
was prescribed by the DFVLR axial compressor test rig. The stagm pressure ratio of
1.51 is predicted to occur at a temperature rise of 15.4 % of th inlet total value.
Fi I demonstrates the compressor annulus geometry. Hub and outer wall are
shapep-eadapt the flow path to the stage pressure rise, to achieve nearly constant
axial velocity over the annulus height, and to balance rotor and stator diffusion
)
factors properly. MCA-profiles were selected for the rotor blading from hub to tip.
NACA-65 profiles with a circular arc camber line were used throughout the stator
blade height as shown in FiR.2. 28 blades with a tip chord length of about 60 mm were
selected for the rotor, 60 blades for the stator yielding usual blade solidities
between 1.34 and 2.0 for the rotor and 1.5 to 2.4 for the stator. The maximum inlet
Mach numbers to rotor and stator bladtng reach up to 1.37 and 0.76, respectively.
The maximum diffusion fact•u i esttimtated to be 0.53 for the rotor and 0.48 for the
stator. Further spanwise blode data including the axial location of blade leading
and trailing edge are givin .n 'Jj and 4.

Instrumentation and Test Results

The compressor has been investigated u3ing both conventional and advanced mea-
suring techniques. The instrumentation planes just upstream and downstream of the
stage (Fig.1) were equipped with probe rakes or radial traversing probes to analyse the
spanwise distribution of the total pressure and temperature, and of the flow direction.
The compressor performance map - adiabatic efficiency and total pressure ratio veroum
mass flow - is shown in Fig.5.
The rotor flow field (ahead, within and behind the blading) was studied in great
detail using advanced laser velocimetry k6). These tests carried out at design
(20 260 rpm) and off-design speeds yielded quite complete information on the span- and
gapwise velocity profiles, on the 3-dimensional shock waves, on the flow separations,
and on the blade wakes. Up to 15 circumferential measuring positions over one blade
spacing were set in each measuring locus designated by circular symbols in i6
The diagram demonstrates that measurements could be performed also in the v fnchty
of the hub and outer casing walls. The complete geometry of the rotor and stator as well
the velocity data given in tabular form for the rotor blade section at 45 span are
available on reqjest from AGARD-Headquarters.

Plane Cascade

The plane ci3cade equivalent to the 45 % rotor blade section was designed by pro-
jecting the rotor blade coordinates defined on stream-surfaces on en axis-parallel plane
radially fixed by the blade stacking point. Thus, the cascade solidity corresponds to the
rotor blade solidity at the stacking point. The 45 % blade section has been designed for-
an inlet Mach numbqr of 1.09 assuming the upstream shock wave to be attached to the blade
leading edge.
The primary cascade design data are:
Blade inlet angle'): 55.10
Blade outlet angle"): 40.10
Stagger angle : 48.50
Chord length 90 mm
Solidity : 1.61
Aspect ratio : 1.88
Leading edge radius : 0.4 5 of chord length.

Wangles related to axis.

M~U1M PU~8~~BL".1ar ]MUMv

 222

Instrumentation and Test Procedure

As illustrated in Fig.7 wall static pressures were measured 63 % of chord ahead and
41 $ behind the cascade to control inlet flow conditions and cascade flow periodicity.
Downstream total pressure and air angle were measured versus blade pitch (mid-span) by
probes positioned 47 % of chord behind the cascade, to determine the cascade performance
characteristics. Detailed blade surface pressure measurements were carried out at the
following cascade flow settings:

1st run 2nd run 3rd run

Inlet Mach number 0.911 0.827 0.813

Inlet air angle*) 64.3 58.5 54.5
Outlet air angle 45.5 46.5 44.9
Axial velocity density
ratio 1.249 1.098 1.06

*)angles related to axis.

During the tests the axial velocity density ratio (AVDR) was not controlled but
fixed by windtunnel side wall boundary-layer displacement.

Detailed geometric data particularly blade thickness distribution and test results
are available on request from AGARD-Headquarters.

List of References

(1) WEYER, H.B. Compressor Design and Experimental Results.

AGARD-CP-195 (1976).
(2) DUNKER, R., Experimental Study of the Flow Field within a Transonic
STRINNING, P., Axial Compressor Rotor by Laser Velocimetry and Compa-
WEYER, H. rison with Through-Flow Calculation.
J.Eng. for Power, ASME Series A, Vol.100 (1978).

(3) MC DONALD, P., A Comparison between Measured an Computed Flow Fields

BOLT, C., in a Transonic Compressor Rotor.
DUNKER, R., ASME-Paper No. 80-GT-7 (1980).
WEYER, H.

(4) STARKEN,H., Invetigation of the Axial Velocity Density Ratio in a

BREUGELMANS, F., High Turning Cascade.
SCHIMMING, P. ASME-Paper No. 75-GT-25 (1975).
(5) SCHREIBER, H.A. Untersuchung des geraden Verdichtergitters L030-4 bei
schallnahen Zustrdmmachzahlen.
DFVLR-IB 352-79/10 (1979).
(6) SCHODL, R. Laser-Two-Focus Velocimetry (L2F) for Use in Aero
Engines.
AGARD-LS-90 (1977).

"*1
 223

Annulus Geometry

zZ
NABE
IHub)
OEHXUSI
(3hrou)
NABE
(Hub)
R
OE9KU83
(Shroud)
N
0-0 0 901000 201,870 as 106,073 19,1179
2 90,083 3 02,000 90 "6,638 195,?7
4 90,000 20,730 92 107,204 195,1?8
6 90,091 202,01. 94 107,770 194,872
8 90,159 2021000 962 18,336 194,570
36 90,200 201,578 98 108,902 194,267
3& 90,349 201,400 100 109,468 193,965
14 90,670 201,392 102 IO,04 190,662
16 90.611 201,993 106 110,600 190,260
18 90,770 201,917 106 I1l,655 193,068
20 90,947 201,904 108 111.725 192.Z5
22 •, 91,181
91,354 201,q3
201,914 d-D0
a 10935
Ito 112,121
1•2) 192,527
19.5
.95,7 201,873 112 112,743 192,149
S28 91,835 201,834131,210 It 191,554
30 92,099 201,785 116 113,64
5 19.568
32 92,3 201,723 118 114,046 191,296
34 92,684 2ol,655 120 11.420 191,51,?7
3
ff0 93,339
93,693 201,92
ZU1,395 1
122 117,260
114,7b9 190,771
190,771

42 94,o65 201,89 126 117,308 190,291

84 94,4 201,172 168 117,650 190,066
46 94,859 201,0 160 115,902 189,852
48 95,283 2W ,904 132 116,127 189.650

me 96,131 200_590 13_ 116.514_,_ I_9__

56 97,147 2uij,•22 136 116',7
55510
58
6o 97,655
98,18U 200,021
199,804• 40[ 116,823 188.q•2
142 116.950 188,•.O2
62 9R.72 19'),574•,6 4 7o6
.12 ...199 .2 -9- 1 ,1
IA 9.5?6 149.•I 117,15A I98,5414
• , - 9 6, ,52 ) 19 9 b i 48 1 17 , 2 40 18 8 , 43 3
'§6
9,149 Ill9,i 150 117,_09 1_8,_5
i. 68 100.415 19m,792
tn t o.rIr
152 117.-s
72 1U1,545 19S, 19 O h-H 153 15 117.!FO• I •19.1"

76 Wu2,678 197,593 158 117,470 W8,%66

78 1U3,244 197,290 160 117,487 188,031
80 IO),MIu 196,9ý8 162 117,497 188,009
82 104,376 196,686 164 117,500 188,0¢0O
C 196a383o 117.
5v7-"I•.U 196 0Si 168 1 117,500 1 188,000

Instrumentation

Fig. I Annuhlus geometry DFVLR I-stage transonic compressor

 224

Stage No. 1 1

Blade Row Rotor Stator

Profile
Type MCA NACA 65

I Number of 28 60
Bladelades

Fig.2 Profile type and number of blades -- DFVLR I-stage transonic compressor
 225

4.
CN to
N (n 0 (n LO m m U
bo W 1- Ln * *- V *4mOD
C
0~~t to n 0
Ln 0 m 004
m 04

Pt d4P

W "0
W p C- CV) m r4 (D 4
rý 0 C(4 () 01 *r-4 CV) OD co 0 o mo
V) m1 C14 -t cl0 4.
C14. CYl 4D () c
oý a) aY to Ln.. m

4-)-
4) ~ L~
~ o~~~ ~o ~~ L -. l n 4.
_
C4 bfC 0 U') - CN mI (NI m~ (N U) (N (.0

4.-) 0) a a a a a a V . V
V "
N 0 '-( m mI (N (N
04 (0
(D
Q) cl ) r4 T-4 C(4 (Y) zr 4. Ln LO (0 Ln4

4<
4- * LO) LI O 4o
-t 4o
z O (.o
CA 00 f) LI

__ ) _ a)- - C L O O Cd)

(
_-r* LO) 4.) CO 01 U) 0O CO) (D 01 to

LOr- co c' V m n co LO a) LO

a~~Q I) 0)
(7 o 0 , C) IV) to C" CN 4. 0D
4Jr-N 0N CV) _4r L± LO LO) L( LI) 4.D C0Dd

4. 0 ) mI 0 N Lf) (
00 to C14C, LO V
co m4 r- to U)4 LI) N N (
C
0N1 %-A CIH 01i q-0 1 v4H- %-4 C4 4-)

-t mVC) tr. C W. LO T4VS- o

LI (0i r4 r- (N (V 4.O - L)
VI N D P
'0 4-) M. W,
U, U, ) %I4 (0 U, m~

0 10

q 4-)
3:

F (NC1
01-
OD
N
cl
CO C)
_-r
)
r.N
0) U'
L
4 (N--V
)
N
_r
(f
TV
4
0
(D)
0C

0 00 r-N- 4.0 (t i- C, a) co ( ~ (I
.rA Cat C,0 1
C C) C' C' a)a0)
01 01 01
4-) 'C ri r-4 '-4
V .- 4 T-4 r-I H.4
ft 4J.

N:r C14 CO (D N- (N U,) (f 0) r4 N- -,

(N4 N- 01 -40 C, 1-4 co (D to U,) CO
(4 *'- N 01 (N N, CY) 01 4t ) N- M) C4
0 IV bý Loa O r- OD a
OD 0) a toa

O) 10 Lo't (r, (C U, to (D U, Nl N- N- N
cy- x

)C, IC C c c c), 0 C. 0 C, C)0

z T- c C1 m _-r U' LO r- OD 01) C) 0
a) q4 - T4
I W- rl r-4 -4 rI T4 T'-C1
rt l
 226

oo b

Q) 0V ~

ru () C n .-

4) c.' ' 04 C'.J (N ('. C'. (V v

Z~ , Nc. to to to t M

T, C)r - n t

b,: C14 i- C4 co r-. to L) j-

w (N (N4
C(4
w

1- LA i V ~
LO.C- ~ ~ i *
i, m- m* C4 C'l CN
V)C
fo o 0 1:1o'

4
r-4 vi4 q-4 4 - - - i 14 0
4) r- _ - - - - - - - - - -0 -

00

~
0) ( ~ (L 4in (D cV)
a, 0) 0) ) 0)0 r-) 0f) =7)

C)4 C) C)4 CN r- vi V- T-4 0

1-4 (O c) t CA 00 (n 0- C04

0 C E - Y 7) () m( H
.aQH V N C 4 C4 C 1 1

,~ ~'u
 227

[• l O D e sig n

S1.7
----

100%
1.5- 92.S%

t,', IA ,as*/.

S12--N-
1.3 7
1.2-10 - ,

6 10 12 14 16 18 20
rnred kg/s ---

Fig.5 Performance map of single-stage transonic compressor

Ft

__-, ,HUB

. Axial Direction

Fig.6 Stage annulus and laser velocimnetry test locii

 228

I
i.

011
Po2 (7)

WI
I)
p Wall

Pl

I ~ Fig.7 Plane. cascade installation and instrumentation

I
 229

A.II.3.2. BBC/SULZER - 4-stage Transonic Compressor

Design speed 15000 rpm

Inlet total pressure .975 bar
Inlet total temperature 193.150 x
Design pressure ratio 3.0G

Test data available at following speeds (percent of design speed) 68,5 90 100 and
105 %.

The data were taken at the following inlet total pressures at reduced speed of
68,5 % Pt . 0.94 bar
90 % Pt . 0.60 bar
100 % Pt = 0.57 bar
105 % Pt - 0.48 bar
The overall performance map and some traverses of stagnation pressure and temperature
at compressor outlet are presented in Fig. 3.8. to 3.26.

Annulus Geometrv
Hub Tip

""m,
mm m R, mm

-156.8 130.00 274.00

-125.0 130.00
- 87.5 130.00
- 34.1 134.26
0.0 143.40
36.8 152.93
74.3 160.97
101.0 165.57
130.7 170.23
165.6 175.71
188.0 179.22
211.1 182.85
245.1 188.18
266.0 191.46
288.8 195.03 274.00
321.5 198.17 273.63
340.0 198.20 272.12
361.3 270.39
391.3 267.94
4 8.0 266.75
477.2 265.01
455.6 262.69
472.0 261.35
490.5 259.84
518.2 257. 58
532.0 256.46
547.6 255.18
622.0 198.20 254.00

instrumentation

Z=O Z

Fig. 3.8. Annulus Geometry - BBC/SULZER 4-Stage Transonic Compressor

230

Stage No. I1V 1 1 2 2

Blade Row Stator Rotor Stator Rotor Stator

Profile NACA DCA DCA DCA DCA

Number of 18 17 28 27 34
Blade

Stage No. 3 3 4 4 OGV

Blade Row Rotor Stator Rotor Stator Stator

Profile
Type DCA DCA DCA DCA DCA

Number of 31 40 37 43 36
Blades

Fig. 3.9. Profile Type and Number of Blades - BBC/SULZER 4-Stage Transonic Com-
pressor -

"i1

11
 231

CO n 04 0) 0 -t

fn cn Ln C1 0 al )

C) W) :- cr co to (D
M- C; 0)
C;C C.) C.) C;)

WC 0v C) 0) , 0n
00
0l -4- LO -t 0m c) cn

0) 0c C. 0) N

C41

4 o co n r- a

WC.) 0' cl I" o" C4i

- - - - - - - - - U)to.t
.,4 x
X 0
4) CDI t- ý U) Cý
0) N4 4
fd tc ) 0 n- c,. to C4 c) 0
0U) m) zr
C.- Ini to)
N W0
U) 4)'.

C4)
0 0A
1
c4 4 '4 -4 '-4 -4 4J U

.*-4 *r 4)0

4j a~) * ) N 0) N C U)

c.4) (71 kr 0) . ) U '

0bN co 00 C' co
C) N

1-4 0~* 'MC t

41 0 )

0 C') C'. N N

0x

-------------------------------------
 232

In ;4 O~ N to * 0 0
0 0 0 to) . C)
.,qe c In -- cý 0ý o Cý

tU) o' 0 0; 0ýC C)

x.0 40 m 0 C C
xO C 00

IV en C4~ C") )

__c 1 L1-
111 ,

m 0
Lo*) Lfl Cn H
N * N C)
U U) )
H U
'U

M4) he I') HD 0

CN - 1 ) *~ 0 InC
r-0 L4-'"0.
1 41
0.- 0 CJ cf U)I
0 4 0

4 ) 0

C0 M) U3 th
N-V4J F. tc C 04) C
0 r_~q 0) p toH

4 C) c$) 10 f4)
'U (0' H)

9Lna L G C1 0..

H.H C1 N1 C14 C1 CN C1 C14 0

4- U.U) U U )

0 r C) 4J 1

~ 0U

OD
00
tLcn
0t-
p) *')
cV 3 *IV4
'U H 'U

'.C4 4) 1ýC;4)L
 233

po ' 0 0 0 0 0 0
;4 C 4N 0 0 0 0 0
p r- 0 w fln LI) LI
'~.,-4 c 0" N C4 NI

01 0 0

CA 'a. LI) C 0 0
.0 0 01 co to LI LI LI) LI)
X.X 00 C- 0 0 0 0 0
r... o C'~ 0 0 0 0ý ;

LO C- C') N

LI) 0

CO4-4 C- m o-

I~~bO ~ (D ~ w H
N c: ' ~ i

4IVW 0) N0 N o *
W' O4 0
z'

Lo 4)i.
4 0
0
C-4~ 0 L 4 - " ( 0 (d) ;i
~In -V m -
*~~W 00 6.
-4 -4 -1 4J r, 6)
'V 04
NH~4 b L) r-

C) 0 9IU I) J f ;4 4) toL
C; a. r-4
ta- - - - - #m- - E.

Ln coLI C z E
(dbe 0' N 0
In) rN C

04 00 m e M en N 0

N~~~ ' N L ~ a 0'

0)4 "-4 CI enb )

or N H N HNNN
_ _ _ _ C..
 234

cor 0 0::0
0a0 0
N 0 0 0
'u w 0 0

:3 0 w O
*. 4.
A~

.4. ?4. S
rd 4

-4 1 0

4) )b

'A 4) av Z S I

f~~
L4
x 4
tics4 ~ 4

bC 4)
4Jr 00
I 0a3. ý4
_____ ) -40

0 4~
C
0 0

C) 4J4
44 0
4)0

0 0r 4

m Go -T C%4. CID

0 bZ cr n C o O o o0f
-r 4tz -4

4-1V E-4WH C
44
u

0 0o VO c4 4 ~ 4) U
'.4 uV 61 kn *n W 4

0 ~ d -t
40 _r zt N toN

'.4
0 9
 235

n I I L ID

w door o

~bC) N' C4*.1

inv~~
N ~r0wlN 0 N

- t44 c;N
So C 0 c C, c C
*4J

'o4 "bOe4 1

CO4J < i CD M 4m) cn m

w r- C, N n m

"-4 -r

-44 U

0 z C3 in -4

W$4 b!
M ' *C 0 * ( 4

4.)~ N c-4 in zr) Ln Go ) i

,j ~ c ko t C44 I.,

14 OE.-

w. V c i
4.J
N 0)a
N 4.1 4 0U

:t u: r t D . o in
ol+4 a. C
00
o m Lo 0 0

U m.J4 4

* ~~~~
O ~ J U i+1
0Cm m co a $4
r -q ) *N Nl r4 04 m 0 (n V
to1 4.0 C
4.1 ' w.

9 CIn
N r4j .NC n 0 N e V O41
to) w) 1' Na ,4 a
+j x 10 I

o~~ ' cob I

3
. b) r- & M ) 14
zr C4 =r r, Q
m
H. VVU
~ WNn
14 T4
" c-4 N cNm ( (

hV *4 'bk~~41) C) ,~ c )
236

(n n
L)~ N, 0 0) fLO o I

CC bCs Uj- 1 ý C
wV1 ( 0 C. a c 0 0 C)

4 dw

to L 0 (0 ,4 On C N- 0
w 0 LO s-I C- 1-1 ýl 0)
0-t~ 0n 00
o 0 0 COa

CO) 0 0

., 0) Cý C; * 0 )
C 4J

4) m) OD(0
( (

co
(0 i to N
M) n N 0
f4bC a) a; z C 0;V

:$C1 eq C14 CN C14 C-4n

_4. C' ) ;

*,r) e LO (3 - cn n 0
LD m V coL ) ) 4
+j r. N N N (' ' ' CSJ 0 I

4-4. $ 0) 1
4.' 0 m rl C 7 D ( n.
0 C04 t C400 0)
bZ Cý Cý 0)
0 41- r* U
4'~ W .

'V 4J c o EN C1 co 00 N *

C' *r al) (0
W '-4) 0

E NV ' 0- 4J*

H U"0_ N- 4 Hý C') (0 N 04 CL' 0

r4i
4 _ _

bt4-
f m -- - 4J w-
 237

H C" 0) CD) U) (n In
U)
to cn~ 0 ND w
P 0 =., C't
0 ~ C co
V0 14 4
.Cý 1
L)~ 0 0 C) C) C,

0) 'L' D U co m 4
V) -r
* C ) cn Nl
0) 0L _r Lo ol 0
to 0) 0I
0 C 0 0 0 0 0

4-' Go Lo N 14 0) LD
4) . ml 0n Cl ) c' -
H bo *ýuý1
4-' 4) co 'n
U) m' r,)
Z V10 Ný
04 C1 0 N H v4

HO)) co a -
Iz 9 N
C. zt to0
I-I * * * ) V) CC' Ln
U)

bý -4 *.-4 4-
beX 0
4Jr-'0 *y N.
ml 0)
cy C)
m'
*~~a' H U) C * I 0

V' C') (Y) m' CC) 0' V) V)

- - : - a) O-
~ -t OD
(1)
-O
t ~ ~
0
0
u C
4.3 CC- N! 0) C' 0 ~ .) 0)
C-~~~4 ) ~ CC * )0 I L)

CI) 4 0
-- - -t - -D -i - -' D.

.U 4J.4)

61 r co co 14 o F 0 4J

*c Cý C
m mDU )
0fa

tob ta N ' ) N0 u E N
+j u. C) U C- C 00
V) 0 I

C' N t" a) zr r* a,)

0 -- - - - '1 0 0
v~ ) 0 4-'
(7) " a)) a) cl
N Ho
< H) ) C N C

0n * *0

* o* L* e* m, w, 0
10 "a _ m z w o 4c3 @3W

H H' C-4 N- Nq N) *N N C
 238

'- 0 0' 0 Lo C U

0 0 ' C C ' .
'a0a).1 C c . .ý ý.ý v
0 0 C) C C) 0 0

a) 0 aj:s s
X.~U
to
4-'

4-' 4)
a) Ca)U

C)

Q))

x t
W 4a) *0
L k -4 E~ aS ?I S L'
Sa4 4.4
4J r- a0 0 H
U) ;4

F) 4j- O C n OD _ tI . 4
0 m 0
'4 04 N4 14 4-4C
0) 0

-t 1 o - 0l. w 0
'04 4 C(4 0) (0 (N C) to M,

'o( 'r_ c4 0ý 0
C)
m~ C') m' C') ;r -t
- ~~ -- ~4
- - 0a

m co r- OD c) Cr- 04 V 0 4J

U) (0c)C co ) 0

0 dW ul) f) Ur) LO U) U) U)

It ;.
4J w~
lC U) C I) I' U I I- I~ e

4- 0 a ) r - - wa wG 0'
rU) V)) a) U) U) ' U)
-4H
0,-

41 1 ) a)
Co U) [(' C) Q)

4-) x W1- In In ) (n 0) L0 u) a) to O

'0c C-- co a) ( ') U o)C

'04 4 -4 C-4 (4 (N4 ( 04 b6V

 239

- - -N - - - -

rz

4Jbr 0

+i C ' N1 r- C4 0 .

x.~ o

o 0)
E-..

vq 14 f, 4J r.0

N
+j
u
-4
4J 'r4N
V a~b~c~ a~0 c~

Sm

*~~ 4j~0

r-4 V 4 . - N Ný Ný V ,4

4) fa V

0 z 0 O D C4 0

C; ;

10 04 CA N 4 CN (14

I I
240

.H 0

4) 0 (D

c 41

0j 0g

1-4 ýr ý 1

Im co -1
z~- 1- 4 0
0 C-1
) v ) )

* n N
m' n m C') 0 I 0"

tita)
_ tc
c 04
r-

I.
cn

0
'

c
D

o to
)'

0
0)
0)
-41

0
I

411 W, o-

CA -t 10 C - 0 to
Nj .t0 0 C4 00

0
oi 4J-14
C) U r.
0 )

0 10 r V
I

>~~- r t = -4-

V) 41 E) 4-

0~ 4)

4-' rN N 0) t- Ur) m' N- , r t

-4 0)V0 4' 4C Ný N4 cJ N4 4
4-' -4 'o t: r 0 0 0 0 C7) c) 0-1 V r

0 < ~x I

0 wi ý C C 8 '-C4

0O w) 0) 0 44 r1
14 C', co C:
 241

4 %0
1,05
7t*

3 • oQ90 , '4

Q
0.685
Equivalence condition:
=.975 bar
PO:
= 293.15 K

15 20 25 rhkg/s 30
kg/

0,8
0.6 • 0*6-1

15 20 25 m.kg/s

Fig. 3.20 : BBC/SULZER 4-Stage Transonic Compressor, Performance map

 242

r
[mmi
260\\P
Total
P temperature
24 0 S
I.ý

Static " fTbtal

220 pressure J pressure '
a I, ,
200 7/' /.///
//2
010\, / r/

P4 N

1.5 2.0 2.5 1.3 1.4

P; Pi
Ip' T'Tro

Fig. 3.21 BBC/SULZER 4-Stage Transonic Compressor

Outlet Radial Distribution of Pressure and Temperature at a Pressure
Ratio of 2.12 (Point 1)

r
(mm]
i ~~~~~260' \I• \ \\ \\
PT
*r' \ P* T*"

240 PP
0 TO 11
Static I \Total i Total
pressurel" I pressure T temperature
220 - I
2 / \
2o00///• ' /// " '7773""'

2.0 Pjpo*-2.5 3.0"

P/IF5,To 13 1.4"

Fig. 3.22 z BBC/SULZER 4-Stage Transonic Compressor

Outlet Radial Distribution of Pressure and Temperature at a Pressure
Ratio of 2.68 (Point 2)
 243

r
[mm]
S~~~260-\_\4
%\ * \ \ \ \.

240-
P P* 11,
o1 PC. *0
Static \Total Total
S~~220
pressure,*
-J
• pressure r temperatur

200/,/ '// 7777

P4N
2.5 3.0 3.5 1.4 1.5
P~p* P/p* T•To

Fig. 3.23 BBC/SULZER 4-Stage Transonic Compressor

Outlet Radial Distribution of Pi essure and Temperature at a Pressure

Ratio of 3.02 (Point 3)

260\\ 4\ ,
P F* T
240- p
Static 0 Total Total
220 pressure I pressure temperatur

200 ,,,
Pr 7
77 7 7 7 7~
4N

2.5 3.0 3.5 1.4 1.5-

Fig. 3.24 BBC/SULZER 4-Stage Transonic Compressor

Outlet Radial Distribution of Pressure and Temperature at a Pressure
Ratio o0 3.19 (Point 4)

-At
244

r
[[mm] P
260\\/ G \ \/.LG

0 0 T
240 P T*
- I• PO
220 220Pressure
static Total ! Total
pressure 0 temperatur

200 _ /_. . , ,.-

P4N

3.0 3.5 i476'- 1.57T1.

Fig. ?.25 BBC/SULZER 4-Stage Transonic Compressor

Outlet Radial Distribution of Pressure and Temperature at a Pressure
Ratio of 3.54 (Point 5)

r
260- 14G\ ,
* 0
240 P P T
Po • PO TO
Static Total Total
pressure I 1 pressure 7 temperature
94 N

3.0 00 3.5 4.0 1.5 Tý

1.6

Fig. 3.26 BBC/SULZER 4-Stage Transonic nompressor

Outlet Radia&. Distribution of Pressure and Temperature at a Pressure
Ratio of 3.78 (Point 6)
 245

II.4. Results of Calculations

Two types of calculations ware performed by various participants, each using his own
method for through-flow calculatione and blade-to-blade calculations. An important as-
pect of the Working Group's task was to demonstrate that the use of systematic blade-to-
blade calculations could provide a support for the confirmation of an existing correla-
tion or the development of a new one. It is indeed believed that the present development
and state of the numerical tools available in order to compute complex flow structures,
should be used more and more extensively in order to develop a deeper understanding of
the flow mechanism of loss production and of turning.

The possibilities for blade-to-blade calculations, with inclusion of radius variation and
stream-tube contraction and coupled to a boundary layer calculation could lead to serious
improvements in the knowledge of the influence of these parameters, if some systematic
series of such calculations were performed. That this is indeed a clear possibility is
illustrated by Koch and Smith's correlation for design point, derived through systematic
series of boundary layer calculations. Some limitations still exist however, due to the
difficulties in the computation of separated flow reg 4 ons. Due to time and cost limita-
tion only a limited analysis was performed within the present work.

Since the central aerodynamic variable in loss correlations is the difusion factor, bla-
de-to-blade calculations were used on two blade sections in order to compare various
existing formulas with the values obtained from the blade-to-blade flow distribution and
from experimental data. This is discussed in the second part of this chapter.
The first part discusses results of through-flow calculations as obtained by the various
participants who accepted to perform some calculations. Most of the calculations were
done with the authors own correlation system and in several of these cases, complete in-
formations could not be disclosed. Particular mention
provided by E. Benvenuti from Nuovo Pignone (Italy) who has to be made for the results
presented results (for the four
stage compressor) with three different approaches, using the same code.

This is a very important aspect in the meaningful assessment of differences between cor-
relations since their interaction with the through-flow computa t .on method is strongly
non-linear and can have a large influence on the results obtain'd. It is a well known
fact that, even with the same correlation, different through-flow programs will lead to
non-identical results on overall performance, see [1 1 . These aspects are also noticed
by R. Novak in a communication to the Working Group. As R. Novak states

"In using my program recently I uncovered some problems which, I believe, will be pre-
sent with any of the currently used techniques. They also have bearing on the reliabi-
lity of published cascade data, particularly for high Mach number cascades derived from
running stage tests.

Most, if not all, off-design prediction schemes work with stations just at the L.E. and
T.E. of blade rows. Streamline curvatures aro iteratively computed from radii progres-
sively established using these two stations. Thý blades, however, do have a finite
thickness. If a station is included at a point near the maximum thickness locus, there
can be a measurable difference in curvature computed at the L.E. and T.E. stations.
The concave upward curvature which is generated by the thickness at this station raises
the outer diameter through flow velocity, and lowers that at the inner diameter as com-
pared with computations without the internal station.

My checks so far have indicated that a single station near the maximum thickness point
traps the main effect, and that the intrablade blockage term is the most important.
More than one intrablade station does not significantly add to the picture (in regard,
most importantly, to the effect on L.E. and T.E. incidence and deviation computations);
the introduction of other terms, such as blade lean entropy gradient, are of secondary
importance.

The problem shows up most clearly fur high-through-flow configurations where the velo-
city-squared coefficient on the curvature term becomes large.

I made some checks recently, using published data on a P&W 1600 ft/sec stage (NASA
CR-72694). Using first, the measured rotor discharge pressures and temperatures, and
no intrablade stations, I was able to duplicate the PaW vector diagrams. The inciden-
ce angle computed at the tip was, for example, seven degrees (this should have been a
tip-off, since the tip relative Mach number was in the range of 1.6). A similar compu-
tation, using the same measured pressures and temperatures, but with an intrablade sta-
tion with blockage, resulted in a tip incidence of three degrees. The tip incidence
was lower and the hub incidence higher, as would be expected.

Since the blade was an MCA with small leading edge camber, and had a thin leading edge,
the three degrees number is cc isiderebly more realistic than the seven degree.
The above has two immediate consequences in regard to compressor off-design prediction
for fan stages (high Mach number) and multistage inlet stages at least one intrablade
station must be included; nearly everyone have reduced stage test data to effective
incidence/loss/turning angle cascade data reduction program using just leading and
trailing edge stations. To the extend that I am correct above, this says that all data
must be re-investigated with intrablade stations.
246

Some caution is therefore needed in the discussion of the results of through flow calcu-
lation in order to be able to make distinction between the various influencason the over-
all results. In particular, the- influence of the intrablade chordwise distribution of
losses and angular momentum and che inclusion of blade wake blockage can be non-negligi-
ble on the overall pressure ratio prediction. Although very important, a deteiled dis-
cussion of those points was conbidered outside the scope of the Working Group.

The second part of this chapter contains a discussion of the accuracy and degree of vali-
dity of diffusion factors and their relation to the loss coefficient.

Based on two test cases, NAGA-65 blades and a DCk blade, of e8' camber, diffusion factors
as calcul4ted by the velocity ratio obtained from a blade-to-blade subsonic calculation
are compared with the values obtained from various correlations and from experiments.

It appears, within the scope of the limited amount of calculations performed, that the
potential flow calculations give an acceptable prediction of diffusion factors.

However, large discrepancies appear when the various loss relations are compared to the
experimental data.

REFERENCE

1 ] W.R. DAVIS, D. MILLAR, "A Comparision of the Matrix and Streamline Curvature Methods
of Axial Flow Turbomachinery Analysis from a User's Point of view", ASME Paper
74-WA/GT-4, November 1974

I÷

II
 247
11.4.1. The Through Flow Calculations

Calculations of the three test cases were performed by various authors belonging to in-
dustrial or research organizations. Their contribution is particularly acknowledged and
added a great deal to the progress of the Work undertaken. We did benefit from contribu-
tions from NASA Lewis Center, Dr. Novak (G,E,, USA), D. Bohm (KWU, Germany), Prof. Macchi
(Italy), Dr. E. Benvenuti (Nuovo Pignone, Italy) and Ch. Hirsch (Brussels, Belgium).
These authors will be refered to respectively bu the letters A, B, C, D, 9 and F.

Four Stage Transonic Compressor

Results for this test case were provided with several correlations of which we have limi-
ted knowledge in some of the cases.

The basic program and prediction system developed by R. Novak, based on (1 1 is used by
E. Benvenuti and KWU. However, no information was provided for the values of the adjus-
table coefficients in this correlation.

No information is available on the correlations of the code used by NASA Lewis.

Fig. 11.4.1.1. presents a comparison between the results obtained by A, B, E i the re-
sults of KWU (C) being practically identical to these of R. hOVAK (B). Dr Benvenuti's
calculation contain some in-house modifications of the basic correlation (which is the
one used by B). It is quite clear that non-negligible differences appear, but this in-
formation is to be considered as an indication of the considerable differences existing
between different methods. The rnsults of E in this figi res are obtained with the corre-
lation of Koch and Smith (9 ] for the design point and maintaining the original off-de-
9
sign contribution.

The Fig. 11.4.1.2. to 5. present under graphical forms the differences between the corre-
lations of Smith and Novak as obtained by E for the design point of this 4 stage compres-
sor. It should be kept in mind that the curves indicated "Smith" contain also off-design
contributions.

Clearly, the minimum losses, as predicted by Koch and Smith are higher than the minimum
losses used by R. Novak for rotating blade rows. Also the way the various contribution,
Mach number effects, tip losses (which are not included in Smith's approach where a cor-
relation for the efficiency drop due Lo the end wall boundary layers is used) are accoun-
ted for is quite different. For stators, the Smith correlation gives appreciable lower
values for the loss coefficients. Unfortunately, no experimental data are available to
make a clear decision in favor of one of the other approach.

It is to be noted in particular, -hat Smith's correlation for the design minimum losses
show a stronger Mach number influence and have a shock loss contribution which is lower
than the one predicted by the model of Lewis and Hartman applied in Novak's correlation.
However, it is important to note that the final overall performance, predicted by the two
methods, are very close, especially in view of the large differences in the detailed loss
distributions as they are produced by the two sets of correlations. ''his raises the ques-
tion of the level of accuracy which is required in the evaluation of losses and deviation,
in order to predict an overall performance map, and of the definition of the most sensi-
tive parameters. Is it the deviation (as in the case for turbines), or a loss gradient,
or another property ?

The results obtained by E. Macchi (D) are derived with a program based on Vavra's 122
technique of solution for axisymmetric and steady flow in axial turbomachines; it solves
the complete radial equilibrium equation - accounting for enthalpy and entropy gradients
and streamline curvature effects - in a specified number of stations ahead of and after
the blade rows. An orthogonal curvilinear system of coordinates having the meridional
coordinates coincident with the generatrises of the flow stream surfaces is used in the
solution. Iterative methods are used both for the numerical solution of radial equili-
brium and continuity equation in single stations and for streamlines location and curva-
tures over the whole machine.

Incidence, deviation and losses produced by the blade rows at "reference" conditions are
computed according to NASA rules [3 1 the only difference being the assumption that the
streaw ine radial position has no influence on them (i.e., in making use of figs. 201,
202, :103 of [3 1 , the curves given for 50 % of blade hei.ght are assumed to be valid fox
all streamlines).

Off-references deviations and losses are computed according to the correlations proposed
by Creveling and Carmody 14 ) as a function of incidence angle and inlet relative Mach
number. *4o AVDR influence on losses is considered; the simple Pollard & Gostelow 1 5 1
correlation is used for deviation. A modified version of the Miller at al. [ 6 1 coxrela-
tiun is urodi shock losses are accounted for also at high subsonic values of the veloci-
ty ahead of the blade row.

The original version of Mellor's theory on casing boundary layer described in 1 7 1 is in-
corporated in the computer program.

The Mellor equations are integrated across every blade row; this calculation is carried
out every time the continuity equation is solved. The displacement thicknesses are

4|
248

accounted for in locating stroamlinee, while all other pertinent boundary layer parameters
yield the enthalpy and entropy increase caused by secondary losses.
The losses are then mass-averaged, i.e., uniformly distributed in the whole section. The
following values of the four "arbitrary" coefficients in Mellor's theory are assumed, for
both hub and tip of rotor and stator blades i
c; - 0.003 C - 0 K - 0.65 H - 1.4

The Reynolds number influence on C1 is accounted for by an exponential rule t Flow angle
changes due to secondary effects are not considered.

The comparison of the computed overall performance with experiments is represented in

Fig. 11.1.6. While a good agreement is found at the design point, large differences are
found at off-design, mainly at low flow rates. The differences df calculated and experi-
mental flow rates for the same pressure ratio vary from about ± I % at choking, and at
design point, up to as high as 10 % near the surge line. As expected the differences are
larger at speeds different from the design one.

A good agreement between calculations and experiments is found at nominal speed and at
90 %, while about 5 % erxor is obtained at low speed, where the calculation is optimistic.
The computed end-wall boundary layer thicknesses are represented in Fig.
11.4.1.7. It
can be seen that, as the design point, the predicted values are close to the suggested
ones. For this reason, it was thought unnecessary to repeat the calculation with the sug-
gested values of blockage and a large increase of the blockage factor (more than 20 1)
near surge can be observed.

Fig. 11.4.8 to 10 show a comparative calculation by author E between three correlations

their own basic correlation, the Smith and Koch correlation for the design point with the
original off-design calculation including Smith's EWBL efficiency corrections, and a
third approach including iteratively with the through glow, Hirsch and Do Ruyck's EWEL
calculation method (8 .

This result is particularly interesting in that it has been obtained with the same code
and the same set of correlations with, in the third case, the inclusion of the end-wall
boundary calculations. The trend of these results clearly supports an approach based on
the end-wall boundary layer representation of secondary flow effects on overall perfor-
mance.

The Tw3 Stage Compressor

This test case has been calculated by NASA (author A) . Author A ran their calculations
with in-house correlations, of which the following elements b•tve been communicated.

The total loss coefficient is the sum of a number of components. These include a profile
loss,63 , plus a shock loss, 0 , plus a secondary flow loss,; ,c plus a part span dam-
per 1o~s, S Tesroil
o S-Ict The profile loss component, 0,; currently is basically a function of
D-fctor bu has corrections for axial-velocity ratio and incidence angle. The shock
loss component, Me, basically uses the flow model developed in the reference "Shock Los-
ses in Transonic Compressor Blade Rows" by Miller, Lewis and Hartmann (see ASME J. Eng.
P'~wer, July 1961). However the shock losses computed from the above model are reduced
as a function of the inlet Mach number. The part span damper loss component, Mpsd, fol-
lows the flow model in the reference "Off-Design Correlations for Losses Due to Part-
Span Dampers on Transonic Rotors" by Roberts, Crouse and Sanderkock, NASA TP 1963, July
1980.

The deviation angle prediction method is basically an extension of Carter's Rule. Seve-
ral of these extensions are noted in the paper for the AGARD WG-12 "Axial Compressor Per-
formance Prediction Deviation/Turning Angle Correlations" by George K. Serovy. One ex-
tension makes the variable M a function of both setting angie and location of maximum
camber as indicated by the equation on pg. 22. The second extension uses an equivalent
circulation fluid turning angle as discussed in A.ppendix E. This accounts for the radius
changes and the range of axial velocity ratio which occurs across the real compressor
stages, Finally, the code increases the deviation angle an arbitrary amount when the D-
factor is above a value near 0.7. For off-design operation the deviation angle predic-
tion method also applies a correction which is a function of incidence angle and flow
Mach number. It is also desired to add a correction term that accounts for the secondary
flow effects. Tne results obtained for the overall performance are shown on Fig. 11.4.1.
11. The shifting of the design speed curve towards higher values than design has been
attributed by the authors of the experiments to a locally uncambering (opening) of the
balde leading edge in the tip region of close to 1.2" more than had been allowed for in
the original, beam-analysis design. According to the authors, this off-design uncambe-
ring may have allowed the fan to overflow and explain also the reduced stall margin.
Author A also produced the performance predictions separately for each blade row and sta-
ge (Figs. 11.4.1.12 and 13). These results are a clear indication that even when over-
all performance is predicted in an acceptable way, blade row performance can be poor,
and that global compensations of errors can occur.

This appears to be the case at part speed, where the correlations do not appear to be
well adapted in this low Mach number region. Better results are obtained at design
speed, as can be seen on Figs. I1.4.1.14 and 15 showing the first stage radial distribu-
tions of the main variables.
 249

Single Stage Compressor

This test case in extensively documented and has been calculated by Hirsch (author F).
The correlations used by author t are based on Carter's rule for deviations, following
the method of SP 36 for the design deviatiuns and incidences. Corrections for Mach num-
ber and axial velocity ratio are introtiuced as well as off-design deviations, following
the method outlinied in[l by Novak. The profile and shock losses are estimated follo-
wing the approach by Koch and Smith 19 1 and an off-design contribution is calculated
based on the estimation of incidence working ranges for stall and i•hoking. A parabolic
variation with incidence for off-design losses is then considered with respect to the
corresponding range.

All secondary etfocts are taken into account by an ond-wall boundary layer approach fol-
lowing 18 1 in an itmrative calculation with the through-flow computation.

The predicted overali performance curve is shown in Fig. 11.4.1.16 and radial transverses
of total pressure (Fig. 11.4.1.17), losses (Fig. 11.4.1.18), axial Mach number (Fig.
11.4.1.19) are shown at two mass flows for the design speed curve.

Although the overall porformance curve is reasonably well predicted, clearly detailed flow
distributions, although acceptable, are not as well predicted.

The full distributions of Mach number and flow angles (Figs. II.4.1.20 and 21) as obtai-
ned through the end-wall boundary layer interaction, show the prediction of the high eke-
wing of the flow in the wall regions, and the boundary layer decrease of velocity. The
influence of the EWBL is also to be seen on Fig. 11.4.1.17, where the efficiency obtained
only on basis of the profile losses is disclosed.

The predicted blockage factors compared with the experimental values, give the following
result
Exp. Calc. (2 t blockage at inlet)
Inlet rotor 2.55 1.82
Outlet rotor 4.10 3.95
Outlet stator 5.1 5.45
It appears clearly that the end-wall boundary layer calculation incrt, ses strongly the
prediction capability of through-flow calculaticna.

REFERENCES

[ 1 J.A.
R NOVAK, "Flow Field and Performance Map Computation for Axial-Flow Compressor
and Turbines" in Modern Prediction Methods for Turbomachine Performance, AGARD-LS-
83-1976
12 1 H.H. VAVRA, "Aero-Thermodynamics and Flow on Turbomachines", John Wiley & Sons, Inc.
New York 1960
(3 The members of the Staff of Lewis Research Center, Cleveland, Ohio, "Aerodynamic
Design of Axial-Flow Compressors", NASA-SP-336, 1965
[4 H.F. CREVELING, R.H. CARMODY, "Computer Program for Calculating Off-Design Perfor-
mance of Multistage Axial-Flow Compressors", NASA CR
[5 0. POLLARD, P. GOSTELOW, "Some Experiments at Low Speed on Compressor Cascades", J.
Eng. Power, (1967), 427-436
[6 G.R. MILLER, G.W. LEWIS, M.J. HARTMAN, "Shock Losses in Transonic Compressor Blade
Rows", J. Eng. Power, July 1961, 235-242
[7 G.L. MELLOR, "secondary Flows in Turbomachines", Von Karman Institute for Fluid Dy-
namics Lecture Series", Jan. 1975
[8 J. DE RUYCK, Ch. HIRSCH, "Investigations of an Axial Compressor End-Wall Boundary
Layer Prediction Method", ASME Paper 80-GT-53 (1980)
[9 C.C. KOCH, L.H. SMITH, "Loss Sources and Magnitudes in Axial Flow Compressors",
Trans. ASME, J. Eng. for Power, July 1976, 411-424
250

+ Experimental points
Calculated points:
4- 0 Novak- Smith Method (E-) nn/e 1.0
* Novak original (B des
Tt. o NASA %A)
(no internal blockage)

3-
+I
I.

+
n÷ II
I0

rdes
* .,0.6• 85,,,++

+ Equi'mlence condition:
15 9075 bar
•o-293.15

0,6.

0.8 ÷•- .•.

i
0.6

15 20 25 rh,kg/s3
Fig. 11.4.1.1 Four Stage Transonic Compressor
Comparison of Calculaten and Experimental Results, Authors A,B,E
 251

aC)

CC

CL I

CN

-a+

CL

oE"o >

0N0

Fig. 11.4.1.2 Four Stage Transonic compressor

Radial Distribution of Losses and Incidences for the Novak and
Smith correlations -First Stage Rotor -Author E
 252

I8

22

I MC
M il/
w/o Tihp losses CL /N

!H 0 50 100

.our stage as ic cmpressJr

-AtOr" }
Stage •tR
Second th I. oa
s_- th-crre•s-
Smith correlatiOng butlOn Of LosSeq "d ,Inc.). Au
 253

ideg.

5 S

1 ii
W8%
 254
i

i,deg.
110
8

0\ti~
09 Stoq CmY~sO
IoF

~
Rj~l ~~rbt
o oae in¶ n~dneAf~ h Nv2 n

5 -it Cr1amTJ ip, loss~t ~aes Rtr-

* .
1
 255

- Calculations
0 Experiments
1.0
0.9 6
i,- 0.9 5

3-
2

. ' 2 ''In
Q685'
des

0,008

. 1'

ZQ.
0860

0,4--
_ I , , _1 ,I ..
0' 15 20 25 rin kg/s
Fig. 11.4.1.6 : Four Stage Transonic Compressor

'
Performance Map Prediction - Author D

.2i 4 . -•
256

-- -

_ _

I,
- ID ...
, II
___

CN CN

C
O C)
OO C)

Fig. 11.4.1.7 7 Lour Stage Transonic Compressor

End-Wall Boundary Layer Presiition - Author D
 257

+ Experimental points
O Calculated points

4 o des

+0
1.0

te
3 1

2-
* > des
0.685
Equivalence condition:
00P = .975 bar
0T = 293.15 K
1- ... I ., I , -_I

0,8 7 10

0,6- ee

0,41 S. .. I _ . .I .. .1
0.415 20 25 m, kg/s 30
FIig. 11.4.1.8 1 Four Stage Transonic Compressor
Performance Map Predtction - Original Correlation Set - Author Z
 258

+ Experimental points
0 Calculated points
1.05

lTE > 1.0 +

0.90
[ ÷ I

+\. '+ I I

•2-- rfde

0+ 0,685
Equivalence condition:
SPO = .975 bar
T, = 293.15 K
1- I

+~2~
0,8 0 0 0

+I
0,84'-
0,6 -- \/ - "
04 I

,15 20 25 mh.kg/s 30
3
Fig. 11.4.1.9 g Four Stage Transonic Compressor
Performance Map Prediction - Original Correlation and End-Wall
Boundary Layer Correlation of C.H. Smith - Author E
 259

+ Experimental points
O Calculated points

4,

0.9
3-

rdes
2--• 0.685
+
+=""= Equivalence condition:

P0 .975 bar
T* 293.15 K

040

0,6
15 20 25 m. kg/s 30

Fig. 11.4.1.10 Four Stage Transonic Compressor

Performance Map Prediction - Original Correlation with Hirsch's
End-Wall Boundary Layer method - Author E
260

F IE-C 7 1
*~~- . IL.21...

......... I. a

L7"
a1;ji:t.j 14'I;4~ ___

* OTA 14-I{
'ý'TEMP
Z
TIOi
11A ~-

7t t

P ESSURE_. .....
RAL TIO. j ~j~ ...-
.... 7'.7-

- - -- *44...* .... ....

... -4--
, .. ......

Porfo77an7 P:7,ci Autor

7Map - .

Open Symbols are experimental data

 261

.
.~- .. . ..

-7..

............ >*-"

7T'-

TMPRATUR*k .

. 7, -7

P E
SURE ...
.... ::T : -

:PATIO: A.-5 -

;:A*~ .. . .. .

7 _7. ... ,

.. . . .....
.. Iq ... ..

, 1 7t*
71 1 ... . .. .. ...-

4+ 44 1
... ... .. . . ... I..
. .

EQUVAEN WEGT7OLS/E.- ~
F~g~II
41 1 TwoStag Copreso7

...
ra. Perormnc
.ap : fFiAtStge::uto
.at.
..... Ope are. e.er.eta

----------------------------------------------------------- t - - -
 262

PEFF1 CIENCY fz 7

::4:7
1.:
n.I

T7: __
Ti t_
... :

- IT

%
EPRESSURE ATAL
u.-H.!i: 2
114
T 417
ly- A 7. 4g.

2W:"77

i....:7. .......
 263

TI

F- ANGLE '- 7

MERWIONAL

r1
- _ ICMfENCE
EF ICIENCY-

LOSS

-~~T
- --- :OEFFICIENT

MIt

P ESSI
+-7ti~
- - T ~ H I-H~-:0

-A
214

7 7 7 -

.2 ~ ±. T.. ....
....
..
... : -..

-Lo IT

7 =
Iiz77-7.1

0
NGLE - ':1:71...

--

.7-
INCIDENCE FO- 1 ~ . EIEN PA RO I

o Sta.
4ig Com7r_7 .1... .1 ...

..............
....... ~
 26S

17:.
4-d

4 14
L::h 42K.
IAI

2..

. ......
Ff11V A~

Fig. 11.4. 1,6 Single Stage compressor

Perforrnanc7 -map Prediction - Author F

calculations with and without End-Wall Boundary Layer
266

or utlet )

C.I
.173'14 /s'

14.

20 40 0! 8

0 20 40 60 80 100-
Percentage. annulus height

F4 .g._ 11.4.1.171 Single Stage Compressor

Total Pressure Ratio Distribution -Author F
 267

__x

7: -....

Ix.

..
........

n .II.41i %1ige1tg2omrso
Flo Losse Prdc n Autho F
268

id C

I .8k 6.

ci St~o lai

4V .. _....._. . ....

40' t .80

P ercentaige annutut heidht

r'ig. 11..1.1.9: Single Stage Cumpressor
Mach Numuber Pr edict~on - Author F
7a1culatior s with and 4 itbout End-Wal.l Boun-dary Layers
 269

11.4.2 EVALUATION OF PROFILE LOSS PREDOCTIONS BASED ON DIFUrSION FACTORS

SUMMARY

A comparison between theoretically predicted and measured profile losses has been
made, in order to evaluate the precision of the differe,.t preiction methnds listed in
reference 1.
These prediction methods are based on the relation betweet. suctlon side diffusion
and profile losses. The evaluation concerns es well the predictior of diffusion facto.-
as predicted values of profile losses.

It is further shown that these correlations are not universslly valid for all blade
shapes and that some correlations perform as well at design as at off design if the.Su;-
tion side diffusion is correctly predicted.

11.4.2.1 introduction

Profile loss prediction methods are generally based on the following observbtionsr
a. wake momentum thickness at the trailing edge is mainly due to the flow deceleration
along the blade suction side
b. profile losses are functions of the blade wake and cAscade neometry.

A first part of the loss predictions therefore consists of the definition of a dif-
fusion factor as a measure of thp suction side diffusion between Vmax and V2. This dif-
fusion factor is normally expressed in function of cascade geometry and inlet and outlet
conditions.

A second part of these loss prcdictions consists of defining a relation between waKe
momentum thickness or profile losses and the diffusion factor.
The first correlations based on this principle were valid only for incompressible
two dinmensional flow at design incidence. 3orrections for off design incidence and com-
pressibility effects have been proposed.

This was done by changing the value of the diffusion fActor, or by adding some extra
loss terms.

The validity of this loss prediction has been evaluated by applying these ýorrela-
tions tl two test cases and by comparing the theoretical predictions with experimental
results.

Because of the limited number of test cases, no general conclusions can be nTbtained.
However, this comparison allows to obtein some indications about the level of confidence
one can have in these correlations, when applied to a typical design. They also illus-
trate the shortcomings of some correlations.

In order to evaluate both parts of this correlation, !ome ralculations nave been per-
formed with a diffusion factor derived directly from Vmax/Vk obtained from pitent:al flow
calculations or from experiments.
11.4.2.2 Test cases

The first test case concerns a cascade of NACA 65 (12A10) 10,blades with the follow-
ing parameters

350
o= 1.25
C = 14.1, 17.1, 20.1, 23.1

results at different inlet Mach numbers are shown in figures 29 to

32 of The experimental
reference 2. .

The second test case consists of 480 camber DCA blades with the following pay4imeters

13 = 34.2, 38.5, 40.0, 42.0

S= .614
y = 13.58
M1 = .6

The experimental results are published in reference 3.

11.4.2.3 Correlations
In order to avoid any confusion concerning thd cor-)lations bsed in this e.aluation,
we will shortly define the different calculations that have beer. per'Pormed. Morn deteils
about these correlations are giver, in referrence 1.

1. Lieblein's correlation based on equivalent di~fusion faictorI

The definition of diffoision factors used here is valid only for incompridisible 'Tlow
et design and off design angles of attack.
COS32 r . * 1o.43 !O
D _ - I . 12 + a~-+ .61 (tg01-tg02))

The wake momeitum thickness is obtained from a curve fitting of figure 11 of

reference 4.

ez 2
- ,OU497 + .011921 (Deq".1.) - ,078935 (D eq-1.)

3 4 5
+ ,21399 (D1,)
eq20:: (eq-1.) + .070112 (Deq-a1.)d)

lhe total pressure losses are calculated by following approximated equation:

22cosl
W 2 (2)
c cos0 2 1CoS6 2 j

2. Lieblein's correlation based of Vmax/Vi

obtained from potential flow c3lculations

Vmax
Oeq 2

This diffusion factor accounts for variation as a function of angle of attack and
compressibility effects. Combined with equations (1) and (2) it should give an indica-
tion of profile losses for compressible flow at design and off'design.

The potential flow calculation is described in reference 5.

3. Lieblein's correlation based on Vmax/Vl obtained from experiments

Vmax
Deq V

Combined with equations (I) ard (2) one should again obtain the real losses for compres-
sible flow at design ind off design.

Calculations 2 and 3 are made in urder to verify the idea that losses can be related
to the suction side diffusion.

A comparison between The diffusioni factor of calculation 1 with the one of calcula-
tions 2 and 3 also allows to investigate the precision, with which the Vmax/V2 is
predicted.

4. Swan's correlation
Profile losses are first calculated for incompressible flow at design incidence in
the same way as proposed by Lieblein (i=i*). However, the relation between 0/c and D*q
is given as a function of radial position in rotor and stator (Hub to tip). Ir this com-
parison with cascade results, the loss correlation for a hub section is used and radial
shift of streamlines has been neglected.

In order to account for off design operation and inlet Mach number, a correction
term is added to the trailing edge momentum thickness.

It
is important to notice her3 that
e* at design incidence, a change of inlet Mach
0
number has no influence on losses = i D q = D%)
If
c - eq)-
5. Correlation of Jansen and Moffat
This correlation is based on the design diffusion factor for incompressible flow
defined by
V2 r2Ve 2 -riVel
D* Idna + a d s
V, !r 1 +r 2 ) V1 a

In this com•parison with cascade data a constant radius was used.

This correlation contains a correction of losses for inlet Mach number above the cri-
Stical one and an incre-se of losses at off design.
 271

6. Correlation of Fottner

This correlatlon contains an additional term in the definition of Vmax/Vi to account

for blade thickness.

Inlet Mach number influence and correction for off design is the same as in previous
correlation.

7. Strinning correlation
This correlation is valid only for design operation at different Mach numbers. The
influence of inlet Mach number is introduced by modifying the diffusion factor.
The relation between losses and the modified diffusion factor is the same as the one
used by Fottner for incompressible flow.
8. Smith's correlation
This correlation is valid only for design losses. ihe description given in reference
1, is incomplete and has been completed according to reference 7.
At design angle of attack Vmax/V2 is defined according to reference 1.
The trailing edge momentum thickness e/c and shape factor HTE are a function of
Vmax/V2 and Reynolds number. The value of Re used in this calculation is 106. These
values are further corrected in function of M, and Vmax/Va. In this comparison, no cor-
rection is made for streamtube height.
Losses are calculated according to the mixing theory of reference 8.
9. Smith's correlation using Vmax/V2 from potential calculations

In order to use this correlation also for off design loss predictions, where Vmax/V2
is not predicted by the correlation, calculations have been performed with Vmax/V2
defined by potential calculations. The correlation is used only to define 6/c and HTE
in function of Vmax/V2.
10. Smith and Koch's correlation using Vmax/V2 experimental results
This calculation is similar to previous one but Vmax/V2 is defined from experiments.
The value of this last calculation is in the verification of the relation between losses
and Vmax/V2.
11.4.2.4 Comparisons
Comparisons between correlations and experimental data are made on two levels, cor-
responding to the two parts of the loss correlation. In a first part the predicted values
of Vmax/Vi are compared to the experimental ones. In a second part predicted losses are
compared to the experimental ones.
The flow conditions corresponding to the different calculations are given in tdble 1.
Tests numbers 001 to 016 concern the NACA cascade at different inlet Mach numbers and dif-
ferent angles of attack (Ref. 2). Test numbers 101 to 104 concern the DCA cascade at dif-
L ferent inlet Mach numbers (Ref. 3).
Comparison of Vmax/Vi

From all the correlations used in this comparison only those of Lieblein and Swan
define a diffusion factor at off design, but for incompressible flow. For test cases
numbers 001 to 008 and 013 to 016 a comparison of predicted and experimental Vmax/Vi is
therefore limited to incompressible flows.
For compressible flow, only the results of the potential calculation can be compared
to the experimental value. This is shown on figure I for test numbers 001 to 004. Calcu-
lated and experimental values of Vmax/Vi agree fairly well at different Mach numbers and
the extrapolated values for M=O are close to the value predicted by Lieblein.
Figure 2 shows the same comparison at design angle of attack (test Nos. 009 to 012).
A comparison with the correlations of Smith & Koch, Jansen-Moffat, and Fottner are pos-
sible only for incompressible flow.
The correlation of Strinning allows calculation of Vmax/Vy for compressible flow at
design conditions. However, this correlation underestimates the influence of compres-
sibility on Vmax/Vj. The prediction of Jansen-Moffat and Fottner agree very well with
Lieblein's prediction for incompressible flow.
Comparison of losses
Losses predicted by the different correlations are compared with experimental values
of reference 2 on figures 3 to 7.
272

The conversion of Cw experimental to w has been made according to reference 6.

Concerning the NACA blade cascade, most of the correlations predict losses which are
higher than the experimental ones.

The losses predicted by Lieblein's correlation in combin3tion with Vmax/Vi obtained

from potential flow calculations or experimental results show good agreement.

Also Swan's correlation is very good for these blades especially at off design inci-
dence. Close to design incidence and at higher inlet Mach numbers, the losses are under-
estimated.

This is due to the absence of a correction for compressibility when Deq - Deq, as
pointed out at the description of this correlation.
The correlations of Jansen-Moffat and Fottner predict losses which are much too high
for the NACA blades at off design.
Smith and Koch's correldtion also predict higher losses thaoi the experimental ones
and the use of Vmax/V2, obtained from calculation or experiments does not show any impro-
vement. However, one should keep in mind that in this correlation, losses are predicted
at complete mixed-out conditions, and experimental values are measured one chord down-
stream of tvailing edge.
For the cascade of DCA blades, test numbers 101 to 104, the situatinn is quite dif-
ferent. Most of the loss predictions are in a close band around the experimental values.
Only the predictions of Jansen and Moffat are out of this range. However they are much
closer to the values obtained in a real stator.

11.4.2.5 Conclusion
Previous comparisons indicate that the classical correlation of Lieblein's type are
fairly good if one can estimate correctly the value of Vmax/Vj at off design and dif-
ferent values of the inlet Mach number. Using a value of Vmax/Vj obtained from potential
calculations gives a correct trend for the variations at a correct level. Swan's cor-
relation could be improved by introducing a better compressibility correction at design
incidence.

The correlation of Fottner cannot be recommended for NACA blades but gives good
results for DCA blades. The correlation of Jansen and Moffat indicates even higher
losses for both cases.
Possible explanations could be that these correlations predict losses at complete
mixed out conditions, where the experiments are the ones at one chord downstream of the
cascade.

Furthermore, some correlations are based on experiments in rotors and stators, and
could contain losses which are not present at the mid section of a cascade.

REFERENCES

1. DUNKER, R. & WEYER, H.B.: Axial compressor performance prediction. Survey on dif-
dusion factors and profile losses.
Chapter 11.2.1

2. EMERY, J.C. & DUNAVANT, J.C.: Two dimensional cascade tests of NACA 65 (12A10) 10
blade sections at typical compressor hub conditions for speeds up to choking.
NACA RM L57H05, 1959.
3. Transonic flows in turbomachinery.
VKI LS 59, 1973.

4. LIEBLEIN, S.: Loss and stall analysis of compressor cascades.

ASME J. Basic Engineering, September 1954, pp 387-400.

5. VAN DEN BRAEMBUSSCHE, R.: Calculations of compressible subsonic flow in cascades

with varying blade height.
ASME J. Engineering for Power, Vol. 95, No. 4, 1973, pp 345-351.

6. LIEBLEIN, S.: Analysis of experimental low-speed luss and stall characterist-ics of

two dimensional compression olade cascades.
NACA RM E57A28

7. KOCH, C.C & SMITH, L.M.: Loss sources and magnitudes in axial flow compressors.
ASME Trans., J. Engineering for Power, Vol. 98, No. 3, 1976, pp 411-424.
8. SCHOLZ, N.: Aerodynamik der Schaufelgitter.
Translated and revised by A. Klein,
AGARDograph 220, 1977.
 273

TEST NUMBER 82 1M1 02 3

001 350 20?1 .307 9?45 1.25

002 .509 9?34
003 .706 9?19
004 .747 9?16

005 350 17?1 .299 12?66 1.25

006 .511 1263
007 .696 12?62
008 .762 12?62

009 350 14?1 .293 lb?88 1.25

010 .507 15791
011 .679 16?00
012 .742 16?05

013 350 23?1 .308 5?26 1.25

014 .500 5?09
015 .618 9?89
016 .750 10?50

101 36?2 21.6 .6 -4?7 .614

38?5 23.9 -4?5
102
103 40?0 25.4 -4?7
104 42?0 27.4 -4?7

SK
 274

oCI 200

* LIEBLEIN SWAN
o POT FLOW CALCULATION
,Vmtx/Vi Vmx/ EXPERIMENTAL
1.8-

i, 1.6 /.
A

1.4

1.2

M-
1.0
.2 .4 .6 .8

FIG. 1
 275

! LIEBLEIN & SWAN ¶7 STRINNING

, POT FLOW CAL-C. JANSEN --MOFFAT
SA EXPERIMENTAL FOTTNER
Vmax /Vl i,
1.8

1.0
1.0 .2-.1-M
1.2

.2 .4 .8 .10

FIG. 2
276

001 -004
.I =:20"1

" EXPERIMENTS
x LIEBLEIN INC.
0 LIEBLEIN Vmax/Vj from POT. CALC.
A LIEBLEIN Vmax/Vl from EXP
v SWAN
a JANSEN MOFFAT
+ FOTTNER
A STRINNING
SSMITH & KOCH
* SMITH & KOCH Vmax/Vl from POT. CALC.
* SMITH & KOCH Vmax/Vl from EXP
.08G3

.06

.04

* I

.02 I _

0 M
.2 .4 .6 .8 1.

'.C__
FIG. 3
 27'7

005 008
(%i =17?1

EXPERIMENTS
x LIEBLEIN INC.
0 LIEBLEIN Vmax/V1 from POT. CALC.
LIEBLEIN Vmax/V1 from EXP
SWAN
1 JANSEN MOFFAT
+ FOTTNER
A STRINNING
SMITH & KOCH
* SMITH & KOCH Vmax/Vl from POT. CALC.
* SMITH & KOCH Vmax/ V from EXP.
.08 -
Wi

.06

,04

.02

M
0
.2 .4, .6 .8 1.
FIG. /4
 278

009 -012
vI = 14"1 DESIGN

* EXPERIMENTS
x LIEBLEIN INC.
o LIEBLEIN Vmax/Vl from POT. CALC.
LIEBLEIN Vmax/Vl from EXP.
1" ' v SWAN
JANSEN MOFFAT
S+ FOTTNER
SA STRINNING
SMITH & KOCH
* SMITH & KOCH Vmax/Vi from POT. CALC.
SS~1MITH & KOCH Vmax/Vl from EXPF
.08 -

I.06

.04 - _ __ __ __(3

•~06

.2 .4 .6 .8 1.
FIG. 5
 279

013-016
Oa.1=23,1

* EXPERIMENTS
x LIEBLEIN INC.
0 LIEBLEIN Vmrax/Vl from POT. CALC.
A LIEBLEIN Vmax/V1 from EXP
7 SWAN
o JANSEN MOFFAT
+FOTTNER
A" STRINNING

v SMITH & KOCH

0 SMITH & KOCH Vmax/Vlfrom POT.CALC.
0 SMITH & KOCH Vmax/Vl from EXP.
.10 0I

.08

.06-
A

.0 + 00

.0 2 -..... --

.2 .4 .6 .8 1.

FIG. 6
•I•• ........
280

101-104

*CASCADE LOSSES M1 .6 AVR. = 1

6 STATOR LOSSES Mi = .65
x LIEBLEIN INC.
o LIEBLEIN Vmax /Vl from POT. CALC.
& LIEBLEIN Vmax /V1 from EXP.
• SWAN
i JANSEN MOFFAT
+ FOTTNER
* STRINNING
v SMITH & KOCH
* SMITH & KOCH Vmax/Vl from POT. CALC.
a SMITH & KOCH Vmax/VI from EXP
.10

.08

.06

++

.02 L-
, -

I I

0 34. 36. 38. 40. 42. 44.

FIG. 7
II.5. Conclusions

Many a'ipects of axial compressor correlations have been reviewed by N.rious members of
the compressor Subgroup and are contained in chapter II, together with the answers to the
questionnaire from industry and research organisations.

Obviously, many gaps are to be filled in order to achieve a satisfactory situation with
reqard to loss and turning correlations.

It is apparent however, that no set of correlations will ever be of general validity, sin-
ce the philosophy behind it is to describe deviatioi.s and losses .s functions of a very
limited number of variables. The complexity of axial compressor flow phenomena, even
without taking unsteady effects into account, is such that only trends can be defined
without a description and a knowledge of the details of the flow behaviour. Therefore it
is challenging to define a valid set of correlations of sufficient generality at design
conditions. As it appeared also from the previous chapters, the degree of validity of
existing correlations at off-design is still more limited.

Hence, the Compressor Subgroup members considered that progress could only come from an
improvement in the basic physical understanding of the phenomena leading to the loss gene-
ration and to the flow deviation.
It is advocated that extensive use should be made of the present day tools, namely sophis-
ticated experimental tools, in particular laser anemometry, and viscous flow calculation.

With regard to two dimensional phenomena, an extensive use should be made of the present-
ly available blade-to-blade calculation metbods,including boundary layer interaction in
order to define general trends with regard to the influence of parameters (such as Mach
number for instance) on losses and deviations. This approach, illustrated by the results
of Koch & Smith at design conditions, is considered as particularly promising. A parti-
cular effort in this direction shou.ld be made, especially at off-design.

On the other hand, very little appears to be know about specific phenomena such as shock-
boundary layer interactions. This is a typical instance where well selected and applied
basic experiments should be encouraged in order to form a basis for a correlation approach.

More generally, a lack of valid multistage data has been noticed. It is of fundemental im-
portance, if the transition is to be made from basic cascade data to real life situations,
to dispose of a reliable set of data taken in multistage axial compressors, and especially
of interstage and end-wall boundary layer data.

From the calculated test cases, it was clear that, even if overall performance could be
predicted satisfactory (by adjustment of some free parameters) detailed flow distributions
or (and' individual stage data are much more difficult to predict with confidence. This
indicates -that in many correlations, either the physical phenomena are not taken into ac-
count and ad-overall compensation of errors occurs,or different assumptions with regard
to the loss mechanism lead to different contributions of the various loss sources - cfr
chapter 11.4. - leading to large differences in the predicted details of the flow beha-
viour, even if the overall performance is obtained correctly.

This observation illustrates the strong need for a very clear physical understanding of
the basic pehnomena, which cannot be stressed enough.

More specific conslusions, with regard to the content and the validity of various approa-
ches to loss and turning correlations are 3ummarized in the following remarks :
The level of accuracy required is sensibly higher for deviations than for losses. The
influence on the predicted flow field of a small variation in outle't angle (one or two
degrees can be very important while a variation in loss level does not appear to be
very significant below, roughly, 10 % variation.
The validity of the diffusion factor is generally accepted and several of the expres-
sions to be found in the literature qive a valid estimation, when compared with the va-
lues derived from blade-to-blade potential flow calculations.
Although a large scatter exists between the various predicted profile loss coefficients,
most of them do have the right trends in function of Mach number or incidence. However,
no -!orrelation could be shown to behave significantly better than the others at off-de-
sign conditions.
Care should be taken to define a consistent set of relations between diffusion factor,
trailing edge momentum thickness, and loss coefficient. Moreover, ambiguities can ap-
pear when comparisons are made with methods which take into account mixing losses in the
through-flow calculation or in the loss definition.
This consistency problem is of particular improtance when .oombining various partial cor-
relations from different source..•
The available information on Mach number effects, on shock-boundary layer interaction
and shock induced separation is very limited. An important effort in this direction is
strongly needed
Secondary flow effects can be taken into account through correlations or through a more
sophisticated and physically based approach of end-wall boundary layers. It is the
Subgroup's opinion substantiated also by the results of the test calculations, that
this latter approach should be developped and generalized, since it offers the possibi-

L IZii _ - -_
 282

2lity of a consistent description of secondary and clearance losseq and deviations as

generat-ed by the end-wall layers. Moreover due to the coupling with a through-flow
calculation, the important parameters of annulus blockage is predicted.
The cirrelation for deviations are generally very unsatisfactory. This is already the
case at lesign conditions and hence, even more so at off-design, especially at high
loading conM .tions.
There will always be a t.eed for simple correlations, even with the increasing
development
of viscous flow calculations. The improvement can only be obtained from a coherent and
systemattc use of basic experiments together with 2D and 3D viscous calculations, in or-
der to understand better the physical behaviour of the flow and to be able to derive the
most meaningful parameters to be used in a correlation.

Sj
rI
 283

PART III

REVIEW OF CALCULATION METHODS

 285

III.1. Axial-Flow Turbomachine Through-Flow Calculation Methods

INTRODUCTION

In this review a through-flow calculation method is considered to be a computational technique for

prediction of fluid velocities and thermodynamic properties at designated locations in the internal flow
pathi ot a turbomachine. Defined in this way, a large part of the total spectrum of turbomachine flow
field analysis methods is included. The objectives of this paper are to trace the development of several
classes of through-flow calculation, to suggest the importance of understanding the assumptions underlying
representative examples of each class, and to outline the principal problems in application of through-
ilow methods to current axial-flow compressors and turbines.

t should be understood that all of the analysis methods described have been developed as useful work-
ing too% over a period of about 30 years. This period roughly coincides with the period of rapid
improveme of electronic digital computer systems. It will become evident that progress realized in
analysis me Nds has come about through improved and more realistic modeling approximations in the equa-
tions describ * the flow field combined with efficient numerical methods for solving the approximate
equations. With he addition of ingenious methods for representation of those features of the experi-
mentally observed w field not adequately modeled by the equations, for example, flow detachment,
,supersonic
"clearance and corner ows, passage cho ing, shock waves, and density variations in the transonic and
flow r.egimes, ~substant ial gain'a.have been made in practical resolution of a problem too complex
for most human and comput memories to sto'ý, recall and assemble.

BAS FOR THROUGH-FLOW CALCULATION METHODS

The overall structure of a turbomain design system is exemplified by Figure 1. Necessarily, the
initial phases of design involve "artisti" judgment based on experience and system technical requirements.
The final phases include both extensive experimental verification and detailed prediction of critical flow
field zones. The AGARD/PEP Working Group 12 study has been focused on the region of Figure 1 enclosed by
a shaded boundary. Within this region, the lowest levels of elegance are represented by mean-line per-
formance prediction methods and the highest by advanced computational procedures roughly corresponding to
three-dimensional, steady relative flow.
In a mean-line performance prediction method for an axial-flow turbomachine, a single computation
location, at the mean flow passage radius is specified at the entrance to and exit from each row of blades.
In the more advanced methods, computation points may be specified at locations throughout the flow path,
both in the axial spaces between rows and within the blade rows at arbitrary coordinate locations.

Characteristics of the Turbomachine Flow Field

The internal flow fields encountered in current turbomachines are viscous, compressible and unsteady.
The flow passage geometry guarantees a th.oroughly three-dimensional flow, and a significant fraction of the
flow occurs in rotating passages.

As implied in the introduction, the situation is so complex that it is difficult to attain a compre-
hensive "mind's-eye" conception of the flow, a necessary prerequisite for development of a good engineering
approximation. Some of the important phenomena influencing flow fields in modern axial-flow compressor
blade rows are shown in Figure 2. These features have a significant effect on the overall flow field but
they are not the only influential factors. What Figure 2 does not emphasize 13 that the phenomena shown
take place in a strong through-flow or primary flow which is generally understandable in terms of the
physical laws and corresponding equations of fluid dynamics. Many of the features of the flow field shown
in Figure 2 might therefore be called secondary flows or secondary effects. This suggests the attractive,
but not necessarily correct, possibility that such flows may be accounted for separately after determina-
tion of the primary flow field in n given case. How these flow features are included in individual
through-flow calculation methods should be a central point in distinguishing between them.

Turbomachine flow fields occur in passages characterized by some unique geometric and kinematic
features, which do not occur in other typical complex flows, either internal or external. These are shown
in Figure 3, which also introduces the idea of looking at the flow field in a representative meridional,
hub-to-tip section of a turbomachine.

An initial unique feature is the sequence of rotating and atationary internal flow passages. This
introduces a recurring need to transfer frames of reference from inertial to non-inertial coordinates,
considering relative flows in rotating rows and absolute flows in stationary blading. The effects of
rotation on the fluw within rotor blade rows should be considered.

A second unique feature existing in turbomachine flow fields is the transfer of energy in the flow
process as work done on or by the fluid as it traverses the rotating elements.

A third unusual characteristic of turbomachine flows is the frequent close proximity of moving and
stationary surfaces, clearly indicated in Figure 2. This complicater any attempts to evaluate the flow
field using traditional boundary layer-main flow iterative methods.
286

Fluid Dynamics Supporting a Flow Model

The equations representing the physical laws controlling the flow field were clearly set forth with
reference to turbomachine applications many years ago. Examples are found in work of Lorenz (1905,
Bauersfeld (1905) and Stodola (1927). Because of the multiple forms and coordinate reference frames in
which the equations are written in through-flow analyses, no attempt will be made to suggest a represen-
tative or recommended set in this review. The relevant physical laws and equation forms are listed in
Table I.
It should again be noted that in development of flow models for the individual through-flow mathods
within the general classes described in the following sections, not all of tha laws are necessarily used
(satisfied), nor are all of the terms in the generalized equations representing each law used. It is of
prime importance, but frequently very difficult, to determine what simplifications are made in each method.

Whether a specific through-flow method is based on the application of differential equation forms,
integral equations or both, use is generally made of control surfaces which subdivide the flow passage by
intersecting both the hub and tip (casing) boundaries. These control surfaces may be simple plane surfaces
perpendicular to the rotational axis for some axial-flow compressors and turbines, or they may be conical
or curved surfaces to conform to a leading or trailing edge trace in the meridional plane (see Figure 3),
or to reach into some particular region of interest. The control surfaces may be, in particular through-
flow methods, called computation surfaces, calculation stations, and channel solution surfaces. In this
review the term control surface will be emphasized. )
General Modeling Assumptions Used in Through-Flow Methods

A few assumptions are almost universally made in the through-flow calculation systems currently in
use. The first of these is that the relative flow at the entrance and exit cf each blade row is steady.
Exceptions occur for some special analyses relating to non-uniform entrance flow (inlet distortion) and
row interactions. The exceptions were considered outside the scope of the Working Group 12 effort.

The assumption of steady relative flow permits the definition of fluid particle paths by means of
streamlines, stream surfaces and stream tubes. This simultaneously simplifies the through-flow problem
and eliminates a feature of the real flow which is of unquestionable aerodynamic significance. Many
through-flow procedures utilize the concept of S1 and S? stream surfaces, generally attributed to Wu
(1952a). S1 surfaces are stream surfacus which cut the flow passage in the blade-to-blade direction as
shown in Figure 4a. These surfaces are not surfaces of revolution but are a second kind of control surface
which may be used to subdivide the total flow field into blade-to-blade stream tubes. S 2 surfaces are
through-flow surfaces which intersect both the hub and casing as shown in Figure 4b. Notice that families
of S 2 eurfaces might be used as a third kind of control surface to divide the blade-to-blade flow into
hub-to-shroud (casing) stream tubes.

A second nearly universal assumption in through-flow calculations, except for Pome cases concerned
with cooled turbines, is that the flow is adiabatic so that no heat is transferred between fluid elements
or across stream surfaces.
The terms axisymmetric flow and circumferential-average flow are frequently encountered in,.descrip-
tions of assumed conditicns in through-flow toethods. The term axisymmetric flow is interpreted to mean
that for any given set of axial and radial coordinate values in the flow field, no velocity component or
fluid property varies in the circumferential direction. The term circumferential-average flow is usually
used in reference to the flow on some designated S 2 surface (or approximation to an S2 surface). It is
taken to mean that the fluid velocity and properties at all coordinate points on that surface are correct
circumferential averages of the values across the blade-to-blade passage at the corresponding radius and
axial location. The significance of the two assumptions and the distinction between them ij discussed by
Wu (1952a), Smith (1966). Horlock and Marsh (1971) and Sehra (1979). One practical reason for ei~her
assumption is to permit or justify the subsequent modeling of S1 surfaces as surfaces of revolution
(axisymmetric S1 surface approximations). This leads further to support of the use of linear cascade data
in axial-flow turbomachine design, flow passage circumferential measurement surveys at constant radius,
and the use of stream surface intersections with the meridional plane to show the nature of the hub-to-
casing flow pattern.
A second reason for assuming the flow to be axisymmetric or for considering the flow on a particular
approximation to an S2 surface to be the circumferential average is that this allows one hub-to-shroud
surface to be considered as representative. This obviously reduces the computational complexity of the
through-flow analysis, along with the potential for prediý:ting the true complexity of the flou.

CLASSIFICATION OF THROUGH-FLOW METHODS

The sequence of development of through-flow calculation systems is noi: easy to follow. There is
evidence of numerous apparently independent efforts leading to rerArkably similar approaches to the prob-
lem. In this section a brief review of the overall pattern of progression will be given. The emphasis
is on the development of calc',lation systems primarily directed toward determinatin' of the hub-to-shroud
variations in the flow field. The elements involved in blade-to-blade flow field de~ermination are con-
tained in other contributions to the Working Group 12 report, and they are automatically included in
those through-flow methods that are three-dimensional in character.

In studying through-flow calculation methods, it quickly becomes evident that some confusion may
arise be,'ause of minor differences in the terminology used to classify individual methods into groups.
In the following, six principal classes are used:
 287

(1) Mean-line method. For the flow passage under consideration, a mean line is defined. By one-
dimensional calculations, heavily supported by empirical correlations of experimenevý data,
fluid properties and velocities are estimated at specified locations on the mean line. These
conditions are assumed representative of the passage at that location along the mean lih•.

(2) Simple radial equilibrium method. A meridional flow passage is subdivided by a set of plane
control surfaces perpendicular to the rotational axis. In each control surface the effects of
any radial component of velocity are assumed negligible. The governing physical laws and
empirical data are used to determine distribution of velocity and properties. This method is
limited to assumed axisymmetric flows in axial turbomachines.

(3) Streamline curvature method. This analysis method class Is based on the direct determination,
through iterative calculation, of the traces or projections of streamlines on a prescribed
surface, for example on a meridional section of a turbomachine. An initial set of streamlines
is assumed, and by successive iterative calculations, with each iteration improving the degree
of satisfaction of the physical laws and empirical data correlations used in the flow model, a
final set of streamlines is determined for a given flow rate, rotational speed, and entrance
condition combination (see Figure 5). Velocities and fluid properties are predicted at loca-
tions on the streamlines corresponding to control surface intersections.

(4) Finite-difference method. This analysis grouping is based on the initial definition of a system
of grid points or nodes throughout the flow passage under study (see Figure 6). For the combina-
tion of physical laws and empirical information to be satisfied in the method, difference
equations are written for each grid point. The equations are solved to determine the distribu-
tions of fluid properties and velocities at the nodal points. These distributions can be used
to construct streamline patterns for the given passage operating condition.

(5) Finite element method. The flow passage to be studied is subdivided by a network of lines into
elements. Nodal points are located on the lines forming the boundary of each element. The
physical laws and empirical input are formulated so that the fluid properties and velocities
may be determined at each nodal point by iteration. Figure 7 shows an example set of elements.

(6) Finite volume method. The flow passage is subdivided into a large number of small volume ele-
ments, fitted to the passage geometry as in Figure 8. The physical laws and empirical informa-
tion governing the calculation are used to generate integral conservation equations for each
small volume. The equations are solved, typically by an iterative time-marching process to
reach a converged distribution of velocities and properties for the finite volume surfaces.

The vast majority of through-flow calculation methods may be placed within these classes.

SUMMARY OF THROUGH-FLOW CALCULATION DEVELOPMENT

This section will briefly describe the pattern of advancement leading toward current and projected
through-flow calculation systems. The BIBLIOGRAPHY contains references to more extensive and detailed
supporting documents.

The earliest through-flow calculation methods to be routinely used were mean-line analyses, and this
class of methods continues to be developed and used for preliminary estimation of performance [e.g. Howell
and Calvert (1978)]. The one-dimensional nature of mean-line analysis, with the necessary empiricism in
estimation of the turning and loss characteristics of blade rows, eliminates the possibility for realistic
flow field prediction.

Simple Radial Equilibrium Methods

As mentioned earlier, the differential equations of motion and the other relationships necessary for
a through-flow field calculation were formulated in reference to a turbomachine problem during the early
part of this century. However thp first attempts to compute axial-flow turbomachine flow patterns used
the simple-radial-equilibrium approximation. The idea of inviscid simple radial equilibrium was certainly
understood in relation to the design point flow field before 1930 [Howell (1976)], and it was in common
but not universal use for design by 1945. In design point application the equation was used without
allowance for radial variation in blade row loss [Serovy (1966), Voit (1953)]. The significance of the
cumulative loss effect as measured by the radial gradient of entropy at a control surface was evaluated
by Hatch, Giamati and Jackson (1954), and later application to off-design flow field performance map pre-
diction for compressors and pumps was bnsed on non-isentropic simple radial equilibrium [Seroy (1958),
Serovy, Kavanagh, okiishi and Miller (1973)].

Streamline Curvature Methods

The influence on the hub-to-shroud flow patterns due to the terms in the radial componcnt of the
equation of motion representing streamline slope (due to radial velocity) and curvature (due to radial
velocity change) was first described by Traupel (1942), and later by Wu and Wolfenstein (1950) and
Hamricl., Ginsburg and Osborn (1950). Mechanisms for quantitative inclusion of the streamline shape
effect in compressor and turbine design were reported by Smith, Traugott and Wislicenus (1953), Wright
and Kovach (1953), and Hatch and Bernatowicz (1957). In design procedures used during this period there
was little if any combined accounting for streamline shape and radial entropy gradient, although the pos-
sibility was mentioned by J. W. McBride in a discussion to the Smith et al. (1953) paper and by Giamati

Sj
288

and Finger (1965). However, in performance prediction Swan (1958) included both streamline shape and loss
distribution.

The term streamline curvature method became generally associated with this class of through-flow cal-
culation following the work of Novak (1967) and Jansen and Moffatt (1967).

Although Wu (1952a) provided the basis for a hub-to-shroud flow field calculation including locations
internal to blade rows and some design procedures tracing streamlines through blade rows were reported
during the 1950s, performance prediction capability was first developed in about 1959 [Smith (1966)],
Smith also discussed the axisymmetric/circurferential-average flow question. Figure 5 shows a typical
design point axial-flow compressor streamline set as computed by Smith's method.

Between 1960 and 1976 a oubstantial number of zomputer code; for s'dal-flow turbomachine flow field
analysis were developed using streamline curvature methods. MOSL of these codes were based on assumption
of axisy-metric, inviscid, adiabatic flow. Some of these codes permitted computation control surfaces
internal to blade rows, but almost all depended on correlations of experimental linear and annular cascade
results, including correlations of compressor and turbine test data for blade row turning and loss.

The assumption of inviscid flow in streamline curvature through-flow methods generally meant only
that the local fluid sheat stress effects were neglected by eliminating the corresponding terms in the
momentum equations. However, the cumulative effect of fluid sheer as meas-red by the entropy increase
through a blade-to-blade stream tube was included through the hub-to-tip entropy variation.
Wu (1952a and 1980), Smith (1966), Horlock (1971), Wennerstrom (1974), and Bosaman and Marsh (1974)
have suggested similar mechanisms for removal of the inconsistency restilting from the local inviscid
assumption by introducing correction terms to simulate the effect of fluid shearing stress in generation
of irreversibility along fluid particle paths.

Finite Difference Methods

Finite difference methods for turbomachine through-flow analysis were first suggested and utilized
between 1950 and 1960 (Wu (1950a, 1952a,b,c, and 1953)]. Figure 6 shows a grid point system used to
predict a design point flow field by a finite difference method using relaxation solution techniques.
Theterm matrix through-flow method began to be commonly used to describe one group of finite dif-
ference methods following the work of Marsh (1968). Additional results from finite difference through-
'low methods have been reported by Davis and Millar (1975 and 1976a), Biniaris (1975), Lindley et al.
(1970) and Smith (1974).

Finite difference methods have demonstrated some advantages in calculations where internal blade row
calculation locations are needed, and where three-dimensional effects are under study [Bosman and El-Shaarawi
(1977)]. However the methods have not been as effectively developed for transonic flow cases as have some
uf the streamline curvature methods.
The finite difference methods developed up to about 1972 were of the inviscid type, and like the
streamline curvature methods, included only the accumulated effects of viscous effects as an entropy dis-
tribution in the computed flow field. Loss correlations based on experiments were the basis for estimation
of the entropy variation. More recent finite difference work [Bosman and Marsh (1974)) has included a
mechanism for local viscous shear stress accounting.

Finite Element Methods

Finite element methods initially appeared as usable through-flow analysis techniques during the middle
1970s [Hirsch and Warzee (1974), Adler and Krimerman (1974)]. A key feature of the finite element method
is its potential for development into a three-dimensional analysis system. Considerable progress has been
made in this direction by Hirsch and Warzee (1979 and 1980).

The finite element mesh can be adapted readily to quite complex turbomachine geometries as shown in
Figure 7. The same forms of correlation for turning and loss used in streamline curvature methods can and
I
have been used in finite element analyses.

Finite Volume Methods

Along with the finite element methoc.s, finite volume methods have a relatively short history. First
proposed for blade-to-blade flow field analysis, recent investigations have focused on the hub-to-tip flow
field, and on the strong potential ior extension to an effective three-dimensional analysis ý.,thod.
Results have been reported by Denton and Singh (1979), Farn and Whirlow (1977), and Ivanov and Kimasov
(1978).

The finite volume regions can, as in the case of the finite element mesh, be generated to conform to
complex passage geometries. Figure 8 shows a three-dimensional example case.
 "N289

Advanced Through-Flow Approaches

A number of additionel"through-flow analysis methods are in current use or under study. While the
five classes of methods discussed above might be located in the LEVEL II category of Figure 1, all of the
methods referenced in this section should be classiied in Level III.

Dodge (1977) and Dodge and Lieber (1977 and 1978) have reported substantial use of a group of blade
row flow field analyses programs. These techniques are useful in configuration development where detailed
flow patterns not ordinarily available from Level II methods are needed internal to blade rows. Moore and
Moore (1980a,b), Sovrano (1980) and Walitt (1980) have developed through-flow calculation systems for
centrifugal compressor sttidies which appear to offer alternative capability in axial-flow turbomachine
blade rows.
For unsteady through-flow calculations Erdos, Alzner and McNally (1977) have developed a time-
dependent, inviscid computation system with capability in the transonic flow regime.

PROBLEM AREAS COMMON TO ALL THROUGH-FLOW METHODS

Aerodynamic Questions

Almost all of the through-flow calculation methods which have reached the status of routine use are
in the Level II category of Figure 1. Except for a very limited number of finite element and finite
volume methods, they are quasi-three-dimensional in character. This means that iteration occurs between
two-dimensional hub-to-shrcud computation and either two-dimensional blade-to-blade computation or corre-
lations of experimental turning and loss data. The term two-dimensional as used here is not as restrictive
as is ordinarily implied in, for example, linear cascade studies. In the two-dimensional through-flow case
a blade-to-blade stream tube may change radius and vary in radial thickness. Variation in properties and
velocity occur in the 0 and x directions (Figures 4a and 4b), but the flow in each blade-to-blade stream
tube is assumed to be independent of the flow in adjacent tubes. For this reason, the quasi-three-dimen-
sional methods are not well adapted to accounting for passage three-dimensional aerodynamic flow features
including flow separation, choking, shock waves, and end-wall, clearance and corner flows. Abundant
evidence of the deficiencies in quasi-three-dimensional methods is available in reported work on non-
axisymmetric streamline curvature and finite difference computation [e.g. Novak and Hearsey (1977)]. Some
unusual strategies have been developed such as the introduction of an additional set of computational
surfaces by Katsanis and McNally (1977) as shown in Figure 9. Although a considerable history now exists
in the development of through-flow methods, it should be noted that flow models and equations remain con-
troversial in some areas. This is recalled to attention by recently published work of Wu (1980). concerned
with the modeling of the effects of shear stresses within blade rows.

For quasi-tIree-dimensional and three-dimensional methods in the Level II cotegory of Figure 1,

questions arise when fluid turning and loss data is used at blade row internal locations. These questions
originate in both the scarcity of data and the mechanism for use of correlations.

Compitational Questions

Reports on through-flow computation code development and application frequently do not pvovide ade-
quate information concerning numerical methods and code organization. Because the solutions are iterative
in character, the iterative strategy is important and methods used to improve stability at convergence
significantly influence the cost and quality of results.

Table I. Equations for turbomachine through-flow calculation

PHYSICAL LAW EQUATION FORM EXAMPLE SOURCES

Conservation of linear momentum Differential equations (3 components) Wu and Wolfenstein (1950)

Wu (1952a)

Integral equation for fluid volume Denton and Singh (1979)

element Farn and Whirlow (1977)
Conservation of moment of momentum Differential equation (I component) Wu (1952a)
(Euler turbine equation) Integral equation -
fluid volume element Denton and Singh (1979)
stream tube Wennerstrom (1974)
Hearsey (1975)
Conservation of mass Differential equation Wu (1952a)
Marsh (1976)
-
Integral equation
fluid volume element Denton and Singh (1979)
Farn and Whirlow (1977)
stream tube Wu and Wolfenstein (1950)
control surface Hearsey (1975)
290

Table I. Continued

PHYSICAL LAW EQUATION FORM EXAMPLE SOURCES

First law of thermodynamics Differential equation Wu (1952a)

Integral equation -
finite volume element Denton and Singh (1979)
Farn and Whirlow (1977)
stream tube Wu and Wolfenstein (1950)
Hearaey (1975)
Second law of thermodynamics Differential equation Wu and Wolfenstein (1950)
Wu (1980)

)I

!I
*v x
 291

BIBLIOGRAPHY

Adler, Dan. 1979. Status of centrifugal impeller internal aerodynamics: Experiments and calculations.
U.S. Naval Postgraduate School. NPS67-79-004.

Adler, Dan. 1980a. Status of centrifugal impeller internal aerodynamics. Part I: Inviscid flow pre-
diction methods. Journal of Engineering for Power, Transactions of the ASME 102: 728-737.

Adler, Dan. 1980b. Status of centrifugal impeller internal aerodynamics. Part II: Experiments and
influence of viscosity. Journal of Engineering for Power, Transactions of the ASME 102: 738-746.

Adler, D., and Y. Krimerman. 1974. The numerical calculation of the meridional tlow field in turbomachines
using the finite elements method. Israel Journal Tech. 12: 268-274.

Adler, D., and Y. Krimerman. 1977. Calculation of the blade-to-blade compressible flow field in turbo
impellers using the finite-element method. Journal of Mechanical Engineering Science 19: 108-112.

Adler, D., and Y. Krimerman. 1980a. Compurison between the calculated subsonic inviscid three-dimensional
flow in a centrifugal impeller and measurements. Pages 19-26 in Performance prediction of centrifugal
pumps and compressors. New York. ASME.

Adler, D., and Y. Krimerman. 1980b. On the relevance of inviscid subsonic flow calculations to real
centrifugal impellers flow. Journal of Fluids Engineering, Transactions of ASME 102: 78-84.

Araki, T. 1974. Three-dimensional flow through turbines. Inneat and Fluid Flow in Steam and Gas Turbine
Plant. IME Conf. Proc. 3: 38-45. (Discussion J. D. Denton, pp. 290-292.)

Balsa, T. F., and G. L. Mellor. 1975. The simulation of axial compressor performance using an annulus
wall boundary layer theory. Journal of Engineering for Power, Transactions of the ASME 97: 305-318.

Bauersfeld, W. 1905. Zuschrift an die Redaktion. V.D.I. Zeitschrift. 49: 2007-2008.

Bauersfeld, W. 1922. Die Grundlagen zur Berechnung schnellaufender Kreiselrader. V.D.I. Zeitschrift.
66: 461-465, 514-517.

Biniaris, S. 1975. The calculation of the quiasi-three-dimensional flow in an axial gas turbine. Journal
of Engineering for Power, Transactions of the ASME 97: 283-294.

Bosman, C. 1973. The occurrence and removal of indeterminacy in flow calculations in turbomachinery.
ARC Rep. and Memo. 3746.

Bosman, C. 1980. An analysis of three-dimensional flow in a centrifugal impeller. Journal of Engineering

for Power, Transactions of the ASME 102: 619-625.

Bosman, C., and M. A. I. El-Shaarawi. 1977. Quasi-three-dimensional numerical solution of flow in turbo-
machines. Journal of Fluids Engineering, Transactions of the ASME 99: 132-140.

Bosman, C., and J. Highton. 1979. A calculation procedure for three-dimensional, time-dependent,
inviscid, compressible flow through turbomachine blades of any geometry. Journal of Mechanical
Engineering Science 21: 39-49.

Bosman, C., M. Hill, and L. Deshpande. 1974. The effect of damping factor on the behavior of flow cal-
culations in turbomachines. ARC Rep. and Memo. 3766.

Bosman, C., and H. Marsh. 1974. An improved method for calculating the flow in turbomachines, including
a consistent loss model. Journal of Mechanical Engineering Science 16: 25-31.

Bryans, A. C., and M. L. Miller. 1967. Computer program for design of multistage axial-flow compressors.
NASA CR-54530.

Burger, G. D., D. Lee, and D. W. Snow. 1979. Study of blade aspect ratio on a compressor front stage.
Aerodynamic and Mechanical Design Report. NASA CR-159555.
Calmon, J. 1969. L'Ecoulement d'un fluide compressible dans ua htage de turbomachine. Entropie. No.
27, pp. 15-20, Mar-Juin. No. 28, pp. 38-44, Juillet-Aout. No. 29, pp. 19-27, Septembre-Octobre.

Carter, Anthony F. 1976. A critical review of turbine flow calculation procedures, In Through-flow
calculations in axial turbomachinery. AGARD-CP-195. Paper 9.

Carter, A. F., and F. K. Lenherr. 1969. Analysis of geometry and design-point performance of nxial-flow
turbines using specified meridional velocity gradients. NASA CR-1456.

Carter, A. F., M. Platt, and F. K. Lenherr. 1968. Analysis of geometry and design point performance of
axial flow turbines. I - Development of the analysis method and the loss coefficient correlation.
NASA-CR-1181.

Caruthers, John E., and Theodore F. McKain. 1976. Through-flow calculations: Theory and practice in
turbomachine design. In Through-flow calculations in axial turbomachinery. AGARD-CP-195. Paper 4.

Celikovský, Karel. 1965. The solution of the direct problem of the viscous and compressible flow through
an axial turbomachine under the axial-symmetry conditions. Kandiditsk" Dizertacni Price-VAAZ Brno.
(In CZECH).

p. N
292

Celikovský, Karel. 1970. Model investigati a of subsonic stages of an axial compressor (summavy report).
Zpr~va VZLO, ARTI Report Z-14.

Celikovsk4, Karel. 1974. Research into the axia flow transonic compressor stages. Zpriva VZLO, ARTI
Report Z-19.

Cox, H. J. A. 1976. Through-flow calculation procedur a for application to high speed large turbines.
In Through-flow calculations in axial turbomachiner AGARD-CP-195. Paper 7.

Creveling, H. F., and R. H. Carmody. 1968a. Axial flow com ossor design computer programs incorporating
full radial equilibrium. Part I - Flow path and radial d tribution,of energy specified (Program
II). NASA CR-54532.

Creveling, H. F., and R. H. Carmody. 1968b. Axial flow compressor'design computer programs incorporating
full radial equilibrium, Part I1 - Radial distribution of total pressure and flow path or axial
velocity ratio specified (Program I11). NASA CR-54531.
Creveling, H. F., and R. H. Carmody. 1968c. Compressor computer program for calculating off-design per-
formance (Program IV). NASA CR-72427.

Dallenbach, F. 1961. The aeroeynamic design and performance of centrifugal and mixed-flow compressors,
In Centrifugal compressors. SAE Tech. Progress Series, No. 3. pp. 2-30.

Daneshyar, H. 1970. The off-desiga analysis of flow in axial compressors.

University of Cambridge. CUEr!A-Turbo/TR 19.
Department of Engineering,
9
Davis, W. R. 1968. A comparison of different forms of the quasi-three-dimensional radial equilibrium
equation of turbomnchines. Carleton University, Div. of Aerothermodynamics Report No. ME/A 68-1.

Davis, W. R. 1970. A computer program for the analysis and design of the flow in turbomachinery. Part
B - Loss and deviation correlations. Carleton University. Div. of Aerothermodynamics Report No.
ME/A 70-1.

Davis, W. R. 1971. A computer program for the analysis and design of turbomachinery--Revision. Carleton
University. Div. of Aerothermodynamics Report No. ME/A 71-5.
Davis, W. R. 1975. A general finite difference technique for the compressible flow in the meridional
plane of centrifugal turbomachinery. ASME Paper 75-GT-121.

Davis, W. R., and D. A. J. Millar. 1972. A discussion of the Marsh matrix technique applied to fluid
flow problems, C.A.S.I. Trans. 5 k2): 64-70.

Davis, W. R., and D. A. J. Pillar. 1973. Axial tlow compressor analysis using a matrix method--revision.
Carleton University. Div, of Aernthermodynamics Report No. NE/A 73-1.

Davis, W. R., and D. A. .. Miller. 1975. A comparison of the matrix and streemline curvature methods of
axial flow turbomachinery analysis, from a user's point of view. Journal of Zngineering for Power,
Transactions of che ASME 97: 549-560.

Davis, W. Roland, and D. A. J. Millar. 1976a. Through flow calculations based on matrix inversion: Loss
prediction. In Through-flow calculations in axial turbomachinery. AGARD-CP-195, Paper 3.

Davis, W. R., and D. A. J. Millar. 1976b. Axial flow compressor analysis using a streamline curvature
method. Carleton University. Div. of Aerothermodynamica Report No. ME/A 76-3.

Denton, John D. 1978. Throughflow calculations for Lransonic axial flow turbines. Journal of Engineering
for Power, Transaction& of the ASME 100: 21k .218.

Denton, J. D., and U. K. bingh. 1979. Time marching methods for turbomachinery flow calculations, von
Karman Institute. Lecture Series, Application of n'Americal methods to flow calculations in turbo-
machines.

DeRuyck, J., C. Hirsch, and P. Kool. 1979. An axial compressor end-well boundary layer calculation
method. Journal of Engineering for Power, Transactions of the ASME 101: 233-249.

Deshpande, R. B, 1975. A new algorithm for the solution of turbomachinery flow problems. Journal of
Fluids Engineering, Transactions of the ASME 97; 372-374.

Dodge, Paul R. 1976. A non-orthogonal transonic relaxation method for cascade flows. ASHE Paper 76-GT.-
63.

Dodge, P. R. 1977. Numerical method for 2D and 3D viscous flows. AIAA Journal 15: 961-965.

Dodge, P. R., and L. S. Lieber. 1977. A numerical method for the solution of the Navier-Stokes equations
with separated flow. AIAA Paper 77-170.

Dodge, P. R., and L. S. Lieber. 1978. 3-D flow analysis of compressor cascade with splitter vanes.
AFAPL-TR-78-23.

Dorfman, L. A. 1974. Chislennyye Metcdy v Gazodinamike Turbomashin (Numerical Methods in Gas Dynamic
Turbomachines). Izd-vo Enetgiya, Lenitugradskoye Otdeleniye.
 293,
I Doyle, H. P. C. 1971.
Advanced through-flow analysis applied to a low speed axial flow compressor.
Intl. Jour. Mech. Engr. Sci. 13: 833-842.

Doyle, V. L., end C. C. Koch. 1970. Evaluation of raiig" and distortion tolerance for high Mach number
tran3onic fan stages, design report. NASA CR-72720.

Dreyfus, L. A. 1946. A three-dimensional theory of turbine flow and its application to the design of
wheel vanes for Francis and propeller turbines. Acta Polytechnia, Mech. Engr. Series, Vol. 1, No. 3.

Duncombe, E. 1964. Aerodynamic design of axial flow turbines. In Aerodynamics of turbines md compres-
sors. Princeton, Princeton Uaiv. Press. pp. 433-523.

Dunker, R. J., P. E. Strinning, and H. B. Weyer. 1978. Experimental study of the flow field within a
transonic axial compressor rotor by laser velocimetry and comparison with through-flow calculations.
Journal of Engineering for Power, Transacticns of the ASME 100: 279-286.

Eckert, E., and G. K. Korbacher. 1967. The flow through axial turbine stages of large radial blrde
length. NACA TM-1118.

Erdos, John I., and Edgar Alzner. 1977. Computation of unsteady transonic flows through rotating and
stationary cascades. I - Method of analysis. NASA CR-2900.

Erdos, John, Edgar Alzner, and Paul Kalben. 1977. Computation of steady and periodic two-dimensional
non-linear transonic flows in fan and compressor stages. In Transonic flow problems ir turuomachinery.
Washington, D.C. Hemisphere Publ. Corp. pp. 95-111.

Erdos, John, Edgar Alzner, Paul Kalben, William McNally, and Simon Slutsky. 1975. Time-dependent
transonic flow solutions for axial turbomachinery. in Aerodynamic analyses requiring advanced
computers. NASA SP-347, Part 1. pp. 587-621.

Erdos, J. I., E. Alzner, and W. McNally. 1977. Solution of periodic transonic flow through a fan stage.
AIAA Journal 15: 1559-1568.

Farn, C. L., and D. K. Whirlow. 1977. Application of time-dependent finite volume method to transonic
flow in large turbines. In Transonic flow problems in turbomachinery. Washington, D.C., Hemisphere
Publ. Corp. pp. 208-227.
Flagg, E. E. 1967. Analytical procedure and computer program for determining the off-design performance
of axial flow tLrbines. NASA CR-710.

Frost, D. H. 1970. A streamline curvature through flow computer program for analyzing the flow through
axia] flow turbomachines. ARC Rep. and Memo. 3687.

Fruehauf, Hans-Heiner. 1977. Applicability of axisymmetric analysis in predicting supersonic flow

through annular cascades. Journal of EngJneering for Power, Transactions of the ASME 99: 115-120.

Giamati, G. C., Jr., and H. B. Finger. 1965. Design velocity distribution in meridional plane. In
Aerodynamic design of axial-flow compressors. NASA SP-36. Chapter VIII,

Glenny, D. E. 1974. An application of streamline curvature methods to the calculation of flow in a

multistage axial compressor. Aere. Res. labs (Australia). Note ARL/M.E. 346.
Glenny, D. E. 1975. Periormance ejaluation of a multistage axial compressor using a streamline curvature
model. ASME Paper 75-WA/GT-IO.
Coulas, A. Flow in centrifugal compressor impellers--A theoretical and experimental study.

Goulas, Apostoles, and Roger C. Baker. 1978. Through flow analysis of viscous and turbulent flows.
Aero. Res. Council C.P. 1382.

Goulas, A., and R. C. Baker. 1978. Through flow analysis of centrifugal compressors. ASME Paper 78-GT-
110.
Grahl, K. G. 1972. Prediction of compressor aerodynamic performance for multistage compressor. Report
29. Dynalysis of Princeton, Princeton, NJ.

Grahl, Klaus G. 1972. Teillasbherechnung fur Axialverdichterstufen. Zeit. Flugwiss. 20: 42-51.

Grahl, Klaus G. Uber den Stand Der Kennfeldberechnung Mehrstufiger Axialverdichter. Zeit. Flugwiss.
Weltraumforsch. 1: 29-41.

Guiraud, J. P., and R. Kb. Zeytounian. 1971-2. Sur la structure des 4coulements tourbillonnaires dana
les turbomachines axiales. La Rech. Aeron. No. 1971-2, pp. 65-87.
Guiraud, J. P., and K. Kh. Zeytounian. Analyse de l'4coulement V'entree et la sortie d'une roue. La
Recb. Aeron. No. 1971-5, pp. 237-256.

Hajek, T., and S. Gopalakrishnan. 1980. Prediction of flow in centrifugal impellers. In Performance
prediction of centrifugal pumps and compreEsors. New York. ASME pp. 27-31.

Hamrick, Joseph T., Ambrose Ginsburg, and Walter M. Osborn. 1950. Method of analysis for compressible
flow through mixed-flow centrifugal impellers of arbitrary design. NACA Report 1082.
294

Hatch, J. E., and D. T. Bernatowicz. 195;. Aerodynamic design and overall performance of first spool of
a 24-inch two-spool transonic compressor. NACA RM E56L07a.

Hatch, J. E., G. C. (iamati, and R. J. Jackson. 1914. Application of radial-equilibrium condition to

axial-flow turbemachire design including consideration of change of entropy with radius downscream
of blade row. NACA RM L54A20.

Hearsey, R. ' 1975. A revised computer program for axial compressor design. Volume I: Theory, des-
criptions a.'4 user's instructions. AR Tr 75-0001t Volume I. 1975. Volume II: Program listing
and program usr ARL TR 75-0001, Volume 1975. axample. fo.
Herzig, H. Z., and A. G. Htieen. 1965.
Three-dimensional compressor flow t:heory and real flow effects.
nAerodynamic design ol ayial-flow compressors.
n NASA SP-36. Chapter XIV.

Herzog, Josef. 1974. Calculation of flow dnstribuoton in large radius ratio stages of axial flow turbines
and comparison of theory and experiment. In Fluid mechanics, acoustics, and design of turbomach1nery.
NASA SP-304, Part II, pp. 565-580.

Hetheringron, R. 1970. Computer calculations of the flow in axial compressors. In Internal aerodynamics
(turbomachinery). London. IME. Paper 6, pp. 57-63.

Hirsch, Ch. 1974. End-wall boundary layers in axial compressors. Journal of Engineering for Power,
Tranbactions of the ASME 96: 413-426.

Hirsch, Ch. 1976a. Dusteady contributions to stehdy radial equlibrium flow equations. cnUnsteady c
In
phenomena in turbomacilnes. AGARD-CP-177, Paper 13.

Hirsch, Ch. 1976b. Finite element method for through-flow calculatlons. In Through-flow calculations in
axial turbomachilnery. AGARD-CP-195. Paper 5.
Hirsch, Ch. 1976c. Flow prediction in axial flow compressors including end wall boundary layers. ASME
Paper 76-GT-72.

Hirsch, Ch. Lacor, and G. Warzee. 1980. Threem-nimensional inviscid calculations in centrifugal com-
Jpressors. Centrifugl compressors, flow phenomena and performance.
Idn AGARD-CP-282. Paper 7.

Hirsch, Ch., and G. Warzee. 1974. A finite element method flow calculations in turbomachinen . Vrije
Universiteit Brussel, Dienst Stromingsmechanica. Report VUB-STR-5.

Hirsch, Ch., and C. Warzee. 1976. A finite element method for throngh-flowc alculations in turbomrchones.
Journal of Fluids Engineering, Transactions of the ASME 98: 403-421.
Hirsch, Ch., and G. Warzee. 1977. Blade-to-blade subsonic fluw calculation with finite elements. VriJe
Universiteit Brussel, Dienst Stromingsmechanica. Report VUB-STR-7.

Hirsch, Ch., and G. Warzee. 1979. An integrate-I quasi-3D finite element calculation program for turbo-
machinery flows. Journal of Engineering for Power, Transactions of the AsME 101: 141-148.

Hirsch, Ch., and G. Warzee. 1980. Quasi 3-D finite element computation of flows in centrifugal compres-
sore. In Performance prediction of centrifugal pumps and compressors. New York. ASME. pp. 69-75.

Holman, F_ F., J. R. Kidwell, and T. C. Ware. 1976. Small axial compressor technology program. Volume
I. NASA CR-134827.

Holmquist, C. 0., and W. D. Rannie. 1956. An approximate method of calculating three-dimensional compres-
sible flow in axial turbomachines. Journal of Aeronautical Sciences 23: 543-556, 582.

Horlock, J. H. 1971. On entropy production in adiabatic flow in turbomachines. Journal of Basic Engi-
neering, Transactions of the ASME 93: 587-593.

Horlock, J. H., and H. Marsh. 1971. Flow models for turbomachines. Journal of Mechanical Engineering
Science 13: 358-368.

Howell, A. R. 1976. Griffith's early ideas on turbomachinery aerodynamics. Aeronautical Journal 80:
521-529.

Howell, A. R., and W. J. Calvert. 1978. A new rtage stackirg technique for axial-flow compressor perform-
ance prediction. Zournal of Engineering fcr Power, Trinsactions of the ASME 100: 698-703.
Ivanov, M. Ya., and Yu. I. Kimasov. 1978. Numerical solution of the problem of determining the average,
axisymmetric flow of an ideal gas througl, turbomacl.ine stages. Fluid Mechanics--Soviet Research 7:
4: 143-152. July-August.

Jansen, W., and W. C. Moffatt. 1967. The off-design analysis of axial-flow compressors. Journal of
"Engineering for Power, Transactions of the ASME 89: 453-462.

Jansen, W. 1970. A method for calculating the flow in a centrifugal impeller when energy gradients are
present. In Internal aerodynamics (turbomachinery). London. IME. Paper 12, pp. 133-146.

Japikse, David. 1976. Review--Progress in numerical turbomachinery analysis. Journal of Fluids Engi-
neering, Transactions of the ASME 98: 592-606.

.4 ..
 295

Johnsen, I. A. 1952. Investigation of a 10-stage subsonic axial-flow research compressor. I - Aerody-

namic design. NACA RM E52B18.

Karlsson, T. 1953. On the influence of radial components of blade forces in axial turbomachines. M.E.
Thesis. Cal. Inst. Tech.

Kateanis, Theodore. 1964. Use of arbitrary quasi-orthogonals for calculating flow distribution in the
meridional plane of u turbomachine. NASA TN D-2546.

Katsanis, Theodore. 1965. Use of arbitrary quasi-orthogonals for calculating flow distribution on a
blade-to-blade surface in a turbomachine. NASA TN D-2809.

Katsanis, Theodore. 1966. Use of arbitrary quasi-orthogonals for calculating flow distribution in a
turbomachine. Journal of Engineering for Power, Transactions of the ASME 88: 197-202.

Katsanis, Theodore.incompressible
dimensional, 1967. A computer program for calculating velocities and streamlines for two-
flow in axial blade rows. NASA TN D-3762.
Katsanis, Theodort. '968. Computer program for calculating velocities and streamlines on a blade-to-
blade stream surface of a turbomachine. NASA TN D-4525.
Kateanis, Theodore. 1969. FORTRAN program for calculating transonic velocities on a blade-to-blade
stream surface of a turbomachine. NASA TN D-5427.

Katsanis, Theodore. 1971. FORTRAN program for qu~si-three-dimensLonal calculation of surface velocities
and choking flow for turbomachine blade rows. NASA TN D-6177.
Katsanis, Theodore, and William D. McNally. 1969. Revised fortran program for calculating velocities and
streamlines on a blade-to-blade stream surface of a turbomachine. NASA TM X-1764.

Katsanis, Theodore, and William D. McNally. 1977. Revised fortran program for calculating velocities and
streamlines on the hub-shroud midchannel stream surface of an axial-, radial-, or mixed-flow turbo-
machine or annular duct. I - User's Manual. NASA TN D-8430. II - Programmer's Manual. NASA TN
D-8431.
Khalil, I., W. Tabakoff, and A. Hamed. 1980. Viscous flow analysis in mixed flow rotors. Journal of
Engineering for Power, Transactions of the ASME 102; 193-201.
Kholshchevnikov, K. V. 1970. Teoriya i raschet aviatsionnykh lopatochnykh mashlin (Theory and design of
aircreft turbomacaines). Moscow. Mashinostroyeniye Press.

Kramer, James J., Norbert 0. Stockman, and Ralph J. Bean. 1962. Non-viscous flow through a puwo impeller
on a blade-to-blade surface of revolution. NASA TN D-108.

Krimerman, Y., and D. Adler. 1978. The complete three-dimensional calculation of the compressible flow
field in turbo impellers. Journal of Mechanical Engineering Science 20: 149-158.

Kurmanov, B. I., G. L. Podvidz, and G. Yu Stepanov. 1977. Analysis of two-dimensional gas flow in turbo-
machinery cascades by means of integral relations. Fluid Dynamics. No. 4: 560-566. July-August.
Lenherr, F. K., et al. The analysis of geometry and design point performance of axial-flow turbines using
specified meridional velocity gradients. Part 2: Design examples. NASA CR-72585.
Lieblein, Seymour. 1954. Review of high-performance axial-flow compiessor blade element theory. NACA RM
E53L22.

Lindley, D., J. Westmoreland, and J. F. T. Whybrow. 1970. Some problems in the apDlication o- meridional
flow programs to thc2design of large steam turbines. In Axial and radial turbomachinery. Proc. IME.
184, Part 3G (II): 38-48.
Liu, H. C., T. C. Booth, and W. A. Tall. 1979. An application of 3-D viscous flow analysis to the design
of a low-aspect-ratio turbine. ASME Paper 79-OT-53.
Lorenz, H. 1905. Theorie und bereehnung der volturbinen und krieselpumpen. V. D. 1. Zeitschrift. 49:
41: 1670-1675.

Luu, T. S., M. M. Abdelrahman, and Y. Ribaud. 1980. Transonic flow analysis in au impeller equipped with
splitter blades. In Performance prediction of centrifugal pumps and comrressors. New York, ASME
pp. 47-60.
McBride, Mark W. 1977. A streamline curvature method of analyzing axisymmetric axial, mixed and radial
flow turbomachinery. Penn. State University, Applied Research Laboratory. TH 77-219.
McBride, Mark W. 1979. The design and analysis of turbomachinery in an incompressible. steady flow using
the streamline curvature method. Penn. State University, Applied Research Labor8tory. TM 79-33.

McDonald, P. W., C. R. Bolt, R. J. Dunker, and H. B. Wayer. 1980, A comparison between measured and
computed flow fields in a transonic cempressor rotor. ASME Paner 80-GI-7.

Macchi, E. 1970. Computer program for prediction of axial flow turbine performanze. 'J.S. Naval Post-
graduate School. NPS-57Ma 7008 1 a.
 296

Harsh, H. 1968. A digital computer program for the through-flow fluid mechanics in an arbitrary
turbomachine using a matrix method. (NGTE Report R282. 1966.) ARC Rep. and Memo. 3509. 1968.

Marsh, H. 1970. The through-flow analysis of axial flow compressors. In Advanced compressors. AGARD-
LS-39-70. Paper 2.
Marsh, H. 1972. The uniqueness of turbomachinery flow calculations using the streamline curvature and
matrix through-flow methods. Journal of Mechanical Engineering Science 13: 376-379. 1971. Also
Hearsey, R. M., and H. Marsh. Communication. Journal of Mechanical Engineering Science 14: 294-296.
1972.

Marsh, H. 1976. Through-flow calculations in axial turbomachinery: A technical point of view, In

Through-flow calculations in axial turbomachinery. AGARD-UP-195. Paper 2.

Messenger, H. E., E. E. Kennedy. 1972. Two-stage fan. I - Aerodynamic and mechanical design. NASA CR-
120859.

Miller, M. L., and A. C. Brians. 1967. Parametric study of advanced multistage axial-flow compressors.
NASA CR-797.
Monearrat, N. T., M. J. Keenan, and P. C. Tramm. 1969. Design report. Single stage evaluation of highly-
loaded high-Mach number compressor stages. NASA CR-72562.

Moore, J., and J. G. Moore. 1980. Three-dimensional, viscous flow calculations for assessing the
thermodynamic performance of centrifugal compressors--Study of the Eckardt compressor.
fugal compressors, flow phenomena and performance. AGARD-CP-282. Paper 9.
In Centri- I
Moore, J., and J. G. Moore. 1980. Calculations of three-dimensional, viscous flow and wake development
in a centrifugal impeller. In Performance prediction of centrifugal pumps and compressors. New
York. ASME. pp. 61-67.

Mukherjee, D. K. 1976. Design of turbine, using distributed or average 7osses; effect of blowing. In
Through-flow calculations in axial turbomachinery. ACARD-CP-195. Peper 8.

Ngo, Vinh-Hai, and D. A. J. Miller. 1973. The design and performance prediction of axial flow turbines.
Carleton University Dept. of Mech. and Aeron. Engrg. Rep. No. ME 73-3.
Novak, R. A. 1967. Streamline curvature computing procedures for fluid-flow problems. Journal of Engi-
neering for Power, Tr. 3actions of the ASME 89: 478-490.

Novak, Richard A. 1976. Flow fivld and performance map computation for axial-flow compressors and
turbines. In Modern prediction methods for turbomachine performance. AGARD-LS-83. Paper 5.

Novak, R. A., and R. H. Hearsey, 1977. A nearly three-dimensional intrablade computing system for turbo-
machinery. Journal of Fluids Engineering, Transactions of the ASME 99: 154-166.
Oates, Gordon C. 1972. Actuator disk theory for incompressible highly rotating flows. Journal of Basic

Engineering, Transactions of the ASME 94: 613-621.

Oates, G. C., and C. J. Knight. 1973. Through-flow theory for turbomachines. AFAPL-TR-73-61.

Oates, G. C., and G. F. Carey. 1974. A variational formulation of the compressible through-flow
problem. AFAPL-TR-74-78.

Oates, G. C., C. J. Knight, and G. F. Carey. 1976. A variational formulation of the compressible
through-flow problem. Journal of Engineering for Power, Transactions of the ASME. 98: 1-8.

Oldham, R. K. 1968. A computer program for predicting the performance of axial flow turbomachines by
the streamline curvature model. NGTE NT717.

Papailiou, Kyriacos D. 1971. Program for the design of an axial compressor stage based on the radial
equilibrium equations. U.S. Naval Postgraduate School. NPS-57PY 71091A.

Platt, M., and A. F. Carter. 1968. Analysis of geometry and design point performance of axial flow
turbines. II - Computer Program, NASA CR-1187.

Podvidz, C. L. 1971. Calculation of the qjasi three-dimensional flow of a gas in the interblade channel
of an axial turbine. Izv. Akad. Nauk SSSR, Mekh. Zhidk. i Gaza. No. 4: 92-101,

Renaudin, A., and E. Somm. 1970. Quasi-three-dimensional flow in a multistage turbine--Calculation and
experimental verification. In Flow research on blad°ig. Amsterdam, Elsevier Publishing Company.

Ribaut, H. 1968. Three-dimensional calculation of flow in turbomachines with the aid of singularities.
Journal uf Engineering for Power, Tvansactions of the ASHE 90: 258-264.

Ribaut, M., and R. Vainto. 1975. On the calculation of two-dimensional subsonic and shock-free transonic
flow. Journal of Engineering for Power, Transactions of the ASME 97: 603-609.

Ribaut, H. 1977. On the calculation of thret-dimensional divergent and rotational flow in turbomachines.
Journals of Fluids Engineering, Transactions of the ASME 99: 187-196.

Richter, W. 1970. Discussion. In Internal aerodynamics (turbomachinery). London. IME. pp. 67-79.

---------.-.--. ,.. - -
 297

Ruden, P. 1944. Investigation of single stage axial fans. NACA TM 1062. (Trans. from Luftfahrtformchung,
14: 325-326, July 20, 1937 and 14: 468-473, Sept. 9, 1937).

Sandercock, D. M., K. Kovach, and S. Lieblein. 195M. Experimental investigation of a five-stage axial-
flow research compressor with transonic rotors in all stages. I - Compressor Design. NACA RM
E54F24.
Scbilhanr-, M. J. 1965. Three-dimensional theory of incompressible and inviscid flow through mixed flow
turbomach.nes. Journal of Engineering for Power, Transactions of the ASME 87: 361-373.

Schr~der, H. J. 1972. Die Wirbelscheibenmethode in Anwendung auf die Auslegung Axialer Turbouaschinen.
VDI-Fortschr~tts-Bericht. Reihe 7. Nr. 28.
Schr~der, H. J., and P. Schuster. 1972. Actuator disc flow calculated by relaxation. An approach to the
aralysis problem of turbomachinery. ASME Paper 72-GT-26.

Sch~tz, J. 1971. Beitrag zur berechnung der str~mung in stufen axialer thermischer turbomaschinen.
Dissertation. TU Darmstadt.

Sehra, Arun K. 1979. Boundary layer and wake modifications to compressor design sys;tems: The effect of
blade-to-blade flow variations on the mean flow field of a transonic rotor. AFAPL-TR-79-2010.

Sehra, A. K., and J. L. Kerrebrock. 1979. The effect of blade-to-blade flow variations on the mean flow
field of a transonic compressor. AIAA Paper 79-1515.

Seippel, C. 1958. Three-dimensional flow in multistage turbines. Brown Boveri Review. 45: 99-107.

Senoo, Y., and Y. Nakase. 1972. An analysis of flow through a mixed flow impeller. Journal of Engineer-
ing for Power, Transactions of the ASME 94: 43-50.

Serovy, George K. 1958. A method for t'e prediction of the off-design performance of axial-flow compres-
sors. Ph.D. Dissertation, Iowa State University, Ames, Iowa.
Serovy, George 'A. 1966. Recent progress in aerodynamic design of axial-flow compressors in the United
States. Journal of Engineering for Power, Transactions of the ASME 88: 251-261.

Serovy, G. K. and E. W. Anderson. 1959. Method for predicting off-design performance of axial-flow
compressor blade rows. NASA TN D-110.
Serovy, G. K., and J. C. Lysen. 1963. Prediction of axial flow turbomachine performance by blade-element
methods. Journal of Engineering for Power, Transactions of the ASME 85: 1-8.

Serovy, George K., Patrick Kavanagh, Theodore H. Okiishi. and Max J. Miller. 1973. Prediction of overall
and blade-element performance for axial-flow pump configuration. NASA CR-2301.

Seyler, D. R., and L. H. Smith, Jr. 1967. Single stage experimental evaluation of high Mach number com-
pressor rotor blading. Part I - Design of rotor blading. NASA CR-54581.

Silvester, M. E., and R. Hetherington. 1966. Three dimensional compressible flow through axial flow
turbomachines. In Numerical analysis: An introduction. New York. Academic Press.

Sinnette, John T., Osmar W. Schey, and J. Austin King. 1943. Performance of NACA eight-stage axial-flow
compressor designed on the basis of airfoil theory. NACA Rep. 758.

Sirotkin, Ya. A. 1961. Calculation of the axisymmetric, vortex flow, inviscid and compressible flow in
axial-flow turbomachines. Izv. Aknd. Nau% SSR. Mekhanika i Mashinostroenie. No. 2. pp. 78-80.
March-April. (Ii Russian.)

Sirotkin, Ya. A. 1972a. Calculation of the twist of long blades in axial turbines and compressors with
an electronic computer. Energomashinostroenie. 18: 4-7.

Sirotkin, Ya. A. 1972b. Aerodnamicheskiy raschet lopatok osevykh turbomashin. Izd-vo Mashinostroyeniye.
Moscow.

Smith, D. J. L. 1974. Computer solution of Wu's equations for compressible flow through turbomachines.
In Flvid mechanics, acoustics and design of turbomachinery. NASA SP-304, Part I: 43-74.

Smith, D. J. L., and J. F. Barnes. 1968. Calculation of fluid motion in axial flow turbomachines. ASME
Paper 68-GT-12.

Smith, D. J. L., and D. H. Frost. 1969-70. Calculation of the flow past turbomachine blades. In Axial
and radial turbomachinery. Proc. IME. 184, Part 3G (II): 72-85.
Smith, L. H., S. C. Traugott, and G. F. Wisltcenus. 1953. A practical solution of a three-dimensional
problem of axial flow turbomachinery. Trans. ASME. 75: 789-803.

Smith, L. H., Jr. 1966. The radial-equilibrium equation of turbomachinery. Journal of Engineering for
Power, Transactions of the ASME 88: 1-12. (Discussion in 88: 281-283. 1966.)
Sovrano, R. 1980. Calcul de llecoulement transaonique dans un compresseur centrifuge par une method
pseudo-instationnaire. In Centrifugal compressors, flow phenomena and performance. AGARD-CP-282.
Paper 5.
 298

Spurr, A. 1976. Progress in the development of a time marching method for through flow calculations.
CEGB •O4/ECH/TG/9. ARC 36907.

Stechkin, B. S., P. K. Kazandihan, L. P. Alekeeyev, A. N. Govorov, Yu. N. Nechayev, and R. M. Fedorov.

1956. Teoriya reaktivnykh dvigateley (Theory of Jet Engines). State Press of the Defense Industry.
Moscow.

Stolts, W. G. 1980. Steam turbine blade path design considerations. In Steam turbines for large power
outputs, von Karman Institute. Lecture Series.

Stepanov, G. Yu. 1962. Gidrodinamika reshotok turbomashin (Hydrodynamics of cascades of turbomachines).

Moscow. Fizmatgiz.

Stockman, Norbert 0., and John L. Kramer. 1963. Method for design of pump impeller using a high-speed
digital computer. NASA TN D-1562.

Stodola, A. 1927. Steam and gas turbines, Volume 11. New York. McGraw-Hill. pp. 990-997.

Struve, E. 1943. Theoretical determlnation of axial fan performance. MACA TM 1042.

Swan, W. C. 1958. A practical engineering solution of the three-dimensional flow in transonic type axial
flow compressors. WADC Technical Report 58-57.

Swan, W. C. 1961. A practical method of predicting transonic-compressor performance. Journal of Engi-

neering for Power, Transactions of the ASME 83: 322-330.

Thiaville, Jean-Marie. 1976. Modeles de calcul de l'ecoulement dens les turbomachines axiales. In
Through-flow calculations in axial turbomachinery. AGARD-CP-195. Paper 1.

Thompkins, William T., Jr., and David A. Oliver. 1976. Three-dimensional flow calculation for a tran-
sonic compressor rotor. In Through-flow calculations in axial turbomachinery. AGARD-CP-195. Paper
6.

Topunov, A. M., and R. D. Ioaifov. 1973. Classification of methods for solving the direct problem of
calculating axisywuetric flow in turbomachines. Soviet Aeronautics. 16: 2, 116-121.

Traupel, W. 1942. Neue allgemeine theorie der mehrstufigen axialen turbomaschinen. ZUrich, Leeman.

Traupel, W. 1977. Thermische turbomaschinen. Volume I. Third Edition. Berlin/Heidelberg/New York.

Springer-Verlag.

Vanco, M. R. 1972. Fortran program for calculating velocities in the meridional plane of a turbomachine.
I - Centrifugal compressor. NASA TN D-6701.

Vazsonyi, Andrew. 1945. On rotational gas flows. Quarterly Appl. Math. III: 1: 29-37.

'vazsonyi, Andrew. 1948. On the aerodynamic design of axial flow compressors and turbines. Journal of
Applied Mechanics 15: 1: 53-64.

Veuillot, Jeer-Pierre. 1977. Calculation of the quasi three-dimensional flow in a turbomachine blade row.
Journal of Engineering for Power, Transactions of the ASME 99. 53-62.

Voit, C. H. 1953. Investigation of a high-pressure-ratio eight-stage axial-flow research compressor with

two transonic inlet stages. I - Aerodynamic Design. NACA RM E53124.

Walitt, Leonard. 1980. Numericsa analysis of the three-4imensional viscous flow field in a centrifugal
impeller. In Centrifugal compressors, flow phenomena and performance. AGARD-CP-282. Paper 6.

Wasielewski, Eugene W. 1940. Design and preliminary tests of NACA axial-flow blower. NACA Memo. Rep.
Dec. 18. 1940.

Wennerstrom, A. J. 1965. Simplified design theory for highly loaded axial ompressor rotors and experi-
mental study of two transonic examples. Mitt. Institute far Thermische Turbomeschinen, No. 12. ETH,
Zurich.

Wennerstrom, Arthur J. 1974. On the treatment of body forces in the radial equilibrium equitior of
turbomachinery. In T:aupel-Festschrift. ZUrich. Jurns-Verlag. pp. 351-367.

Wiggins, J. 0. 1963. A procedure for determining the off-design characteristics oý multistage axial flow
compressors. MS Thesis. University of Cincinnati.

Wilkinson, D. H. 1969-1970. Stability, convergence and accuracy of cwo-dimensional streamline curvature

methods using quasi-orthogonals. In Steady and unsteady flow. Proc. IME. 134, Part 3G (I): 108-119.

Wilkinson, D. H. 1972. Calculation of Llade-to-blade flow in a turbomachine by streamline curvature.

ARC Rep. and Memo. 3704.

Wright, Linwood C., and Karl Kovach. 1953. Design procedure and limited test results for a high solidity,
12-inch transonic impeller with axial discharge. NACA RM E53B09.

Wright, L. C., and R. A. Novak. 1960. Aerodynamic design and developieent of the General Electric
CJ805-23 Aft Fan Component. ASME Paper 60-WA-270.

- -.- - - _ _
 299

Wright, L. C., 11.C. Vitale, T. Ware, and J. R. Erwin. 1973. High-tip speed, low-loading transonic fan
stage. (Part I - Aerodynamic and Mechanical Design). NASA UR-121095.
Wu, Chung-Hua. 1950a. A general through-flow theory of fluid flow with subsonic or Oupersonic velocit).
in turbomachines of arbitrary hub and casing shapes. NACA TN 2302.
Wu, Chung-Hua. 1950b. Formulae and tables of coefficients for numerical differentiation with function
values given at unequally spaced points and application to the solution of partial differential equa-
tions. NACA TN 2214.

Wu, Chung-Hua. 1952a. A general theory of three-dimensional flow in subsonic and supersonic turbomachines
of axial-, radial-, and mixed-flow types. NACA TN 2604.

Wu, Chung-Hua. 1952b. Matrix and relaxation solutions that determine subsonic flow in an axial-flow gas
turbine. NACk TN 2750.

Wu, Chung-Hua. 1952c. A general theory of three-dimensional flow in subsonic and supersonic turbomachines
of axial-, radial-, and mixed-flow types. Transactions of the ASME 74: 1363-1380.

Wu, Chung-Hua. 1953. Subsonic flow of air through a single-stage and a seven-stage compressor. NACA TN
2961.

Wu, Chung-Hua. 1976. Three-dimensional turbomachine flow equations expressed with respect to non-
orthogonal curvilinear coordinates and methods of solution. In Third International Symposium oilAir
Breathing Engines. Munich. pp. 233-252.

Wu, Chung-Hua, and Curtis A. Brown. 1951a. Method of analysis for compressible flow past arbitrary turbo-
machine blades on general surface of revolution. NACA TN 2407.

Wu, Chung-Hua, and Curtis A. Brown. 1951b. A method of designing turbomachine blades with a desirable
thickness distribution for compressible flow along an crbitrary stream filament ol! revolution. NACA
TN 2455.

Wu, Chung-Hua, and Curtis A. Brown. 1952. A theory of the direct and inverse problems of compressible
flow past cascade of arbitrary airfoils. Journal of the Aeronautical Sciences 19: 183-196.

Wu, Chung-Hua, Curtis A. Brown, and Eleanor L. Coatilow. 1952. Analysis of flow in a subsonic mixed-flow
impeller. NACA TN 2749.

Wu, C-H., C. A. Brown, and V. D. Prian. 1952. An approximate method of determining the subsonic flow in
an arbitrary stream filament of revolution cut by arbitrary turbomachine blades. NACA TN 2702.

Wu, Chung-Hua, and Eleanor L. Costilow. 1951. A method of solving the direct and inverse problem of
supersonic flow along arbitrary stream filament of revolution in turbomachines. NACA TN 2492.

Wu, Chung-Hua, John T. Sinnette, Jr., and Robert E. Forrette. 1950. Theoretical effect of inlet hub-tip-
radius ratio and design specific mass flow on design performance of axial-flow compressors. NACA TN
2068.

Wu, Chung-Hua, and Lincoln Wolfenstein. 1950. Application of radial-equilibrium condition to axial-flow
compressor and turbine design. NACA Report 955.

Wu, Wen-quan, Rong-Guo Zhu, and Cui-E Liu. 1980. Computational design of turbomachine blades. Jour.
Aircraft 17: 319-325.

Wu, Zhonghua (Chung-Hua). 1980. Fundamental aerothermodynamic equations for stationary and moving coor-
dinate systems. Engineering Thermophysics in China. 1: 59-94. (Translation cf paper first published
in China in 1965.

Wysong, R. R., T. C. Prince, D. T. Lenahan, et al. 1978. Turbine design system. AFAPL-TR-78-92.

Zhukovskly, M. I. 1967. Aerodinamlcheskiy raschet potoks v osevikh turbomashinakh. (Aerodynamic Calcu-

lation of the Flow in Axial-Flow Turbomachinery). Leningrad. Izd-vo Mashinostroyeniye.

Zhukovskiy, M. I., Ye. Ye. Karyakin, and 0. I. Novikova. 1974. Application of the finite-difference method
for solving the inverse problem of designing turbine blades. Fluid Mechanics - Soviet Research. 3:
6: 15-21. Nov.-Dec.

IALU-~- ---.
 SYSTEM/CYCLE
STUDIES

DESIGN POINT VALUES

SYSTEM OPERATING REQUIREMENTS

DESIGN POINT DESIGN POINT

MEAN LINE ANALYSIS DEVELOP EXISTING UNIT

VELOCITY DIAGRAMS SCALING/SIMILARITY
PRE/BLADE ROW GEOMETRY MODIFY/RESET BLADING
I TOADD
THROUGH-FLOW LEVEL I OR SUBTRACT STAGES

SIMPLE RADIAL EQUILIBRIUM

VELOCITY DIAGRAMS
PRE/BLADE SECTIONS

SI MEAN LINE ANALYSIS

LEVEL I ALTERNATIVE GEOMETRIES

THROUGH-FLOW CALCULATION - LEVEL II

AXISYMMETRIC HUB-TO-TIP SOLUTION

ITERATIVE REDEFINITION - BLADE ROW
GEOMETRY AND AIRFOIL SECTIONS
EXPERIMENTAL CORRELATIONS -
FLUID TURNING
LOSS - DIFUSION LOADING
•rOCK
CAVITATION
MIXING
END-WALL TURNING AND LOSS CORRECTIONS
CHOKING, STALL, CAVITATION MARGINS
i I
LEVEL II - DEVELOPMENT GEOMETRY

THREEG-F0 NSNLCALCULATION
FHRULL-LEOR LEVEL III
SAE OWNAYE

. . °........
COMPUTATIONAL SUPPORT EXPERiMENTAL SUPPORT
BLADE-TO-BLADE FLOW FIELDS t LINEAR CASCADE TESTS
AIRFOIL SECTION IMPROVEMENT ,STAGE AND STAGE-MATCHING TESTS

iFULL-SCALE OR SCALE MODEL BUILD

Figure 1. Typical structure of a desion system for axial-flow turbomachires.

BOUNDARY
VO RTEX
SHOCK .
NF R

l ot or f~
A E DR

STATIONA
1

CRNERL~
ENDE LL
~ ~Figur 2. of aA 7
1 o YER~
Axial~'low
t~pr s--rCoto
hape aor .

IR
UM
ETA
AIE TI
NSUV~
yL
302

MI I
.MM
 303

COMPUTING STATION
(CONTROL SURFACE) NUMBERS ..

.07 .08 .09 .40 .42 .45 .50 .95 1.60 1.65 2.00 2.20 3.00 4.00

2
3TI

v) 6. HUB

7 "ROTOR
9
W-INLET GUIDE'I,
VANE

COMPRESSOR CENTERLINE

Figure 5. Intersection of stream surface approximations with meridional plane section of

axial-flow compressor stage [from Doyle and Koch (1970)]

(in.)
r

+ + + .+ ,,,
l- /•'+ + + + + ++++ + ++ ++ +
t+l.6 6.480
S~~+ ,- ,+ + + • 5.900
',, ',5-585 44U4= 9"/i/llL,/i/ Z / /./ Z ./// .". . . .

F .3 + • + + + ÷
+ +b +•
'9 1.876 ti t" . . t + + + 4+ +" + + +
4.522 + + • + + + ++ + + +- + + + +
4.168 + + + +- " ÷ +÷ "" •

4-3.813 '+ + + +4 + + 4.- + +'

3.!104 4'.4

2.5
0 , 5 . _. . . ._ . + 4 + I- I
2 . - N toQ QDN4aW D4

p~~~~
.4 444NN~w ni
W.J LO W4. 0 to - r- 0

-5 -4 -3 -2 -1 0 1 2 3 4 6 7 8 9
Axial distance, a, in.

Figure 6. Calculation grid points for an axial-discharge, mixed-flow impeller [from Wu. Brown, Coatilow
(1952)].
304

JI-LI

STRE~STATOR

CAIN
BLD
~
Figure
8 ~
CotrtinffntEDGluE
TAIIN eeet o he-ieuoa fo
fiLdAoDuE to fecnan ~g 17)
LEDNGEG
r 305

Flow Bae

Blade -~channel

Figure 9. Channel control surface [Kateanis and McNally (1977)]

 307

Ill.? BLADE-TO-BLAnE COMPUTATIONS AND BOUNuARY LAYER CORRECTIONS

IN AXIAL COMPRESSORS AND TURBINES

111.2.1 Introductior

It is a common practice in industrial firms to evaluate the losses and deviations of

compressor blade cascades by means of empirical correlation formulae derived from experi-
mental results, and improved to a certain extent on the basis of compressor tests.

Due to the development of theoretical studies on cascades, we can now contemplate

associating the calculations carried out in perfect fluids with boundary layer corrections,
to allow for viscous effects on the suction and pressure surface. With such a method, a
more efficient optimization of blade profiles, as well as higher machine performance can
be achieved.

This is the reason why in this paper, we shall survey the existing methods. with par-
ticular emphasis on the association of calculations in perfect fluids and in viscous fluids.

In fact, the proposed method is an extension of Prandtl's boundary layer theory to the
case of cascades. The general Navier-Stokes equations are solved here is a two-step
process
- within the flow, viscous effects are disregarded;
- near the wall, it is admitted that velocity and pressure gradients are imposed by the
inviscid flow in the absence of a boundary layer separation.

Despite this dissociation, the interdependence of viscous and non-viscous effects is

non negligible in a blade cascade channel. As a result the influence of viscosity cannot
be determined "a posteriori" by means of a simple calculation in a perfect fluid. There-
fore, it will be necessary to resort to iterations, to take the interaction effects into
account.
As the methods described in the technical literature are numerous, such a survey will
necessarily be incomplete, and will only address the main families of methods.
In Chapter 11.2.2, we shall review the methods currently recommended for blade-to-
blade calculations in the following configurations

(i) Subsonic conditions

the flow is subsonic in the whole field.

(ii) Supercritical conditions

the flow is quasi-subsonic to the whole field, with, at the utmost, supersonic pockets
which do not extend from the upper surface of a blade to the lower surface of the next
blade; in this case, the entropy variations associated with the passage through a possible
shock wave can be disregarded.

(iii) Supersonic conditions

the supersonic area of the flow occupies at least one full section of an interblade
channel. In this area, recomprassic,•i hock waves are more intense, and the corresponding
entropy jump cannot be disregarded. As the methods described are to be applied only to
existing cascades, we shall only review the direct, inviscid, blade-to-blade computation
methods, and shall not be concerned with the'inverse methods.

Chapter 11.2.3 will be devoted to the description of boundary layer and wake computa-
tion methods. Obviously, these methods are still empirical to a great extent; however,
they apply to more general flow patterns than those which can be solved by means of empiri-
cal loss correlations.
In the next chapter (Chapter 11.2.4), the association methods are described, together
with the reciprocal influence of calculations in perfect fluids and in viscous fluids.

A survey of methods to solve the Navier-Stokes equations is given in Chapter II.Z.5.

111.2.2 Blade-to-blade flow calculations

There are essentially five categories of the blade-to-blade flow calculation methods;
they differ by the numerical techniques used.

111.2.2.1 Singularity method

Singularity methods are based on the prinicple of the superimposition of potential

flow solutions; the potential from which the flow is derived is considered as the sum of
elementary potentials corresponding respectively to
- the basic uniform flow;
- sources, sinks or vortices located in adequately selected points of tha flow field.

A-I -I
L ..... . . ............. ......... ...
 308

The various singularity combinations proposed in the published literature are as

follows I
(1) Singularities located within the blade contour; this method applies essentially to
low camber blades (1).
(2) Singularities made up by vortices located on the blade contour (2). These methods
lead generally to more accu. ate results than the previous ones. They were developed for
flows in incompressible fluids, but can be extended to the case of subsonic flows of com-
pressible fluids, by the addition of sources and sinks in the inter-blade channels, r3p-
resenting the compressibility effect (3), (4), (5).
, With such methods, the precision of the inviscid solution on the trailing edge makes
it possible to determine the direction of the flow issuing from the cascade, with the
help of Kutta-,loukowski or an equivalent condition. This condition can be easily intro-
duced in the computation program. In the case of moderate viscous effects, for instance
in the case of turbine blades at the design point, the theoretical and experimental
results obtained by this method correlate very well (Fig. 1). This mehtod offers an ad-
ditional advantage: due to its high degree of accuracy, it permits a fine analysis of high
velocity gradient zones (in the vicinity of the leading edge, fo instance) which is im-
portant in view of the prediction of boundary layer transition. Its major drawback lies
in the fact that it is limited to the case of subsonic flows with shock-waves.
111.2.2.2 Methods based on the curvature of streamlines
Streamline curvature methods are quite commonly used. Their starting point is a
family of pseudo streamlines deduced, by similarity, from the profile geometry. The
transverse pressure gradients are connected to the curvature of these streumlines; a
transverse velocity distribution is derived, and, by iteration on the continuity equation,
the shape of the streamline is changed until a convergence of the proccss is reached (6),
(7).
The main advantage of this method is the rapidity of the computation, also for sub-
sonic compressible flows. For transonic flows, difficulties arise because of the discon-
tinuity of the streamline curvature at the shocks and because of the ambiguity in the
choice between a subsonic or supersonic solution. Another drawback Pies In the lack of
accuracy in areas of strong curvature (leading and trailing edges). As a result, local
velocity peaks are smoothed out, which limits the capability in predicted boundary layer
transition (Fig. 2).
111.2.2.3 Finite difference or volume methods
It is necessary to distinguish between methods where entropy is assumed to be cons-
tant, and methods where entropy variations are taken into account.

A - Stationary irrotational flows

(i) Relaxation methods using the stream function
The continuity equation allows definition of a stream function which combined with the
condition of irrotationality provides a second order, non linear equation for compressible
flow. This equation is generally discretized in an orthogonal grid, by means of a scheme
suited for an elliptic type problem (subsonic flow).
The finite difference equation is solved by a relaxation technlque (8), (9) or by a ,
matrix technique (10).
This well proven type of method is relatively fast and shows good agreement with ex-
periments (Fig. 3). A second advantage is the ease of extension of these methods to
rotational flow on non cylindrical stream surfaces. The major drawback is the necessity
to specify the outlet flow angle. This difficulty has been overcome at the cost of a
more complex iterative calculation as discussed in Chapter 4. These methods are also
limited to subsonic flows. Extensions to transonic flow, in combination with ;' streamline
curvature technique have not been very successful (11).
(ii) Relaxation methods using the velocity potential

This type of method has first been developed to calculate transonic flow around iso-
lated airfoils. However, the number of references about this method are too numerous to
be all listed here.
The ambiguity between the subsonic and supersonic solution, when using a stream func-
tion has been avoided by the use of a potential function. However, this requires the
assumption of irrotational flow. The continuity equation provides a non-linear, second
order, partial derivation equation which is of the elliptic type in the subsonic field,
and of the hyperbolic type in the supersonic field. Discontinuous solutions can be adop-
ted 'or this equation; although they do not satisfy the Rankine Hugoniot relations for
shock waves. The methods recommended in the published literature differ essentially by
the discretization method adopted :
- it is possible to use mixed discretization schemes, that is centered meshes in the sub-
sonic flow, and excentric meshes in the supersonic field (12), (13); (14)i with this
method, it is necessary to subject to a special treatment the points where discontinuities
appear (shock waves), for the scheme to remain conservative.
- it is possible to put the scheme off center, systematically, by using a term of artifi-
cial viscosity; this type of method leads generAlly to a conservative scheme (15).
 309

-an original method (16) consistz in using a discretizatlon method of the finite
volumc type
Numerical results obtained oy this methou are very satisfactory if the Mach number
upstream of the shock is only slightly supersonic. However, the use of a potential func-
tion makes it difficult to control the mass flow through the flow channels (conservation
problem) which puts a limitation to the prediction of choking. Furthermore, the use of a
potential function does not allow to apply these methods to rotational flow on non cylin-
drical stream surfaces. A comparison of calculated and experimental results Is shown in
figure 4 for a compressor cascade for slightly transonic flow.
(iii) Pseudo non-stationary, isentropic method
'
FThe iterations of the relaxation method can be replaced by
tions time dependent derivatives which do not necessarily have
introducing in the equa-
a physical significance.
The entropy is assumed to be constant in the whole field, even during transient periods,
so that only the stationary asymptotic solution has a physical significance (17).
While this method, which requires a longer computation time, ha's a more extessive
field of application than the previous methods, it is limited to the case of motions with
only low intensity shock waves.
B - General solution to Euler equations under steady conditions
Until now, the problem raised by the mixed elliptic-hyperbolic nature of the Euler
equations for steady transonic motion with intense shocks, could only be solved by non-
steady type methods (18, 19). Here again, we find the characteristic (ýmentioned under
A(iii)J of the steady flow considered as the asymptotic state of a non-steady [notion.
Such methods offer the great advantage of being applicable to any transonic flow;
however, they require very long computation times for the asymptotic condition to be
achieved. As far as cascades are concerned, the computation time depends on rapid~ty
with which disturbances get damped and disappear at the borders of the computation field.
As the asymptotic solution is the only solution retained, the intermediate states do not
need to have a physical significance, and the non-steady terms can be modified in order

V to accelerate the achievement of the final solution.

With these methods, shock waves are usually dealt with by means of a shock capturing
method, with which they appear quite naturally, owing to the dissipative properties of
the numerical schemes used (dissipative properties wii0ch are either natural or reinforced
by an artificial viscosity term).
While this technique is simple, it is not accurate enough as regards the description
of shocks; these do not appear as true discontinuities, but are somewhat spread out.
The use of a two-step calculation process (20), (21) offers the advantage of second
order accuracy; however, the calculation time is doubled for each time step, which leads
to very long computation times.
Owing to an improved computation method (22), a single step method, with a damping
term whose intensity can be modified in the course of the computation, has been developed
(23) :this method, which is definitely quicker, offers also a second order accuracy for
the converged solution.
The computation time was further reduced using finite volumes, either for the isentro-8
pic scheme (22), or for the scheme which takes entropy variation into account (24). The
latter method was even generalized to the cascades drawn on surfaces of revolution with
an evolution of radius and a varying thickness of the stream tube.
A variant of this method (25) uses upwind differences for mass and momentum and down-
wind differences for pressure gradients. ]his procedure acts somewhat like a smoothing
and together with the finite differentiation of the fow it tends to smear out high local
gradients as seen in the leading edge region of figure 5.
This method is first order accurate and requires a large number of calculation points
or else corrective terms to attain a sufficient degree of accuracy.

This method has also been extended to three dimensional flows (26).
111.2.2.4 Finite element method
Finite element methods are of increasing interest for computations using the streamI
function, as well as for those using the velocity potential.
This method is based on an approximation of dependent variables in the form of poly-
nomials, and on an integral definition of the problem (27), (28).
The benefits expected from this technique are as follows
-possibility of giving an optimum design to the grid of T.he plane, especially by using
curvilinear meshe's;
- automatic treatment of natural boundary conditions;
- preservation of symmetry in the discretization of differential operators;
310

On the other hand, we risk incrcising the computation time due to the complexity of
the meshes used and to the number of terms of the polynomials.
This mehtod seems now to have reached an adequate level of efficiency for problems
of the elliptic or parabolic type. Its application to transonic problems of the locally
hyperbolic type requires further developments, although results have already been obtained
in the case of transonic, irrotational flows (29), (Fig. 6).

111.2.2.5 Methods of characteristics

As supersonic flows lead to equatiors of the hyperbolic type, one can think about
calculating the supersonic flow directly from the upstream boundary conditions. The
field of dependence of each point is then defined by the characteristics issuing from it.
This very classical problem is described is numerous references (30) and the applica-
tion to two dimensional or three dimensional blades is described in (31), (32) and (33).
The main difficulties encountered are as follows:
the supersonic field must be perfectly well defined (fully supersonic flow), and up-
stream conditions (supersonic flow or sonic line) must be known;
- only oblique shock waves can be taken into account, since the flow is subsonic down-
stream of normal shocks; and the method is no longer applicable;
the computation cannot be carried out unless the upstream boundary conditions are set.
Thus, as regards turbine cascades, the sonic line has to be calculated first, for instance
by a method of series development (34 to 37) : in this case, the subsonic portion of the
flow is calculated by any method which can use the downstream boundary conditions as de-
fined by the calculation of the sonic region.

For all these regions, the method of characteristics Is only applicable to cascades
with a well-defined minimum section, where the sonic line can be calculated in a suf-
ficiently accurate manner.
In addition, the use of other computation methods suited to calculate the subsonic
flow field increases markedly the complexity to the computation process.
111.2.3 Boundary layer calculations
The comparisons of the previous chapter between calculated and experimental results
have mainly been made for turbine cascade because in this case, viscous effects are much
less important than in compressor cascades.
However, in calculation of the flow in compressor cascades the influence of boundary
layers can no longer be neglected.
The description of the many methods developed during the last decades for boundary
layer calculation does not fall within the scope of this survey. We only shall classify
them according to the assumptions on which they are based and the techniques they use.
(4) The oldest methods are the integral methods based on von Karman's integral equation
(-ntegral form of the equation of motion written along a streamline). In these methods,
tnere are only two unknown quantities left (boundary layer thickness and shape factor),
owing to the additional relations which provide the shape of the velocity profile, or tc
results obtained on similar boundary layers. Consequently, it is necessary to have a
second (or auxiliary) equation, and the integral methods proposed differ mainly by the
formulation of this equation. A first formulation consists in using the energy equation
(38 to 42). A second formulation uses an equation based on the entrainment principle
(43, 44, 45). The system of equations is then integrated along the wall streamline allow-
ing calculation of the boundary layer parameters.
(ii) More recently, methods have been developed to solve the boundary layer equations by
a finite difference technique (46, 47, 48) : In this approach, the discretization of
partial derivative equations and the numerical technique to solve them constitute the
difference between the methods proposed by various authors. Moreover, when the boundary
layer is turbulent, an additional assumption is necessary to define the turbulent shear
stresses, whose order of magnitude exceeds that of viscous stresses by one unit. With
integral methods, it is sufficient to relate turbulent magnitudes, in particular through
a mixing length (49, 50, 51).
The data necessary for this calculation, particularly velocity profiles, can be
derived from similar solutions (52).

In finite difference methods, three different approaches can be used

- a zero-equation model (use of mixing length resulting from previous calculations),
- a one-equation model, for instance the convection equation of one of the turbulent
magnitudes (turbulent kinetic energy, Reynoldr stress) (53).
- a two-equation model, using two convection equations, for instance the model (Kt r) des-
cribing the evolution of the turbulent kinetic energy and that of the turbulent scale
(54, 55, 56).
The last two models require long computation times, which limits their applicability
for industrial calculat'ions. They require also a large number of constants and para-
meters for which empirical models have to be established. On the other hand, they provide
an improved description of the turbulence structure and are more generally applicable then
integral methods. They are quite suited to the case of non equilibrium boundary layers -
 for instance with a high 1oressure gradient

I
- or to the %_se of blowing.31
On the contary, integral methods are applicable to boundary layers in equilibrium,
where the evolution of external conditions is sufficiently progressive for turbulent
stresses to adapt to the mean velocity profile. This can be further facilitated by
using a delay in the calculation of local dissipation energy (40, 44), or by relating the
latter to the shape of the velocity profile obtained during the previous cal culation steps
(57).
Excellent results have been obtained for the calculation of the boundary layer on flat
plates. A few problems remain to be solved
transition from laminar to turbulent boundary layer, or else Its return to laminar,
-the

according to the evolution of external parameters, still requires complementary studies

(58, 59)
- the streamline curvature and the Coriolis turces have an impact on the turbulent level.
For the time being, this is not taken into account except by corrective terms which are
functions of Richardson's or Rossby's number (60 to 63).
So far, we have considered only the case of boundary layers without separation. The
calowing p obl sems ain ymaso onaylyr ehd assesnilytefl
c alculatio of sepnaryatinye mheoyansuoftboundaylyrmtosriesesnilytef,
-validity of Prandtl equations,
- validity of turbulence models.

The boundary layer theory Is indeed at fault when a separation takes place, since it
does not allow for the strong interaction with the external flow. It is possible to
overcome this difficulty, together with the related numerical singularities, by associa-
ting the boundary layer computation with the external flow computation, and, especially
by using the inverse methods described in 111.2.4. This way, if the separation extends
over a limited space, Prandtl theory continues to be applicable. However, the turbulence
models which are currently used do not seem to provide an adequate representation of
actual physical phenomena, and need to be improved.
Generalizing the boundary layer theory had led to the development of methods for
culculating the evolution of wakes downstream of the blades (64 to 68).
In the case of tendemi blades, it may be extremely useful to know the blockage of the
inter-blade channel by the wake of the first blade, so that it may be taken into account
for the calculation of the second blade (69). However, one is mainly interested in the
downstream, stagnation pressure, velocity and flow angle conditions, and the detailed
calculations of the wake can be replaced by a global calculation which eventual ly provides
the exit angle and the losses.
111.2.4 Association of inviscid flow and viscouts flow calculations
in the boundary layer
The association of inviscid flow calculations with boundary layer calculations takes
place at several levels
the mass flow deficit in the boundary layer, quulified by a displacement thickness,
leads to a displacement of the walls by an equiv~~lent quantity. As a result the distri-
bution of velocities in the external flow is modified when this flow is confined within
two relatively close walls. This effect is evidently considerable in the case of separa-
ted boundary layers.
direction of the exit flow is a function not only of the blade and cascade geometry,
-the

but also of the viscous effects on the trailing edge. The Kutta-Joukowski condition, or
any equivalent condition takes these various conditions into account. However, the

I. presence of the wake downstream of the blade and the shape of this wake exercise a con-
siderable influence on the exit angle.
-
-
The energy dissipation in the boundary layer and the wakes leads to pressure losses.
In transonic and supersonic flows, the pressure jump of a shock wave creates a strong
interaction with the boundary layer.
The resulting boundary layer thickness brings about a displacement of the shock wave
and a modification of its intensity.
111.2.4.1 Evaluation of the exir Inl
As mentioned above, the exit angle is defined by the cascade geometry and the
viscous effects.
As long as the boundary layer remains thin along the trai ling edge, it can be taken
into consideration in a very simple way in the potential flow calculation, in particular
by maintaining a stagnation point on the trailing edge itself (Kutta-Joukowski condition).
However, it has been demonstrated (7U) that a small displacement of this stagnation point
is sufficient to indurce a non negligible variation in the exit angle.
When the blade does not have a sharp trailing edge, the expression "trailing edge"
hardly applies, and there is no real stagnation point, but an inert fluid area, located
at the separation point of the boundary layer, almost at the point where the trailing edge
curvature starts. Therefore, a more realistic criterion consists in imposing an eq ual
312

pressure on buth sides of the blade, at the points where the tailing edge curvature
starts (71).
This criterion yields good results in the case of turbine blades (72) %nd (73) but is
less accurate in case of compressor blades, where the exit angle is predicted with an
accuracy of 1 36 only.
The difficulty we are faced with, results from the thickening of the boundary layer on
the suction or pressure surface of compressor blades which often leads to a small sepa-
ration zone.
An attempt to improve the equal pressure criterion through a correlation between the
position of the corresponding points and the pressure increase between the highest over-
speed point and the trailing edge (74), did not prove successfu.
The only method which looks promising consists in allowing for the boundary layer
sep3ratinn in the inviscid flow (75) (Fig. 7) : Correct results were obtained, even for
rather large separations near the trailing edge (76), as well as for operating modes with
a laminar separation bubble (77). The singularity methods used in references 75 and 76
have also the advantage that they easily allow for a wake downstream of the blade.
In fact, the exit angle is Influenced not only by the upper surface separation but
also by the development of the wake downstream of the blades.
111.2.4.2 "Blockage" effects
The association of the perfect fluid with the viscous fluid is currently done for
simple cases (cascades) in the following way : A first calculation in a perfect fluid,
by a method suited to the test case, provides a velocity distribution which, through a
boundary layer calculation, given an evolution of the displacement thickness. Two
methods are generally used then to take into account the presence of the boundary layers.
(1) Either one adds the boundary layer displacement thickness to the profile and, behind
the trailing edge, one extends this blade further downstream in a way which, although it
is somewhat arbitrary, is physically plausible (78).
(2) Or one injects, at the blade wall, a mass flow which is a function of the evolution
of the displacement thickness. Downstream of the trailing edge, one sucks back the mass
flow injected along a sink line, with an arbitrary distribution (79, 80).
One starts a new calculation with a perfect fluid, either on the thickened blade or on
the real blade, but with a wall mass condition, and so on, until the convergence is reached.
It is to be noted that there are cases where this convergence is not ensured. The main
problem consists in preventing the fictitious portion downstream of the trailing edge
(thick segment or sink line) from contributing to the lift (81). If necessary, this con-
dition may lead to adjusting either the shape of the segment added to the profile, or the
sink distribution.
The influence of the wake is globally txken into account. For this purpose, one
writes that the total enthalpy and the mass flow remain unchanged between the plane of
the fictitious trailing edges and the infinite downstream. The calculation of the result.-
ing exit angle (82, 83) must be combined with that of the losses. A more detailed calcula-
tion of the wake itself is given in reference 84.

These various calculation methods are relatively simple and yield good results, at
least in the absence of a clean-cut separation resulting essentially from the interaction
of the boundary layer and n high intensity shockwave, figure 8.
Specific methods have been developed to take this last phenomenon into account.

111.2.4.3 Shock-wave-boundary layer interaction

lhe shock-wave-boundary layer interaction can vary in intensity according to the value
of the Mach number upstream of the shock, and to the pressure ratio. It may lead to a
high amplitude separation, with a modification of the external flow (X shock). To take this
interaction into consideration, the inviscid flow calculation must be carried out in paral-
lel with the boundary !ayer calculation.
The main methods currently recommended are the following
A - Global methods, or control volume methods which relate a boundary layer characteristic
before and after the shock, to the overall shock intensity.
B - Integral methods described previously have to be adapted to boundary layers with
rapidly evolutive properties. Among these integral methods, direct methods have to be
distinguished from inverse and semi-inverse methods.
- A direct boundary layer method has been used in (85) for modest shock strength. This
was possible by using small integration steps and by limiting Cf to small but non zero
velues. The result shown in figure 9 shows a remarkable improvement in comparison to a
pur% invisrid calculation. A more sophisticated method (86) calculates the variation of
boundary 'ayer thickenss and shape factor on the basis of an abrupt discontinuity in func-
tion of pressure ratio and shape factor of the upstream boundary.
- Inverse methods can overcome the difficulty generally encountered in integral methods
due to the separation of the boundary layer induced by a high intensity shock : the singu-
larity corresponding to the separation point can be eliminated in the inverse method.
For this purpose, t;,e prp.....'e distribution at the external border of the boundary layer
 313

must be chosen is the unknown quantity, and a longitudinal evolution oF the frirtion co-
efficient or boundary layer displacement thickness must be set. Then, the pressure dis-
tribution in the external flow and the boundary layer characteristics are provided by the
calculation (87 to 95). This procedure is not limited to shock induced separation, and
has been sucessfully applied (Fig. 10) to calculation of the trailing edge separation
zone in a coirressor cascade (84).
- The methods, which take full account of inviscid and viscous flow interactions, make it
nossible to ca;rulate boundary layers displacement thickness, as well as the external flow
(96). This method is not only applicable to the shock wave-boundary layer interaction; it
gives also good results in the case of an interaction with a laminar separation bubble, as
well as for trailing edge separation, and for wakes. Initially used for supersonic flows,
these methods are now being generalized to the case of subsonic or transonic flows, in
association with either direct or inverse methods.
C - Different level methods compare the boundary layer to multiple layers fitted in each
other, with different aerodynamic properties (the simplest case in a laminar sub-layer
and a turbulent layer). In each layer, the flow is described by means of a system of
partial differeriLial equations which is obtained from the Navier-Stokes equations by re-
taining only the terms which are essential for the layer used. It is even possible to use,
in some of these layers, the integral equations corresponding to the type of boundary
layer which develops within them. In order that the solutions may be compatible with each
other, adjusting conditions are written on the interfaces of the various layers (97 to 102).

111.2.5 Solution of the Navier-Stokes equations

The influence of viscosity on the flow has been calculated in the previous section by
associating a boundary layer model with an inviscid calculation, and results in an inter-
active procedure.
However, the most rigorous way to solve this problem is by solving the complete
Aavier-Stokes equations. With this method, the external flow an'4 the boundary layer are
calculated simultaneously; thus, the close association between the two conditions is
directly taken into account (103 to 106). Flow predictions for stationary blade rows are
quite promising (Fig. 11).
For transonic flow, it is general practice to solve the instationary Navier-Stokes
equations by a time step progression until complete convergence of the solution is
attained (107 to 119).
The main difficulty with this method arises from the necessity of having a very fine
grid in the area where the fluid viscosity plays a preponderant part, so that each term
of the Navier-Stokes equation may be of the same order of magnitude. For numerical sta-
bility conditions this leads to very small time steps and long calculation durations.
Therefore, the essential difference between the various numeri;..al methods proposed results
from the scheme retained, and the partial derivative equation discretization best suited
to the problem.

The schemes currently retained vary from the fully explicit scheme to the fully impli-
cit scheme, with every possible intermediate solution. The shortest calculation times
have been obtained with mixed schemes.
"Evidently, the modelling of turbulence poses additional probierrs, which are discussed
above, and, here we find again the same three models (algebraic model, one-equation model
and two-equation model), with their respective advantages and drawbacks.

111.2.6 Conclusions
The number of inviscid calculation methods, available today, is quite impressive, and
they allow an accurate prediction of inviscid incompressible, subsonic compressible and
transonic flow. Recent progress has al'so resulted in a remarkable decrease of required
computational effort, which makes them even more attractive to industrial applications.
However, as these methods do not account for viscosity, the field of application is
quite restricted, and good predictions can be obtained only for flow configurations where
viscous effects can be neglected (f.g. turbines and low loaded compressor cascades).
A similar conclusion can be drawn about boundary layer methods, with a limitation on
shock boundary layer interaction, where complete solutions still require a large amount of
computational effort.
The combination of these two types of calculatiors has started quite recently, and a
lot of progress can be expected in the near future. It is the opinion of the authors that
the improvements which can be expected from a better combination of these two types of
calculations will be more important than an improvement in the inviscid calculations.
Based on the present state of the art concerning shock boundary layer interaction cal-
culations, it appears that integral or inverse methods are the best suited to industrial
calculations.
However, in view of the considerable efforts currently made towards the development
of the Navier-Stokes equation solution, it is more likely that, in the near future, this
method will lead to more reasonable computation times and permit practical applications.

.' ,'~.,
314

From now on, it can be recommended to use Navier-Stokes methods for flow analysis in limi-
ted fields sucl, as the strong :hock-boundary layer interaction field, and to associate
their with simpler methods for other fieltis.
REFERENCES

1. SCHLICHIING, 'I.: Berechnung der reibungslosen incompressiblen Strsming fUr ein

vorgegebents t-henes Schaufelgitter.
VKI Forschungsheft 447 - B/21, 1955.
2. MARTENSEN, E.: Berechung der Druckverteilung an Gitter profilen in ebener Poten-
tialstrdmung
Archive mit einerMechanics
for Rational Fredholmschen integralgleichung.
& Analysis, Vol. 3, N0 3, 1959.

3. IMBACH, H.: Die Berechnung der kompressiblen, relbungsfreien Unterschallstramung

durch raumliche Gitter aus Schaufeln auch grosser Dicke und Wolbung.
ETH Zurich, Prom 3402, 1964.
4. VAN DEN BRAEMBUSSCHE, R.: Calculation of compressible subsonic flow in cascades
with varying blade height.
ASME Trans., J. Engineering for Power, Vol. 95, No. 4, 1973, pp 345-351.
5. LUU, T.S.: Mithode d'6quattons integrales en mecanique des fluides,
Computing Methods in Applied Sciences and Engineering,
R. Glowinski & J.L. Lyons Eds., Lecture Notes in Economic and Mathematical
Vol. 134, Springer, 1976, pp 359-373.
System. 9
6. WILKINSON, D.M.: Stability, convergence and accuracy of two dimensional streamline
curvature methods using quasi orthogonals.
Thermodynamics and Fluid Mechanics Ccnvention 1970,
Inst. Mech. Engrs., London, Vol. II.

7. BINDON, J.R.: Stability and convergence of streamline curvature flow analysis

procedures.
Int. J. for Numerical Methods in Engineering, Vol. 7, 1973, pp 69-83.
8. KATSANIS, T.: Computer program for calculating velocities and streamlines on a
blade
NASA TNto D blade
4525, streamsurface
April 1968. of turbomachines.

9. FENAIN, M.: Met, odes de relaxation pour Ia r&bolution d'6quations elliptiques dans
les domaines de frontiCres quelconques. Applications au calcul d'dcoulements
subcritiques.
J. de Mecanique Appliquee, Vol. 1, 1977, pp 27-67.
10. CALVERT, W.J. & SMITH, D.J.L.: A digital computer program for the subsonic flow
past turbomachine blades using a miatrix method.
ARC R&M 3838, 1976.
11. KATSANIS, T.: Fortran program for calculating transonic velocities on a blade to
blade streamsurface of a turbomachine.
NASA TN D 5427, 1969.
12. DODGE, P.: A transonic relaxation method for cascade flow systems,
in Transonic Flows in Turbomachinery,
VKI LS 59, May 1975.
13. LUU, T.S & COULMY, G.: Calcul de l'Ecoulement transsonique avec choc A travers
tine grille d'aubes.
ATMA 1975.
14. RAE, W,J. & HORNICJ, G.F. : A rectangular coordinate method for calculating non
linear transonic potential flowfields in compressor cascades.
AIAA Paper 78-248, 1978.
15. IVES, D.C. & LIUTERMOZA, J.F.: Second order accurate calculations of transonic flow
over turbomachinery cascades.
AIAA Paper 78-1149, 1978.
16. CASPAR, J.R.: HOBBS, D.E. & DAVIS, R.,,: The calculation of t.4n dimensional compres-
sible potential flow in cascades using finite area techniques.
AIAA Paper 79-007, 1979.
I1. VEUILLOT, J.P. & VIVIAND, H.: A pseudo unsteady methou for the computation of
transonic potential flows.
AIAA Paper 78-1150, 1978.
18. MAGNUS, R. t YOSHIHARA, H.: Steady inviscid transonic flow over planar airfoils -
A search for simplified procedure.
NACA CR 2186, 1972.
 315

19. MAGNUS, R. & YOSHIHARA, H.: Inviscid transonic flow over airfoils.
AIAA J., Vol. 8, i1o. 2, 1970, pp 2157-2162.
20. GOPALAKRISHNAN, S.: Fundamentals of time Marching methods,
in Transonic Flows in Turbomachinery,
VK; LS 59, 1973.
21. VIVIAND, H. & VEUILLOT, J.P.: Mithodes pseudo-instationnaires pour le calcul
d'6coulements transsoniques.
ONERA Publication No. 1978-4
22. McDONALD, P.: The computation of transonic flow through two dimensional
gasturbine ascades.
ASME P 71 Go 89, 1971.
23. COUSTON, M.: McDONALD, P. & SMOLDEREN, J.J.: The damping surface tochnique for time
dependent solutions to fluid mechanic problems.
VKI TN 109, March 1975.
24. COUSTON, M.: Mithode de calcul de lI'coulement inter-aube pseudo-tridimensionnel
VKI ULB Ph.D. Thesis, September 1976.
25. DENTON, J.D.: Extension of the finite area time marching mkthod for three
dimensions,
in Transonic Flows in Axial Turbomachinery,
VKI LS 84, February 1976.
26. DENTON, J.D. & SINGH, U.K.; Time marching methods for turbomaZhinery flow
calculations,
in Application of Numerical Methods to Flow Calculations In Turbomachines,
VKI LS 1979-7, April 1q79.

27. KRIMERMAN, Y. & ADLER, D.: Calculation of the blade to blade compressible flow
field in turbo impellers using the finite element method.
Fac. Mech. Eng. Technion, Israel Inst. of Technology, Haifa, 197b.
28. MORICE, Ph.: Une mithode numerique basie sur les principes variationnels pour des
6coulem.!nts avec frontitres libres.
ONERA TP 1978-40.

29. HIRSCH, Ch.: Transonic flow calculations in blade rows with an optimal control -
Finite element formulation.
Bat-Sheva Seminar on Finite Element for No,-elliptic Problems.
Tel-Aviv Univ., July 1977.
30. SHAPIRO, A.: TI.e d.inamics and thermodynamics of compressible fluid flow.
The Ronald rress Company, New York, 1953.

31. LAWACZLCK, 0.- Verfahren zur Ermittlung der AsLtrdmqrossen transsonischer

Turoinegitter.
AVA Gdttingen Bericht 68 A 62.
32. FRUEHAUF. H.M.: A method of characteristic for three dimensional steady supersonic
flow in rotating and stationary annular cascades.
ESRO TT-44, July 1975.
33. MARTINON, J.: Use cr Lht characteristics method for the prediction of the three
dimensional flow field in high transonic compre-sors.
ASME ýaper 79 GT 34, 1979.
34. MEYER, T.M.: Forschungsheft 62, 1908.
35. HI.L, I.M.: Transonic flow in two dimensional and axially symmetric nozzles.
ARC 23,347, 1961.
36. MOORE, A.W.: The transonic flow in the throat region of a two dimensional nozzle
with walls of arbitrary smooth profile.
ARC R&M 3481, 1965.
37. DECUYPERE, R.: Berekening van de bidimensionele, wrijvingsloze, schok-golfvrije
stroming doorheen subkritische, kritische en superkritische turbineschoepenrijen.
B(:uw van een bidimenslonele transsonische stoomtunnel.
VKI-RiIG Ph.D. Thesis, August 1973.
38. LEFOLL, J.: A theory of representation of the properties of boundary layers on
"aplane.
Proc. Seminar on Advanced Problems in Turbomachinery,
VKI LS 1, March 1965.
39. PAPAILLIOU, K.: Blade optimization based on boundary layer concept.
VKI CN 60, March 1967.
316

40. NASH, J.F. & McUONALD, l A.G.: A calculation method for the incompressible turbulent
bojundary layer, inc uding the effect of upstream history on the turbulent
shear stress.
ARC 29,088, NPL Aero R 1234, 1967.
41. KUHN, G.0. & NIELSEN, J.N.: Prediction of turbulent separated boundary layers.
AIAA Paper 73-663, 1973.
42. FLAMERTY, R.J.: pressure
A method gradient
for estimating turbulent boundary layers and heat transfer
in arbitrary
United Aircraft Research Labs., Rep. UAR-G-51, August 1968.
43. HEAD, M.R.: Entrainment in a turbulent boundary layer.
ARC R&M 3152, 1960.
44. GREEN, J.E.; WEEKS, D.J. & BROOMAN, J.W.F.: Prediction of turbulent boundary layers
and wakes in compressible flow by a lag-entrainment method.
RAE TR 72231, January 1973.
45. MICHEL, R.; QUEMARD, C. & COUSTtIX, J.: Mithode pratique de previslon des couches
limites turbulentes bi et tridimensionnelles.
La Recherche Atrospatiale, No. 1972-1, pp 1-14.
46. BRADSHAW, P. & FERRISS, D.H.: Calculation of boundary layer development using the
turbulent energy equation ; compressible flow on adiabatic walls.
J. Fluid Mechanics, Vol. 46, Part 1, 1971, pp 83-110.
47. PATANKAR, S.V. & SPALDING, D.B.: Heat and mass transfer in boundary layers.
Intertext Books, London (2nd edition), 1967.
48. KELLER, H.B.: Numerical methods in boundary layer theory.
Annual Review of Fluid Mechanics, Vol. 10, 1978, pp 417-433.
49. DE LA CHEVALERIE, A,: Application d'une mithode intigrale au calcul d'interaction
onde de choc-couche limite turbulente en *coulement faiblement supersonique.
Ph.U. UniversitO de Poitiers CEAT-ENSMA.
50. SHANG, J.S.; HANRAY, W.L. & WOYER, D.: Numerical analysis of eddy viscosity models
in supersonic turbulent.
AIAA Paper 73-164, 1973.
51. HORSTMAN, C.C.: Turbulence model for non equilibrium adverse pressure gradient flows.
AIAA J., Vol. 15, No. 2, February 1977, pp 131-132.
52. MICHEL, R.: QUEMARD, C. & DURANT, R.: Application d'un schema do longueur do
melange a l'#tude de,; couches limites turbulentes d'equilibre.
ONERA NT 154, 1969.
53. RUBESIN, M.W.: Numerical
in Computational Fluid turbulence
Dynamics, modelling,

AGARD LS 86, 1976.

54. CHAMBERS, T.L., & WILCOX, D.C.: Critical examination of two equation closure models.
AIAA Paper 76-352, July 1976.
55. COAKLEY, T.J. & VIEGAS, J.R.: Turbulence modelling of shock separated boundary
layers flow.
Turbulent Shear Flow Conference, Penn. State, 1977.
56. MICHARD, P. & DUTOYA, D.: Calcul des coefficients d'echanges sur les aubes de
turbines A haute temperature.
La Recherche Alrospatiale 1978-3, pp 133-137.
57. HUO, S.: Optimization based on the boundary layers concept for compressible
decelerating flows.
VKI TN 111, June 1975.
58. MORKOVIN, M.V.: Technical evaluation report of the Fluid Dynamics Panel Symposium
on Laminar Turbulent Transition.
AGARD AR 122, 1978.
59. Proceedings of AGARD Fluid Dynamics Panel Symposium on Laminar Turbulent Transition.
AGARD GP 224.
60. SO, R.M.C. & MELLOR, G.L.: An experimental investigation of turbulent boundary
layers along curved surfaces.
NASA CR 1940, 1972.
 317

61. TFTERVIN, N.: An exploratory theoretical investigation of the effect on longitudinal

surface curvature on the turbulent boundar-y layer.
AOL TR 69-22, February 1969
62. BRADSHAW, P.: Fffects of streamline rurvature on turbulent flow.
AGAROograph 169, 1973.
63. COUSTEIX, J. & HOUDEVILLL, R.: Methode int6grale de calcul d'une couche limilte
turbulente sur une paroi courbte longitudinalement.
La Recherche A~rospatiale 0977-1, pp 1-13.
64. TOWNSEND, A.A.: 1he structure of turbulent shear flow.
Cambridge University Press, 1956.
65. HILL, P.G.; SCHAUB, U.W. & SENO0, Y.: Turbulent wakes in pressure gradients.
ASME Trans., J. Applied Mechanics, 1963, pp 518-624.
66. SOLIGNAC, J.L.: Calcul du sillage turbulent bidimenslonnel en aval d'un obstacle
mince.
CANCAN 73; also ONERA TP 1253, 1973.
67. QUEMARD, C. & J.P. ARCHAMBAUD, J.P.: M6thode intigrale de calcul d'un sillage
avec gradient de pression longitudinale.
CERT-DERAT Fiche technique No. 1/76.

b8. HODGE, J.G.; STONE, A.L. & .ILLER, T.E.: Numerical solution tu-r :;ifoils near
stable in optimized boundary-fitted curvilinear coordinate.
AIAA J., Vol. 17, No. 5, 1979, pp 458-464.
69. LOUDET, C.L.: Contribution thiorique et exp6rimentale do 1'etude des grilles
d'aubes en tandem A forte d6flexion et & forte charge.
IVK-ULB Ph.D. Thesis, April, 1971.
7U. MILLER, N.J.: Estimation of deviation angle for axial-flow compressor blade
sections using inivscid-flow solutions.
NASA TN D 7549, 1974.

71. WILKINSON, D.: A numerical solution of the analysis and design problems for the
flow past one or more aerofoils or cascades.
ARC R&M 3545, 1967.

72. VAN DEN BRAEMBUSSCHE, R.: Calculation of compressible subsonic flow in cascades
with varying blade height.
ASME Paper 73-GT-59, 1973.
73. DODGE, P.: The use of a finite difference technique to predict cascade stator and
rotor deviation angles and optimum angles of attack.
ASME Paper 73-GT-10, 1973.
74. MILLER, M.J. & SEROVY, G.K.: Deviation angle estimation for axial flow compressors
using inviscid flow solutions.
ASME Paper 74-GT-74, 1974.
75. GELLER, W.: Berechnung der Druckverteiling an Gitterprofiler in ebener inkompres-
sibler StrUmung mit Grenszschichtablsung in Bereich der Profilenden.
DLR FB 72-62, 1972
76. GOUJON-DUBOIS, A. & VAN DEN BRAEMBUSSCHE, R.: Calcul d'un 6ooulement incompressible
d~colli dens une grille d'aubes.
U-VKI 1972-72.
77. GEROPP, D. & GRASHOF, J.: Berechnung von Strdmungen mit ablosenblasen bei
grossen Reynoldszahlen.
OLR FS 76-52, 1976.
78. MrAUZE, G.: Comparaison calcul-expirience de l'tcoulement transsonique bidimension-
nel dens une grille d'aubes & forte deviation.
La Recherche Alrospatiale 1978-4, pp 175-180.
79. HANSEN, E.C.; SEROVY, G.K. & SOCKUL, P.M.: Axial flow compressor turning angle and
loss by inviscid-visc-us interaction blade-to-blade computation.
J. Engrg. for Power, Trans. ASME, Vol. 102, 1980, pp 28-34.
80. DECONINCK, H. & HIRSCH, C.H.: A finite element method solving the full potential
equation with boundary layer interaction in transonic cascade flow.
AIAA Paper 79-0312, 1979.
81. GOSTELOW, J.P.; LEWKOWICZ, A.K & SHALAAN, M.R.A.: Viscosity effects on the two
dimensional flow in cascades.
ARC CP 872, 1965.
318

82. SCHOLZ, N.: Aerodynamics of cascades.

AGARDograph 2?0, 1977.
83. I'RAUPEL, W.: Thermische TurLomaschiioer.
Springer Verlag, 1968.
84. CALVERT, W.J. & HERBERT,'M.V.: An inviscid-viscous interaction method to predict
the blade to blade performance of axial compressors..
NGlE M 80012
85. SINGH, U.K.: Computation of transonic flows in cescade with shock boundary layer
interaction,
in Shock-Boundary Layer Interaction in Turbomachines,
VKI LS 1980-8.
8i6. PANARAS, A.: Calcvlation of a boundary layer interacting with a normal shock by
discontinuity analysis.
VKI TN 121, 1976.
87. DELERY, J.; CHATTOT, J.J. &LE BALLEUR, J.C. Interaction visqususe avec d~col-
lament transsonique.
ONERA TP 1975-15.
88. THIEDE, P.G. : Emn inverses Integralverfahren zur Berechnung abgel~ster turbulenter
Grenzschi chten.
DIR F8 1977-16, pp 245-255.
89. LE BALLEUR, J.C. : Couplage visqueux - nor visqueux : mithode num~rique et applica-
tions aux 6coulements bidimensionnels transsoniques et supersoniques.
La Richerche Afirospatiale 1978-2.
90. LA BALLEUR, J.C. : Couplaqe visqueux - non visqueux analyse du problems incluant
dicollements et ondes de chocs.
La RlIcherche Abrospatiale 1977-6.
91. KLINEBERG, J.M. & STEGER, J.L. : Calculation of separated flow at subsonic avid
3rd Int. Conf. Num. Methods in Fluid Dynamics. Springer 1973.
92. LEES, L. & REEVES, B.L. : Supersonic separated and reattaching liminar flows.
AIAA J.. Vol. 2, No, 11, 1964, pp 1907-1920.
93. NIELSEN, J.M. : LYNES, 1.1. & GOODWING, F.K. : Calculation of laminar separation
with free interaction by the method of integral relations.
AFFDL-TR 65-107, June 1965.
94 CARRIERE, P.; SIRIEIX, M. & UELERY J.: M~thodeS de calcul des ficoulements turbu-
lents dicollis en supersonique.
Progress in Aerospace Science., Vol. 16, No. 4 1975, pp 385-429.
95, GEORGEFF, M.P.: Momentum integral method for viscous inviscid interactions with
AIAA J., Vol. 12, No. 10, 1974, pp 1393-1400.
96. LE BALLEUR, J.C. : Calculs couples visqueux - non visqueux incluent dicollements et
ondes de choc en 6coulem~ent bidimensionnel
in Three Dimensional and Unsteady Separation at High Reynolds Numbers.
AGARD IS 94, 1978.
97. TU. K.M. & WEINBAUM, S.: A nonasymptotic triple deck model for supersonic boundary
layer interaction.

98. LJOU,
AaAJVl
M.S.: Asymptotic
4 oanalysis
,Jn of 96 interaction
p7775 between a normal shock wave anAi
turbulent boundary layer in transonic flow.
Ph.D. Thesis, University of Michiga~n, 1978,
99. BRIALLIANT, N.M. & ADAMNsON, TC. : Shockwave boundary layer inte:'lictions in laminar
transonic flow,
AIAA J., Vol. 12, No. 3, March 1974, pp 323-329
100, STEWAPRIS0N, K.: Multistructured boundary layers on flat plates and related bodies.
Advances in Applied Mechanics, Vol. 14, 1974 pp 145-239.
101. PANARAS, A.G. : Normal shock boundary layer interaction in transonic speed in the
presence of a streamwise pressure gradient.
VKI-ULB Ph.D. Thesis. May 1976.
1U2. INGER, G.R. & MASON, W.I. : Analytical theory of transonic normal shock - turbulent
boundary layer interactions.
AIAA J., Vol. 14, No. 9, September 1976, pp 1266-1272.
 319

103. DOUGE, P.R.: A numerical method for two and three dimensional viscous flows.
AIAA Paper 76-425, 1976.
104. McDONALD, H.: Computational aspects of viscous flows,
in Application of Numerical Methods to Flow Calculations in Turbomacninery
VKI LS 1979-7.
105. MOORE, J. & MOORE, J.G.: A calculation procedure for three dimensional. viscous
compressible duct flow. Parts I & 11.
ASME Trans., J. Fluids Engineering Vol. 101, No. 4; 1979, pp 415-428.
106. WALLIT, L.: Numerical analysis of the three dimensional viscous flow field in a
centrifugal impeller.
AGARD CP 282, 1980.
107. MacCORMACK, R.W.: Numerical solution of the interaction of a shock wave with 1
laminar boundary layer.
Lecture Notes in Physics. Vol. 8, Springer-Verlag. 1970.
108, MATEER, G.: BROSH, A. & VIEGAS, J.R.: A normal shock wave turbulent boundary
layer interaction at transonic speeds.
AIAA Paper 76-161, 1976.
109. hdN!;, R.W. & MacCORMACK, R.W.: Numerical solutions of supersonic and hypersonic
laminar flow over a three dimensional compression corner.
AIAA Paper 77-4694, 1977.
110. BEAM, R.M. & WARMING, R.F.: An implicit factored scheme for the compressible
Navier-Stoket equations.
AIAA Paper 77-645, 1977,

111. WAGNER, B. & SHMIDT, W.: Theoretical investigation of real gas effects in cryogenic
wind tunnels.
AIAA Paper 77-669, 1977.
112. SHANG, J.S.: !aplicit-explirit method for solving the Navier-stokes equatior..
AIAA Paper 77-646; AIAA J.. Vol. 1b, No. 5, 1978, pp 49b-502.
113. SHANG, J.S. & HANKEY, W.L.: Supersonic turbulent separated flows utilizing
the Navier-Stokes equations,
in Flow Separation, AGARD CP 168, 1975.
114. BALLHAUS, W.F.: Some recent progress in transonic flow computation,
in Computational Fluid Dynamics,
VKI LS 87, March 1976.
115. MacCORMACK, R.W.: An efficient explicit implicit characteristic method for solving
the compressible Nnvier-Stokes equations.
SIAM-AMS, New York, April 1977.
116. LI, C.P.: A mixed explicit implicit splitting method for the compressible Navier-
Stokes equation.
Lecture Notes in Physics, Vol. 59, SDringer Verlag 1976 pp 285-292
•.7. VIVIAND, H.: Traitements des problfmes d'interaction fluide parfait, fluide
visqueux en 6coulernent bidimensionnel compressible A partir des equations de
Navier-Stokes,
in Three Dimensional and Unsteady Separation at High Reynolds Numbers,
AGARD LS 94, 1978.
118. PEYRET, P. & VIVIAND, H.: Computation of viscous compressible flows based on the
Navier-Stokes equations.
AGARDograph 212, September 1975.
119. HALLANDERS, H. & VIVIAND, H.: The numerical treatment of compressible high Reynolds
number flows,
in Computational Fluid Dynamics.
VKI LS 1979-7; also ONERA TP 1979-22.

II
 -

320

1.0 ___

0.6

0.4

0
0 0.2 0./. 0.6 0.B5 1.0
/Smax

FIG. 1-NASA TN 0-3751 STATOR MEAN SECTION

M1=.23
o EXPERIMENTAL
-CALCULATED REF (4)

1.0

b
0.8

0.6

0.2 o---- -4 --- I

0 1.0
0 0.2 0.4 0.6 0.8 5 1SmQx
FIG.2-NASA TN 0-3751 STATOR MEAN SECTION
Ml .231
EXPERIMENTAL
CALCULATED REF(6)
 321

0.4 _ __ FS~a

0 0.2 0.4 0.6 0.8 1.0

FIG. 3 -NASA TN 0- 3751 STATOR MEAN SECTION

M I1 . 23
0 EXPERIMENTAL
- CALCULATED REF (10)

1.4.

1.2

10.4..

0.6 -

01

0 0.2 0.4 0.6 0,8 1.0

Sismax
FIG. 4-TRANSONIC STATOR
M 1 =.82
o EXPERIMENTAL
--- RELAXATION METHOD
-TIME MARCHING METHOD REF(24)
 322

1.0 -_ _ _

0.8

0.6

2 0.4, _

0.2 . . .

0I
0__.....

0 0.2 0.4 0.6 0.8 1.0

5/3 max
FIG.5-NASA NOZZLE BLADE
SEXPERIMENTAL
-SOLUTION WITHOUT CORRECTION
-STANDARD SOLUTION REF (25)

1.4,

k 1,2 . . . .. . . . . ...

1.0
1.0 . ... ...
...
..

0.6
4
So~~a --. ........ . . ......... '

0.62o,0

0 0.2 0., 0.6 0.8 t.0o1

51
Smax
FIG. 6-VKI LS59 TURBINE BLADE
o EXPERIMENTAL
- CALCULATED REF(29)
• , .,L • , "
 323

2,0A

s s-

ofi
0- Q2 o, 6 o0o , 0 0o 0,2 o,4x0,o ,6 l1,0

1,60 4$1,6"
a.be 1%* I !o qs ,,,

S SO
s•2 -,2 J

- - 0,8

00,2 •hA° ,8
• 00o 2 o/.xAo ,8 0,
•

20 . . .

1,2 - 1,2

0 0 "

00,2 /,0Ao,6 4,0 %0 0 ,2 44,AQ6 0,6% oo

FIG. 7- COMPARISON OF MEASURED AND CALULATED

PRESSURE DISTRIBUTIONS.
o MEASUREMENT (UPPER SURFACE)
a MEASUREMENT(LOWER SURFACE)
- CALCULATION
ref. (75)

P/P, P1P1

0.6

0.7 a. se
0.7 -
o 53t

0.80.

0.9

0 0.2 0.4 0.6 0.8 1 S/Sa 0 0.2 0.4 0.6 0.8 1 SiSe

FIG. 8 - EXPERIMENTAL PRESSURE DISTRIBUTIONS OVER

"A BLADE AT MID-SPAN(103).
0 EXPERIMENT
-- COMPUTATION WITH THICKENING.
COMPUTATION WITHOUT THICKENING
- ref.(78)

p lS:
 324

iAjuo a)n3jfls uopflS - c.Lx W)/9

0 C

ýxoJ
oo cW

L06
.4e-

0 a C-

~~px

61 A x%
tLi

00

:1. 0 0
C
CT

INI
- xe-

*o tD

g J9 _ O
 325

Vi CASCADE M, 0.8 54.5

1-3
EXPRIMNT &SOLOUS TUNNEL WALL AL a 1-228
TUNNEL WALL ftL- 1.228
POROUSOD

1-2- PREDICTION - - -INVISCID- VISCOUS

I----------INVISCID

0.9

07-

0-6-

0.5-

0 10 20 30 40 50 60 70 80 90 l00
DISTANCE - PERCENT OF CHORD

FIG.10-MACH NUMBER DISTRIBUTION FOR

Vi CASCADE ref (84.)

--.--------....--.....
t 326

-' SURFACERI

2 ~~ ~ ~ ~ I FACEDSTNEDGRE

CrUAFC SURFACE 4ASSA

2 S-IS CIRCUMFERENTIAL
3 16-16 DISTANCE, DEGREES
* 11-20
9 20.21
S 261-U

Fig.dlonssaniteMaredio
fors TN 287 costn
NfCAcontours of
turbin
tutrbin seta(103
 327

PART IV

GENERAL CONCLUSIONS AND RECOMMENDATIONS

AiL-wi
 329

Part IV - General Conclusions and Recommendations

Faced with the task of comparing the overall performance and detailed flow predictions
with experimental data, it appears that the outcome of this exercise is dependent on the
accuracy and consistency of three components i the available experimental data, the
through-flow calculation method and finally the loss and turning correlations.

Although the main task of the Working Group was to investigate the reliability of exis-
ting loss and turning correlations at design and off-design conditions, it coald not be
separated from the two other components as soon as comparisons with experimental results
were to go beyond simple cascade data.

The first observation was the iack of reliable and publicly available data especially for
multistage compressors. Although some data can be found showing overall performance,
there is a strong need for detailed flow field data in multistage machines, representa-
tive of present day technology.

Moreover great care should be given to the accuracy of the data recording and more parti-
cularly to the data reduction programs. Generally, experimentalists deduce detailed flow
data from a reduced set of measurements taken behind the stators. The determination of
the losses and deviations behind the rotors for instance are obtained from the data reduc-
tion code based on some assumtions with regard to the streamline shape (usually the de-
sign shape) and eventually on some model for the loss and turning distributions within
the blade rows. The data reduced in this way are aubject to an uncertainty and make any
comparison with calculated data, espe:ially at off-design, doubtful. Hence, we suggest
the most extreme attention to this problem and advice to establish comparisons of data
reduced from the measurements in an unambigous and objective way.

The second problem is concerned with the interaction betv.aen a particular through-flow
method and a set of loss and turning correlations. For the same system of correlations,
it is well known that different results can be obtained with different through-flow pro-
grams. This is not connected to the numerical aspects of the methods (although, due to
the strong non-linearity of the equations, the particular iterative procedure used could
have an influence on the overall converged results), but to the variius assumptions which
enter in each program. The need and the influence on the calculated data of internal
stations in streamline curvature methods has been stressed by Novak and already reported
in Chapter 11.4. This has also a strong influence on reduced experimental data and adds
to the caution to be kept in mind as mentioned above.

With additional internal stations in through-flow calculations, assumptions have to be

made with regard to the law of variation, with chordwise distance, of losses and devia-
tions. Presently, there is no systematic analysis available of how the profile losses
(connected to the streamwise variation of blade boundary layer thickness) and the stream-
wise variation of angular momentum, behave under various conditions of incidence, Mach
number and blade profile geometry. Informations about the angular momentum could easily
be obtained from two dimensional inviscid blade-to-blade calculations and the rate of in-
crease of the profile losses could be obtained from a boundary layer calculation. This
is a typical example of how some general trend could be obtained from systematic blade-
to-blade calculations and correlated in order to be used in through-flow calculations.
It is our experience that the assumptions on the law of variation of these two variables
inside the blade row can have strong effects on the predicted results. For instance,
going from a linear to a parabolic variation of angular momentum in function of axial dis-
tance inside the blade rows, can lead in some cases to variations of the order of 10 %
on the predicted overall pressure ratio.

Another parameter influencing strongly the results of through-flow calculations is the

inclusion of the tangential blockage due to the blade boundary layers inside the blade
row and to the wakes at blade trailing adge which have to be added to the blade blockage.
Besides the effect on the calculated data, this introduces an additional interaction be-
tween the kinematic flow parameters, such as the blorkage, with the loss correlations,
since the wake blockage is proportional to the prof le loss coefficient.

A parameter of fundamental importance is the end-wall blockage which, even calculated

with an end-wall boundary layer theory of the type described in Section 11.2.3., adds a
strong non-linear interaction between the kinematics and the loss and turning corralation.

It was the Working Group's opinion that these sources of ambJguity should be kept in mind
when it comes to assess the influence of thi chosen set of correlations on the predicted
flow and overall performance data.

However, both Subgroups - Turbine and Compressor - came to the observation that with the
present level of experimental accuracy, the existing methods of through-flow were compa-
tible with the level of accuracy of the correlations, especially with regard to the over-
all performance prediction. In bosh fields, detailed flow predictions such as spanwise
flow angle of loss distribution at a given blade row outlet are however far from being
satisfactory for compressors as well as for turbines. This is true even if the overall
performance is satisfactorily predicted.
The level of accuracy needed for performance prediction is not equivalent for the losses
and the deviations. It appeared clearly that, in the turbine field as well as in the
compressor field, the required accuracy for devlations is appreciably higher that for
losses. Indeed, one or two degrees on deviation angle have a considerable influence on

P M M N%
P AG
I SLma -lN 71Inm
330

the predicted flow behaviour and on the machine performance. On the other hand, the pre-
dicted flow behaviour in loss sensitive to small variations in the loss coefficient.

Present day designa are faced with two complementary requirements : the need for rapid,
reliable and simple correlations in order to optimize stage design and, on the other hand,
the need for moreaccurate detailed flek predictions in the advanced stage of the design
and conception work.

These requirements are indeed complementary# since the way to the improvement of simple
correlations lien in the better understanding of the basic physical phenomena. It was the
working Group's concensus that this could be achieved by a systematic use of two- and
three-dimensional blade-to-blade calculations in order to derive simple and sound physical
correlations. This implies the development and use of 2D and 3D viscous flow calculations
and, on the other hand, the definition of basic experiments.

Hence, in addition to the specific observations to be found in the conclusions of the

Parts I and II of this report, the following rocnmmendations came forward out of the acti-
vity of the Working Groupt
An important effort should be made to obtain good and reliable multistage compressor
and turbine data.
A deep physical understanding is needed with regard to the various loss sources and
their particular generation mechanism. Basic experiments allowing one to separate and
understand the various contributions should be considered.
Care should be taken to define a consistent set of relations within the correlations in
order to avoid the duplication of effects and influence of parameters.
A particular effort is required with regard to secondary flow deviations and end-wall
boundary layer affect&. When possible, In axial compressors and high aspect ratio
turbines, end-wall boundary calculations, coupled interactively with the through-flow
calculations, should be used.
Not withstanding the continuous development of three dimensional viscous flow calcula-
tion and the increase of computing power of the digital computers, it is noL likely
that through-flow calculations in multistage machines will be replaced by full viscous
calculations in the near future. Therefore the need for good correlations will conti-
nue to be of great importance.
 331

WORKING GROUP 12
MEMBERSHIP

PEP MEMBERS

BELGIUM M. le Professeur Ch. HIRSCH CHAIRMAN

Vrije Universitelt Brussel
Dienst Stromingsmechanica
Pleinlaan 2
1050 Brussel
FRANCE M. le Professeur J. CHAUVIN
Universit6 d'Aix Marseille I1
Directeur du Laboratoire Associf LA 03
Institut do M6canique des Fluides
1 rue Honnorat
13003 Marseille
M. J. FABRI
Office National d'Etudes et do
Recherches A6rospatiales
29 Avenue de la Division Leclerc
92320 Ch&tillon sous Bagneux
GERMANY Professor Dr-Ing. G. WINTERFELD
Institut fUr Antriebstechnik
DFVLR
Postfach 90 60 58
5000 K5ln 90
TURKEY Professor Dr. A. UqER
Middle..East Technical University
ODTU
Makina Muh. BlUmU
Ankara
U.K. Dr. J. DUNHAM
National Gas Turbine Establishment
Pyestock
Farnborough, Hants GU14 OLS
U.S. Mr. J. ACURIO
Director, Propulsion Laboratory
US Army Research and Technology
Laboratories (AVRADCOM)
21000 Brookpark Road
Cleveland, Ohio 44135

NON-PANEL MEMBERS
BELGIUM M. le Professeur R. VAN DEN BRAEMBUSSCHE
Von Karman Institute for Fluid Dynamics
72 Chausl6e do Waterloo
1640 Rhode Saint Genise
M. le Professeur K. SIEVERDING
Von Karman Institute for Fluid Dynamics
72 Chaussae do Waterloo
1640 Rhode Saint Genbse
FRANCE M. G. MEAUZE
Chef de Groups do Recherches
Office National d'Etudes et do
Recherches A6rospatiales
29 Avenue do la Division Leclerc
92320 Chftillon sous Bagneux
 332

ME4LGERSHIP (Ct:1)

GERMANY Professor Dr-Ing. L. FOTTHER

WE 12 Str8mungsmasohinen
Hlochochule der Sundeswvhr
Warner-Heisenberq Weg 39
8014 Ntubiberg
Dr-Ing. H.B. WEYER
Institut fUr Antriebatechnik
DFVLR
Postfach 90 60 58
5000 Kaln 90

ITALY Dott. Ing. E. BENVENUTI

Responsabile Sezione Fluidodinamica
dell'Ufficio Studi a Ricerche
Soc. Nuovo Pignone
Via Matteucci 2
50100 Firenze
Professor E. MACCH!
Istituto di Macchine
Politecnico di Milano
Piazza Leonardo da Vinci 32
20133 Milano

U.K. Dr. J.D. DENTON

Science Research Council
Whittle Laboratory
University Engineering Department
Madinfley Road
Cambridge CB3 eEL

U.S. Dr. R.A. NOVAK

Aircraft Engine Group MS A24048
1000 Western Avenue
Lynn, Massachusetts 01910

Dr. A.M. PFEFFER

Pratt a Whitney Aircraft
Engineering Building 2 C
400 Main Street
East Hartford, Connecticut 06108
Dr. D.M. SANDERCOCK
NASA Lewis Research Center
21000 Brookpark Road
Cleveland, Ohio 44135
Professor G. SEROVY
207 Mechanical Engineering Building
Iowa State University
Ames, Iowa 50011
OBSERVERS
ITALY Mr. G. DORIA
ANSALDO
Via Pacinotti 20
Genoa-Sampierdareva 16151
Ing. B. INNOCENTI
Soc. Nuovo Pignone
Via Matteucci 2
50100 Firenze

Dr. P. ZUNINO
ANSALDO
Via Pacinotti 20
Genoa-Sampierdareva 16151

_ _
L
 333

THROUGH FLOW CALCULATIONS IN AXIAL TURBOMACHINE8

Editors
Ch. HIRSCH
J.D. DENTON

LIST OF CONTRIBUTORS
Na ma Address Chapter*
VAN DEN BRAEMBUSSCH:P, Von Karman Institute for Fluid Dynamics 11.4.2
Prof. R. 72 Chauss~e do Waterloo 111.2
1640 Rhode Saint Genfee, Belgium
CHAUVIN, Prof. J. Universit6 d'Aix Marseille II 11.2.4
Directeur du Laboratoire Associ6 LA 03 11.2.9
Institut do M6canique des Fluidss 11.2.10
1 rue Honnorat
13003 Marseille, France

DENTON, Dr. J.D. Science Research Council Notation Part I

Whittle Laboratory 1.1
University Engineering Department 1.4
Madinfley Road 1.5
Cambridge CB3 OEL, UK A.1
DUNKER, Dipl-Ing. R. Institut fdr Antriebstechnik 11.2.1
DFVLR
Postfach 90 60 58
5000 K81n 90, Germany
FOTTNER, Prof. Dr-Ing. L. WE 12 Strdmungsmachinen 11.2.2
Hochachule der Bur.deswehr 11.2.6
Werner-Heisenberg Wag 39
8014 Neubiberg, Germany

HIRbCH, Prof. Ch. Vrije Universiteit Bruasel Gen.Introduction

Dienst Stromingsmechanica 11.1
Pleinlaan 2 11.2
1050 Brussel, Belgium 11.2.3
11.3
A.II.3.2
11.4
a 11.4.1
11.5
IV

MACCHI, Prof. E. Istituto di Macchine A.I

Politecnico di Milano
Piazza Leonardo da Vinci 32
20133 Milano, Italy
McDONALD, Mr. P.W. Pratt & Whitney Aircraft Group 1.3
Commercial Proucts Division
400 Main Street
East Hartford( Connecticut 06108, US

MEAUZE, M. G. ONERA 111.2

Chef de Groups de Recherches
29 Avenue de la Division Leclerc
92320 ChAtillon sous Bagneux, France

NOVAK, Dr. R.A. Aircraft Engine Gioup MS A24048 11.2.7

1000 Western Avenue
Lynn, Massachusetts 02139F US
ROBERTS, Mr. W.B. FAR 11.2.5
1543 Vernal Avenue 11.2.7
Fremont, California 94538, US
334

LIST OF CONTRIBUTORS (Ctd)

N a me Address Chapters

do RUYCK, Mr. Vrije Univeraiteit Bruhuel 11.2.3

Dienst Stromingsmeahanica
Pleinlaan 2
1050 Brusbel, Belgium
SCHAFFLER, Dipl-Ing. A. Motoren und Turbinen Union 11.2.6
Dachauarutramse 665
8000 Mtcohen 80, Germany
SIEVERDING, Prof. K. Von Yarman Institute for Fluid Dynamics 1.2
72 Chaues6e do Waterloo
1640 Rhode Saint Genlue, Belgium
SEROVY, Prof. G. 207 Mechanical Engineering Bldg 11.2.8
Iowa State University 111.1
Amen, Iowa 50011, US
UVER, Prof. Dr. A Middle East Technical University A.I
ODT U
Makina Muh. B8l1mU
Ankara, Turkey
WEYER, Dr-Ing. H.B. Institut fUr Antriebstechnik 11.2.1
DFVLR 11.3
Postfach 90 60 58 A.II.3.1
5000 K81n 90
ZUNINO, Dr. P. ANSALDO A.I
Societe Generale Elettromecanica
Via Pacinotti 20
Genoa-Sampierdareva 16151, Italy

if
I
 335

LIST OF CONTRIBUTIONS

Author Chapter Author

Chapter

HIRSCH 11.2.6 SCHAFFLER

Gen. Introduction FOTTNER

Notation Part I DENTON

11.2.7 ROBERTS
NOVAK
I. 1 DENTON
11. 2.8 SEROVY
1.2 SIEVERDING
1.7?. 9 CHAUVIN
1.3 McDONALD

11.2.10 CHAUVIN
1.4 DENTON
HIRSCH
DENTON WEYER
1.5

A.II.3.1 WEYER
A.I DENTON
MACCHI
UCER A.II.3.2 HIRSCH
ZUNINO
HIRSCH 11.4 HIRSCH
II

HIRSCH
H1.•. 11.4.1 HIRSCH

HIRSCH II.4.2 VAN DEN BPAEMBUSSCHE

11.2

DUNKER II.5. HIRSCH

11.2.1
WEYER
111.1 SEROVY

11.2.2 FOTTNER 111. 2 MEAUZE

VAN DEN BRAEMBUSSCHE
11.2.3 HIRSCH
de RUYCK
IV HIRSCH

11.2.4 CHAUVIN

II.2.5 ROBERTS
 REPORT DOCUMENTATION PAGE
1.Reciplent's Reference 2.Originator's Reference 3. Further Reference 4. Security Classification
AGARD-AR-175 ISBN 92-835-1400-9 UNCLASSIFIED
5.Originator Advisory Group for Arrospace Research and Development
North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuilly sur Seine, France
6.Title PROPULSION AND ENERGETICS PANEL WORKING GROUP 12 on
THROUGH FLOW CALCULATIONS IN AXIAL TURBOMACHINES
7.Presented at

8. Author(s)/Editor(s) 9. Date
Edited by Ch.Hirsch and J.D.Denton October 1981

10.Author's/Editor's Address 11. Pages

See Flyleaf 342

12.Distribution Statement This document is distributed in accordance with AGARD

policies and regulations, which are outlined on the
Outside Back Covers of all AGARD publications.
13. Keywords/Descriptors
Axial turbines Through flow calculations
Axial compressors Performance estimation

14.Abstract
In 1977, the Propulsion and Energetics Panel of AGARD had set up its Working Group 12 on
'Through Flow Calculations in Turbomachines' after having found in the 47th (B) Meeting that
-,,\ the prediction of off-design performances, especially for axial flow compressors, was not fully
satisfactory.
The objectives vere to review the existing information on blade performance and wall effect
prediction, and to extend this information by systematic application of numerical methods to
representative geometries.

In its performance period Working Group 12 had confined to axial turbomachines only and
split into a Turbine Sub-Group and a Compressor Sub-Group. In the Turbine Sub-Group five
correlations were reviewed and evaluated against the test cases. Each correlation had its
strengths and weaknesses and room for further improvements.
The Compressor Sub-Group report begins with a comprehensive survey of the various loss and
deviation mechanisms. For comparison of the prediction methods to the test cases five
authors have used their own correlations, while the sixth employed a single code in
conjunction with three correlations for the four stage compressor. The results of the
evaluation are similar to those of the Turbine Sub-Group, but the spanwise parameter
distribution is often poorly predicted.
This Advisory Report was prepared at the request of the Propulsion and Energetics Panel of
AGARD.
 0 0

*z 11 '-a
bz
2-0
0
a.;
_r
1 gS
1 1 1I

00
LID~C~
0 , z

U R

00
Z t0 0Z

Zc r. 4 9 C
04)0 0 eC)04

cco 0 <o

cc 0 r
., ,

oz3z- tI! -
z c 1. x

Cccca

C43*r U.Cu r

f t:o
a~ba
Woz
0 (
0, 0 '0 z 0, z_____
 co w

0~-10

0 . 0, I 4.

10 0 z 42 0

0 ~ 0-0
S ~ 00 o 0a

0~ me
Ca.

0 0 V 0 ~
0
U O' . . :0

P. 0 or~

0u 0 " a-
'02 0.1 ' 0 > <
> op 0

0 0 0\.
w,

ou~~ . 0
o -0 Am am. o

.0 rh .0u ro-
'O.o 0 0 5 4.,5

0 cts

10 E

$.~~ ~k 8 52U
0 .0~ 2:.s a-

4)0~. 0
U 0 0 ..
b . z
to, -P.
0. 42

__Ir,:

rq0-04;4
!
 - ~ ~ NT -+.- rr;..N

NATO
OTANDISTRIBUTION OF UNCLASSIFED
7 RUE ANCELLE .92200 NEUILLV-SUR-SEINE AADPBIAIN
FRANCE
Telephone 745.08.10 . Telex 610176

AGARD does NOT hold Itocks of AGARD publicatione at the above a4dress for general distribution. Initial distribution of AGARD
publications is made to AGARD Member Nations through the following National Distribution Centres. Further copies arn sometimes
available from these Centres but if not may be purchased in Microfiche or Photocopy form from the Purchase Agencies listed below.
NATIONAL DISTRIBUTION CENTRES
BELGIUM ITALY
Coordonnateur AGARD - VSL Aeronautic& Militate
Etat-Major de Ia Force Adrienne Ufficio del Delegato Nazionale alIAGAP.D
Quartier Ruine Elisabeth 3, Piazzale Adenauer
Rue d'Evere, 1140 Bruxelles Roma/EUR
CANADA LUXEMBOURG
Defence Science Information Services see Belgium
Department of National DefenceNE ERAD
Ottaa, Otari KiA0K2Nethedlands Delegation to AGARD
DENMARK National Aerospace Laboratory, NLR
Danish Defence Research Board P.O. Box 126
9sterbrogades Kaserne 2600 A.C. Delft
Copenhagen 0 NORWAY
FRANCE Norwegian Defence Research Establishment
O.N.E.R.A. (Direction) Main Library
29 Avenue do !a Division Leclerc P.O. Box 25
92320 Chitilion sous Bagneux N-2007 Kjelier
GERMANY PORTUGAL
Fachinformatlonszentrum Energie, Direc~lo do Servi#o de Material
Physik, Mathematik GmbH da Forca Acres
Kemnforschungazentrurn Rua da Escola Politdcnica 42
D-7514 Eggienstein-Leopoldslbafen 2 Lisboa
Attn: AGARD National Delegate
GREECE IRE
Hellen~c Air Force General Staff DUepateto eerhadDvlpet(RE
Research and Development DirectorateDeatetoRsarhndevlp
Holaros, AhensMinistry of National Defence, Ankara nt( GE
UNITED KINGDOM
ICELAND Defence Research Information Centre
Director of Aviation Station Square House
c/o Flugrud St. Mary CrayF
Reykjavik Orpington, Kent BRS 3RE
UNITED STATES
National Aeronautics and Space Administration (NASA)
Langley Field, Virginia 23365
Attn: Report Distribution and Storag Unit
THE UNITED STATES NATION4AL DISTRIBUTION CENTRE (NASA) DOES NOT HOLD
STOCKS OF AGARD PUBLICATIONS, AND APPLICATIONS FOR COPIES SHOULD BE MADE -
DIRECT TO THE NATIONAL TECHNICAL INFORMATION SERVICE (NTIS) AT THE ADDRESS BELOW.
PURCHASE AGENCIES
Microfiche or Puhotocopy Microfiche Mka'ol ek
National Technical Space Documentation Service Technolo~ Reports
Information Service (NTIS) European Space Agency CentreaUH
5285 Port Royal Road 10, rue Mirio Nikis Station Square House
Springfield 7S015 Paris, France St. Mary C2ray
Virginia 22161, USA Orplirton, Kn R R RR
Enfeand
Requests for microfiche or photocopies of AGARD documents should include the AGARD serial. nunbiw, title, author or editor, and
publication date. Requests to NTIS should include the NASA accession report number. Full blbIlogesphicel references and abstracts
of AGARD publications are given in the following journals:
Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements (GRA)
published by NASA Scientific and Technical published by the National Technical
Informaton Facilit Information Service, Springfield
Post Office Box 877Virginia 2216 1, USA
Baltimore/Washington International Airport
Mtryland 21240; USA

Printedby Technhzl Ediftinaed Rqrodwdon Ltd

Harford Romge, 7-9 CUwltte At London WJP !ADL

3 ISBN 924835-1400.9