AGARDAG328

AGARD
QD
ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT

7 RUE ANCELLE 92200 NEUILLY SUR SEINE FRANCE
Advanced Methods for

Cascade Testing
(M&hodesA~&pour
les Essais des Grilles #Auk)
AGARD-AG-328
ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT

7 RUE ANCELLE 92200 NEUILLY SUR SEINE FRANCE
AGARDograph 328
Advanced Methods for

Cascade Testing
(MCthodes AvancCes pour
les Essais des Grilles d’Aubes)
Edited by
Professor Charles Hirsch

Vrije Universiteit Brussel
Dienst Stromingsmechanica
Pleinlaan 2
1050 Brussel
Belgium
This AGARDograph was prepared at the request of the

Propulsion and Energetics Panel of AGARD.
- North Atlantic Treaty Organization

Organisation du Traite de I’Atlantique Nord
I
The Mission of AGARD
According to its Charter, 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:
- Recommending effective ways for the member nations to use their research and development capabilities for the
common benefit of the NATO community;
- Providing scientific and technical advice and assistance to the Military Committee in the field of aerospace research and
development (with particular regard to its military application);
- Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence posture;
- Improving the co-operation among member nations in aerospace research and development;
- Exchange of scientific and technical information;
- Providing assistance to member nations for the purpose of increasing their scientific and technical potential;
-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.
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.
Published August 1993

Copyright 0 AGARD 1993
All Rights Reserved
ISBN 92-835-0717-7
Set and printed by Specialised Printing Services Limited

40 Chigwell Lane, Loughton, Essex IGlO 3TZ
ii
Recent Publications of
the Propulsion and Energetics Panel
CONFERENCE PROCEEDINGS (CP)
Heat Transfer and Cooling in Gas Turbines
AGARD CP 390,September 1985
Smokeless Propellants
AGARD CP 391,January 1986
Interior Ballistics of Guns
AGARD CP 392,January 1986
Advanced Instrumentation for Aero Engine Components
AGARD CP 399,November 1986
Engine Response to Distorted Inflow Conditions
AGARD CP 400,March 1987
Transonic and Supersonic Phenomena in Turbomachines
Advanced Technology for Aero Engine Components
Combustion and Fuels in Gas Turbine Engines
AGARD CP 422,June 1988
-
Engine Condition Monitoring Technology and Experience
AGARD CP 448,October 1988
Application of Advanced Material for Turbomachinery and Rocket Propulsion
Combustion Instabilities in Liquid-Fuelled Propulsion Systems
AGARD CP 450,April 1989
Aircraft Fire Safety
AGARD CP 467,October 1989
Unsteady Aerodynamic Phenomena in Turbomachines
AGARD CP 468,February 1990
Secondary Flows in Turbomachines
Hypersonic Combined Cycle Propulsion
AGARD CP 479,December 1990
Low Temperature Environment Operations of Turboengines (Design and User’s Problems)
AGARD CP 480,May 1991
CFD Techniques for Propulsion Applications
Insensitive Munitions
AGARDCPSll,July1992
Combat Aircraft Noise
AGARD CP 512,April 1992
Airbreathing Propulsion for Missiles and Projectiles
Heat Transfer and Cooling in Gas Turbines
iii
ADVISORY REPORTS (AR)
Suitable Averaging Techniques in Non-Uniform Internal Flows (Results of Working Group 14)
AGARD AR 182 (in English and French), June/August 1983
Producibility and Cost Studies of Aviation Kerosines (Results of Working Croup 16)
AGARD AR 227, June 1985
Performance of Rocket Motors with Metallized Propellants (Results of Working Group 17)
AGARD AR 230, September 1986
Recommended Practices for Measurement of Gas Path Pressures and Temperatures for Performance Assessment of
Aircraft Turbine Engines and Components (Results of Working Group 19)
AGARD AR 245, June 1990
The Uniform Engine Test Programme (Results of Working Croup 15)
AGARD AR 248, February 1990
Test Cases for Computation of Internal Flows in Aero Engine Components (Results of Working Group 18)
AGARD AR 275, July 1990
Test Cases for Engine Life Assessment Technology (Results of Working Group 20)
AGARD AR 308, September 1992
Terminology and Assessment Methods of Solid Propellant Rocket Exhaust Signatures (Results of Working Group 21)
AGARD AR 287, February 1993
LECTURE SERIES (LS)

3-D Computation Techniques Applied to Internal Flows in Propulsion Systems
AGARD LS 140, June 1985
Engine Airframe Integration for Rotorcraft
Design Methods Used in Solid Rocket Motors
AGARD LS 150, April 1987
AGARD LS 150 (Revised), April 1988
Blading Design for Axial Turbomachines
Comparative Engine Performance Measurements
AGARD LS 169, May 1990
Combustion of Solid Propellants
AGARD LS 180, July 1991
Steady and Transient Performance Prediction of Gas Turbine Engines
AGARD LS 183, May 1992
Rocket Motor Plume Technology
AGARDOGRAPHS(AG)
Measurement Uncertainty within the Uniform Engine Test Programme
AGARD AG 307, May 1989
Hazard Studies for Solid Propellant Rocket Motors
AGARD AG 316, September 1990
Advanced Methods for Cascade Testing
AGARD AG 328, August 1993
REPORTS (R)
Application of Modified Loss and Deviation Correlations to Transonic Axial Compressors
AGARD R 745, November 1987
Rotorcraft Drivetrain Life Safety and Reliability
AGARD R 775, June 1990
iv
Preface
Cascade testing has been, and still is, an essential component for the understanding of the basic flow mechanisms of axial
turbomachinery blade rows.
Notwithstanding the considerable development of numerical prediction techniques, the experimental investigation of cascade
flows, under well controlled initial and boundary conditions, still remains an essential element in the understanding of the
complex flow interactions inside blade rows.
Many experimental facilities and measuring techniques have been developed over the years with the objectiveof a more realistic
and accurate determination of flow quantities.
This AGARDograph was set up by the Propulsion and Energetics Panel in order to collect the large amount of know-how and
expertise accumulated over the years at various institutions, on various aspects of modem cascade testing techniques.
It is subdivided into seven chapters, including an introduction chapter on the basic concepts and definitions in modelling
turbomachinery flow conditions.
The second chapter, on linear cascades, deals separately with subsonic, transonic and supersonic conditions. Annular cascades
are considered in a third chapter, followed by a description of particular aspects of hot and transient cascades.Unsteady flows in
cascades are treated in the sixth chapter, while the seventh chapter deals with the specific aspects of 3D laser anemometry in
annular cascades.
Since this AGARDograph is written by many authors, it benefits from the amalgamation of a wide range of experience. This
advantage is offset by the difficulty of achievingthe kind of unity of a text originating from a single author. Consistent efforts have
been made to minimize overlaps, but it was felt that the unique collection of know-how communicatedin this text was worth this
small penalty.
We have been fortunate in being able to call upon some of the best experts in this field, who have agreed to communicatemuch of
their experience. We are grateful to all of them for their willingness to contribute to this joint effort and for their availabilityfor
the necessary coordination of collaborative work.
Prof. Ch. Hirsch

Editor
V
Prbface
Les essais des grilles d‘aubes ont it6, et demeurent, un iliment essentiel pour la comprihension des micanismes d’icoulement
de base des grilles d‘aubes des turbomachines axiales.
Malgrk le diveloppement considirable des techniques de pridiction numtriques, l’itude expirimentale des icoulements dans
les grilles d’aubes, effectuie dans des conditions initiales et limites bien diterminies, reste un Clement essentiel pour la
comprihension des interactions complexes d’ecoulement.
Un grand nombre d‘installations expirimentales et de techniques de mesure ont it6 diveloppies au fil des anntes avec pour
objectif de diterminer de faGon plus pr6cise et plus rialiste le volume de ces icoulements.
Cette AGARDographie a it6 proposie par le Panel AGARD de Propulsion et d’Energitique afin de rassembler dans un seul
volume I’expertise et le savoir-faire appriciables qui ont it6 accumulis sur de longues annCes par diffirents organismes sur
divers aspects des techniques modernes d’essais des grilles d’aubes.
L‘itude est subdivisie en sept chapitres, dont un chapitre d’introduction sur les difinitions et les concepts de base utilises pour la
modilisation des conditions d’ecoulement dans les turbomachines. a
Le deuxibme chapitre, qui traite des grilles liniaires, considbre s6pariment les conditions subsoniques, transsoniques et
supersoniques. Les grilles annulaires sont examinies au chapitre 3, qui est suivi de la description de certains aspects particuliers
des grilles chaudes et transitoires. Le chapitre 6 porte sur les icoulements instationnaires dans les grilles, et le septibme chapitre
traite de certaines questions spicifiques a I’animomitrie laser dans les grilles annulaires.
Cette AGARDographie ayant it6 ridigie par divers auteurs, elle binificie d’une grande itendue d’expkrience. Toutefois, cet
avantage se trouve contrebalance par les difficultis rencontries pour arriver 1 I’homogCnCitiqui caractirise un texte Ccrit par un
seul auteur. Un effort soutenu a 6tC fait pour minimiser les redondances, mais il a Cti considirk que le recueil unique du savoir-
faire que reprisente ce texte compensait largement cet Cventuel et liger inconvknient.
Nous nous estimons heureux d’avoir pu faire appel B quelques uns des spicialistes les plus iminents dans ce domaine, qui ont
bien voulu accepter de nous communiquer une bonne partie de leur expkrience. Nous sommes trks reconnaissants envers eux
tous pour la bonne volonti dont ils ont fait preuve dans leurs contributions 1 cet effort commun et pour leur disponibiliti en ce
qui conceme le travail de coordination.
Prof. Ch. Hirsch

Editeur
vi
Contents
Page
Recent publications of the Propulsion and Energetics Panel iii
Preface V
Ch. Hirsch, Editor
PrCface vi
Ch. Hirsch, Editeur
List of Authors viii
Chapter 1 Modelling Turbomachinery Flow Conditions 1

0 1.1 Flow Characteristics in Real Machines
L. Fottner
1
1.2 Modelling Turbomachinery Flows 1

C.H. Sieverding
1.3 Simulating Cascade Flow Parameters 3
L. Fottner
1.4 Definition of Cascade Blade Sections 7
C.H. Sieverding
1.5 Types of Cascade Tunnels (continuous, blow dowdsuction, short duration 7
open closed loop) and References to Existing Literature on Instrumentation
C.H. Sieverding
Chapter 2 Linear Cascades 11

2.1 Basic Fluid Dynamics Boundary Conditions of 11
Cascade Wind Tunnels (including turbines)
H. Starken
2.2 Subsonic Cascades (compressor, turbines) 12
L. Fottner
2.3 Subsonic-Choked (turbine) Cascades 22
0 C.H. Sieverding
2.4 Transonic and Supersonic Cascades (including turbines) 35
H.A. Schreiber, H. Starken and W.Steinert
2.5 Testing Techniques for Straight Transonic and Supersonic Cascades 60
M.M. Hoorelbeke, Gaillard and Losfeld
Chapter 3 Annular and Rotating Cascades 73

3.1 Annular Cascades 73
H.P. Hodson and R.G. Dominy
3.2 Rotating Annular Cascades 85
J. Amecke and F. Kost
Chapter 4 Particular Aspects of Hot Cascades 96

C.T.J. Scrivener
Chapter 5 Transient Cascade Testing (including shock tubes) 103
T.V. Jones, M.L. G. Oldfield, R. W. Ainsworth and T. Arts
Chapter 6 Unsteady Flow in Cascades: Flutter & Forced Response 153
D.H. Bufim and S.Fleeter
Chapter 7 Three-Dimensional Laser Anemometry in Annular Cascades 170

L.J. Goldman
vii
List of Authors
Chapter 1
Prof. L. Fottner 1.1,1.3,2.2
Universitat der Bundeswehr Munchen
Fakultat fiir Luft- und Raumfahrttechnik, Institut fiir Strahlantriebe
Wemer-Heisenberg-Weg39, D-85579 Neubiberg, Germany
Prof. C.H. Sieverding 1.2,1.4,1.5,2.3

von K i n n t h Institute
72, Chaussie de Waterloo, B-1640 Rhode-St-Genbe, Belgium
Dr H. Starken 2.1,2.4
Dr H.A. Schreiber 2.4
and Dr W. Steinert
DLR
Postfach 90 60 58, D-51140 Koln, Germany
Mr Hoorelbeke, Mr Gaillard and Mr Losfeld 2.5
ONER4
B.P. 72, F-92322, Chltillon Cedex, France
Dr P. Hodson 3.1
Whittle Laboratory, Department of Engineering, Cambridge University
Madingley Road, Cambridge CB3 ODS, United Kingdom
Dr R.G. Dominy 3.1

School of Engineering and Applied Science, University of Durham
South Road, Durham DH13LE, United Kingdom
Dr J. Amecke/F. Kost 3.2
DLR
Bunsenstrasse 10, D-37073 Gottingen, Germany
Dr C.TJ. Scrivener 4 I
Rolls-Royce plc
P.O. Box 3 1, Derby DE2 8BJ, United Kingdom
Prof. T.V. Jones, Dr M.L.G. Oldfield and Dr R.W. Ainsworth 5

University of Oxford, Department of Engineering Science
Parks Road, Oxford OX1 3PJ, United Kingdom
Dr T. A r t s
von Kim611 Institute
72, ChaussQ de Waterloo, B-1640 Rhode-St-Genbe, Belgium
Dr D.H. Buffum
NASA Lewis Research Center
Cleveland,OH 44135, United States
Prof. S. Fleeter
Purdue University, School of Mechanical Engineering
West Lafayette, IN 47907, United States
Dr LJ.Goldman 7
NASA Lewis Research Center, Aerodynamics Branch, Internal Fluid Mechanics Division
Cleveland, OH 441 35, United States
1
Chapter 1
Modelling Turbomachinery Flow Conditions
1.1 FLOW CHARACTERISTICS IN REAL MACHINES rotation on the flow within rotor blade rows and the effect of
L. Fottner unsteady i d o w conditions due to rotor blade wakes to the
The internal flow fields encountered in current turboma- stator blade rows should be considered. A second unique fea-
chines are viscous, compressible and unsteady. The flow ture existing in turbomachine flow fields is the transfer of
passage geometry provides a completely three-dimensional energy in the flow process as work done on or by the fluid as it
flow, and a sigmficant fraction of the flow occurs in rotating traverses the rotating elements.
passages. Some of the important phenomena influencing flow
fields in modem axial-flow turbomachinery blade rows are 1.2 MODELLING TURBOMACHINERY FLOWS
shown in Figure 1.1-1. The flow is C.H. Sieverding
- three-dimensional,the flow channel being given by the hub The progress in the understanding of the flow in real machines
and casing contour and the profile contours on suction and has come mainly through the use of simplified models allow-
pressure surface; ing the systematic study of selected aerothermodynamic and
- viscous, losses being generated by friction, mixing and
I
0 shockhoundary-layer interaction;
- compressible,with supersonic inlet (compressors) and exit
geometric parameters and the in depth investigation of parti-
cular flow phenomena. The models range from simple ducts
to linear cascades and fixed and rotating annular cascades,
(turbines) conditions; each step increasing the complexity of the flow conditions.
- unsteady due to the unsteady inlet conditions of rotating
and stationary blade rows of stages and multi-stage mach- Ducts
ines. Duct models are used for basic boundary layer studies, most
at low speed, looking at effects of surface roughness, surface
curvature, velocity gradients, Reynolds number, turbulence
and coolant flow injection on boundary layer development,
with emphasis on boundary layer transition.
Linear cascades
Contrary to duct flows, linear cascade models introduce the
aspect of flow periodicity because of the alignment of a finite
number of blades of constant blade sections at constant stag-
ger angle and separated by a constant pitch. The research
areas cover blade optimization,shock boundary layer interac-
tion, coolant flow effects, trailing edge flows and wake mixing.
The linear cascade model offers several attractive features
e like:
- geometric simplicity;
- simple adjustment and variation of cascade geometry and
flowfield;
- large blade sections compared to annular cascades;
Fig. 1.1-1 Axial-flow compressor rotor flow phenomena - clean optical flow path,
- independent variation of different flow parameters such as
inlet (outlet) Mach number M,,(M2), inlet flow angle 8,
These features have a significant effect on the overall flow pressure ratio P2ff1,Reynolds number RE and degree of
field, but they are not the only influentialfactors. What Figure turbulence Tb;
1.1-1 does not emphasize is that the phenomena shown take - possibility to generate approximate quasi-three-
place in a strong through-flow or primary flow which is gener- dimensional (Q3D) flowfields;
ally understandable in terms of the physical laws and - easy setup for basic investigation on wake-blade interfer-
corresponding equations of fluid dynamics. Many of the fea- ence and forced blade vibration.
tures of theflowfieldshowninFigure 1.1-1might therefore be The value of linear cascade tests has been for years a point of
called secondary flows or secondary effects. This suggests discussion because of the difficultyin establishing truly peri-
that such flows may be accounted for separately after deter- odic and two-dimensional flow conditions. Truly periodic
mination of the primary flow field in a given case. conditions require an infinite number of blades. However,
Tbrbomachine flow fields occur in passages characterized by because of mass flow restrictions, the infinite cascade model
some unique geometric and kinematic features, which do not is simulated by a cascade with a relatively small number of
occur in other typical complex flows, either internal or exter- blades, typically between 6 and 15, the higher numbers often
nal (Ref. 1.1-1). An initial unique feature is the sequence of being obtained at the expense of a reduced chord size. This
rotating and stationary internal flow passages. The effects of leads to boundary condition problems and much effort is
2
spent on minimizing the degree of non-periodicity. At trans- The first type of unsteady flow conditions may be simulated
onic inlet velocities, the sensitivity to variations of the by moving a cylindrical rod through the inlet flowfield parallel
boundary conditions reaches a maximum whereas the setup to the cascade front. This setup allows the simulation of wake
of periodic inlet conditions becomes progressivelyeasier with blend interferences and, at very high speeds of the rod, shock-
further increasing Mach number. blade interferences. Regarding unsteady aerodynamics
Because of the developmentof the boundary layer on the cas- related to blade vibrations, linear cascade research has gener-
ally been associated with analysis validation and direction of
cade tunnel side walls, all linear cascade flows are inherently
three-dimensional. In general, these effects are bigger in flutter and the aerodynamic portion of the forced vibration
problem, in contrast to developing a flutter data bank or gust
decelerating than in accelerating cascades. The combination
induced unsteady aerodynamics. This is because it is
of small inlet boundary layers, high aspect ratios and high
accelerations are conditions which favour two-dimensional extremely difficult to generate proper cascade inlet gust forc-
ing functions in a linear cascade. However, techniques to
flow conditions.
oscillate cascaded airfoils and acquire data over a range of
In recognition of the problem to establish truly two-dimen- intrablade phase angles have been developed. The proper
sional flow conditions and in regard to the convergence and simulations of the reduced frequency is nevertheless very dif-
divergence of the stream surfaces in real machines, cascade ficult to achieve.
wind tunnels are used more and more as a tool to provide
quasi-three-dimensionalblade to blade data, either for direct
application of losses and turning angles or to validate numeri- Annular cascades
cal codes. Abandoning the conservative two-dimensional According to the exact meaning, an annular cascade consists
flow concept for linear cascade tunnels has opened the way of an annular space between two concentric cylinders which
for a fruitful cooperation between experimentalists and contains a scale model of a turbomachine blade row. The
numerical analysts. The reliability of the experimental data is annular cascade therefore simulates the flow conditions in a
improved by using numerical calculations for the interpreta- turbomachine much more closely than a linear cascade.
tion of the data and the experimental data contribute in Though the majority of annular cascades do have cylindrical
defining the proper boundary conditions for numerical cal- endwalls, conical and contoured endwalls are also frequently
culations including the selection of appropriate turbulence used. The blade profiles may themselves be exact replicas of
models. those in a turbomachine, including lean and sweep, blade
twist and changes of section along the span, but many low
Over the last 15 years, linear cascade tunnels have been used speed annular cascade facilities use scaled up versions of
very extensivelyfor basic investigations of the generation and actual blade rows to ease detailed three-dimensionalflow sur-
evolution of secondary flows through turbine and compressor veys and to allow correct simulation of Reynolds number.
bladings including the effects of endwall contouring and
Compared to linear cascades, the main advantage of annular
blade stacking (effect of lean, sweep, dihedral). Also tip clear-
cascades is that they enable the simulation of the radial static
ance flows, including the effect of relative motion between
pressure gradient field which exists as a result of the need fo
blades and endwalls, are studied in linear cascades in view of a establish radial equilibrium in swirling flows. While such gra-
better understanding of the role of inviscid and viscous forces
dients have little effect on the main stream (inviscid) flow, low
driving the tip clearance flow and its interaction with the
momentum fluid, experiencingthe same gradients will tend to
mainflow. The experimental data on secondary and tip clear-
migrate towards the hub in the direction of the gradient. By
ance flows in linear cascades are very valuable for code
their nature, these gradients are absent from linear cascades
validation but they cannot be used directly for calculating the
and so a potentially important three-dimensional aspect of
flows in turbomachines where the radial static pressure gra-
the flow cannot be modelled.
dient field plays a particularly important role with respect to
the spanwise distribution of losses and outlet flow angles. A second advantage,compared to linear cascades, is the truly
periodic character of the flow. Shock and expansion waves
The need for very detailed three-dimensional flow measure- can propagate up- and downstream without being reflected
ments, with attempts to measure turbulence quantities in from any boundaries.
addition to the mean values, has led many researchers to use
very big models with blade chord lengths of up to 300 mm, Annular cascades, however, suffer from several limitations
limiting the number of blades to as few as three to five. Under like: reduced accessibility for instrumentation, backward
these conditions it is difficult to set up periodic flow condi- scatter mode rather than forward scatter mode for laser
tions. One way out of this dilemma is to act on the anemometry, difficultyin changingthe inlet flow angle, expen-
downstream boundary conditions to make the experimental sive models and higher operating costs for comparable model
blade velocity distribution coincide with the theoretical one. sizes. Because of power and massflow constraints, high speed
This is another example of an intelligent integration of theor- models are often smaller than their linear counterparts
etical prediction methods into an experimental procedure. causing increased probe blockage effects and higher measure-
ment errors due to stronger pressure gradients.
Finally, linear cascade test rigs are also used to model With regard to unsteady aerodynamics, the annular cascade
unsteady flow phenomena. Here we distinguish between the offers some attractive features.Wake-blade interferences are
following types of flow conditions: simulated easily by generating wake with spooked wheels or
rotating airfoils upstream of a stationary blade row. Similar
a) generation of unsteady blade forces due to periodic inlet setups may be used to investigate forced response gust aero-
and/or outlet flow uniformities, not inducing any blade dynamics. Furthermore, annular cascades are also suitable
motion; for developing a flutter data bank. The problem of non-
b) aerodynamic induced blade vibration in the form of flutter uniform airfoil phasing found in free flutter tests in linear
or forced vibration. cascades is eliminated in annular cascades.
3
Annular cascade sectors The performance of axial turbines and compressors generally
Annular cascade sectors are used occasionally as a compro- deteriorates as Reynolds number is decreased. Low Reynolds
mise between the objectives of (1) simulating properly the numbers are most commonly encountered in the last stages of
radial pressure gradient field and (2) reducing model and test- turbines and in high-flying aircraft or small-engine compo-
ing costs. It appears that this solution is chosen in cases where nents. As the Reynolds number of axial compressor blading
use can be made of existing engine parts. Hot turbine vane drops below a critical value, the decrease in efficiency, flow
cascade testing seems to be a typical application. However, and pressure ratio are particularly pronounced. A major part
the problem of flow periodicity is even worse than in a linear of cascade tests on subsonic cascade tunnels have been per-
cascade because of the difficulty to act on the boundary con- formed in this critical Reynolds number region (approx. lo5).
ditions in the presence of strong radial pressure gradients. If Laminar separation is almost inevitable under these circum-
used at all, the use of annular sectors should be limited to sub- stances, resulting in lower flow deflections than would be
sonic flow conditions. attained in an operational compressor. For cascade tests on a
given cascade ( k e d blade chord length) the Reynolds num-
Rotating cascades ber is proportional to the flow velocity and also dependent on
In general, a rotating cascade is defined as an annulus the temperature. Since Mach number is also a function of the
between two cylindrical or conical endwalls containing cylin- velocity and temperature, in most of the cascade tunnels it
drical blades. Contrary to annular stationary cascades, proved difficult to separate Mach and Reynolds number
rotating cascades initially were designed to provide two- effect on cascade performance. The solution is to systemati-
dimensional cascade performance data. The blade aspect cally vary the ambient pressure level, and hence the density,
ratios are usually chosen very high to minimize three-dimen- within the tunnel. In this case the temperature has to be con-
sional flow effects. Compared to the stationary cascade, the trolled at a constant value (as for Mach number) by cooling
rotating cascade offers two major advantages: upstream of the cascade.
a) it allows an easy change .of the inlet flow angle by changing The flow through a turbomachine is characterized by a strong
the rotational speed; turbulence. Periodic velocity changes are superposed by sto-
b) for turbine rotor blade cascades with high supersonic rela- chastic velocity fluctuations and the degree of turbulence
tive outlet velocity, the outlet flow is subsonic with respect achievesvalues up to 20%. Free-streamturbulence has a major
to a stationary downstream probe. effect on the flows that occur in turbomachinery (see Ref.
At present, the main interest in rotating cascades is their abil- 1.3-1). This is mainly due to its influence on the transition
ity to simulate the effect of centrifugal and coriolis forces on from laminar to turbulent boundary-layer, its further develop-
the blade flow. Tip clearance flows is a further area of inten- ment and hence its separation. Furthermore, the
sive research. Finally, flutter data can be obtained under boundary-layer displacement thickness affects the free-
realistic rotor flow conditions. stream flow and thus alters the conditions that would be
expected from potential flow theory. The nature of the turbul-
ence in a wind tunnel of conventional design, i.e. with proper
1.3 SIMULATING CASCADE FLOW PARAMETERS settling chamber plus an aerodynamically-designed contrac-
L. Fottner tion, can be expected to be homogeneous and nearly
If the results obtained in cascade (model) testing should be isotropic. In any series of tests in which the turbulence level is
representative of the flow in prbomachines, the similarity to be systematically varied the most convenient procedure is
rules have to be considered, i.e. viscous effects by Reynolds to ensure that the basic turbulence level of the “clean”tunnel
number and degree of turbulence, and compressibility effects is equal to or lower than the lowest desired test value. It is then
by Mach number similarity. Thus, cascade measurements necessary to insert suitable devices which will give the
have to be performed at typical turbomachinery values of required level of near-isotropic turbulence.
Mach number, Reynolds number and degree of turbulence.
This is not only true for design condition but also for off- Various types of devices have been used for this purpose.
design conditions of aero-engines. During flight operation a. These include turbulence-generating fans mounted in the
turbojet engine is faced with a large variety of operating con- settling chamber, rows of small flags oscillating in the inflow
ditions. A change of flight level, for instance, results in a and continuous chains of transverse wires passing upstream
change of the Reynolds number which is mainly caused by the of the cascade. However the most common and reliable
different density; but is also leads to a change of the Mach device is the simple bi-planar grid of bars mounted at the
number, as temperature is changing with the flight level, too. upstream end of the working section (Fig. 1.3-1and 1.3-2, see
On the other hand the Mach number is also influenced by the Ref. 1.3-2). It is the bar dimension which is the significant
engine speed. The various combinations of Reynolds num- parameter for the strength of shed vorticity and hence the tur-
bers and Mach numbers severely influence the flow through a bulence level at downstream locations (Fig. 1.3-3, see 1.3-1).
blade row. Therefore, a cascade wind-tunnel should have the
Because the effects of turbulence level and Reynolds number
capability of independently changing these two parameters to
are closely related it has been suggested by some workers that
make the cascade measurements transferable to turboma-
turbulence effects may be accounted for by postulating an
chines. In addition, the correct simulation of secondary flow
“effective Reynolds number” - a multiplier on the actual
effects should be considered.
Reynolds number. It has been experienced that the rela-
The Mach number effect can easily be simulated by varying tionship is not a linear one, thus, although closely related, the
the flow velocity and temperature. Thus, for subsonic flow Reynolds number and turbulence level must be treated as
conditions, the maximum Mach number is dependent on the separate experimental variables.
power of the air supply installed in the cascade test facilities.
In the case of closed circuit tunnels, heat exchangers have to For correct simulation of secondary flow effects side wall
be installed, to control the temperature of the air in the test boundary-layers are required, which are equivalent to those
section. on the hub or casing of the axial turbomachine (e.g. Fig. 1.3-4,
4
Ref. 1.3-3). Special tailored boundary-layerfences have been side-walls of the cascade wind tunnel (e.g.Fig. 1.3-5, Ref. 1:3-
used to generate prescribed boundary-layer profiles on the 4).
Turbulence Grid IXgK
Turbulence Grid VIK
21
$
e m - - - - - -
I)
I
91
-I
1ffit-- ml
Y
Turbulence Grid VIIIK

Fig. 1.3-1 Turbulence grids (see Ref. 1.3-2)
5
. .
7.4 7.21 1.b 0.8 0.6
-
-
Referenz Turbulenzgmd
fur p, 0.L5 bar
Re, - 5 . W ’
Ma, ‘ 0 . 2 9
H, - 6 L 2 m m
3.1 %
56%
Fig. 1.3-2 Degree of turbulence generated by turbulence grids (see Ref. 1.3-2)
Fig. 1.3-3 Turbulence attenuation downstream of a bi-planar grid (see Ref. 1.3-1)
6
Variable
Side-Wall
n n
'----+It
I
I Inletchannel
'----U
L---Yi?
Fig. 1.3-4 Cascade with vadable aspect ratio (see Ref. 1.3-3)
7
r 7 1 . 2 --t-71,2 7 7 2 2 - -7121
+
Bkch 70x8
360 _
76872
I
Q3
-
1 ' \
Blech 60 x 6,s
I33 0 --
Fig. 1.3-5 Boundary-layergenerators (see Ref. 1.3-4)
1.4 DEFINITION OF CASCADE BLADE SECTIONS faces. In incompressible flow conformal mapping may be
C.H. Sieverding used to map a conical stream surface onto a cylindrical sur-
In the case of an annular blade row with cylindrical endwalls, face (Ref. 1.4-2). The requirement of a constant pitch in
the corresponding linear cascade is obtained by developing a straight cascades results of course in a significant change of
cylindrical cut through the blading onto a plane surface. This passage geometry. Hence the two-dimensional performance
simple cylindrical cut neglects of course that (a) the stream- of these sections cannot be used without correction factor to
line in the meridional plane is in general not a straight line and evaluate the performance of the corresponding stage.
@) the initial axisymmetric stream surface becomes twisted
when passing through the blade row (Figs. 1.4-1 and 1.4-2). 1.5 TYPES OF CASCADE" E L S
The endwalls of compressors and turbine blade rows are, C.H. Sieverding
however, in general neither cylindricnor parallel. Under these Depending on their running time we distinguish basically
circumstances it is much more difficult to define appropriate three types of cascade facilities:
blade sections for straight cascades; see Figure 1.4-3 (Ref. running time
1.4-1).There is no general agreement as to whether the blade 1. Continuous facilities
several hours
sections should be defined on a cylindrical or a conical sur- oDen or closed circuits
face, or along a streamline. At subsonic velocities and 2. Intermittent facilities: from several seconds
moderate radius change, the cylindrical cut seems to be blowdown or suction type to several minutes
applied most widely. At supersonic flow and increased radius 3. Transient facilities:
exhaust to vacuum tank
<2 seconds
change, preference is given to blade shapes from conical sur-
8
Fig. 1.4-1 Streamline in meridional plane
Blade surface SB
rotated to P
Fig. 1.4-2 Twisting of stream surface through blade row
\ Developed Axisymmelric Section
Projection Surfacr
Cascade Projection
Fig. 1.4-3 Development cross section and projection of an annular cascade

9
Transient facilities form strictly speaking a subcategory of the sometimes appear to be unfavourable, considering the high
intermittent type facilities but their very short running time installation and operating costs on one side and the limitation
and their characteristic feature of producing a sudden onset of the linear cascade model on the other side. However, most
of the flow put them into a distinct class both from the point of compressor cascade tunnels are equipped with rather sophis-
view of measurement and instrumentation technique.Catego- ticated endwall boundary layer control systems which need a
ries 1 and 2 use basically the same instrumentation long time for proper adjustment. Also,certain measurement
techniques. techniques like extensive three-dimensional LDV measure-
Except for hot cascades testing facilities, most facilities of ments are extremely time-consuming and could hardly be
categories 1 and 2 are cold flow facilities with total tempera- done in intermittent facilities. Finally, it may also be argued
tures near to ambient conditions to ensure adiabatic flow that, in many cases, the cascade tunnel is only one out of many
conditions. In continuous running facilities, a heat exchanger different users of the same central air supply system.
reduces the compressor outlet temperature to the desired test The reason for intermittent facilities is usually that the power
section inlet temperature. requirements of equivalent high speed continuous facilities
For the selection of the most appropriate type of facility the are prohibitively large. This is very relevant in the university
followingaspects are to be taken into account: environment whose operating costs are to be kept small.
Intermittent cold flow facilities appear to be used more for
turbine cascade than compressor cascade testing. The reason
1. the-desired level of flow simulation: e.g. high speed or low is that turbine cascades require less refined and easier to con-
speed, need of Re variation or not, simulation of tempera- trol endwall boundary layer control systems (Ref. 1.5-12).
ture ratios in the case of cooled turbine blades, size of Intermittent facilities make use of a high pressure air supply
blades, etc.; e.g. VKI (Ref. 1.5-13) or of a big vacuum tank e.g. DLR Gott-
2. the financial needs for installation of the facility and its ingen (Ref. 1.5-14). In the Gottingen suction type tunnel, the
oDeration; downstream Mach number is set by a plug valve in the outlet
3. the measurement techniques to be used in the facility and duct connecting the test section with the vacuum tank. In
the data acquisition techniques required by them. blowdown facilities the cascade exit pressure is generally
close to atmospheric pressure, however, a diffuser may lower
From the point of view of scientific return to capital invest-
ment ratio, the continuous open loop low speed cascade
facility with air supply with a big ventilator producing air velo-
-
this pressure. For very high Mach number turbine nozzle
blades, the required pressure ratio is so high (e.g. x 10 for
M, = 2.5) that blowdown tunnels are the only feasible solu-
cities in the test section of 30 to 100 m / s is probably the tion to test such cascades. The big inconvenience of many
optimal research tool for aerodynamic measurements. The intermittent cascade facilities is that they do not allow for an
possibility of using large blades developing relatively thick independent Reynolds number variation.
boundary layers makes these facilities particularly well suited
For more information about cascade testing facilities the
for detailed boundary layer studies. Linear cascade facilities
reader is referred to the followingchapters and Reference 1.5-
with blades of 200-300 nun chord length are not unusual
16).
(Refs. 1.5-1,1.5-2) and in one case (Ref. 1.5-3)a three bladed
cascade with a chord length of 530 nun was used. Of course, The transient facilities can be divided into three subcatego-
the absence of Mach number effects sets definite limits to the ries:
application of low speed investigations to high speed cascade
flows. Neither the effect of shock induced pressure gradients 1. the MIT blowdown tunnel, which consists of a pressurzied,
on boundary layer characteristics, nor the strong Mach num- heated vessel which is suddenly vented through the test sec-
a ber dependence of the base pressure of turbine blades

operating in the transonic domain can be simulated in low
speed facilities.
tion (Ref. 1.5-16);
2. the isentropic light piston tunnel developed at Oxford (Ref.
1.5-17) in which the air is compressed, and thereby heated
Compared to low speed cascades, the power requirementsfor up, through a light piston and then released through the test
continuous high speed facilities are at least one order of mag- section by a fast acting valve;
nitude higher. Mass flow requirements of 5 to 10 kg and 3. the shock tunnel, used mainly at Calspan (Ref. 1.5-18) in
pressure ratios of 2 to 4 require multistage axial compressors which the air is compressed through a shock wave travelling
with power consumptions of up to 3 MW. Most facilities are of along a tube before expandingthrough a nozzle into the test
the closed loop type to allow easy Re number changes and section via a small diaphragm ruptured under the impact of
limit the power requirements, e.g. at DLR (Ref. 1.5-4), the shock.
ONERA (Ref. 1.5-5), EPFL (Ref. 1.5-6), Whittle Lab (Ref. These tunnels lend themselves particularly well for heat trans-
1.5-7). In some cases the cascade tunnel is installed in a big fer measurements using transient measurement techniques
tank allowing testing at pressure levels between 0.1 and 1bar but can also be used for aerodynamic measurements. All
in the facility of the German Armed Forces University, three tunnels have the capability of simulating not only Mach
Munich (Ref. 1.5-8) and between 0.1 and 4 bar at GEC (Ref. and Reynolds numbers but also the correct temperature
1.5-9). ratios between main gas, blade surface and coolant gas.
Contrary to most continuous facilities, continuous running
hot cascade facilities are of the open circuit type. These facilit- REFERENCES
ies require heavy insulation to allow safe operation and steady 1.1-1 SerovyG
state flow conditions. The measurement techniques are IW.1 Axial Flow 'krbomachine Through-Flow Cal-
altogether very different from those used in cold flow tunnels; culation Methods. In Hirsch Ch, Denton J (Eds.)
see (Refs. 1.5-10,1.5-11) and Chapter 4. Through FIow Calculations in Axial Turbomachines,
The use of continuous high speed cascade facilities may AGARD AR 175,1981.
10
1.3-1 Gostelow JP bomachinery Research, ASME Paper 76-GT-62.

Cascade Aerodynamics, Pergamon Press,. Oxford, 1.5-8 Sturm W, Fottner L
1984. The High Speed Cascade Wind n n n e l of the Ger-
1.3-2 Kiock R, Laskowski G, Hoheisel H man Armed Forces University Munchen. Paper in
Die Eneugung hoherer Turbulenzgrade in der Proceedings of the 8th Symposium on Measuring
Mepstrecke des Hochgeschwindigkits Gittenuindkan- Techniquesfor Transonic and Supersonic Flows in
als Braunschweig zur Simulation turbomaschine- Cascades and Turbomachines,Genoa, 1985.
niihnlicher Bedingungen,DFVLR, FB 82-85,1982. 1.5-9 Forster VT
1.3-3 Watzlawick R, Fottner L Turbine Blade Development Using a Tmnsonic Var-
Investigationson the Effect of Flow Parameters in Sec- iable-Density Cascade Wind Tunnel, Proc. Inst.
ondary Loss Correlationsfor Different Blade Aspect Mech. Engrs, P6/65,1964.
Ratios, MTZ Motortechnische Zeitschrift 53, 6, p. 1.5-10 Tobery EW, Bunce RH
276-284,1992. Cascade Heat Transfer Tests of the Air Cooled
1.3-4 Hoheisel H W 501 D First Stage kine, ASME Paper 84-GT-114.
Verlustmessungen an einem Verdichtergitter mit 1.5-11 Rogers HF,Graham C, hlicholas K
Schaufelspalt im Hochgeschwindigkeits-Gitterwind- Advanced Techniques Employed in Blade Cooling
kanal, DFVLR-Ber. Nr 0680,1971. Research,AGARD CP 421,1987.
1.4-1 Starken H 1.5-12 Kiock R, Lehthaus F, Bames NC, Sieverding CH
Comparison between Flows in Cascades and Rotors The TmnsonicFlow through a Plane Turbine Cascade
in the Transonic Range. I - Basic Considerations.In as Measured in Four European Wind Tunnels, J. of
Ma LS 84 Transonic Flows in Axial Turbomachin- Engineering for Gas lbrbine and Power, Vol. 108,
ery, 1976. No 2, April 1986.
1.4-2 Lewis RI, Fischer EH, Saviolakis A 1.5-13 Sieverding CH
Analysis of Mixed-Flow Rotor Cascades, ARC R & M Test Case E/CA-8: Transonic Turbine Cascade. In
3703,1971. AGARD Advisory Report 275 on Test Cases for
1.5-1 Langston LS, Niu ML, Hooper Rh4 Computation of Internal Flows in Aero Engine Com-
Three Dimensional Flow within a Turbine Blade Pas- ponents.
sage, ASME J. of Engineering for Power, Vol. 99, 1.5-14 Lawaczeck 0
No. 1,Jan. 1977. Der Prifitand fiir ebene Schaufelgitter der Aero-
1.5-2 Gregory-Smith DG, Graves CP dynamischen Versuchsanstalt Gottingen, DFVLR-
Secondary Flows and Losses in a Turbine Cascade, AVA-Bericht 70 A 46,1970.
AGARD CP 351,1983. 1.5-15 Sieverding CH
1.5-3 Han LS, Cox W R Aerodynamic Development of Turbomachinery
A VisualStudy of Turbine Blade Pressure Side Boun- Bladings - Testing Techniques and Steady State Per-
dary Layers, ASME Paper 82-GT-47. formance Measurements. In Thermodynamics and
Fluid Mechanics of Turbomachinery, Vol. I. NATO
1.5-4 Starken H ASI, Series E, No 97B, Martinus Nijhoff Publishers,
Untersuchungder Stromung in ebenen ifberschallver- 1985.
zogerungsgittem,DLR FB 71-99,1971.
1.5-16 Epstein AH, Guenette GR, Norton RIG
1.5-5 Leynart J The MIT Blow Down Turbine Facility, ASME Paper
Cascade Test Methods in Wind n n n e l at ONERA. 84-GT-116.
Symposium on Measuring Techniques in Tmnsonic 1.5-17 Ainsworth RW, Schultz DL, Davies MRD, Forth
and Supersonic Flows in Cascades and Turboma- CPJ, Hilditch MA, Oldfield MLG, Sheard AG
chines, EPFL Lausanne, Nov. 1976. A TmnsientFlow Facilityfor the Study of the Thermo-
1.5-6 Bolcs A fluid-Dynamicsof a Full Stage Turbine under Engine
A Test Facility for the Investigation of Steady and Representative Conditions, ASME Paper 88-GT-
Unsteady Transonic Flows in Annular Cascades, 144.
ASME Paper 83-GT-34. 1.5-18 Dunn MG
1.5-7 Gostelow JP,Watson PJ Heat-Flux Measurements and Analysisfor a Rotating
A Closed Circuit Variable Density Air Supplyfor Tur- Turbine Stage, AGARD CP 390,1985.
13
systematic variation of the ambient pressure level, and accuracy the blade chord length should not be smaller than 60
hence the density, within the tunnel. In this type of tunnel a mm. Thus, depending on pitcWchord ratio, stagger angle,
heat exchanger downstream of the air supply compressor blade chord length and cascade tunnel height, a minimum
has to be installed to cool the air to laboratory conditions. number of five blades will be recommended: one as measure-
ment blade, two above and below this measurement blade.
2.2.1. Control of Periodic Flow Conditions For high pitch cascade testing (s/c>2) for Propfan bladhgs,
Number of blades cascades consisting of only three blades have been tested.
In principle, cascade measurements are mostly performed for
one blade passage (Fig. 2.2-2) i.e. the blade velocity distribu- Flow periodicity
tion on pressure and suction side of one blade and the wake The flow periodicity will be influenced by the upper and lower
behind the blade averaged over one pitch. Thus, the number walls of the cascade tunnel in front of the cascade, the cascade
of blades of the cascade has to be chosen in such a way, that itself and the flow field downstream of the cascade. In front of
the addition of further blades would not seriously affect the the cascade there are straight walls of the cascade tunnel (Fig.
flow field in the region of the measurement blade. Only in this 2.2-3, see Ref. 2.2-2). In the range of the blade passages very
case, the infinite cascade model will be simulatedby a cascade often an intermediate streamline between suction and pres-
of a limited number of blades. Therefore, a cascade should sure surface is proposed as end-wall shape in connection with
have a large number of blades. On the other hand, a cascade the possibility of suction at points A and D to account for this
wind tunnel has limited dimensions.Because of measurement approximation.
lboundary layer suction]
' lboundary liyer suction1
Fig. 2.2-2 Cascade test section in a cascade tunnel

14
Fig. 2.2-3 Control of periodic flow condition (see Ref. 2.2-2)
The essential condition for good flow periodicity is, that the inlet flow decays very rapidly in upstream direction. At an
stagnation streamlines of the outer blades of the cascade are axial distance of half a pitch in front of the cascade the distur-
identical. The shape of these stagnation streamlines,however, bancies are less than the measurement accuracy. The
are very strongly affected by the flow in the outer by-pass experience in cascade wind-tunnel operation has shown that
c h F e l s A-B and C-D (Fig. 2.2-3) normally installed in cas- it is sufficient to check the velocity distributionin the plane B -
cade wind tunnels. If there is in addition a common back C'by means of side-wall static pressure tappings for one pitch
pressure in the plane E-H the static pressure rise in the blade and the flow angle at one point by means of a single
passage has to be equal to that in the by-pass channels. This directional probe in the same plane. The flow is adjusted in
condition is especially severe in the bottom passage A-B such a way that the side-wall static readings are constant and
where the static pressure rise has to be achieved at thick boun- the flow is parallel to the wind-tunnel axis. A similar proce-
dary layers along the suction surface A-E. The experience in dure can be applied to the exit flow if the measuring plane
cascade wind-tunnel operation showed that a flexible wall F'-G is placed between half to one pitch downstream of the
along A-E together with some kind of boundary-layer con- cascade measured in axial direction. In this distance exit flow
trol, for instance porous wall with suction along A-E or angle and static pressure are .nearly constant. According to
blowing in a backward facing step at A, is sufficient to control theory and experience, the flow becomes more and more sen-
the stagnation streamline B-B. The problems with the upper sitive to disturbances with increasing flow Mach number. In
by-pass passage are not so difficult because the influence of cascade investigations this means that the flow periodicity
this channel on the inlet and exit flow field is much less com- adjustment becomes more difficult. Depending on the cas-
pared to the boffom one. Very often a flexible wall along D-H cade geometry this may result for instance in a non-periodic
is sufficient. Sometimes however, a possibility for suction flow angle variation ahead of the cascade, although the static
along D-H or through a protruding slot at D would give a use- pressure distribution may be constant. Especially high turning
ful and additional parameter for the flow adjustment. In the compressor cascades show this effect. But also the problem in
cascade arrangement of Figure 2.2-3, the flow would leave the achieving a periodic flow upstream of high staggered turbine
plane E-H in a guided channel or a guided exhaust. But it is cascades is probably due to the same reason. Good inlet flow
also possible to increase the area downstreamof E-H, thereby conditions can only be achieved by careful adjustment of the
applying a dump diffusion downstream of the cascade. In this by-pass wall shape, of the boundary-layer suction at A and D,
case the exit flow develops as a free jet and the flow angle and the horizontal and vertical position of points A and D,
adjusts itself to the downstream back pressure. Periodicity is thus reducing the flow angle gradient to zero.
achieved by shifting the points A and D and by varying the The measurement of the exit flow field over one pitch is gener-
suction rate through slots arranged at these points. A third ally performed half to one chord downstream of the cascade
method would be a dump diffusion in the area F-G together measured in axial direction. In this plane the flow angle varia-
with the by-pass channels connected to a separate suction sys- tion is normally small and also the static pressure distribution
tem. This would allow an independent control of the mass is nearly constant. This facilitates the flow traverses because
flow through the by-pass channels and would probably also the probes are generally sensitive to gradients in flow quantit-
reduce the difficulties. ies. For this reason any measurement in the plane F-G is
extremely difficult although the influence of the wall bounda-
After the discussion of the different methods for the periodic- ries is much less in this position. To facilitate measurements
ity control in cascade wind-tunnels the question arises, how close to the trailing edge, i.e. for measuring mixing losses in
the desired periodicity can be checked. From incompressible the wakes, special probes have to be used (e.g. Fig. 2.2-4, Refs.
theory it can be found that the influence of cascades on the 2.2-2 and 2.2-3).
11
Chapter 2
Linear Cascades
2.1 BASIC FLUID DYNAMIC BOUNDARY not, depending on the related velocity range or velocity com-
CONDITIONSOF CASCADEWIND TUNNELS ponents. The fluid dynamic background to this behaviour is
described in the literature (e.g. Refs. 2.1-1 to 2.1-S), and can-
H. Starken not be repeated here.
Due to the fmite size of the cascade model the fluid dynamic The variability of the @et flow angle, PI, exit flow angle, P2
boundary conditions are not as simpleas in the case of a single and static pressure ratio, p2/p1, as a function of the inlet and
body in an infinitely large flow field. Especially at high velocit- exit flow conditions of a cascade are summarized in Figure
ies, inlet flow angle, exit flow angle, and exit Mach number are 2.1-1. The relations are shown for a compressor cascade, but
sometimes independent variable quantities and sometimes they apply also to turbine cascades.
\ Exit supersonic Supersonic

Subsonic Transonic axial axial
Inlet subsonic supersonic
subsonic
Subsonic
choked
1
Transonic
Transonic/
Supersonic
choked
Supersonic
axial
subsonic
Supersonic
axial
supersonic
-
1) free In the meanlng of an Independent varlable parameter between certaln Omb
2) flxed due to vIxx)us eFFecLs (AVDR constant)
3) determined by passage “at
4) determlned by unique Incidence
Fig. 2.1-1 Classificationand variability of inlet and exit boundary conditions of cascade flows
12
Subsonic flow: The subsonic inlet and exit flow condition is 2.2 SUBSONIC CASCADES (COMPRESSORS,
characterized by the fact that inlet velocity and inlet flow TURBINES)
angle can be specified independently and the exit flow angle is L.Fottner
obtained due to viscous effects (Kutta condition). A very customary simplification for axial turbomachines is
the linear cascade model. It can be obtained by geometrically
Subsonic choked inlet: The subsonic choked inlet flow condi- developing a cylindrical blade-to-blade surface which inter-
tion is generally a limit of the operating range of a compressor sects the interestingblade row (see Fig. 2.2- 1).This procedure
cascade as shown in Figure 2.1-1 but it is also the operating results in a mathematically infinitely long radial projection of
point of transonic and supersonic turbine cascades. At this the considered problem (i.e. Ref. 2.2-1). The main geometri-
condition there exists only one inlet flow angle at each inlet cal parameters of a linear cascade are:
Mach number. The related exit flow conditions ranging from - profile geometry (camber, thickness distribution);
subsonic to axial supersonic velocity are obtained by varying - cascade geometry (pitch/chord ratio, stagger angle);
the downstream pressure or back-pressure. With the back and the aerodynamic parameters are defined by
pressure the exit flow angle is also changing at transonic and - flow angles at cascade inlet and exit;
supersonic exit flow. However at axial supersonic exit condi-
tions theexit flow angle is fixed by inlet velocity and blade
- Mach number and Reynolds number.
geometry (unique deviation) (Ref. 2.1-4).
Transonic inlet: At transonic inlet flow conditions the inlet
flow angle is specified by the back pressure at subsonic and by
the unique incidence at transonic and supersonic exit flow
conditions.
Transonidsupersonicchoked inlet: If the flow experiences at
supersonic inlet velocity a minimum cross section within the
blade passage either due to the blade geometry or due to
boundary layer separation, it may also choke. This inlet condi-
tion is therefore called “transonidsupersonic choked”. It is
identical to the so-called ”unstarting”or transonic or super-
sonic cascades.
Supersonic axial-subsonicinlet: Starting is generally achieved
by increasing the Mach number and the inlet flow angle. The
related flow cases are shown in the next line of Figure 2.1-1.
They are characterized by the unique incidence condition for
the inlet flow which means blade geometry and inlet Mach
number are furing the inlet flow angle. This applies also to
impulse type supersonic turbines or compressors which suf-
fered in the past quite often on detached bow waves caused by
a relatively thick leading edge or a large wedge angle leading
to a considerable increase in incidence (Ref. 2.1-7).
The exit flow is again dependent on the back pressure at axial
subsonic exit conditions and on blade geometry and inlet flow
at axial supersonicexit conditions.
Axial supersonic inlet: If the axial component of the inlet flow

is increased to supersonicvalues, as considered in the last row
of Figure 2.1-1,the inlet flow angle is, between certain limits,
again independent of the inlet Mach number and a free para-
meter. The exit flow behaviour is identical to the axial
subsonic case described before.
All the above described dependencies lead to severe conse- Fig. 2.2-1 The cascade model
quences in cascade wind tunnel design, operation and
measurement techniques. For instance, the inlet flow angle of
a transonic or supersonic compressor cascade may be com- Subsonic cascade tunnels are usually continuous-running
pletely different from the upstream nozzle direction. rather than of the intermittent type because their power
Therefore, it has to be determined somehow, either experi- requirements are modest. There are different possibilities:
mentally or theoretically or in a combination of both ways. a) the tunnel is fed from an air supply upstream of the cascade
and has a dump discharge to the ambient condition;
The complexity is increased further by the possible variation b) the tunnel consists of an intake to the cascade direct from
of the inlet and exit flow angle by changing the back pressure, the ambient condition with suction of the air downstream
as described in Figure 2.1-1. In order to account for all of of the cascade by the use of suction pumps.
these varying boundary conditions, the high speed cascade c) The most usual solution is the closed circuit variable dens-
wind tunnel has to be highly flexible with regard to the geo- ity cascade tunnel. In this case, the independent variation
metry as well as to the operating conditions. of Mach number and Remolds number is Derformed bv
15
Wedge Probe
measurement locations
@ flow direction (0,)

static pressure (p,)
Neptun Probe
(aI-
(b)
(a)-(a) probe center-line
(b)-(b) axis of rotation for D2-adjustment
5 - Hole Probe
Fig. 2.2-4 Probes used for wake measurements(see Ref. 2.2-3)
Three-dimensionaleffects it is observed that for a given cascade configurationand inlet

The derivation of good passage periodicity has been asso- flow and with a given incoming side-wall boundary-layer the
ciated with the use of end-wall boundary-layer bleed. secondary flow influences are independent of aspect ratio.
Another fundamental criterion,the two-dimensionality of the Hence, the dominating three-dimensional flow feature is the
flow is associated with the side-wall boundary-layer and the commonly observed comer stall which emanates from the
aspect ratio of the blading (Ref. 2.2-4). In any cascade wind- first region of reasonably unfavourable pressure gradient in
tunnel with straight side-walls the measured flow may not be the end-wall region. It is this comer stall which causes consid-
two-dimensional even in the mid span section. This is due to erable blockage and therefore a higher axial velocity density
the increase in side-wall boundary-layer thickness which ratio in the mid-span section of the passage. Because the phy-
leads to a contraction of the cross-section area. In addition, sical area of comer stall is unaffected by changes in aspect
the side-wall boundary-layer is responsible for the develop- ratio (unless the aspect ratio is so low that the stalled regions
ment of secondary flow effects caused by the pressure from the comers actually meet at the mid-span) the axial
difference between suction and pressure surfaces of the density velocity ratio in solid side-wall tests is strongly
blades. In systematic tests on cascades having solid side-walls dependent upon the aspect ratio.
16
Tho-dimensional flow conditions at a given aspect ratio are 4.4. By this means the rotational speed of the axial compres-
usually defined by the following conditions (Ref. 2.2-5): sor within the pressure tank can continuouslybe altered in the
- equal pressures, velocities and directions exist at different range of about 1500 to 6200 rpm. Its maximum air flow is
spanwise locations; about 30 m3/s and a maximum total pressure ratio of 2.14 can
- the static pressure rise across the cascade equals the value be achieved. Since each cascade works like a throttle and,
associated with the measured turning angle and wake; therefore, corresponds to a different working point in the per-
- no regions of low energy flow other than blade wakes exist. formance map of the compressor, a variable by-pass helps to
These wakes are constant in the spanwise direction; prevent the compressor from stalling.As mentioned above,all
- the various performance values do not change with aspect the compressor work must be cooled out of the air flow again.
ratio. This is achieved by a lamella cooler for the main flow and by
two smaller ones for the by-pass air. To increase the cooling
To fulfil these conditions, careful side-wall suction has to be efficiencyof the main flow cooler, a diffusor is built in in front
applied using protruding slots upstream of the cascade and of the cooler. Although the main flow cooler already acts like
continuous suction through porous side-walls. With this a straightener,a settlingchamber followsthe cooler for mixing
degree of control over wall boundary-layers, flow two-dimen- out pressure and temperature non-uniformities.After this the
sionality is no problem regardless of aspect ratio. It was flow is reaccelerated within a nozzle until it reaches the
therefore possible to use an aspect ratio value of unity. desired upstream Mach number of the cascade to be mea-
sured.
Three-dimensional effects occur not only in cascades but are
basic flow phenomena in turbomachines. Therefore, the The rectangular test section is 300" broad for verifymg
objective is to control them so that they are similar in the high aspect ratios. The height is variable between 250" and
wind-tunnel and in the turbomachine.The necessary contrac- 500" depending on the upstream flow angle of the cascade.
tion or divergence of the cascade flow area can either be The latter is mounted between two rotatable discs, which
achieved by an adequate shaping of the side-wallsor by appli- allow a continuous variation of @1and 250 and 155 with @1
cation of suction or blowing through porous walls. However, measured against the circumferential direction. Especially if
both methods do not guarantee that the desired distribution measuring compressor cascades the end-wall boundary-lay-
of the flow contraction or divergence from cascade inlet to ers cannot be neglected. Therefore, the wind-tunnel is
exit really exists in the mid-span section where the measure- equipped with a special suction system, which is based on a
ments are generally taken. Techniques for obtaining a centrifugal compressor of 155 kW power. Suction of the
prescribed axial velocity density ratio in cascades often upper and lower end-wall boundary-layers (see Fig. 2.2-2)
include the use of porous side-walls.Even with the best side- provides the necessary inlet flow periodicity. Suction of the
wall suction arrangements it would be prudent not to use left and right end-wall boundary-layers allows control of the
aspect ratios below two. With low aspect ratios even if desired axial velocity density ratio. In experiments with blade boun-
axial velocity ratios are obtained the suction surface boundary-layer control by blowing or with simulating the cooling
dary-layer may still be strongly three-dimensional.A problem of turbine blades an additional air supply is necessary. For this
exists if it is desired to use Schlieren visualisation as well as a 1000 kW screw compressor has been built up. By using dif-
side-wall suction. ferent types of turbulence generators various degrees of
turbulence between about 0.3% and 6% are simulated.As the
generators are inserted within the nozzle (see Fig. 2.2-5)
2.2.2 Variable Mach NumberAteynoldsNumber Cascade which is some distance away from the cascade, a nearly
Test Facility uniform upstream velocity profile can be achieved.
A unique facility for independent systematic variation of
Mach and Reynolds number at different degrees of turbul-
ence has been installed at the Jet Propulsion Institute of the 2.2.3 Advanced Measuring Techniques
University of the Federal Armed Forces in Munich, which For the interpretationof profile pressure distributions and the
was formerly designed and operated at the DLR in Braunsch- boundary-layer behaviour it is important to know the inlet
weig (Refs. 2.2-6,2.2-7). This High-speed Subsonic Cascade flow conditions, i.e. total and static pressure, turbulence
Wind-Tbnnel is a continuously operating closed loop test intensity and total temperature distributions as accurately as
facility with an open test section. It consists of two main parts possible. It is essential that the pressure and temperature dis-
(see Fig. 2.2-5), the driving unit and the wind-tunnel itself, of tribution of the inlet flow show good constant distribution in
which the latter is installed within a cylindrical pressure tank. the relevant area of the inlet flow for the application of adv-
The necessary air flow is delivered by a six-stage axial com- anced measuringrechniques (Ref. 2.2-8).
pressor. By varying its rotational speed, any desired Mach
number between 0.2 and 1.0 can be obtained. Sincea constant
Mach number requires a constant total temperature within Schlieren-measurements
the tank, the whole energy which is equivalent to the compres- For investigation of shocks in transonic cascades, Schlieren-
sor work must be cooled out of the air flow again. For measurement systems are used in the above mentioned
independently setting the Reynolds number the pressure cascade test facility. In order to be able to focus the system
within the tank can be varied between 0,05and 1.2 bar, thus during test runs all optical equipment is installed outside the
achieving a Reynolds number range of ReA 106m-' and
-' vacuum-tank. Therefore two further Schlieren windows in the
1.5.107m-'. The complete performance map of the High- vacuum-tank are necessary in addition to the windows in the
Speed Cascade Wind-Tbnnel is shown in Figure 2.2-6 side-walls of the cascade (Fig. 2.2-7). The optics are built up
including the limitsof operation.The driving unit on the right- in a z-type arrangement and the light source is an optional
hand side of Figure 2.2-6 which is outside the pressure tank, Xenon-lamp or a strobe; documentation can be done with a
consists of an a.c. electric motor of about 1300 kW power, a photo-camera and a video-system. Problems will arise when
hydraulic coupling, and a gear box with a transmission ratio of side-wall boundary-layer suction has to be applied.
17
0 . .
.. I1 II
a-aNa
II I1
.-
c
C
cn v
.. .. ..
3
0 . .
-
'E
VI V I VI VI
18
..
t
.-
t
0
d
L
aJ
a
19
J concave mirror
~ ~~
plane mirror
Schlieren-windows of vacuum-tank
L
-/st-section
T/
/ photo-camera
Granite optical bench
movable mirror
Schlieren-windows of
e cascade side-walls
Schlieren-edge
7
f ocusing screen
Fig. 2.2-7 Schematic view of the Schlieren-measurement system

20
Laser-two-focusvelocimetry A general view of the laser-two-focus system at the cascade

wind-tunnel is shown in Figure 2.2-8: The light from the laser
For various research topics it is essential to measure the veloc- Lexel95 (2 W multiline) is guided via a 13m fibre optic cable
ity in cascades, especially in the blade passages including to the watercooled and ventilated optical head, which is inside
profile and sidewall boundary-layers. Because of the advance the vacuum-tank on a 4-axis traversing system. The tran-
in nonintrusive measuring techniques, laser anemometry can sceiver lens of the optical head has a working distance of
contribute to the solution. The most serious problem of this 270". Light scattered by particles passing through the
experimental set-up was the environmental condition for probe volume is collected by the outer part of the lens and
such a system inside the vacuum-tank of the cascade wind- directed to the two photomultipliers which are connected
tunnel: pressure down to 50 mbar, temperature up to 60'C.A with the signal processing unit outside the cascade wind-tun-
new development for the laser-two-focusvelocimeter avoids nel. The traversing system driven by stepper motors is
this problem by installing the laser outside the vacuum tank controlled by a computer HP 9856, which can also operate as
and using a single mode fibre optic cable to guide the light a controller of the velocimeter. Furthermore, it is used for
from the laser to the optical head inside the tank. data storage as well as graphical presentation of the results.
probe volume
I
-L:
I
air
to Host-computer
Perkin-Elmer 3203
traverse system
Bragg-cell
1 fiber optic coupler
I \r
I
- -
I
laser LEXEL 95
I laser power unit
water
Fig. 2.2-8 Schematic view of the laser-two-focus measurement and signal analysis system
21
Heated thin-film measurements Tunnel has been used for turbulence level measurements. In
The condition of a boundary-layer (laminar, transitional or addition velocity measurements within the blade passage of a
turbulent) can be determined by measuring the mean and turbine cascade are done. Because of temperature variations
fluctuating component of the heat transfer rate. The experi- in the inlet flow, temperature compensated sensors are used
mental set-up of this technique for investigating with a symmetrical anemometer bridge. For setting various
boundary-layers on cascades at the High-speed Cascade Reynolds-numbers the pressure inside the tank can be
Wind-Tunnel is similar to that published in References 2.2-9 adjusted. The hot-wire measuring system is calibrated and lin-
and 2.2-10 (Fig. 2.2-9). Investigations are carried out to earized for each tank pressure. In order to decrease sensor
detect the transition region and laminar separation bubbles damage dust-filters are installed in the suction area of the vac-
on turbine and compressor cascades at the High-speed Cas- uum-tank. The experimental set-up for hot-wire measure-
cade Wind-Tunnel. This technique shows the transition region ments at the High-speed Cascade Wind-Tunnel (Fig. 2.2-10)
quite satisfactorily. is similar to the heated thin-film technique (Fig. 2.2-9). The
system is designed for analog evaluation of the anemometer
Hot-wire measurements output signal.The possibility exists for frequency analysis and
The hot-wire anemometry at the High-speed Cascade Wind- visual display of the signal.
Profile with mounted sensors
Wall of uacuum-tank
Analog Multiplexer 48 to 1
Prema 2024
to Host-computer
Perkin-Elmer 3 2 0 3
Const. Temp. Anemometer

DlSA 5 5 M system
Low-Pass Filter (96dB.lOOkHz)

NF M o d e l FV-665
Fast Fourier
Analyzer
Rockland
5815A
M
Digital
to Host-computer
Plotter Perkin-Elmer 3 2 0 3
IEEE-Bus
-CO-axial cable
Fig. 2.2-9 Schematic view of the heated thin-film measurement and signal analysis system
I 22
Sensor e.g. DlSA 55P61
Const. Temp. Anemometer

DlSA 5 5 M system
symmetrical bridge
I Linearizer
TSI M o d e l 1072
N F M o d e l FV-665
Dual Beam
Dig. Storage
Oscilloscope
Gould
to Host-computer
Perkln-Elmer 3203
IEEE-Bus
-CO-axial cable
Fig. 2.2-10 Schematic view of the hot-wire measurement and signal analysis system
2.3 SUBSONIC-CHOKED (TURBINE)CASCADES ers in transonic turbine cascades undergo an overall

C.H. Sieverding acceleration resulting in a thinning of the boundary layer and
2.3.1 Two Dimensional Flow Conditions therefore in an effective divergence of the blade channel
Basically, the requirements for two-dimensional flow condi- rather than a contraction as in the case of compressor blades.
tions are the same as those described in sections 2.2 and 2.4. Apart from very low reaction blades, the acceleration of the
However, compared to compressor cascades, the require- flow up to the throat is in general shock free. For supersonic
ments for sophisticated side-wall boundary layer control outlet Mach numbers, the flow continues to accelerate down-
systems are somewhat less critical for transonic turbine cas- stream of the throat but the effect of this additional
cades with subsonic inlet and transonic to supersonic outlet acceleration on the side-wall boundary layer is offset by shock
velocities. An upstream boundary layer aspiration slot para- interferences (except for the limit loading case), such that the
llel to the cascade leading edge plane is in general the sole overall effect is relatively small. In the case of strong shocks,
boundary layer control device used in transonic turbine cas- the side-wall boundary layer may undergo a thickening com-
cade facilities.A number of facilities are operated without any pared to its state in the throat, but since it is a thickening of an
such device at all. The reasons for this seemingly carelessness initial thin boundary layer, its effect on the flow at midspan
are outlined below. appears nevertheless to be small. One should of course keep
in mind that side-wall boundary layers are highly three-
Contrary to compressor cascades, the end-wall boundary lay- dimensional due to secondary flows and shock boundary
23
layer interactions and local th"

' ensional effects on the surements nre subject to three types of error inducing flow
cascade flow may be much more important than the overall situations shown in Figure 2.3-2
effects. a) reflected shocks interfere with the suction side boundary
The best evidence of an effective spanwise divergence in the layers of the blades selected for wake traverses;blade 4,
streamtube up to the h a t is the recording of measured Figure 2.3-2;
choking Mach numbers in excess of the thenretical choking b) reflected shocks pass between the blade trailing edge and
Mach numbers for Oao-dimensionalflows.These higher inlet the d o w n " probe, probe position A, Figure 2.3-2;
Mach numbers imply of course also an effect of the blade c) reflected expansion waves pass between the blade trailing
velocity distributionwhich is biggest close to the leading edge edge and the downstream probe, probe position C, Figure
and diminishesgradually towards the throat region. However, 2.3-2.
as demonstrated in a comparison betweenvarious European Incase(a)themeasured bladelossesmaybeupto50%higher
cascade wind-tunnels with inlet duct to blade height ratios than for a neighbouring blade without this interference. In
varying from5 to 20, this does necessarily affect neither the case @) two things happen: 1) the probe measurements
blade losses nor the outlet flow angle in a significantmanner include the total pressure loss of the reflected shock, 2) as a
(Ref. 2.3-1). The reason is that the boundary layer develop result of the reflected shock the probe records the total blade
m a t along the pressure side as well as along the suction side pressure loss at a reduced velocity, k,the loss coef6cient will
up to the throat conaibutea only a minor part of the overall betoohigh.Incase(c)theflowvelocityinfrontoftheprobeis
lossesandtheoutletflowangleisalsolittleinftuencedbywhat higher than in the case of a nonperturbed flowfield;the total
happens upstream of the throat. blade pressure loss is r e f e d to a too high dynamichead, i.e.,
. 'onal &ects small, the relative inlet
To keep three-dtmensl the loss c d c i e n t will be too small. Flow Visualizations are
boundary layer thickness must be small. It is therefore essen- an absolute necessity to control, under which conditions the
0 tial to keep the inlet duct to blade height ratio small and the
aspect ratio high. Ba@ on a limited amount of cascade data,
wakemeasurementsare-ted.
Onedistinguisbesthrectypesofboundariesattheextremities
the curve in Figure 2.3-1 suggests minimum aspect ratios in of the cascade exit flow:
functionof the velocity ratio acrossthe cascade to obtain reli-
able two-dimensional midspan performance data. a) free shear layers developing downs- of the trailing
edge plane due to a sudden pitchwise aren enlargement
behind the upper and lower blade;
b) solid tailboards hinged to the trailing edges of the upper
and lower blades,
3.0 I I 1 1 I c) perforated tailboards.
Schlieren photogmphs of the exit flowconditions behind a
Ih IC Imin trahponic cascade with boundaries of types (a) and @) are
shown in Figure 2.3-3, h m (Ref. 2.3-2). For a correct inter-
2.5 pretation of the flow pa" in the awe of free shear layem at
the cascade flow boundaries, a short comment on the
interaction of shock waw with shear layers may be useful.
In inviscid flowtheory,the interference of an oblique shock
2.0
with a separated flowzone is treated by assuming the transi-
tionbetween the supemuic flowand the separatedzone to be
a discontinuity. Then, to ensure a pressure in the
0 1.5
separated zone, an incident shock is directly reflected as
expaasion wave and vice vem In real viscous flows, both
expansion andcompression waves emanate from the interfer-
ence region of the shock with the free shear layer. 'Ibe
compression wawa owe their existence to the gradual change
1 .o of the incident shock, from an oblique shock to a n o d
0.2 0.4 0.6 0.8 1.0 - shock, through the shear layer, this c k g e necessitating for
Vl'V2 equilibrium conditionsthe generation of a new oblique shock.
.. . .
Hence there exist two tvoes of d e c t e d shock IcomDression)
waw which are related:
Fig. 2.3-1 Recommended minimum aspen ratio for
reliable 20 performance measurementsIn turbine c88uLde8 a) to the m a l u r e of the incident shock inside the shear layer,
and
b) to expansion waw interfering with the free shear layer.
25.2 Out*t~wPerlodicity Itisgenerallyadmittedthatfreeshearlayersasboundariesfor
Due to the limited number of blades, it is impossible to esta- supersonic cascade exit flowsare the wrst solution.They are
blish truly periodic flow conditions for s u p e d c outlet flow not only responsible for strong reflected shocks but induce
conditions. Shock and expansion waves generated inside the a h strong unsteady flow effects.lhiling edge shocks oscill-
cascade interfawith theboundaries limiting thecascade exit ate at a fewhundred Hertz with non-negligible amplitudes,
flowfield and are reflected back into the flowfield deteriorat- such that the blade suction side pressure rise across the
ingtheflowperiodicity.Whatevertypcofboun~esareused impact ofthe trsiling edge shock is smeared out over a certain
at the cascade exit flow extremities, it is not possible to distance, giving incorrect information of the real adverse
eliminate entirely these reflections. As a result, wake mea- pressure gradient.
24
Fig. 2.3-2 ReRected shock end expension waws causing non-periodic cascade outlet flow conditions
Fig. 2.3-3 Schlieren photcgraph of lransonic turbine cascade with d h n t cwcade outlet fiow boundaries
25
The use of solid tailboards as boundaries for the cascade flow the plate. In spite of a quasi-periodiaty in the downstream
is limited to flow regimes characterized by a strong depend- Mach number distribution,the Schlieren photograph clearly
ence of the outlet angle on the outlet Mach number, i.e., it is shows the existence of a reflected shock and its interference
not advisable to use them in the transonic range where this with the central blade. Wake traverses were not made to show
dependence is weak. Due to the unique relation between whether or not this interference influenced significantly the
Mach number and flow angle in supersonic flow, the outlet boundaly layer. Also, the authors did not show a duect com-
Mach number may be varied by the sole variation of the tail- parison with a solid tailboard.
board inclination. The measured average outlet flow angle is
in general within 1to 2’ of the tailboard setting angle. 2.3.3 Pmbe Blockage Effect
It is usually assumed that the presence of the probe in the
The Schlieren photograph in Figure 2.3-3shows veryclearly
flowfield does not signblcantly affect its characteristics. How-
theshockrenectionsfromthetailboard.Theleftrunningtrail-
ever, this does hold only if the blockage area of the probe is
ing edge shock of blade 2 interferes after reflection from the
tailboard with the suction side of blade 3 (retlected shock
small compared to the flow area. The blockage effect
increases rapidly with Mach number and is most pronounced
RS2). The left running trailing edge shock of blade 3 passes
atM- 1.0. WithfurtherincreaseinMachnumberitdecreases
after reflection just behind the trailing edge of blade 4 (ref-
in magnitude. Wyler (Ref. 2.3-5)derived equations for evalu-
lected shock RS3) and reinforces the right mnning trailing
ating the blockage effect of cylindrical probes inserted
edge of blade 4 impinging on blade 5 , etc. It would be hazar-
sidewards halfway into an axisymmetric jet. The author
dous to perform wake traverses without simultaneous visual
arrives at the following equatiom for the change in pressure
flow control.
e
In the past, several attempts have been made to avoid wave
reflections from the cascade boundaries by using perforated
tailboards. HeMe and Gallus (Ref. 2.3-3)report encouraging
results with a 6% open perforated wall with holes inclined at
P
- (7);
- (s)
and Mach number near the probe for closed (open) tunnels:
d
[1.15 + 0.7S(M - 0.2)](-)
D
60’, a configuration recommended by h t h e r t (Ref. 2.3-4) (2.3.1)
for use in supersonic wind-tunnels. Due to Goethert, a com-
plete shock canfellation can be achieved, if the pressure loss
through the porous wall is equal to the pressure rise through
the shock, i.e., the pressure rise inside the chamber (see Fig. _
2.3-4)is equal to the static pressure upstream of the shock. M
6M -
- (k) - [
1+ e M 2
M2 [1.15 ] + d
0.7S(M - 0.2)](-)
D
Since the Mach number varies along the perforated wall, it
appears unlikely that this condition will be satisfied all along (2.3.2)
e
--
-
L
,-, \-, \---5.1
PITCH 3 PITCH 2 PITCH 1

. . . . . . . . .
1.
or]
Fig. 2.3-4 Use of perforated walls for shock cancellation (Ref. 2.3-3)
26
Figure 2.3-5presents the Mach number change in graphical field of a turbine cascade. The cylinder is positioned in the
form. It allows calculation of the maxi" blockage ratio
d/D (d - probe diameter, D -jet diameter) so as not to distance 4 C a -
wake of the upper blade on the schlieren picture at an axial
0.52 from the trailing edge plane (C, -
exceed a certain percentage of the Mach number itself.
-
Although the abcve relations cannot strictlybe extended to M
1, they have the merit of showing the tendencies and point
- -
axialblade chord) and spans the entire flow channel.The ratio
of stem diameter to chord length is d/c 0.046, the ratio of
stem diameter to the cascade exit flow width d/H 0.02. The
out clearly that the probe size has to be minimized by all comparison of the two Schlieren pictures shows that the trail-
means. ing edge shock of the blade behind which the probe stem is
positioned is Little affected while the shock system of the next
blade, which is much closer to the stem, is strongly modified.
A cylindrical probe stem with a short probe head would obvi-
ously give completely erroneous results. Better results would
be obtained with a probe equipped with a probe head extend-
ing far upsweam of the probe stem to point A as indicated on
the right picture.
h b e s with short probe heads were investigated by Koschel
and F'retsch (Ref. 2.3-7)in view of their application for inter-
blade row measurements in turbines where space limitations
dictate the probe design to a large extent. The optimum probe
designresultingfromthisstudyisshowninFigure2.3-8a.The
use of such probes is in principle also attractivefor secondary
flow messurements in cascades because of the advantage of
usingtheminazerobalancingmode,reducingthusgreatlythe
calibration effort.However, their use in the transonic domain
0 .05 .to .PO

at M
probe
-
is not advisable. The Schlierenpicture in Figure 2.3-8b taken
1.2 illustrates the problem of the interaction of the
stem bow shock with the probe head.
DIAMETER RATIO - .Is
dlD The blockage effect ofprobes inserted through the side-walls
increases in general with increasing immersion of the probe
Fig. 2.3-5 The effect of blockage on Mach number at into the flow.Themeasurement problem is then further aggra-
various Mach numbers (Ref. 2.3-6) vated by the fact that the probe stem causes an important
spanwise pressure gradient in particular at transonic outlet
Both equations give only the blockage effect near the probe. Mach numbers. These problems can be overcome by using
&uld (Ref. 2.3-6)showed that the blockage effect decreases probes with stems which exceed the probe head and extend to
upstream in a linear manner with distance. the opposite wall.
Compared tothecaseofacylinderattheexitofauniformaxi- The probe stem being the main cause of the probe blockage
symmetric nozzle flow, the situation of a probe stem behind a effect, it is obvious that single headed multi-hole pressure
cascade is in general quite different, as demonstrated in probes mounted on a long sting and introduced from far
Figure 2.3-6.With the probe stem in position A the blockage behind into the cascade outlet field as in the case of the DLR
effectdepends on the ratio d/H, but in position B the prodm- cascade tunnel (Fig. 2.3-9)are the best solution to minimize
ily of blade 5 augments considerably the blockage effect.The the blockage effect This probe configuration has also the
schlieren photograph in Figure 2.3-7shows qualitatively the advantage that traversing at various span heights does not
upstream effect of a cyhdrical probe stem on the outlet flow- change the blockage effect
Fig. 2.3-6 Probe stem in outlet flowfield of various axial distances

4
PROBE STEM
Fig. 2.3-7 Schlieren pholograph of blockage Mecl of probe stem
Fig. 2 .8 Cone pmbe with prismatic shafi in Rowfieldat M - 1.17 (Ref. 2

28
,
INLET PROBE
POSITIONS. 4
LGol ENOWALL
Fia. 2.3-9 DLR turbine cascade tunnel
2.3.4 Measurements in Flows with StrongPressure To reduce the measurement errors with 5 ie pressure p r o
Crndients bes traversing the outlet flow field of transonic turbine guide
Apart from measurement errors which can directly be attri- vanes with strong trailing edge shocks, Osnaghi and Perdi-
buted to calibration errors or inaccurate functioning of parts chica (Ref.2.3-9) use a procedure in which the probe is
of the measurement chain, the most frequent and most successively shifted into different positions so as to allow all
important measuringerrorsareduetothefactthattotaZstatic readings to be taken on a line parallel to the assumed trailing
and differentialprepsures are measured in different flowfeld edge shock direction. A comparison of cascade data with the
points with non-identical flow conditions. Multi-hole pres- proposed procedure and standard traversing techniques (i.e.
sure.probes like cone and wedge type probes are by definition simultaneousreadings of all pressure values for a given probe
not well suited for use in strong pressure gradient fields position) is shown in Figure 2.3-12.
because of the relatively big distances between the various Dienichs et al. (Ref.2.3-10) went a step further and carried
sensing holes, in particular the distance between the total and out a test program in which the local flow properties behind a
static pressure holes denoted by lin Figure 2.3-10. The mea- transonic turbine cascade were evaluated by interpolating
surement error depends both on the pressure gradient and the measurements from three. closely spaced measurement
distance I, the ermr being of course biggest in the case of a planes to account for the distances between the total and
pressure discontinuity(shock)crossing the probe head, probe statiepressureholesinpitchandaxialdirections(Fii.2.3-13).
position a in Figure 2.3-10. Siewrding (Ref. 2.3-8) has The single headed 5-hole wedge probe was moved in steps
attempted to evaluatethe error forthe case of probes interact- and Schlierenphotographsweretaken at each step to obtain a
ing with expansion waves, probe position in Figure 2.3-10. visual correlation between the flowdeld. the probe position
The curves in Figure 2.3-11 present for given Mach number and the measured values. The authors concluded that, in spite
gradients dM/d(y/g) in the measurement plane the maximum of this time-consuming procedure, the shock losses were not
-
ratios Ug to limit the error in the evaluation of the local effi-
ciency to Aq OS( 1.0)%. Effortsare made to miniaturize the
probe heads but manufachuing problems and excessive
correctly evaluated.
For two-dimensional flowfields, the requirement of point
probe response times set a lower limit. measurements can be relaxed and replaced by the require-
29
meat that all presmea be measured on the same skeam sion wawa to interfere with the probe nose before rea-
surface along a line p d d to the sailing edge of the blade. the pressure holes. "re do not exist any adequate calira-
Under these condiiiom the use of 2- or >finger probes is to tion p m c e d m to account for pressure gradient e&an on
be pnferred to single headed probes. Howew, experience the probe readings but Sieverding (Refs. 2.3-4 d 2.3-11)
showsthataproperslignmentofthepressureholesisnotsuf- SUgBested two simple pressuregradient mts dowing evalua-
ficient to guarantee correct measuretnents. Wbrated m tion of the aptimde ofprobes for use in supersonic &e
uniform flow,theprobeis exposedto highly non-uniform flow flow:
conditions.Thestaticpressunprobeisdwaysthemostsensi-
tiw to react to these diflerent flow conditions,beeause the a) traversing flowfrelds governed by a P d t l - M e y e r expan-
static pressure hdes are necessarily positioned somewhat siollfan;
behindtheprobenose,whi&allmexpansionandcompres- b) traversingobliqne shocks oflcnomintenaity.
Fig. 2.3-10 Shock and expansion waves interfering wilh cone probe head
0.11
11s
0.12 L
0.10
0.08.
0.06 ~
0.OL -
0.02
OO
Fa 2.3-11 Maximum recommended I/g for keeping errors of local ,,-values below 1% (0.5%)
Fig. 2.3-12 Results of shifted and standard traversing techniques (Ref. 2.3-9)
Fig. 2.3-13 Interpolationof data

31
I
I
I I Y €1
Fig. 2.3-14 Enuneous behaviour of needle static pmssura pmbe - Prandtl-Meyer expansion
In the first case,the total p m r e is conserved thmughout the Negative values of 5 indicate an overcorrection for the bow
expansion fan. Hence,any deviation of the measured total wave losses due to the staticpressure being mawred too low.
measurement error which,in general, has to be attributed to -
pressurefromthe total upsaeamp~sureisanindicationofa Near the top wall and outside the expansion fan, the measure-
ments appear faultlesswith 5 0. In between these positions,
an incorrect evaluation of the pitot probe bow wave shock the measurement err01 becomes very signilicant, reaching a
loases resulting from an error in the static pressure measure- maximum of -2.4% just before the needle probe pressure
ment. Figure 2.3-14 illusaates very convincingly the sensing holes start to move out ofthe expansion fan.Based on
inaptitude of a 2-finger probe with a combined totaldirec- this experiment,needle static pressure probes were rejected
tional probe head and a separate needle static pressure probe for use in supersoniccascade flow.
the dash-dotted line. The Mach number Vsried from M 1.4-
head for suchflow conditions.The probe was traversed along In the second case, an oblique shock is set up at the exit of a
supersonic nozzle (Fig. 2.3-15). At WU. a 4.5’ wedge was
near the top wall to M = 2.0 outside the expansion fan. The
results are presented under the form of a kinetic energy loss
usedforanozzleexitMnumbsrM- 1.6.TheMachnum-
ber ratio across the shock was M2 M, - 0.91 with a
corresponding kinetic energy loss coefiieient F = 0.1% only.
The probe tested in this setup was a 2-finger probe with a
combined total-directional probe and a separate 16‘ cone
static pressure probe head. The probe was traversed in steps
(2.3.3) BS well as at various constant speeds across the shock.The raw
32
pressure data for the stepwise probe displacement are pre- pressure probe). The differentpressure rise lengths lead obvi-
sented in Figure 2.3-16. None of the pressure sensing holes ously to errors in the evaluation of Mach number, flow angle
records the shock as a discontinuity. This can of course be and kinetic energy loss coefficient.In terms of the loss coeffi-
expected due to the finite size of the sensing holes. However, cient the maximum error in the shock region is of the order of
the total interference length AI,,, deiined by the pressure 1 point for the above example. Traversing the shock at contin-
rise length is longer than the geometric interference AI in uous speed aggravates the situation because of the relatively
Figure 2.3-17, as indicated in the table below: slow response times of the complete pressure transmission
system, i.e., probe-transmission lines-transducer, which is in
Pitot static Directional most tunnels of the order of several tenths of a second. It is
Almm(mm) 1.2 2.0 3.4 thereforeimportant to adapt the probe traversing speed to the
Algmm(") 0.8 0.6 2.2 probe response time. The tendency to reduce probe heads
and pressure sensing holes increases of course the response
The differences between AI, and AIge,,,,, are due to the time. On the other hand, very low traversing speeds entail high
interferenceof the field shock with the probe shocks (in part- testing costs due totheincreasedairconsumption. Apractical
cular with the pitot probe bow wave) and with the probe setup to test the probe response time is shown in Figure 2.3-
boundary layers (in particular on the cone head of the static 18.
Fig. 2.3-15 Shock generator at exit of supersonic nozzle

0.33 1 1 1 i i i 1
Pcot
- 0.02
Po1
0.3 1 “PLR
p 01
0
0.29
-0.02
0.27
-0.04
02 5
- 0.90 - 0.06
0.23
I I I 1 1 1 1
2.8 3. 6 L.0 5.2 6.0 6.8 7.6 ‘.‘I[mal 9
Fig. 2.3-18 Measured pressure rise through oblique shock
Fig. 2.3-17 Shock crossing pressure sensing holes

34
TRANSOUCER
I
’ TO ATMOSPHERE
I
1 - S M A L L CAVITY OF VOLUME vi
2 - RESERVOIR OF VOLUME v~
3 - SOLENOID:s
4- TUBE CONNECTING CAVITY WITH TRANSDUCER
5 - VALVE
Ap = 200 mm Hg
X :2 m l s Igrad.
Y : 1 V Igrad
PRESSURE JUMP I N CAVITY Vi
Fig. 23-18 Pressure jump generator

35
2.4 TRANSONIC AND SUPERSOMC CASCADES as shown in Figure 2.4-2. However, the drawback of this noz-
H.A. Schreiber, H.Starken and W.Steinert zle is its considerable length compared to a complete Laval
2.4.1 Introduction nozzle resulting in thick wall boundary layers. It is therefore
Within the following sections some basic considerations will better to allow one or two blades being outside of the test
be presented which are especially important for the design of rhomb of a complete nozzle (Fig. 2.4-3) and to take the mea-
a transonic and supersonic cascade wind-tunnel as well as for surements around a blade passage which is inside of the test
preparation, running, and data reduction of the tests. Further- rhomb. The errors are small, because they are only caused by
more, the given information shall be helpful for slightly different bow waves of the first blades connected with
understanding and interpretation of the results obtained. small differences in the upstream total pressure distribution.
Most of the complexities of low speed testing apply to high The first bow wave is anyhow different from the following
speed tests too, but there are some more to be considered. In ones, because the first blade has to establish the periodic inlet
particular the special features of the transonic and supersonic flow. It has to be taken care, of course, that the shock and
inlet flow field have to be taken into account to establish a expansion waves, reflected at the upper wall do not enter the
uniform and periodic flow field and to reduce the experimen- cascade again. This can be accomplished by a sufficiently
tal inlet flow data in a correct manner. The considerations are large upper by-pass passage as shown in Figure 2.4-4 (Ref.
true for high speed compressor cascades as well as for turbine 2.4-2).
cascades with transonic and supersonic inlet velocities. Typi- This figure also explains, where the inlet flow conditions (MI,
cal examplesfor such turbine bladings are supersonic impulse PI) of the related infinite cascade can be measured. This is
type bladings or tip section bladings of large steam turbines. only possible in the triangular area I-K-L limited by the first
The experimental treatment of the transonic and supersonic reflected shock wave provided that the upstream nozzle flow
exit flow fields of the compressor cascades is quite similar to is uniform there. In many of the existing cascade facilities,
the turbine cascades, discribed in the previous sections, but however, the blades are placed further downstream for
the effects of the axial stream sheet contraction (AVDR) and mechanical reasons. The nozzle is thereby extended by a
varying back pressure call for special consideration. slightly diverging channel keeping the Mach number as con-
In addition, at transonic velocities the blockage effect and stant as possible. Examples for this are the DLR (Fig. 2.4-5)
thereby the sensitivity to variations of the boundary condi- and the ONERA Supersonic Cascade Wind-?bnnels (see also
tions reach a maximum. This is especially true for the Section 2.5).
displacement effects of the endwall boundary layers which In the transonic case, the characteristics are nearly normal to
can have a tremendous influence on the flow field. In com- the flow direction and the reflections always enter the blade
pressor cascade tests, with highly loaded boundary layers, the passages. In this case the upper wall has to be made porous in
endwall boundary layer control is one of the most important order to provide the necessary near constant pressure area
subjects. < and to reduce the strength of the reflected shock waves. A suc-
tion system connected to the porous wall improves the
adjustment of the inlet Mach number and inlet flow angle. The
2.4.2 Cascade Inlet Flow DLR Transonic Cascade Wind-Tunnel shown in Figure 2.4-6
An impressive example of what may happen in a transonic is equipped with such a wall, which enables a continuous var-
cascade wind-tunnel has been derived by Lindner (Ref. 2.4- 1) iation of the inlet Mach number from subsonic to low
who calculated the upstream flow field of a solid wall wind- supersonic values including a very stable sonic operation.
tunnel configuration with a cascade at positive incidence Additionally this system provides the possibility to increase
operating at sonic condition upstream of the first blade as the inlet flow angle a few degrees. This is very useful because
shown in Figure 2.4-1. Due to the self-adjustmentof the cas- in the transonic velocity range the inlet flow angle depends on
cade inlet flow to the unique incidence condition combined the back pressure and therefore the inlet flow field has to be
with the reflecting solid upper wall the inlet flow field is con- adjusted according to the throttle setting. Without a suction
tinuously accelerated up to a Mach number of 1.84 ahead of system the whole cascade has to be rotated and set at the cor-
the fifth blade. There exists absolutely no periodicity in the rect incidence prescribed by the back pressure. However, the
inlet flow field and therefore the results would be useless for determination of the inlet flow conditions is extremely diffi-
turbomachinery applications. cult in this case. Possible solutions to this problem are
described later in Sections 2.4.5.1 and 2.4.5.2.
It may be possible to improve the periodicity somewhat by
applying back pressure and thereby generating detached bow
waves. However, the main problems are due to the solid upper 2.4.3 Cascade Exit Flow
wall where shock and expansion waves are reflected entering Compressor cascade tests in the transonic and supersonic
all the blade passages. inlet Mach number regime require a special throttle device in
the downstream area of the blade row. This is necessary
This example illustrates quite clearly the importance of the because at subsonic choked, supersonic unstarted-choked,
right hand running characteristics in the supersonic cascade and supersonic started cascade flow conditions, the back
inlet flow. Theoretical considerations lead to the following pressure can be varied independently of the upstream flow
conclusion: The supersonic cascade inlet flow is periodic if conditions. But also at higher inlet flow angles and unchoked
the right hand running characteristics originate in a uniform conditions, the back pressure must be controlled to achieve
upstream flow field. If, additionally, the upstream flow field is the desired incidence.Besides this, thick wall boundary layers
assumed to be irrotational by neglecting the shock losses in limit the pressure recovery in the outer passages. A detrimen-
this area, the inlet flow field is of the “simple wave” type. tal upstream flow break-down can only be avoided either by
In order to satisfy this condition,the cascade model has to be adding tailboards to the outer blades and throttling only the
placed within the test rhomb of a Laval nozzle. In principle inner region (see Figs. 2.4-5 and 2.4-7) or by connecting the
the half nozzle would be best suited for a cascade wind-tunnel outer passages to a separate suction system.
36
Fig. 2.4-1 a Flow pattern and related boundary conditions of possible cascade flow configurations
I
II 1 1 1 1 31 jy&
I I I I
/ / I / - I
9
Fig 2.4-1 b Calculated entrance-characteristicsand iso-Mach lines of a compressor cascade in a solid wall wind-tunnel
configuration at sonic nozzle flow conditions
test rhomb
---
_c
\
Fig. 2.4-2 Supersonic cascade wind-tunnel with half Lava1 nozzle
test rhomb
Fig. 2.4-3 Supersonic cascade wind-tunnel with complete Lava1 nozzle

37 I
O P
Fig. 2.4-4 Inlet flow field in a supersonic cascade wind-tunnel
Test Section Data Air Supply
-
Mach number
Reynolds number
Turbulence level
1.3 2.4
5*105- 3.10'
< 1%
- -
Closed loop. continuous operation
Mass flow (M, 1.5 pt, 1.2 bar) 8 kg/s
Endwall boundary layer suction
Total pressure -
0.6 2.5 bar Sillca gel air dryer
Total temperature -
300 325 K
Dimensions 238~152.4mm
Blade span 152.4 mm
Blade aspect ratio (C-80mm) 1.9
Fig. 2.4-5 DLR supersonic cascade wind-tunnel

38
Lower Bypass Suction Flow
Test Section Data Alr supply
Mach number -
0.2 1.4 Closed loop, contlnuous operation
Reynolds number 1 * 1 O S - 1.5*106 Mass flow 2-15 kg/s
Turbulence level -
1% 4% Silica gel air dryer
Totat 'pressure -
0.1 1.8 bar
Total temperature 300 - 320 K
-
Test sectlon helght variable
Blade span
Blade aspect ratlo (C-70mml
150-450 mm
168 mm
2.4
Endwali boundary layer suction
Slotted transontc test sectlon for M -
,
0.9-1.2
Half-symmetrical nozzle for supersonlc flow
Fig. 2.4-6 DLR transonic cascade wind-tunnel
Suclion
~!!-!-
slotted or perforated wall
-I throttle
--9 ---_
tai Iboard
Fig. 2.4-7 Schematic of a transonic cascade wind-tunnel with a tailboard throttle system
39
The latter system was applied for instance in the Detroit Die- by feeding the downstream stagnation pressure below a slot-
sel Allison Supersonic Cascade Wind-Tunnel (Fig. 2.4-8). In ted plate with variable porosity. Only with this tailboard was it
using the tailboard system, the leading tailboard from which possible to achieve an acceptable downstream periodicity at
the downstream characteristics are emanating, should be high back pressure conditions.
porous in order to simullte the uniform far downstream flow
field (Ref. 2.4-4).
Because the back pressure has to be provided along this ONERA uses a very short but flexible tailboard (Fig. 2.4-11)
porous wall, tailboard and throttle may be combined as followed by a free shear layer to adjust the downstream peri-
shown in Figures 2.4-9 and 2.4-10 and developed by DLR in odicity. The free shear layer is thereby providing an
Cologne. Within a hollow tailboard the pressure is increased approximatelyconstant back pressure.
Fig. 2.4-8 Allison supersonic cascade wind-tunnel (Ref. 2.4-3)
Fig 2.4-9 Test section of DLR supersonic cascade wind-tunnel

40
Fig. 2.4-10 Combined tailboard throttle system of DLR
Fig. 2.4-11 ONERA supersonic cascade wind-tunnel
The downstreamflow adjustment is difficultbecause exit flow the back pressure must be increased sufficiently by the throt-
angle and exit velocity are changing simultaneously and throttle system.Consequently, a strong dependency arises between
tle as well as tailboard position have to be set accordingly. inlet flow field, periodicity, and prescribed back pressure.
Increasing the back pressure at supersonic exit velocity for Probably due to this, only very few cascade test results at inlet
instance, increases also the exit flow angle until sonic down- flow angles above the unique incidence were published in the
stream condition is reached. Beyond this point the exit flow past.
angle decreases with back pressure (see also Section 2.4.5.4
and Figs. 2.4-41 and 2.4-43). This behaviour also explains 2.4.4 Sidewall Effects
why it is not possible to use the tailboards as throttle system The linear cascade originally was developed as a two-dimen-
because then only sonic and supersonic exit flow conditions sional model. However, due to the limited dimensions of a
could be achieved and no subsonic ones. cascade wind-tunnel, the two-dimensional flow conditions
can be achieved only approximately. Because of the limited
As long as the cascade is choked and the inlet flow field is blade span the displacement and secondary flow effects
independent of the back pressure, the adjustment of the induced by the interaction of the cascade flow field with the
desired flow conditions is still comparatively easy. Tests at incoming sidewallboundary layers cause a more or less severe
high positive incidences,however, are rather difficult because three-dimensional disturbance.
41
However, this is not always a disadvantage.The adverse pres- sidewall boundary layers. Severe three-dimensional secon-
sure gradients in the sidewall region cause a considerable dary flow effects would disturb the cascade flow field. In these
thickening of the wall boundary layers within the cascade cases it is necessary to use contoured sidewalls as was
blade passages, resulting in a contraction of the midspan performed for example at DDA (Ref. 2.4-8), Fig. 2.4- 14) or at
stream tube and an increase of the axial velocity density value. DLR (Ref. 2.4-9).
This change of the axial velocity density value in tum is a
desired effect. It also happens with the blade element flow of
real turbomachines with converginghub and casing contours,
where it has a considerable innuence on the flow turning, the 1.30 I I I I I I
static pressure increase, and partly on the total pressure loss
development. Therefore, the variation of the axial velocity a adjustment o f AVOR
p 1.25 - by sidewall -
density value has to be simulated in cascade tests as an addi-
tional parameter. -x
.2-
c
1.20 - -
Important, however, is that the corresponding stream sheet m
L
I
I thickness variation in the cascade wind-tunnel can be con- 2 1.15 - -
trolled, at least for the midspan section, where generally the VI
c
W control of AVOR by
measurements are carried out. In most cases this control is U sidewall suction -
performed by a more or less sophisticated sidewall suction 1.10 -
.-U
device. It also should control the secondary flow effects such 0
I as comer stall, shock-wave sidewall boundary layer interac- g 1.05 -

4
-
I tion effects and reduce secondary vortices. In the ideal case a -.-
m
pl-variation with
AVDR=const. is possible
quasi three-dimensional flow field with a continuous stream x
Q. 1.00 - -
sheet thickness variation at midspan position is obtained.
0.95 I I I I I
I
2.4.4.1 Effect ofSidewall Boundary Layer Loading -
AVDR Dependence
The amount of the boundary layer thickening and the result-
ing increase of the axial velocity density value strongly
depends on the blade loading. For subsonic and transonic Limit due to max. achievable endwall
inlet Mach numbers this loading and thereby the loading of boundary Layer displacement
the sidewall boundary layer is primarily a function of the inlet
flow angle. In cascade tests, therefore, an increasinginlet flow Fig. 2.4-12 Possible range of AVDR values of a subsonic/
angle shows a simultaneous increase of the stream sheet con- transonic cascade at midspan depending on the inlet flow
vergence at cascade midspan, which is expressed by the axial angle
I velocity density ratio (AVDR). The shaded region in Figure
2.4-12 shows a typical AVDR dependence on the inlet flow
angle of a given cascade configuration.The level of the maxi- Test range o f supersonic "started"
mum achievableAVDR value depends on the thickness of the cascade flow, MI = const, p1 = const
incoming boundary layer, the blade aspect ratio, the inlet 1.30 I I I I I I
I
1
. tt
Mach number, and particularly on the cascade geometry such
as blade camber,that prescribes the diffusion in the blade pas- adjustment of AVOR
0 sage. Lower AVDR values can only be obtained by the
application of a sidewall suction device. Using an intelligent
$U 1.25 by sidewall
contouring
k. .:,e::$;.
., ~.:..:<::.:. .....
suction system, it becomes possible to keep the AVDR value m
L 1.20
constant while the cascade is tested at different inlet flow
angles.
AVDR values above the shaded region in Figure 2.4-12 can
be achieved only by additional sidewall contractions. This can
be done either by using wedge type inserts (Ref. 2.4-7) or flex-
ible inner sidewalls.
U
h
+
c
al
"'t
1.10
p2/p1 variation with
:I
? 1.05 AVDR = const is possible
In supersoniccascade tegs, when the supersonicflow into the m
blade passage is started, the loading of the cascade blade pas- :: 1.00
sage depends only on the prescribed back pressure. The inlet
flow conditions and the corresponding inlet wave pattern is
not influenced from the rear. In this case the sidewall boun- 0.95
1.2 1.k 1.6 1.8 2.0 2.2 2.4 2.6
I dary layer displacement effect in the blade passage and the
achievable AVDR value is a function of the static pressure Static pressure ratio, p2/p1
ratio as shown by the shaded region in Figure 2.4-13. Again a
sidewall suction system enables tests with a constant AVDR Limit due to max. achievable endwall
parameter but different back pressures, p2. boundary layer displacement
Highly loaded supersonic compressor cascades are often
designed to operate with relatively high AVDR values, which Fig. 2.4-13 Possible range of AVDR values of a started
cannot be generated solely by the displacement effect of the supersonic cascade depending on the static pressure ratio
42
i-'
FLOW E X I T
a--
--
k --.--
BOUNDARY L A Y 3 7
BLEED DUCT
nunE HOW
BOUNDAHY LAYER
BLEED DUCT
SCHLIEREN VIEW
OF INLET WAVE
FORMATION
BLEED SLOTS
i AXIAL FLOW
Fig. 2.4-14 Compressor cascade cross section in axial plane with bleed slots and contoured sidewalls of DDA (Ref. 2.4-8)
2.4.4.2 Sidewall Effects on Blade Passage Flow The first test without suction achieves only a static pressure
Sidewall displacement effects reach a maximum when the ratio of about 1.75 whereas the second one with suction
Mach number level becomes transonic. This is especially true allows a pressure ratio of 2.2. Without suction, the passage
for the blade passage flow of transonic and supersonic cas- shock interacts with the sidewall boundary layer and induces
cades. Especially, cascade configurations with relatively low separation combined with an additional blockage. Due to the
gap chord ratios and a small throat cross section are extremely relatively low back pressure, the flow in the rear blade channel
sensitive to any additional blockage in the sidewall region. In reaccelerates to supersonic velocities and is finally deceler-
supersonic cascades the oblique and normal shock waves ated to a high exit Mach number level by the trailing edge
within the blade passages interact with the sidewall boundary shock waves. In the test case with suction (Fig. 2.4-15b), wall
layer, inducing often boundary layer separations which result boundary layer separation is suppressed and therefore the
in not negligible disturbances of the blade passage flow even passage flow can decelerate to subsonic velocities with a
at midspan position. corresponding high back pressure. The test without suction
(Fig. 2.4-15a) shows a slight disturbance already upstream of
Test results of a supersonic compressor rotor blade cascade the first passage shock impingement point (dashed line in the
with a relatively narrow throat cross section shown in Figure suction surface Mach number distribution) that is probably
2.4-15 shall illustrate this sensitivity on the blade surface related to disturbances which originate in the sidewall boun-
Mach number distribution and the corresponding shock wave dary layer interaction region.
pattern. The first test was performed without any sidewall The above examples show that such highly loaded supersonic
boundary layer treatment and the second one with sidewall cascade tests really have to be operated with sidewall suction
suction in a region where the first passage shock interactswith at least when the thickness of the incoming sidewallboundary
the wall boundary layer. In both tests the cascade has been layer is not negligible. Furthermore, such tests should be per-
throttled in such a way that the first oblique passage shock formed with relatively high blade aspect ratio, so that the
wave meets the blade suction surface at nearly the same sidewall disturbances do not influence the cascade midspan
chordwise position ( d c = 0.65). section (Ref. 2.4-11).
M i i. 1.51 p2/p1 = 1.74
p 1 = 1480 w = 0.101
II
1.8
1.6
1
-
1 t
.e 1.6
c
S
i
Y
0
1.4 - L
gE 1.4
9c
2
-
t.2,- 5 1.2
._ i
Y
1.0 -
e
*
c
Y
s 0.8 -
0.6
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Relative axial chord, t/cax-
Fig. 2.4-15 Blade Mach number distribution and shock wave pat” d e supemk capcade “Mad to nearly Idemid
nrsl passage shock wave position (R&. 2.4-10)
a) lest without sidewall suction
b) lest wkh $dewall suction
2.4.4.3 SidewIIBounda?yL.ajerCowoI slots along the rearpart of the suctionsurface,which are mu-
In order to obtain a specitic distribution of stream-tubecon- nected to a completely indepeadent suction chamber. The
vergence, either the sidewall boundary layer displacement comer slots prevent a spanwise flow of boundary layer fluid
effect is utilized or the cascade sidewalls are contoured. In and suppress the development of secondary vurtices (Refs.
.
both cases the undesired heed"‘onalssondaryflow 2.4-13 and 2-4-12),
effects have to be controlled in some way In the course of the ONERA has good experience with one suction slot in the
years,different sidewall suction systems have been developed entranceregion ofeach blade passage which is aligned nearly
and applied. At firstthese systems were used in subsonic and paralleltotheisobarsofthecascadeflowfidd(ReL2.4-15)as
~sonictestsonly,butveryswnitalsobaameobnousthat shown in Figure 2.4-20. Thistype of suction system has some
supersonic tests had to be performed with special sidewall advantap especially in supersonic d e tests ( F i i 2.4-
suction systems. Zl), because the opening can be placed in a region where the
main passage shock interacts with the sidewall boundary
Various ~~nfisuratioas using either screens,holes and slots
layer.
have been applied to vary the axial velocity density ratio.
Howmr,three-dimensionalssondaryfloweff~causedby The correspondingfluid mechanic mechanism of such shock-
separations and redrntlations cm be reduced only if the wave/waU boundary layer interactions with lateral suction
openings or the porosity are well distniuted within the blade systems is eaphed in detail by Meauze (Ref. 2.4-17).
passage. To avoid recirculation through the suction chamber, Besides the removal of the sidewall boundary layer fluid,
forexample, the suction openines should start aft of the mini- these openings provide an additional positive effect on those
mum pressure region of the passage flow field. In most cases trausonic/supersonic tests which are run with high back pres-
the.openings are thereforelimited to the rear part of the blade sures. Normally,when the back preaou~is inereased and no
passage. Streamwise slotted walls are very effective, which sidewall suction is applied, it is difficult, and sometimes
have beenused for example at DLR and P&W (see Figs. 2.4- impossible, to obtain a periodic caseade flowwithin the blade
16 and 2.4-18). Fortunately, with these systems, Schlieren passages and even u p ” ofthe cascade, because the back
observation is still possible (Fig. 2.4-17) which is extremely pressurcinformationtravelsupstreamthrou~thewaUboun-
important for tests in the tnwsonic and supersonic flow dary layer and influences the rear most blade passages 6rst
regime. The suction slots prevent these unperiodic disturbances
through the boundary layer and the cascade can be throttled
m e most advanced suction systems have additional corner to relatively high static ratios.
44
.-WALL SIAlICS IIYPICALI

ROIYANUE pmw - AIffOIL
LOCAllON nYRCAll 7 ,' , ' ,-MI"
, ,' ,'
- CORPCR SlKllON SLOl
rAIRH)IL
LCORKRSWIION St01
-NON FfRFORAlED
WALL IIVPICALI
Fig. 2.4-16 Sidewall suction slots at DLR for a transonic Fig. 2.4-19 NASA sidewall suction system (Ref. 2.4-14)
cascade
Fig. 2.4-17 Schlieren photograph of a transonic cascade

with sidewall boundary layer auction, M, = 1.0
Fig. 2.4-20 ONERA sidewall suction system (Ref. 24-15)
6 Sldewoll Slots
-
U,
P
(a) ,+:v-q-
I
Corner Slot Parallel

To Suctlon Surface
Fig. 2.4-16 PlLW sidewall suction system with corner slots Fig. 2.4-21 ONERA suewall suction system for supersonic
(Ref. 2.4-12) compressor cascades (Ref. 2.4-16)
45
2.4.5 Experimental Determination of Flow Quantities condition of this method is the observation that surface mea-
2.4.5.1 Inlet Flow Angle of Supercritical and Transonic surements and inviscid predictions agree quite well in the
Cascades front part of the suction surface; Figure 2.4-22 shows as an
At subsonic and supercritical inlet flow velocities, flow angle example the agreement between measured and predicted sur-
probes are successfully used to determine the cascade inlet face Mach number distributions of a supercritical Controlled
flow angle. The inlet periodicity may be checked either by tra- Diffusion Airfoil (CDA) cascade, whereby the inlet flow angle
versing with a single probe or by mounting several probes in was adjusted by probes and the calculation performed by an
the inlet reference plane, where also the inlet velocity is mea- inviscid code. Differences due to viscous effects are observed
sured by sidewall static taps. generally behind 30 to 40%of chord on the suction side and
Above inlet Mach numbers of about 0.8 the application of along the pressure side (caused by displacement effects on the
probes becomes critical and around M, = 1.0 probes would suction side). Therefore, a suction surface position between 5
destroy the inlet flow quality completely. Furthermore, in the to 20%would be best suited for the comparison of static pres-
transonic flow range the inlet flow angle depends on the back sure data resulting from measurements and numerical
pressure and it is therefore possible to achieve different inlet calculations respectively. In Figure 2.4-22, the 10%position is
flow angles at the same geometric position of the cascade in selected with a corresponding suction surface Mach number,
the wind-tunnel. (In transonic compressor blade element denoted M,,.
flow, this unchoked flow condition with subsonic exit flow By varying the inlet flow angle the predicted surface Mach
represents the most common operating condition.) The poss- number distribution changes as shown in Figure 2.4-23. All
ible solution by measuring inlet flow angle and velocity the suction surface curves are crossing nearly at one point.
distributions with a laser system is very time consuming and This is a unique feature of all subsonic and supercritical blade
prohibitive for large test series. In addition, it is nearly impos- sections.
sible to adjust the inlet flow angle in this way to a desired
prescribed value.
Within the last years a new method has been applied success- 1.6I 1
fully which relies partly on theoretical data (Ref. 2.4-18). It is
similar to the method described by Weingold and Behlke --
I
M ss
I I I
SKG-FVV 2.2
1
I
4
(Ref. 2.4-19) for measurements in multistage compressors 1.4 I
and uses the theoreticallypredicted suction surface pressures

to compare those to the experimental surface pressure data.
In fact, it is utilizing the blade as a flow angle probe whereby
2. 1.21
the “probe calibration” is performed by several flow field cal- L
culations. Therefore, the final measurement accuracy also 2 1.0

depends on the quality of the computer code. Another prior 5c
0.8
5
m
.E
1.6t SKG-FVV 2.2

1
.-a 0.6
0
L
2
Q)
0.4
v1
inviscid prediction
M1=0.85 AVDR = 1.15
4 0 0.2 0.4 0.6 0.8

Relative chord, x / c
Fig. 2.4-23 Surface Mach number distributions at different

1.0
inlet flow angles of a Controlled Diffusion Airfoil (Ref. 2.4-20)
Therefore, at these inlet velocities the inlet Mach number may

be determined also from the measured surface Mach number
at this special point. Ahead of this intersection point, the sur-
face Mach number, M,,, changes linearly with the inlet flow
angle. This linear dependency of M,, at the 10%chord posi-
tion is shown in Figure 2.4-24 for different inlet flow Mach
0 numbers. Such a favourable behaviour exists also for the gra-
0 0.2 0.4 0.6 0.8 1.0
dient d&/dM,, of these 0, - M,, curves which varies also
Relative chord, x/c linearly with M, up to a certain value beyond which it
becomes constant (Fig. 2.4-25). The latter effect is due to
Fig. 2.4-22 Measured and predicted surface Mach number “freezing” of the front surface Mach number around sonic
distribution of a Controlled Diffusion Airfoil conditions which is observed also in single airfoil aerodynam-
46
0.7 0.8 0.9 1.0 1.1 1.2 1.3

Suction side Mach number, MSS
Fig. 2.4-24 Inlet flow angle versus surface Mach number of a Controlled Diffusion Airfoil for different inlet Mach numbers
0
I I I I I I I I
0.6 0.7 0.8 0.9
Inlet Mach number, MI
Fig. 2.4-25 Gradient dfl,/dM,, versus inlet Mach number of a Controlled Diffusion Airfoil
ics. It is clearly visible in the surface Mach number Mach number:

distribution of Figure 2.4-26, for inlet Mach numbers beyond
0.87 and even better in Figure 2.4-27 which shows the distrib-
utions of a transonic high gap chord ratio cascade for inlet (2.4.1)
numbers beyond 0.82.
If the suction surface Mach number, M,,, of the supercritical
blade section is plotted as a function of the inlet Mach number whereby M,, ref = f(Ml) (Fig. 2.4-28), and dS,/dM,, = f(Ml)
at a constant inlet flow angle, again a linear dependency is (Fig. 2.4-25).
obtained up to the freezing condition as shown in Figure 2.4- Due to the linear dependencies, only four flow field calcula-
28. By using, for instance, 8, of Figure 2.4-28 as a constant tions are required at two different inlet flow angles and
reference value (together with the corresponding M,, -curve corresponding inlet Mach numbers to establish the “calibra-
as M,, ref) as well as the dependency d&/dM,, of Figure 2.4- tion curves” of Figure 2.4-25 and Figure 2.4-28 below the
25, a simple relationship can be derived for predicting the “freezingpoint” which is attained around M, = 0.85. Beyond
inlet flow angle from measured suction surface and inlet flow the “hezing point” dfll/dM,, and M,, ref remain constant and
47
1.6 . I I I I I 1.6 1 I I 1 I
SKG-FVV 2.2 DLR Propfan 2

1.4 -
1.4
1.2 - .E 1.2
.- v)
x
L- 1.0 - ki
n 1.0
QI
n
5
$ 0.8- =
J=
0.8
J= U
U m
2 0.6- =
.-
U
. 0.6
.-
U a
n 0
E'
c
0.4- 5
c 0.4
c W
W inviscid prediction -
v) p 5 125.6' AVDR = 1.0
I
-
v)
o'2- P1=1L2O AVOR.1.15 0.2
0' I I I I 0 I I I I I I
0 0.2 0.4 0.6 0.8 1.0
Relative chord, x/c
Fig. 2.4-26 Surface Mach number distributions at different Fig. 2.4-27 Surface Mach number distributions of a
inlet Mach numbers of a Controlled Diffusion Airfoil (Ref. Propfan blade section (Ref. 2.4-21)
2.4-20)
1.41 1
1-3t
I I I I I I 1
wl
SKG-FVV 2.2
wl
x 1.2 -
L
1.1 -
QI
n
5c
J= 1.0 - -
U
m
x 0.9 -
QI
.-
U
VI
c 0.8 -
.-c
0
U
3
0.7- o inviscid prediction -
cn
I I I I I I I
Fig. 2.4-28 Suction side Mach number versus inlet Mach number of a Controlled Diffusion Airfoil
two additional calculations are necessary at different inlet

flow angles to derive the constant values for d&/dM,, and
that between about M, = 0.975and M, 1.04 the velocity is
constant along the front suction surface (frozen condition)
-
M,, ref in Equation (2.4.1). and is increasing again at higher inlet Mach numbers. If the
The dependency of the suction surface Mach number on the 8.9% chord position is considered with the respective surface
inlet flow angle at low supersonic inlet flow conditions is Mach number, M,,, the diagrams of Figures 2.4-31,2.4-32
explained with the transonic compressor blade section and 2.4-33can be derived, which show the relations between
LO30-4which has a design inlet Mach number of M, = 1.085. MI, M,,, &, and d&/dM,,. It is quite evident that near linear
In Figures 2.4-29and 2.4-30 the variation of the surface dependencies also exist at supersonic inlet velocities. Thereby
Mach number distribution with inlet flow angle and inlet the gradient d$,/dM,, remains constant about M, > 0.97
Mach number is shown. It can be derived from Figure 2.4-30 (Fig. 2.4-33),whereas the suction surface Mach number, M,,,
1.6 I I I I I
LO30-4 LO30-4
1.2 - e- 1.2
v)
1.4 -
-
-\-
-1.16
-/
--
1.07
1.04
M1
i
.-
v)
L
E al
- 1.0
L 1.0 LI 7
n
01
5c 0.89 I
5c 0.8 '-
.c
U
m
E 0.6 ~-
.-a
U
0
L
c
c
W
inviscid prediction
-
v1
p 1 = 150' AVOR 1.2
I 1 I I I 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 0
Relative chord, x/c Relative chord, x / c
Fig. 2.4-29 Variation of the surface Mach number Fig. 2.4-30 Variation of the surface Mach number
distribution with inlet flow angle of an MCA blade section distribution with inlet Mach number of an MCA blade section
I I I I I I
153'
x/c=0.089 LO30-4
0.8 0.9 1.0 1.1 1.2 1.3 1.4

Suction side Mach number, MSS
Fig. 2.4-31 Inlet flow angle versus surface Mach number of,an MCA blade section
increases again above a certain inlet Mach number which but rapid prediction of the inlet flow angle is possible using
depends on the inlet flow angle (Fig. 2.4-32). In order to esta- again Equation (2.4.1). In applying Equation (2.4.1) the
blish the supersonic dependencies, therefore, only two values of M,, ref and d&/dM,, are either linear dependent on
additional flow field calculations are necessary at constant M, or constant depending - on the velocity- range:
-
inlet flow angle. In Figure 2.4-32 experimentally derived data
subsonic range: M,, ref = f(ml);dp,/dM,, = f w , )
at choked condition as well as the predicted unique incidence LLfTozen,,
range: M,, ref = const.; dp,/dM,, = const.
curve are included as lower boundaries. supersonic range: M,, ref = f(Ml); d&/dM,, = const.
With the aid of one curve sf Figure 2.4-32 (determination of The method is very useful because it offers the possibility to
M,, at a fixed 8, (= fJ1 ref) and the dependency shown in continuouslycalculate and observe the inlet flow angle during
Figure 2.4-33 (determination of d&/dM,,) an approximate cascade tests, by measuring only the inlet and one suction sur-
49 I
:1.5
=-
k 1.4
n
E
1.3
r
:1.2
II
= 1.1
v)
5 1.0
.-
c
Y,
0.9
0.8
0.7 0.8 0.9 1.0 1.1 1.2 13
Inlet Mach number,M
Fig. 2.4-32 Suction side Mach number versus inlet Mach number of an MCA blade section
50
x/c=0.089
LO30-4
I
E
U
v)
v)
$30
c
c
4c
\&
.-
U
W
U
E 20
10
0.7 0.8 0.9 1.0 1.1 1.2 1.3
Inlet Mach number, MI
Fig. 2.4-33 Gradient dfJ,/dM,, versus inlet Mach number of an MCA blade section
face Mach number and by specifying the relevant velocity system and probably also within the accuracy of the
range. The accuracy of this method depends, of course, on the periodicity adjustment of the experimental flow field. A fur-
accuracy of the computer code used to establish the calibra- ther good agreement was obtained between the experimental
tion curves, but also on the linear approximations and on the inlet flow angles of the DLR-L030-4 cascade (AGARD Test
determination of MI. In order to check this method a test case E/CA-4, Ref. 2.4-22), determined by laser measure-
series has been performed with a high gap chord ratio cascade ments and the above method.
at an inlet Mach number of MI = 0.9 (Ref. 2.4-21). Laser
measurements (L2F) just upstream of the leading edge plane In order to proof the general validity of the method, ten com-
at three different incidences resulted in inlet flow angle differ- pressor cascades have been checked theoretically at various
ences between 0" and -0.5" compared to the above described geometric as well as flow conditions. All cascades confirmed
method. These differencesare within the accuracy of the laser the near linear dependencies described above.
50
2.4.5.2 Supersonic Cascade Inlet Flow angle or equivalently maximum mass flow at a given inlet
A cascade with supersonic inlet velocities can operate in four Mach number. Figure 2.4-35 provides a Schlieren photo-
essentially different modes (see Fig. 2.4-34): graph of such a cascade. The mass flow or the inlet flow angle
1. unstarted cascade flow depends on the cascade entrance geometry (blade stagger
2. unstarted and choked cascade flow angle, leading edge thickness and suction surface camber dis-
3. started cascade flow where the axial component of the inlet tribution). Between inlet Mach number and inlet flow angle a
velocity is subsonic fixed relationship exists, which is often referred to as "unique
4. started cascade flow where the axial component of the inlet incidence" and discussed in several references (Refs. 2.4-23
velocity is supersonic. to 2.4-27). In a wind-tunnel the cascade inlet flow conditions
(M, and &) must not necessarily be identical to the nozzle
exit conditions. Using for example fixed nozzle blocks which
are designed for a particular exit Mach number, the cascade
can be tested even at slightly higher inlet Mach numbers by
simply varying the cascade setting angle in relation to the noz-
zle axis. In this case the flow expands around the foremost
blade to slightly higher inlet Mach numbers while the follow-
unstarted ing cascade establishesautomaticallya periodic inlet flow that
p, * f(P, 1 meets the unique incidence condition. The experimental
determination of the corresponding inlet flow quantities is
relatively easy and described below.
unstarted and choked

._
started, Mlax< 1.0

'unique incidence"
Fig. 2.4-35 Schlieren photograph of a compressor rotor

blade cascade with started supersonic flow (M, = 1.18, p2/p1
Fig. 2.4-34 Supersonic inlet flow configurations = 1.45)
The first unstarted mode is obtained, when the cascade oper- At axial supersonic inlet flow condition - which represents
ates with relatively high back pressure. The blade passages do the fourth mode - the wave pattern (shock and expansion
not swallow the mass flow they could at lower back pressures, waves), emanating from the front portion of the blades,
but operate with a certain spillage by which the stagnation remains downstream of the cascade entrance plane. The cas-
streamline is shifted and the inlet flow angle is increased. cade does not transmit any information into the upstream
Under these conditions, a normal shock wave stands ahead of region. Therefore, cascade inlet and nozzle flow are identical
the blade passage entrance. The prescribed back pressure is and no problems do exist to determine the cascade inlet para-
responsible for the "stand off distance of this normal shock meters:
and the corresponding inlet flow angle. MI =M,and 5 &
The second mode occurs when the throat cross section
between the blades is small and the inlet velocity too low to
allow the supersonicflow being started in the passage. Again a Experimental Supersonic Inlet Flow Conditions
normal shock stands ahead of the passage, but its position Unstarted Cascade Flow
now is independent of the back pressure. Only the inlet Mach Besides the transonic inlet flow, the supersonicunstarted cas-
number and the cross section throat area are responsible for cade flow is the most difficult one to establish. Even if the test
the inlet flow angle (i.e. mass flow) and the stand off distance engineer has been successful in obtaining a periodic flow pat-
of the detached normal shock wave. The corresponding inlet tern, the experimental determination of the inlet flow
flow angle represents a minimum flow angle (maximum mass quantities such as inlet Mach number and inlet flow angle is
flow) for the respective inlet Mach number. rather difficult. Again, as in the transonic flow regime, probes
The third started mode with an axial subsonic inlet compo- would disturb the flow pattern and provide unreliable data.
nent also stands for an operation with minimum inlet flow Also, the use of sidewall static pressures obtained from an
51
upstream plane within the periodic wave pattern, is rather Tangential momentum equation
Irl
problematic. Because of the existing sidewall boundary layer, --
K - 1
the static pressure can provide only smeared out data of the S + t
K - 1
real wave pattern and especially at highly loaded flow cases 1, = -
t
M?rl[1 + 7 MfS] * sin PI, cos PI, dll
with increased back pressure and strong shock waves, the
sidewall data often do not represent the conditions of the mid- (2.4.5)
span flow field. This is true especially for the supersonic
unstarted flow cases, whereby the cascade operates with a
Constants defined using I,, I,, and I,
certain "spillage" and strong detached bow shock waves. Like-
wise, the inlet flow angle determination is difficult, because
the direction of the wind-tunnel nozzle exit flow must not
necessarily be identical to the inlet flow direction of the fol-
lowing cascade.
For these unstarted flow cases and for the sensitive transonic (2.4.6)
flow, the nonintrusive laser anemometry is the best suited
technique to provide accurate inlet flow data. The laser anem-
ometer can provide flow velocity and flow angle distributions
in the midspan region of a suitable measurement plane in
front of the cascade (see for instance Fig. 2.4-36). With the
assumption of constant total temperature in the entire flow
B = 1-2K
K + 1
field &e local inlet Mach numbers'can be calculated:
0 - 1
(2.4.7)
The final expressions for the integrated inlet Mach number

2 (2.4.2) and flow angle are
The hypothetical uniform inlet flow quantities far upstream of

the cascade are obtained by integrating the measured (peri-
odic) laser data over a single blade pitch. Under the
assumption of two-dimensional flow, several different averag- (2.4.8)
ing methods can be applied for these data, for example mass
averaging or area averaging (Ref. 2.4-5). The method of inte-
gration, that is derived from the conservation laws of fluid
motion, which also allows total pressure variations between
the uniform flow field infinitely far upstream of the cascade
and the measurement plane, uses the followingequations: (2.4.9)
Continuity equation
In Equation (2.4.7), which provides the flow velocity, the sign
of the inner square root must be negative, if the axial compo-
An + I
nent of the inlet flow velocity is subsonic. A positive sign is
necessary for an axial supersonic component upstream of the
Started Cascade Flow - Unique Incidence

J S
When the supersonic flow into the cascade is started and its
(2.4.3) axial velocity component is subsonic, the wave pattern in the
entrance plane is only a function of the inlet Mach number
and the geometry of the cascade entrance. Between the inlet
Axial momentum equation Mach number and the inlet flow angle there exists a unique
relationship - often called "unique incidence". Any variation
of the exit flow conditions (for example changing back pres-
sures) does not influence the inlet flow properties.
For a cascade with a limited number of blades, that is
--K K- 1 mounted into a wind-tunnel test section, this unique incidence
K - 1 relation is also valid. This is shown and discussed in more
I2 = +-
2 detail for example in (Ref. 2.4-26).
Testing a cascade at a particular nozzle Mach number, the cas-
cade inlet flow angle & should agree with the nozzle flow
direction &, so that nozzle exit conditions and cascade inlet
flow conditions are identical. This however involves the
assumptions that the flow in the entire nozzle exit region and
throughout the wave pattern upstream of the cascade is two-
dimensional, approximately irrotational, and isentropic. As
(2.4.4) the supersonic blades of the cascade usually have relatively
52
156" 1
1.8 c
Il . 7 -
(Lmple-vaue
theory
-0.5 0.0 0.5 1.0 1.5

Tangential location, rl/t
Fig. 2.4-36 Flow angle and Mach number distribution in the entrance region at supersonic inlet flow conditions. Laser-2-
Focus data 2.7 percent of gap ahead of the leading edge plane (Ref. 2.4-6)
thin leading edges and low front camber, the total pressure incidence condition of the cascade. Thereby the cascade
losses of the inlet wave pattern are very small and the above automatically establishes a periodic wave pattern down-
assumptions can be considered to be a good approximation. stream of the first neutral characteristic. A high cascade
Figure 2.4-37 shows a sketch of a test section with nozzle exit setting angle (PN) for example results in an acceleration
flow quantities MN and & and the cascade inlet wave pattern around the leading edge of the first blade and the cascade inlet
with oblique shock waves and the so-called neutral charac- Mach number increases in relation to the nozzle Mach num-
teristics, emanatingfrom the front portion of each blade. Flow ber. This process of expansion around the first blade is
velocity and flow direction along these neutral characteristics sketched in Figure 2.4-38, where the cascade setting angle &
represent the far upstream .cascade inlet flow (Ref. 2.4-26). is greater than the "unique incidence" P1. Under the assump-
Nozzle exit flow and cascade inlet flow, however, are identical tion of irrotational and isentropic flow, the expansion follows
only if MNand & agree with the unique relation of M, and p,. the dashed-dotted line which is calculated from the Prandtl-
Other cascade setting angles (&) would result in a change Meyer relation
between the nozzle exit Mach number and the cascade inlet P I 1" =&+vN (2.4.10)
Mach number. At the leading edge and the front portion of the
first blade the flow is turned and either decelerated or acceler- where VI= v(M1) and V, = v(MN).
ated into the flow direction that corresponds to the unique The parameter vis the Prandtl-Meyer angle:
53
Inlet flow quantities of a supersonic cascade test series using I

the above method are plotted in Figure 2.4-39. It is to be seen
I
that most of the data accumulate along one curve, which rep-
resents the unique incidence condition. These correct data,
1
- arctan(M2 - 1 ) T however, belong to cascade tests with low or moderate back
(2.4.11) pressure only. Tests with high back pressure and thus highly
loaded sidewall boundary layers show slightly reduced inlet
Mach numbers and too high inlet flow angles. At high back
pressures the sidewall boundary layers allow downstream
In the case of a cascade setting angle & that is too low in rela- pressure information to move upstream influencing the pres-
tion to the unique incidence - B,, an oblique shock wave sures of the inlet static taps. Relatively small errors in the
would be generated by the first blade that turns and deceler- “sidewall” inlet Mach number, however, can result in large
ates the nozzle flow to the necessary periodic cascade inlet errors in the calculated flow angle.
flow condition. This shock, however, would produce an
entropy increase too high for the assumption of irrotational The second method avoids the inaccuracy of the sidewall
and isentropic flow. Therefore preferably, cascade tests static measurement: Thereby only the nozzle exit Mach num-
should be run with cascade setting angles that are close to or ber and the cascade setting angle & is used to calculate M,
slightly higher than the unique incidence. and 6,. This, however, requires the knowledge of the rela-
tionship between cascade inlet Mach number and inlet flow
For the experimental determination of the inlet flow quantit- angle (unique incidence) a priori for the particular cascade
ies M, and B1 in a wind-tunnel, two relatively simple methods being tested. This M,-& relationship can be obtained either
shall be introduced here, although several other methods do from some experiments with relative low back pressure using
exist. These methods use either static pressures from the cas-
@ cade sidewall entrance region (Refs. 2.4-25 and 2.4-10) or
the above “sidewall” technique, or from theoretical calcula-
tions for the interesting Mach number range. Such unique
flow directions and surface Mach numbers from the front incidence calculations also should include the leading edge
portion of the blades (Ref. 2.4-28) to calculate the hypotheti- blockage effect and the shock losses of the entrance wave pat-
cal uniform inlet flow conditions far upstream of the cascade. tern (Ref. 2.4-29).
The first method uses averaged sidewall static pressures For the experimental determination, again, a two-dimen-
obtained from a measurement plane in the “periodic” flow sional, irrotational, and isentropic flow is assumed for the
field upstream of the cascade and the total pressure in the entire entrance flow field. The equations used are the Prandtl-
settling chamber to calculate the cascade inlet Mach nhmber Meyer relation and the unique incidence relation:
M, = MlwWAssuming again a two-dimensional, irrotational,
and isentropicflow in the nozzle exit and cascade inlet region,
the cascade inlet flow angle can be calculated by means of the Bi + v(Ml) = & + v(MN) (2.4.13)
Prandtl-Meyer relation
B1= f(Md (2.4.14)
neutml
characteristics representing cascade mlel flow MI , /3,
exit flow
Y
Fig. 2.4-37 Nozzle exit flow and cascade inlet flow with periodic wave pattern behind neutral characteristic of first blade
54
160a I I I I I I
\,p+v(M)=const.
c t
aJ
d 159'1
159 \. m
c1
a- \ Exoansion around
d
m
c
3
158'
unique incidence
-
d
d
0 c
LL
157'1
15T 1.3 1.4 1.5 1.6 1.7
I I I I I I
Inlet Mach number, M,
1.3 1.4 1.5 1.6
Fig. 2.4-40 Experimental cascade inlet flow angle and
Mach number,M Mach number determined from nozzle exit flow quantities
and unique incidence relation
Fig. 2.4-38 Flow direction and Mach number at nozzle exit
( pN,MN)and cascade inlet plane (p,, M,)
does not feel much of a thickness variation that usually starts
downstream of the cascade inlet plane.
A check of the above experimentally determined inlet flow
quantities can be performed by comparing the experimental
160" I I I I I I I I I static pressure distribution on the front part of the blade suc-
:o incorrect data tion surface with theoretical pressure distribution obtained by
- m \ when back- a two-dimensional or quasi-three-dimensional characteristic
1 .' pressure is method or Euler solver. This is sufficient, because three-
6
aJ
159'
\
\
dimensional and viscous effects in the front part of the blade
d
m midspan region are relatively weak.
c
m
2.4.5.3 Exit Flow Quantities
z
d
U-
158" The most convenient and fast technique to measure the exit
flow is to traverse a combination probe, which takes the static
c
Q) \ I ' 't ests with low and pressure, the total pressure and the exit flow direction from a
-c
d
particular measurement plane at mid-span position. The non-

uniform data in the measurement plane can be reduced by the
157'
application of different averaging methods, a comparison of
which is briefly discussed in (Ref. 2.4-5). The most consistent
1.3 1.4 1.5 1.6 1.7 and widely used method that provides the so called mixed-out
values far downstream, uses the conservation equations of
Inlet Mach number,Mlwa,, fluid motion. The method involves the assumption that the
flow is two-dimensional between the exit measurement plane
Fig. 2.4-39 Cascade inlet flow angle and inlet Mach and the hypothetical parallel plane far downstreamof the cas-
number determined from nozzle exit flow and inlet sidewall cade. The equations used are similar to the Equations (2.4.2)
pressures and (2.4.9) in Section 2.4.5.2, and provide the exit velocity,
exit flow angle, density, static pressure, and total pressure.
By using the equation which provides the exit Mach number
(Equations (2.4.7) and (2.4.8)) or the exit velocity, it is
The combination of these two equations can be solved iterat- important to point to the sign of the inner square root in Equa-
ively to find the cascade inlet flow quantities M, and PI. As tion (2.4.7). The sign must be negative, when the axial
Figure 2.4-40 shows, the data now are forced to lie on the component of the exit flow velocity is subsonic (M2= < 1.0)
unique incidence curve, which should be correct for the tests and it must be positive, when the axial component is super-
with started supersonic flow. sonic (M2= > 1.0).Figure 2.4-41-provides an instruction to
The effect of a stream sheet thickness variation on the inlet use the correct sign either for the calculation of the inlet or
flow angle can be introduced approximately in the unique exit flow quantities respectively.
incidence calculation (Refs. 2.4-30 and 2.4-31). But the influ- In cascade tests with transonic and supersonic exit velocities,
ence is relatively small, because the front portion of the the probe measurements become difficult and can lead to
cascade entrance, which is responsible for the M,-&relation, considerable errors. The blockage effect of the probe and the
-+
55
I
~~
signlequ.7 Max, inlet flow outlet flow
negative <I
(-1
<ap 1
positive r a r e f a c t i o n shock
(+I
E l non -existing flow case
rarefaction shock
negative
non-existing flow c a s e
(-1
posilive
(*I
Fig. 2.4-41 Instruction for the choice of the correct square root sign in the data reduction equations for inlet and outlet flow
(Ref. 2.4-5)
probe stem reach a maximum in the transonic range and can trailing edge plane can be assumed as a reasonable approxi-
cause significantmodificationsof the flow field and unperiod- mation for corresponding blade to blade calculations.
icities during probe traversing. Laser anemometry can
overcome this problem, however, pitot-probe measurements 2.4.5.4 Effect of Back Pressure and AVDR on Exit Flow of
still have to be performed to determine the exit total pressure, Supersonic Started Cascades
because the laser velocimeter only provides velocity and flow The fundamental characteristic of a supersonic cascade that
operates with supersonic started flow is that the pressure
0 Eaxiale ? zvelocity
e t and exit flow quantities have been determined, the
density ratio (AVDR) can be calculated from
behind the cascade can be varied independently of the inlet
flow quantities. Such a started supersonic cascade can be
the continuity equation interpreted as a choked cascade whose exit flow quantities
behave similarlyto the respective quantities of a transonic tur-
bine cascade that operates with a choked throat cross section.
In such cascades the downstream values depend on the
prescribed back pressure only (for two-dimensional condi-
(2.4.15) tions) and exit flow angle and total pressure losses are
dependent parameters.
To illustrate this dependency Figure 2.4-42 shows the exit
Unfortunately this simple one-dimensional relation does not flow angle and the corresponding total pressure losses of a
provide any information on the actual flow field and the simple flat plate cascade that is throttled to different exit
stream sheet thickness distribution throughout the blade pas- Mach numbers. The sketches in the figure qualitativelyillustr-
sage. In fact, the flow field is more or less three-dimensional ate the corresponding shock wave patterns at the rear of the
and only appropriate sidewall contouring and sidewall boun- blade passage for some characteristic exit flow conditions.
dary layer suction can provide a reasonable quasi Notice, that in throttling the cascade from low back pressures
three-dimensional stream tube thickness distribution at mid- (supersonic exit Mach numbers) up to higher back pressures
span position. Furthermore, the AVDR value must be quoted with subsonic exit Mach numbers, the exit flow angle reaches
for well defined cascade inlet and exit planes (Refs. 2.4-32 a maximum. This maximum is achieved, when the exit Mach
and 2.4-33), because a varying stream sheet thickness outside number is near unity. (A more detailed discussion of this
of the cascade also changes the inlet or exit flow vectors. The throttle behaviour can be found in (Ref. 2.4-26)).
experience showed, that for most cascade tests, a linear Passing to realistic cascades, this fundamental flow behaviour
stream sheet thickness distribution between leading edge and remains. However, in real blade element flows and in cascade
56
155' I
pT h r=o150'
ttle
I
ML =1.4
Curve
I
Flat P l a t e Cascade
I I I I
-
I
ML (66
1 I
1
I32
e
N 1I 3
d
a-
cs)
c t 0.0 0
.
L
U
c
aJ
.-
.-
U-
Y-
a
0.06 0
1 U
VI
VI
0.04 0
145"- 0.02
I 1 I I I I
0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Exit Mach number, M,
Fig. 2.4-42 Throttle curves of a two-dimensional flat-plate cascade showing the variation of exit flow angle and loss
coefficient
tests the axial velocity density ratio is an additional parameter blade loading are varied, a slight scatter of the data cannot be
that has a strong effect on the exit flow properties and espe- avoided. The calculated curves from the correlation men-
cially on the exit flow angle. For a correct simulation of the tioned above, help to sort out the data and allow a
back pressure- and AVDR-influence,it is important to know considerable reduction of test points and testing time for the
their relative effect on the exit flow. Therefore, a correlation examination of the back pressure and AVDR influence.
was derived from test data of a supersonic compressor cas-
cade that operates with started supersonic flow and a constant 2.4.6 Uncertainty of Test Results
inlet Mach number (Ref. 2.4-6). Design engineersand theoreticians, who use the experimental
The curves in Figure 2.4-43 and Figure 2.4-44 show the varia- results for validating their numerical blade to blade codes,
tions of the exit Mach number and the exit flow angle for should know something about the reliability and accuracy of
different static pressure ratios and the AVDR as additional the experimental data. All data obtained from a wind-tunnel
parameter. The exit Mach number shows a strong but near lin- facility of course are to some extent subject to uncertainties.
ear dependency on the back pressure, whereas the AVDR has There are two main sources:
only a small effect. The exit flow angle dependency on the a) the limited measurement accuracy that is caused by the
static pressure ratio in Figure 2.4-44 is similarly qualitative to type of instrumentation such as probes, laser-anemome-
the B2-M2dependence in Figure 2.4-42. However, the p2-P2/ ters, transducers, calibrations etc.
PI curves of Figure 2.4-44 are shifted to considerably lower b) the uncertainty that is caused by the non perfect modelling
&-values @@er flow turning), when the AVDR is increased. technique, the unknown or non-perfect boundary condi-
Another plot in Figure 2.4-45 shows the strong effect that the tions of the cascade flow.
AVDR has on the exit flowangle. These curves are only a dif- The uncertainties resulting from the first error source can be
ferent presentation of those shown in Figure 2.4-44 and are evaluated relatively correctly from the accuracy of the instru-
each plotted for constant static pressure ratios. In cascade mentation, presupposed the experimenter uses his
tests, when the blade passage and exit flow have to be adjusted instrumentation really carefully.
simultaneously by endwall boundary layer suction, throttle The uncertainties caused by the modelling technique are
and tailboard setting, this strong &-AVDR dependency has to much more complex to be determined and it is nearly imposs-
be considered. ible to define any specific quantitativevalue. For example,the
A comparison of the calculated p2 dependency on the static flow field of a cascade is in reality not two-dimensional even
pressure ratio and AVDR with the corresponding test data is though the AVDR value may be determined to be unity. There
shown in Figure 2.4-46. Because in practice it is difficult to are many factors which provide uncertainties to the reduced
keep the AVDR value constant while the back pressure and boundary conditions of the cascade:
57 I
1.4 ARL-SL19
154' I 1
1.3
152'
N
E 1.2
i
01
150'
n c
5 1.1
-p
W
c 148'
r
v
m
1.0
E
*
.-
4
h
146'
X
w .9 L
lz I440
.8
142'
.7 h I
1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6
0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30
Static pressure ratio, pz/pl
Axial velocity density ratio, AVOR
Fig. 2.4-43 Influence of static pressure ratio on exit Mach Fig. 2.4-45 Influence of AVDR on exit flow angle (see Fig.
number at constant inlet conditions; semi-empirical 2.4-44)
correlation (Ref. 2.4-6)
ARL-SL 19
154' I Mtz1.61 Pc
. . ~147.8'
154' I I I I I 1
AVDR
" I' A \ I
t
1
140' I ' 'I
lk2'
140' I I I I I I 1
1I
1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6
Static pressure ratio, p2/p1 Static pressure ratio, p2/p1
Fig. 2.4-44 Influenceof static pressure ratio on exit flow Fig. 2.4-46 Exit flow angle data depending on static
angle of a supersonic compressor cascade at constant inlet pressure ratio and AVDR compared to a semi-empirical
conditions (M, = 1.6); semi-empirical correlation (Ref. 2.4:6) correlation (Ref. 2.4-10)
nozzle exit flow quality Most of the above mentioned factors are strongly interde-
condensation effects if air humidity is too high pendent, making it difficult to estimate their relative effect on
0 cascade geometry
the cascade flow field and the upstream and downstream
0 number of blades
boundary conditions. In addition the effect can depend
0 flow periodicity
strongly on the operatingpoint. It is obvious,for example, that
0 effect of tailboard and throttle setting on exit flow (in the
tests with a maximum blade loading (very high incidence or
transonic range both also influence the upstream flow field) maximum static pressure ratio with spillage) are associated
blade aspect ratio with the highest uncertainties.
0 sidewall boundary layer effects and induced secondary To provide the user of experimental cascade data with some
flow feeling for the order of magnitude of uncertainties, two test
0 effect of sidewall suction system points have been examined: Tables 2.4-1 and 2.4-2 show the
0 probe blockage effects. estimated and calculated uncertainties of a subsonic and a
58
Table 2.4-1
Transonic compressor cascade at optimal incidence (M, = 0.84, flow turning 16 deg)
Measured
input quantities Estimated uncertainties
(20 : 1 odds)
Inlet Mach number M1 = 0.843 0.0&5

Inlet flow angle = 145.5' 0.5,
Exit flow angle
Total temperature Fi =
Ttl =
129.6O
304.7K
0.5
1.OK
Reference pressure
Inlet total pressure
Exit static pressure
g:r
p2
- 0.992bar
0.953bar
= 0.720bar
0.4mb (0.01%)
1.3mb (0.14%)
2.7mb (0.4%)
Total pressure drop ptl - pt2 = 0 - 0.093bar 0.4mb
Calculated uncertainties (20 : 1 odds) of exit flow

and performance parameters
Exit Mach number M2 = 0.647 20.0048
Static pressure ratio = 1.191 +0.019
Total pressure ratio F2/7b - 0.9906 EO. 0004
Loss coefficient t2 t1 0.0253 20.0813
Exit flow angle
Axial velocity density ratio
f
A4DR
= 129.6
= 1.208
20.5
20.019
Table 2.4-2
Supersonic compressor cascade with relatively high total pressure losses (M, = 1.36, flow turning 2.6 deg)
Measured Estimated uncertainties

input quantities (20 : 1 odds)
Inlet Mach number = 1.3g 0.025

Inlet flow angle
Exit flow angle Ftp2 = 157.9
= 155.3'
0.5
l.oo
Total temperature Ttl = 315.8K 1.OK
- 1.03bar
Reference pressure
Inlet total pressure
Exit static pressure
Ft:r
p2
1.24bar
= 0.63bar
0.4mb (0.01%)
2.5mb (0.2%)
6.3mb (1%)
Total pressure drop ptl -
pt2 = 0.06 - 0.28bar 2.5mb (1.5%)
Calculated uncertainties (20 : 1 odds) of exit flow

and performance parameters
Exit Mach number = 0.921 -to. 0096
Static pressure ratio -t0.0549
Total pressure ratio 0.864 -to. 0020
Loss coefficient -to.0846
Exit flow angle -+1.0
Axial velocity density ratio -+O ,047
59
supersonic cascade test respectively.The uncertainties shown The uncertainty evaluations showed that, for example, the
in the upper part of the tables are estimated from the directly total pressure ratio, ptZ/ptl,and the loss coefficient, y can be
measured quantities. All of these estimated uncertainties determined relatively correctly from wake measurements.
should be considered as “engineering”approximationsbased The calculated static pressure ratio is mainly sensitive to devi-
on experimental observations and operational knowledge of ations of the inlet Mach number. Rather important are the
the measurement system. effects of the inlet and exit flow angles because they strongly
The above estimated uncertainties have been used as input influence the blade surface pressure distribution, the blade
data to a data reduction program. The uncertainties asso- loading and therefore the total pressure losses and the forma-
ciated with the corresponding output quantities were tion of the shock waves within the flow field. Thereby, the
calculated using the uncertainty propagation methods of most severe effects come from errors in pl, especially for the
I
Kline and McKlintock (Ref. 2.4-34),where the second-power unstarted and unchoked flow conditions. For choked and
equation was solved numerically using a so-called “Jitter- supersonic started flow errors in the determined back pres-
, Program” a’s discussed by Moffat (Ref. 2.4-35). sure result in incorrect passage- and rear-shock wave
positions.
The second test with the supersonic inlet Mach number has
been used to show additionally the relative effect of varying The accuracy of the reduced AVDR value again strongly
input quantity uncertainties on some exit flow and perform- depends on the accuracy of the inlet or exit flow angles, espe-
ance parameters. In Figure 2.4-47the uncertainties of the cially when the flow directions deviate from axial direction.
output quantities (exit flow and performance parameters) are The AVDR-sensitivityis particularly high, if the flow angles
calculated for each individualinput uncertainty while the rest are between 110 and 160 degrees (or 20-70 degrees with
, of the input uncertainties have been assumed to be zero. respect to the axial direction).
e
A 1 AVDR = 105 * 0.06 1
AVDR = 1.05
:: I/,
0
0.02
0 ’
-
I I I I 1
0 0.02 0.04 0.08 0 1’ 2’

AB2
AM1
0.012 I- reference test data :

M = 1.36 M2 ~0.92
a 1 = 157.9’ a 2 = 155.3’
p2/p1 = 1.514
AVDR = 1.05
P t ? / P ti = 0.861,
w = 0.203
0 0.02 0.04 0.08 p t i ~ 1 . 2 4bar T i = 315.8 K
AM1
0.12
0.1 0
0.006 k.= 0.864
Ap2
PI 0.08
0.04 0.002 p tl - p ,10.17 bar

0 0
0 0.02 0.04 0.08
AM1
Fig. 2.4-47 Influence of measurement uncertaintieson the uncertaintiesof reduced cascade performance parameters of a
supersonic compressor cascade test
60
2.5 TESTING TECHNIQUESFOR STRAIGHT The boundary layers developing on the top and bottom wails
TRANSONIC AND SUPERSONICCASCADES of this nozzle are captured upstream of the cascade by traps
Hoorelbeke, Gaillard and Losfeld which communicate with cavities B and C. The pressure in
2.5.1 The SSCh Wind-Tunnel these cavities is controlled by valves. The flow rate through
The S5Ch wind-tunnel installed in the Meudon test center of the traps is controlled to adjust the flow upstream of the cas-
ONERA is a continuous closed circuit facility. Generally used cade.
for transonic and supersonic testing, it was chosen to accom- A cascade incidence between 20 and 56 degrees is obtained
modate the cascade testing equipment because its flow rate, by pivoting the nozzle around the leading edge of the two end
approximately 5 kg/s, is well suited to this type of experiment. profiles to comply with the geometry of the boundary layer
Its high compression ratio allows operation of the adjustment traps. This pivoting can be camed out during operation of the
apparatus specific to performing cascade wind-tunnels. wind-tunnel.
The wind-tunnel illustrated in Figure 2.5-1 is equipped with a The cascade includes two separate channels located at the top
12-stage axial compressor driven by a 1.4 M W motor unit. and bottom. The flow rate through these channels is con-
The stagnation pressure, between 0.4 and 1 bar, is held at the trolled by registers forming a sonic throat.
required value by an auxiliary pumping unit. The maximum
compression ratio of the compressor is 5.6. This value allows The periodicity in the center part of the cascade is obtained by
starting of a Mach 3.15 nozzle. A heat exchanger installed simultaneously adjusting the flow rates in these end channels
downstream of the compressor holds the stagnation tempera- and in the boundary layer traps. These local actions allow the
ture at approximately 300 K. To avoid condensation magnitude and direction of the flow velocity to be adjusted at
phenomena in the test section, the relative humidity is con- the ends of the cascade. The upstream periodicity is satisfac-
trolled according to the Mach number by injecting dry air into tory when the boundary conditions are similar to those
the system. measured in the center of the cascade. The experimenter uses
the pressure distribution measured on the lateral walls to
The wind-tunnel includes two branch circuits allowing altem- make these adjustments.The test Mach number is set by two
ate use of the two experimental chambers. These mobile registers located downstream of the walls delimiting
experimental chambers, with access doors, are designed to the useful part of the cascade.
accommodate test fixtures prepared outside the wind-tunnel
as well as their specific control and instrumentation systems. The periodicity downstream of the cascade is obtained by
adjusting the flow direction on the ends of the cascade by
The operational control of the wind-tunnel, supervision of the deformable walls (flexible plates). Because of the low pres-
safety devices and the various regulations are set by a pro- sure in the chamber around the fixture, all the registers and
grammable controller controlled by a microcomputer. control valves are sonic throats which contribute equally to
Control of the measurement apparatus: displacement of pro- complete stability of the flow.
bes, control of pressure measurement manifolds, supervision
of pressure stabilization as well as data acquisition and pro-
cessing are camed out by a Bull SPSS minicomputer. This 2.5.1.3 Supersonic Cascade Test Fixture (Fig. 2.5-3)
equipment allows tests to be conducted with only two people The fine adjustment of the Mach number is achieved by vary-
and supplies the result as quasi-real-time curves. ing a nozzle throat section obtained by remote-controlled
pivoting of the inside block around an axis close to its down-
stream end. Four sets of interchangeable nozzles with
2.5.1.1 Cascade Testing Equipment
nominal Mach numbers of 1.3,1.5,1.7 and 2 allow the com-
The wind-tunnel can be equipped with two separate test fix-
plete range from Mach 1.2 to Mach 2.0 to be covered.
tures. One is designed for subsonic and transonic stator
cascade tests. The other is designed for supersonic rotor Tho boundary layer branch channels are provided at the end
blade tests. Each of the fixtures is a compact assembly which of the nozzle. The geometry of these channels is designed to
is installed in one of the two experimental chambers of the prevent blocking of the flow.
wind-tunnel. The blades are clamped between two parallel disks which
As the wind-tunnel power is fixed, it was necessary to find a limit the width of the test section to 1OOmm. These disks are
tradeoff between the blade dimensions and the number of rotated to vary the incidence of the cascade. The disks are
channels. A cascade consisting of blades whose chord and made of clear plastic.
height are around lOOmm was deemed necessary and suffi- Downstream of the cascade, the flow is channeled by a set of
cient to obtain both a good geometric definition of the profiles hinged flaps which adjust the cascade boundary conditions to
and accessibility to the measuring equipment. These obtain periodicity of the flow. This is achieved by imposing a
considerations led to contemplating the design of cascades streamline downstream of the end blades with a form similar
with five or six blades with adjustment apparatus optimized to to that obtained for an infinite cascade. This adjustment is
obtain a satisfactory periodicity of the flow. obtained by trial and error, using the pressure measurements
made on the lateral walls.
2.5.1.2 Stator Cascade Test Fixture Theoretically, it is impossible to obtain a periodic flow with
This fixture, illustrated in Figure 2.5-2, is laterally limited by such a limited number of blades, but experience has shown
two parallel end plates spaced 120mm apart (blade height). that crossing the aerodynamic field of three or four blades is
The low velocity flow is captured by a nozzle which channels it sufficient to organize the flow downstream of the channel
to the cascade. This nozzle consists of two metal blocks form- studied.
ing a manifold, extended downstream by perforated flat walls The complexity of this upstream aerodynamic field is more-
allowing communication with cavities A, in accordance with over one of the difficulties for analyzing the tests. The
the technique used in transonic wind-tunnels. definition of mean values for the Mach number and incidence
61
Experimental chamber 2
/
/
1400 kW motor
12-stage axial compressor

' Heat exch:m g e r Maximum compression
ratio 5.6
0.4 Pi < 1 bar Ti = 300 K
Fig. 2.5-1 S5Ch wind-tunnel
Fig. 2.5-2 Stator cascade test apparatus
\
\
62
periodicity Control of the incidence

Adapted plate control flaps
Fig. 2.5-3 Supersonic cascade test fixture
of the flow upstream of the cascade must be the subject of structure related to the secondary flows whose importance
conventions during comparisons with numerical computa- depends on the pressure gradients on the lateral walls. For
tions. This difficulty no longer exists with blades that have a large recompressions, the secondary flows may occupy
flat part on the suction surfacejust downstream of the leading almost all the channel and induce undesirable convergence of
edge, when the upstream Mach number is defined as the value the stream sheets in the median plane. Thisphenomenon is
measured on this flat part. illustrated in Figure 2.5-4, obtained by interpretation of the
test results of a stator cascade with a large deflection (wall vis-
2.5.2 Test Techniques ualizations and probing of the flow).
This section describes a few special techniques used to mod- Many processes, such as blowing slots and porous walls, were
ify the flow structure. They are used separately or together experimented with to act on the secondary flows (Ref. 2.5-1).
depending on the type of test. These methods were abandoned because they were inefficient
or too difficult to implement.The solution adopted consists of
2.5.2.1 Control of Lateral Boundary Layers providing suction slots on the lateral walls as shown in Figure
In the fixtures described in Section 2.5.1.1,the flow has a 3D 2.5-5. This method is inspired from the internal boundary
layer traps on supersonic air intakes (Ref. 2.5-2).
These slots are located in the interblade channels in the area
where the unfavorable pressure gradient risks causing separa-
tion of the boundary layer. In particular, in the case of
supersonic cascades, they are in the vicinity of the main shock
impact point for the maximum back pressure. The traps com-
municate directly with a constant pressure cavity provided
between the lateral wall of the fixture and the door of the
wind-tunnel experimental chamber. The pressure level in this
cavity is controlled by a valve on the line connectingthe cavity
to the downstream part of the experimental chamber where
the pressure level is similar to the wind-tunnel compressor
suction pressure.
The efficiency of the boundary layer traps is demonstrated in
Figure 2.5-6, showing the constant pressure curves in a stator
interblade channel with and without control of the boundary
layers. A 2D flow can be obtained on the major part of the
blade height. This condition is a requisite for validation of cer-
tain computation methods. Depending on the results, the
Fig. 2.5-4 Representation of secondary flows traps are provided on one or both lateral walls.
63
vergence of the stream sheets according to the desired law of

variation in the median plane of the cascade (Ref. 2.5-3). The
1
resulting convergence is not exactly as predicted due to the
imperfection of the computation codes, but it can be modified
by action on the lateral suction slots as indicated in Section
2.5.2.1.
Figure 2.5-7 is an example of the shape generally used for a
high subsonic stator cascade.
Cross section b-b
Fig. 2.5-5 Lateral boundary layer trap
a) Without suction
Fig. 2.5-7 Geometric test section convergence
Applied to supersonic cascades, this method is not suffi-

ciently rigorous. In effect, to comply with the convergence
variation law of a real machine, it would be necessary to begin
convergence of the walls very far upstream of the cascade so
that the effects of this convergence are already manifest at the
channel inlet. These considerations would lead to developing
b) With suction I a special 3D supersonic nozzle for each Mach number and
each cascade tested, which is technically very difficult and
economically unfeasible.
Fig. 2.5-6 Influence of suction on the lateral boundary For the tests conducted at O h R A , only global convergence
layer (Constant pressure curves. Probings at one half-chord is achieved. This convergence is carried out only on the sub-
downstream) sonic region of the flow.
For supersonic flows, this process is obviously not suitable for
validation of the computation codes. It was however used to
compare the performance of blades with relatively similar
2.5.2.2 Test Section Geometric Convergence profiles. But the results must now be considered with caution.
The tests in a 2D flow are very interesting and even necessary In all cases, the convergence coefficient is computed from
for basic research, but it is also necessary to be able to simul- measurements made by a five-hole probe in the plane consid-
ate in the wind-tunnel the 3D effects due to convergence of ered.
the stream sheets encountered in real machines.
A good approach is achieved by using specially shaped lateral 2.5.2.3 Simulation of the Blade Tip/CasingClearance
walls of the cascade. This form is computed to induce con- The rotor blade tip clearance causes large losses which are
64
h2es4
-- -i
y+ttt
ill + + + +
, t + t t & 4 Measurement
:tl
+ + + t
1 + +-+ t
'- + + t t
J +: Pressure taps
Fig. 2.5-8 Simulation of blade tipkasing clearance
detrimental to compressor performance. A study is being 2.5.3.1 Wall Pressure

conducted in the S5Ch wind-tunnel by the Energetics Pressure taps distributed on the lateral walls upstream and
Department of ONERA to model this phenomenon. The downstream of the cascade allow the experimenter to control
setup is illustrated in Figure 2.5-8.The tip clearance is simu- the test conditions and check the periodicity of the flow.
lated by slots machined in the blades, as they are too fine to be The pressure measurements on the suction and pressure sur-
cantilevered. The tip clearance (e) is adjusted by lateral face of two blades delimiting a channel are used directly to
movement of the blades. The cavities of the slots in the wall validate the computation codes. They allow the position of
are sealed with compound. the point of impact of the shock waves to be located as well as
The lateral wall opposite the tip clearance is provided with
suction slots to eliminate secondary flows on this side (see
the position of boundary layer separation. 0
2.5.3.2 Aerodynamic Field of the Cascade
Sec.2.5.2.1).
To determine the performance of a cascade, the aerodynamic
For the experiment, the channel is equipped with wall pres- field upstream and downstream of the cascade must be accu-
sure taps: four rows of pressure taps on the suction surface rately known in a very wide range since, in the case of rotor
and the pressure surface. blades, the Mach number downstream of the cascade gener-
The losses are determined from probings made downstream ally varies from Mach 0.8to Mach 1.2with very large pressure
of the cascade using a five-hole probe. Figure 2.5-9gives an gradients.
example of results obtained for a 4mm slot. The probing These difficult conditions led to developing a small five-hole
50" from the wall, i.e. far from the phenomenon studied, probe (dia. 1.5") calibrated from Mach 0.3 to Mach 3.0
allows it to be checked that the test conditions are the same (Ref. 2.5-4).This type of probe is insensitive to the Mach
with and without tip clearance. At a distance of 10" from number in a range around Mach 1.0. The insensitivity
the wall, it is observed that the tip clearance causes a large depends on the cone angle. The range of uncertainty is from
stagnation pressure drop in the subsonic region of the flow. Mach 0.95 to Mach 1.05 for a cone angle of 60 degrees. The
probe is moved by a two-axis actuator and allows determina-
2.5.3 Instrumentation tion of the Mach number, the flow direction and the local
The measurements required for cascade testing are mainly stagnation pressure in each point of the volume studied. Its
pressure measurements which must be computed in quasi- interaction with the flow, even in an interblade channel,is gen-
real time since for the most part, they determine the test erally small. These measurements were validated by
conditions and are necessary for the adjustments.These mea- comparative measurements made by laser velocimetry.
surements are made by pressure transducers via rotary To limit errors in high pressure gradients, the pressure mea-
pressure scanners. surements relative to each of the probe holes were
65
interpolated in a single point. Figure 2.5-9 gives examples of layers, the defects of this setup (flexibilityand vibration) led to
the results obtained for the five-hole probe. designing the system illustrated in Figure 2.5-11.
In this case, the probe is moved by a rod which crosses all the
2.5.3.3 Boundary Layer Measurements blades located undemeath the channel studied. The remote-
Boundary layer probings on the suction surface of a stator controlled actuator and the position measurement
blade with a large deflection were made at Mach 0.7 and 0.85. potentiometer are located outside the cascade. The fine, flexi-
The measurement of such boundary layers whose thickness, bile probe duct is welded to the trailing edge of the blade.
over a large part of the profile, is below one millimeter,
required developing special probing equipment. This method ensures complete rigidity but has the drawback
of disturbing the flow in the lower channels. In addition to
The first apparatus, made from the actuator normally used for being difficult to implement, it does not lend itself readily to
probings with the five-hole probe, is illustrated in Figure 2.5- systematic use. In both the systems described above, the
10. origin of movement is defined as the contact of the probe with
The movement is obtained by rotating the probe support the wall, determined electrically. The validity of these tech-
around an axis with the actuator used to control the rotation niques was verified by discrete measurements made with
and the movement by a potentiometric measurement. The probes fixed on the blade and whose position was rigorously
amplitude of movement is demultiplied in a ratio from 3 to 9. controlled outside the wind-tunnel.
This system was satisfactory in the probing planes located in The probes used are illustrated in Figure 2.5-11. ASan exam-
the downstream part of the profile, where the boundary layer ple, Figure 2.5-12 shows the probings made at 23 percent and
had a thickness of a few millimeters. For thinner boundary 52 percent of the chord.
0 Without tip clearance

e = O m
With tip clearance
e = 4 m
A: Probings 50 mm from the wall
B: ProbingslO mm from the wall

Fig. 2.5-9 Influence of the tip clearance
,
66
xis of rotation
dl
Fig. 2.5-10 Boundary layer probing system on the suction surface of a blade
-I Detail A
Boundary layer probe
Potentiometer i Fig. 2.5-11 Thin boundary layer probing system
I”
I J-
d Q5
A: 2 3 percent of t h e chord B : 52 percent of t h e chord
Fig. 2.5-12 Boundary layers measured on the suction surface of a stator (M = 0.85)
67
2.5.3.4 Hot Film Chcterizalionofthe Boundaryhyer - the turbulence rate

SUE -the convected power.
To determinethe boundary layerLaminar/turbulenttransition The same process applied to supersonic flows did not give
regionon the suction surface of a blade, the technique was to satisfactory results because the noise from the wind-tunnel
use hot films bonded to thesurface tobesNdied (Fig.2.5-13). m e r e d the phenomenon isvestigated.
For this purpose, a set of 12 hot films bonded to a 0.OSmm
thick support was produced by DISA Eleftronique. Bonding 2.5.3.5 Wall Viiualuation
of such a fragile film with a relatively large surface area to the Viualizationsof the sueamlines on a profile were made using
blade was the subject of a procass developed by the Physics a viscous, colored coating. This coating materialized the
Department of ONERA,in particular to obtain a homogene- stresmlinesand located the impact of the shock waves as well
ous glue thickness(0.005mm).During the measurement,the as the point of separationof the boundary layer.
film d d e r e d is held at constant temperature.and the mean The composition of the mating is specific to each installation,
conyeded power (Pc) and rms value of the fluctuating power sinceit depends on the pressure, velocity and temperature. Its
components are measnred. These values are used to define a application requim a certain amount of experience so as not
turhulence rate (?) characterizing the state of the boundary to apply too thick a layer which would disturb the flow.
layer. Each film is used separately, since the thermal wake it
develops does not allow simultamms measurements to be In the Ease of the SSC% wind-tunnel, the coating hardens
made.ThisprocessusedatMach0.85onastatorbladewitha under the effect of temperature.. A photograph CM be taken
after removal pig.2.5-15).
large deflection gave full satisfaetioa Aa an example, Figure
2.5-14shows the variation along the chord of: These visualizationshave only a qualitative character,but are
-thewallpressure very U S for ~interpreting complex now SUUCNXW.
Fig. 2.5-13 Slator blade equipped with hot films on the suction surhca
0 0.s
x/c
1
OJ
.
0
I
0.U a5
- 0 0.W a5
Fig. 2.5-14 Example of resultsobtained by hot films (M - 0.85)

2.5.4 Critical Andyais ofthe Testa and Prospects these problems, with special test fixtures.The SSCb wind-
The results of the tests conducted in the twosetups described tunnel is used for these tests because of its operational
above were compared with those obtained from inviscid fluid flexibilityand adaptability. As an example, we can mention:
computations. -analysis at Mach 1.20 of the shock wavehoundary layer
Taking boundary layers into account by measurements or by interaction in an isolated channel
computation using integral methods (Ref. 2.5-5) allowed - study of the losses of thick leading edges arranged in cas-
good agreement between computation and experiment to be cades due to shock waves.
achieved in high subsonic flows. Finally, an annular cascade test system is being developed.
For supersonicflows the situation is much less satisfactory.To Designed to validate 3D computation codes based on solving
the difficulties already mentioned in this Section, in panicular the Navier-Stokes equations, its application can be extended
the lack of repmentativity of convergence, is added the to research on turbine blade cooling.
imperfection of the computation d e s . The leading edge
These very performing computation codes are under
shoekwavasarenottskenintoaccountnoristheshock/boun- development at ONERk Figure 2.5-16 shows a comparison
dary layer interaction. between computation and experiment obtained at Mach 1.37
In recent years, research at ONERA has been focused on in a 2D flow (Ref. 2.5-6).
I Pressure s u r f a c e
I,: P
Fig. 2.5-15 Wall visualizations by viscous coating

69 I
1.04 f-P/PIO 1.04
1.00.
3. 0.75. 0
0
0
0.50.
0 , 0.5 - 0.25- ~
Fig. 2.5-16 Comparison between computation and experiment in a compressor cascade (-: “Navier-Stokes”
computation; 0:experiment)
REFERENCES 2.2-3 Kiock R

2.1-1 Scholz N Description of a Probe for Measurements of Two-
Aerodynamicsof Cascades, (translated and revised by dimensional Wake Flow Quantities, DFVLR IB
A. Klein); AGARDograph AG220,1977. 151-74/2.
2.1-2 Gostelow JP 2.2-4 Heilmann W
Cascade Aerodynamics, (Thermodynamics and Fluid The Influence of Axial Velocity Density Ratio on
Mechanics for Mechanical Engineers Series), Per- Compressor Cascade Performance, AGARD-AG-
gamon Press, Oxford, ISBN-O-O8-020428-7,1984. 164,1972.
2.1-3 Bolcs A, Suter P 2.2-5 Erwin JR,Emery JC
Tmnssonische Turbomaschinen, Karlsruhe, Braun Effect of Tunnel Configurationand Testing Technique
Verlag, (Wissenschaft und Technik Taschenausga- on Cascade Performance,NACA Rep. L016,1951.
ben), ISBN 3-7650-2036-2,1986. 2.2-6 Scholz N, Hopkes U
2.1-4 Lichtfuss HJ,Starken H Der Hochgeschwindigkeits-Gitterwindkanal der
Supersonic Cascade Flow, Progress in Aerospace Sci- Deutschen Forschungsanstalt f i r Luftfahrt in
ence, Vol. 15,D. Kuchemann, ed., Oxford, New York Braunschweig. Forschg. Ing. Wes. 25 Nr 5,pp. 133-
Pergamon Press Ltd., 1974. 147,1959.
2.1-5 Lichtfuss HS, Starken H 2.2-7 Sturm W, Fottner L
A CriticalReview of the OutletFlow of Supersonic Tur- The High-speed Cascade Wind-Tunnel of the Ger-
bine Nozzles, in Transactions of the Institute of man Armed Forces University,Munich. Paper of the
Fluid-Flow Machinery: PRACE IMP,Gdansk, pp. 8th Symp. on Measuring Techniquesfor Transonic
491-522,1976. and Supersonic Flows in Cascades and Turboma-
2.1-6 Starken H chines, Genoa, 1985.
a 2.1-7
Untersuchungder Stromung in ebenen Uberschallver-
zogemngsgittern,DLR-FB 71-99,1971.
Starken H, Zhong Y,Schreiber HA
2.2-8 Romer N, Ladwig M, Fottner L
Measuring Techniques at the High-speed Cascade
Wind-Tunnel of the University of the Federal Armed
Mass Flow Limitation of Supersonic Blade Rows due Forces, Munich. Paper of the 9th Symp. on Measur-
to Leading Edge Blockage,ASME Paper 84-GT-233,
ing Techniquesfor Transonic andSupersonic Flows in
Amsterdam, 1984. Cascades and Turbomachines,Oxford, 1985.
2.1-8 Schreiber HA, Starken H 2.2-9 Oldfield MLG, Kiock R, Holms AT, Graham CG
Experimental Cascade Analysis of a Tmnsonic Com- Boundary-Layer Studies on Highly Loaded Cascades
pressor Rotor Blade Section, ASME Journal of Eng. Using Heated Thin-Films and a Traversing Probe,
for Gas Turbines and Power, Vol. 106, No. 2, April Journal of Engineering for Power Vol. 103,1981.
1984. 2.2-10 Pucher P, Giihl R
Experimental Investigation of Boundary-Layer
Separation with Heated Thin-Film Sensors, ASME
86-GT-254.
2.2-1 Starken H
Utilisation of Linear Cascades in Turbomachine
Research and Development. Lecture notes of the
ASME Fluid Dynamics of Turbomachinery,Course
held at Iowa State Univ., Ames, 1982. 2.3-1 Kiock R, Lehthaus F, Baines N, Sieverding CH
2.2-2 Pianko M (Edt.) The Tmnsonic Flow through a Plane Turbine Cas-
Modern Methods of Testing Rotating Components of cade as Measured in Four European Wind Tunnels,J .
Turbomachines (Instrumentation), AGARD-AG- Engineering for Gas Turbines and Power, Vol. 108,
207,1975. No 2, April 1986.
70
2.3-2 Sieverding CH 2.4-4 Starken H

Unsteady Flow Measurements in Straight Cascades. A New Techniquefor Controllingthe Exit Flow Peri-
Symp. on Measuring Techniques in Transonic and odicity of Supersonic Cascades, in Proceedings Ed.
Supersonic Cascades and Turbomachines, Lau- A. Bolcs and T. Fransson. Communication de L'In-
sanne, 1979. stitut de Thermique AppliquCe, ETH Lausanne, Nr
5, pp. 83-86,1977.
2.3-3 Heme JM,Gallus HE
The Influence of Tailboards with Various Wall Por- 2.4-5 Schimming P, Starken H
osity and Transonic Turbine Cascade Flow. Symp. Data Reduction of Two-Dimensional Cascade Mea-
on Measuring Techniques in Transonic and Super- surements, in AGARDograph 207 on Modem
sonic Cascades and Turbomachines,Oxford, 1988. Methods of Testing Rotating Components of Tur-
2.3-4 Goethert BH bomachines (Instrumentation), pp. 14-46,1975.
Transonic Wind Tunnel Testing, AGARDograph 49. 2.4-6 lkeedt DL, Schreiber HA, Starken H
2.3-5 Wyler JS Experimental Investigation of the Performance of a
Probe Blockage Effects in Free Jets and Closed Tun- Supersonic Compressor Cascade, ASME Journal of
nels, J . Engineering for Power, Vol. 97, No 4, pp Turbomachinery, Vol. 110, pp. 456-466,October
509-515,Oct 1975. 1988.See also NASA TM-100879,1988.
2.3-6 Gould RWF 2.4-7 Stark U, Hoheisel H
Pitot Stem Blockage Corrections in Uniform and The Combined Effect of Axial VelocityDensity Ratio
Non-Uniform Flow, ARC CP 1175,1971. and Aspect Ratio on Compressor Performance,
Transactions of ASME Jour. Eng. Power; Vol. 103;
2.3-7 Koschel W, Pretzsch P pp. 247-255,Jan. 1981.
Development and Investigation of Cone-Type Five
Probes for Small Gas Turbines. Symp. on Measuring 2.4-8 Holtman RL, Huffman GD, Mc Clure RB, Sinnet GT
Techniques in Transonic and Supersonic Cascades Test of a Supersonic Compressor Cascade (vol. I),
and Turbomachines,Oxford, 1988. Detroit Diesel, Allison Division, General Motors
Corporation, Indianapolis, Indiana, Aerospace
2.3-8 Sieverding CH Research Laboratories, Wright-PattersonAir Force
Pressure Probe Measurements in Cascades in Mod- Base, Ohio, ARL-72-0170,Vol. 1,Dec. 1972.
ern Methods of Testing Rotating Components of
Turbomachines (Instrumentation - Section I - 2.4-9 Fuchs R, Starken H
Aerodynamic Measurementsin Cascades),AGARD- Experimental Investigations of Supersonic Cascadek
ograph 207,1975. Designed for High Static Pressure Ratios, ASME
Paper No. 77-GT-37,1977.
2.3-9 Osnaghi C, Perdichicci A
Aerodynamic Measurements in Turbine Cascades 2.4-10 Schreiber HA, Tweedt DL
at High Mach Numbers in Measuring Techniquesfor Experimental Investigation and Analysis of the
Transonic and Supersonic Flows in Cascades and Supersonic Compressor Cascade ARL-2DPC,
Turbomachines, 10th Symposium. DFVLR-IB-02-87,1987.
2.3-10 Dietrichs HJ, Hourmouziadis J, Malzacher F, 2.4-11 Schreiber HA
Braunling W Three-Dimensional Endwall Disturbances in the
Flow Phenomena in Transonic Turbine Cascades, Entrance Region of a Supersonic Cascade, Proceed-
Detailed experimental and numerical investigation, ings of the 9th Symposium on Measuring
ISABE 87-7031. Techniques for Transonic and Supersonic Flow in
Cascades and Turbomachines, Oxford, 21-22
2.3-11 Sieverding CH, Sferuzza S March 1988.
Standards for Pressure Measurements in Super-
sonic Non-Uniform Flow. Symp. on Measuring 2.4-12 Hobbs DE, Weingold HD
Techniques in Tmnsonic and Supersonic Cascades Development of Controlled Diffusion Airfoils for
and Turbomachines,Genoa, 1985. Multistage Compressor Application, Transaction of
the ASME, Journal of Engineeringfor Gas Turbines
and Power, Vol. 106,pp. 271-278,April 1984.
2.4-13 Peacock RE
Boundary Layer Suction to Eliminate Corner Separ-
2.4-1 Lindner E ation in Cascades of Aerofoils, Aeronautical
Untersuchungder schallnahen Stromung in geraden Research Council Reports and Memoranda R. & M.
Verzogerungsgittern,Thesis T H , Dresden, 1964. No. 3663,Oct. 1965.
2.4-2 Starken H, Lichtfuss HJ
Cascade Flow Adjustment, AGARDograph 207 on 2.4-14 Boldman DR, Buggele AJZ, Shaw LM
Modern Methods of Testing Rotating Components Experimental Evaluation of Shockless Supercritical
of Turbomachines (Instrumentation), 1975. Airfoils in Cascade, AIAA Paper 83-0003,1983.
2.4-3 Fleeter S, Holtman RL,Mc Clure RB,Sinnet GT 2.4-15 Leynaert J
Experimental Investigation of a Supersonic Com- Cascade Test Methods in Wind Tunnels at ONERA,
pressor Cascade, ARL TR 75-0208,Aerospace in Proceedings of the Symposium on Measuring
Research Laboratories, Wright-Patterson AFB, Techniques in Transonic and Supersonic Cascades
Ohio, June 1975. and Turbomachines, Lausanne 18-19 Nov. 1976.
71
Ed. A. Bolcs and T. Fransson. Communication de (Technical Translation of the Report DLR-FB 73-
L'Institut de Thennique AppliquBe, ETH Lau- 101,1973).
sanne, Nr 5, pp. 103-116,1977.
2.4-29 Starken H, Zhong Y, Schreiber HA
2.4-16 Meauze G Mass Flow Limitation of Supersonic Blade Rows due
The ONERA Supersonic Straight Cascade Wind to Leading Edge Blockage, ASME Paper 84-GT-
Tunnel at Chalais-Medon, ONERA T.P. No. 1409, 233, Amsterdam, 1984.
1974.
2.4- 17 Meauze G 2.4-30 Bolcs A
Shock Recompression of a Supersonic Internal Flow Berechnungder Stromung in einem uberschallanges-
in the Presence of a Boundary Layer Bleed, ONERA tromten Schaufelgitter bei variabler Schichtdicke,
N.T. No. 1976-8. Technical Translation ESA-TT- Journal of Applied Mathematics and Physics
369, ESA, Paris, 1976. (ZAMP), Vol. 32,1981.
2.4-18 Steinert W, Fuchs R, Starken H 2.4-31 Bolcs A, Tsarnourtzis V
Inlet Flow Angle Determination of Transonic Com- Quasi-Three-Dimensional Characteristics Method
pressor Cascades, ASME Paper 91-GT-274, for a Supersonic Compressor Rotor, ASME Paper
Orlando, ASME Journal of Turbomachinery, Vol. 91-GT81,1991.
114, pp. 487-493, July 1992.
2.4-32 Schreiber HA, Starken H
2.4-19 Weingold HD, Behlke RF
Evaluation of Blade Element Performance of Com-
The Use of Surface Static Pressure Data as a Diagnos-
pressor Rotor Blade Cascades in the Transonic and
a 2.4-20
tic Tool in Multistage Compressor Development,
ASME Paper 86-GT-3,1986.
Starken H, Jawtusch V
Low Supersonic Flow Range, Fifth International
Symposium on Airbreathing Engines (ISABE), pp.
67-1-9, Bangalore, India, 12-16 Feb. 1981.
Der Machzahl - und Winkeleinfluss auf Umstrii-
mung und Verluste superkritischer Verdichtergitter, 2.4-33 Meauze G
VDI-Berichte 706, VDI-Verlag Dusseldorf, pp. 317- A Method for Testing Cascades with Converging
332,1988. Sidewalls, J. of Engineering for Power, No. 3, Vol.
100, pp. 457-464,1978.
2.4-21 Weber A, Steinert W
Experimentelle Untersuchungz weier Propfan-Gitter 2.4-34 Kline SJ, McClintock FA
im transsonischen Gitterwindkanal, DLR-IB-325- Describing Uncertainties in Single Sample Experi-
08-89,1989. ments, Mechanical Engineering 75, pp. 3-8,1953.
2.4-22 Fottner L
Test Casesfor Computation of Intemal Flows in Aero 2.4-35 Moffat RJ
Engine Components, Experimental Test Cases for Conm'butionsof the Theory of Single-Sample Uncer-
Cascades, Test Case E/CA-4, pp. 81-94, AGARD- tainty Analysis, Journal of Fluids Engineering 104,
AR-275,1990. pp. 250-260, June 1982.
2.4-23 Levine P 2.4-36 Gostelow JP
Two-Dimensional Inflow Conditions for a Super- The Present Role of High Speed Cascade Testing,
sonic Compressor with Curved Blades, Journal of ASME Paper 81-GT-95,1981.
a 2.4-24
Applied Mechanics, Vol. 24, pp. 165-169, June
1957.
Novak 0
2.4-37 Gustavson BA
Some Observationsfiom Low Speed Cascade Tests
Conceming Sidewall Boundary Layer Suction, in
Flow in the Entrance Region of a Supersonic Cas-
AGARD CP 214, Secondary Flows in Tbrbo-
cade, Strojnicky Casopis XIX, C. 2-3, Slovenskej
Academic vied Bratislava, pp. 138-150,1967. machines, Paper No. 19, Sept., 1977.
2.4-25 Starken H 2.4-38 Meauze G

Untersuchung der Stromung in ebenen Uberschall- Effets des Ecoulements Secondaires dans les Grilles
venogemngsgittem, DLR-FB 71-99,1971. dxubes Rectilignes,AGARD Conference on Secon-
dary Flows in Turbomachines, The Hague,
2.4-26 Lichtfuss HJ, Starken H
AGARD-CP-214, pp. 10-1 to 1O-5,1977.
Supersonic Cascade Flow, Progress in Aerospace
Science, Vol. 15, D. Kuchemann, ed., Oxford, New 2.4-39 Schreiber HA, Starken H
York Pergamon Press Ltd., 1974. On the Definition of the Axial Velocity Density Ratio
in Theoreticaland Experimental Cascade Investiga-
2.4-27 York RE, Woodard HS tions, Proceedings of the Symposium on Measuring
Supersonic Compressor Cascades - A n Analysis of Techniques for Transonic and Supersonic Flows in
the Entmnce Region Flow Field Containing Cascades and Turbomachines, Lyon, France, 15-16
Detached Shock Waves, Journal of Engineering for Oct. 1981.
Power 98, No 2, pp. 247-257, April 1976. 2.4-40 Schreiber HA, Starken H
Experimental Cascade Analysis of a Transonic Com-
2.4-28 schimming P pressor Rotor Blade Section, ASME Journal of Eng.
Calculation and Data Evaluation of Supersonic Inlet for Gas Turbines and Power, Vol. 106, No.2, April
Flow of Two-Dimensional Cascades, ESA 'IT-209, 1984.
72
2.5-1 Meauze G, Thibert JJ 82 exit flow angle

Methode d’itude expirimentale de grilles d’aubes AVDR axial velocity density ratio = p2 w2 sin B2/p1w1 sin
transsoniques a forte dkviation, Association Tech- 81
nique Maritime et ABronautique (ATMA) session
1972. TF’ONERA No. 1094,1972.
2.5-2 Meauze G NOMENCLATURETO SECTION 2.4
The ONERA Supersonic Straight Cascade Wind Tun- AVDR axial velocity density ratio =
nel at Chalais-Meudon, Supersonic Tunnel 5 2 w2 sin 82/51 w1 sin 81
Association, Buffalo (N.Y. USA) 1-2 Oct. 1974, TP C blade chord
ONERA No 1409,1974. M Mach number
2.5-3 Meauze G Mi, isentropic Mach number f(p/ptl)
A Method for Testing Cascades with Converging Side MN nozzle exit Mach number
Walk, Engineering for Power, No 3, vol 100, p. 457- MSS suction surface Mach number
464, July 1978. P static pressure
Pt total pressure
2.5-4 Gaillard R R universal gas constant
Etalonnage et utilisation de sondes 5 trous ONERA.
t blade pitch
7th Symposium on Measuring Techniques for Tmns-
Tt 1 total temperature
onic and Supersonic Flows in Cascades and
W velocity
Turbomachines,Aachen, September, 1983.
X/C relative blade chord
2.5-5 Cousteix J, Houdeville R 8 flow angle with respect to cascade front
Mkthode intkgrale de calcul d’une couche limite turbu- BN cascade setting angle measured between cascade
lente sur une paroi courbie longitudimlement, La front and nozzle axis
Recherche Airospatiale 1977-1, p. 1-13. ‘1 coordinate in tangential direction
2.5-6 Cambier L, Veuillot JP K ratio of specific heat
Computation of Cascade Flows at High Reynolds P Mach wave angle
Numbers by Numerical Solution of the Navier-Stokes V Prandtl-Meyer angle
Equations, AIAA 26th Aerospace Sciences Meeting, 5 coordinate in axial direction
Reno-Nevada (US),TP ONERA No 1988-3. YC, relative axial chord
0 total pressure loss coefficient =
(PI1 - Pt2)4Ptl - P1)
LIST OF SYMBOLS Indices:
M, inlet Mach number 1 uniform inlet conditions
p2/p1 static pressure ratio; p2 = backpressure 2 uniform outlet conditions
81 inlet flow angle ref values at reference inlet flow angle
73
Chapter 3
Annular and Rotating Cascades
3.1 ANNULARCASCADES occur as a result of the need to establish radial equilibrium in

H.P. Hodson and RG. Dominy swirling flows. Low momentum fluid, such as that which
3.1.1 Summary occurs in boundary layers, experiencingthese same gradients
This section describesthe testing techniqueswhich are applic- will tend to migrate in the direction of the local gradient. By
able to fixed annular cascades.A brief introduction compares their very nature, these gradients are absent from linear cas-
the annular and linear cascade models. The importance of cades and so a potentially important three-dimensional
establishing the correct inlet and exit conditions in the annu- aspect of the flow cannot be modelled. However, linear cas-
lar environment is then discussed. This is followed by a cades can be used to study the effects of blade stacking in
description of the design of high speed facilities in general terms of the effects of lean (Ref. 3.1-6) and sweep (Ref. 3.1-7).
and, by way of an illustration, reference is made to a specific Stream-tube height variation (flare) can also be adequately
facility. A section then follows on low speed testing. The pro- modelled in a linear cascade (e.g. Ref. 3.1-8), albeit in a sim-
blems and advantages of modelling high speed flows at low plified environment but the effects of blade twist or changes of
a speed are described. A brief discussion of experimental con-

siderations is also provided.
section along the span cannot. Unlike a linear cascade, an
annular model can be simultaneously used to study the effects
of radial pressure gradients, of blade stacking, of changes of
3.1.2 Introduction blading geometry and of twist but the inlet and exit flow fields
In principle, an annular cascade consists of the space between may require careful attention if the correct spanwise distribu-
two concentric cylinders which contains a row of blades tions are to be obtained.
whereas the linear cascade consists of two-dimensional blade vpically, in a linear cascade, there may be six to ten blades.
sections fitted between parallel walls. The annular cascade The number is essentially determined by the mass flow capac-
therefore resembles a part of a turbomachine very much more ity of the air supply to which the cascade is connected. The use
closely than the linear cascade. Annular cascades are gener- of a small number of blades in a linear cascade can cause
ally employed when there is a need to simulate the problems due to a lack of periodicity. In transonic linear tur-
turbomachine environment or when the three-dimensional bine cascades, for example, shock-waves radiating from the
nature of the flow field is of primary interest. Linear cascades cascade exit will reflect from the pitch-wise boundaries of the
are generally used for large scale, detailed investigations (e.g. exit flow (tailboards or free jet boundaries) and so disrupt the
Refs. 3.1-1 and 3.1-2) or occasions when two-dimensional periodicity. The reflections can be controlled (e.g. Ref. 3.1-9)
performance data are required (e.g. Refs. 3.1-3 and 3.1-4). but only at the expense of added complexity (e.g. porous tail-
Though the majority of annular cascades have cylindrical boards) and experimental effort. Similar problems can also
inner and outer walls, there is no fundamentalreason why this exist at inlet when the flow is supersonic. The use of more
should be so. The blade, hub and casing profiles may be exact blades would reduce the problems but would also negate the
replicas of those in a turbomachine. The annular cascade thus advantages of scale which are often offeredby these cascades.
requires few compromises to be made during its design and In the Case of unsteady flow experiments, the problems of
construction. Unfortunately, the annular cascade may also periodicity can become paramount.
present problems of accessibility which are similar to those In an ideal annular cascade, the flow is truly periodic. As a
encountered in turbomachines. In addition, the annular cas- result, any trailing-edge shock or expansion waves should
cade will undoubtedly possess a larger number of blades than propagate without interference from reflections unlike in a
its linear counterpart and so, for a given mass flow and aspect linear cascade. However, at inlet, problems may still be
ratio, it will be smaller. encountered if inlet guide vanes are used to create supersonic
A linear cascade consists of a small number of identical con- flow. In practice, of course, the problems of the accurate
stant section aerofoils. The cascade geometry is produced by machining and the setting of the individual components also
the development of a cylindrical section. Strictly, this section remain but these problems are common to all blade test faci-
should represent a stream-surface within the turbomachine lities.
from which the geometry was taken. In practice, a compro- Annular cascades appear in various guises. They may operate
mise is often made regarding the geometry and flow continuously(e.g. Refs. 3.1-10 and 3.1-11) or for a shortdura-
conditions although in the case of incompressible flow, con- tion (e.g. Refs. 3.1-12 and 3.1-13), they may have low speed or
formal transformations may be used to map a conical compressible flow and they may be configured as part of an
stream-surfaceonto a cylindrical surface (Ref. 3.1-5). Though open or closed loop air system (e.g. Refs. 3.1-10 and 3.1-12).
the linear cascade is further removed from the turbomachine This section will discuss the testing techniques which are
environment than the annular cascade, it remains an invalu- applicable to annular cascades. Particular emphasis will be
able tool for the investigation of many of the phenomena given to the methodology of testing. Details of instrumenta-
present in turbomachines.The popularity of linear cascades is tion and measurement techniques will only be given when
mainly due to their scale, their simplicity of construction and these are specific to the annular environment. A full discus-
the ease of access for probe and optical measurements. sion of complete turbomachine stages is beyond the scope of
Radical static pressure gradients exist in turbomachines.They this section. Nevertheless, there are aspects which are com-
74
mon to both the testing of stationary cascades and rotating layers is easily controlled by a variety of devices. In compres-
cascades and these will be discussed below where it is appro- sors, the nature and thickness of the endwall boundary layers
priate. has received much more attention (e.g. Ref. 3.1-16).
3.1.3 General Considerations The use of suction slots to reduce the thickness of the endwall
In high speed annular cascade testing, it is usual to establish boundary layers (e.g. Ref. 3.1-17) and of trip wires to increase
the correct Mach number and Reynolds number so that it (e.g. Ref. 3.1-18) are both well established techniques,
compensation for the effects of compressibility is not neces- though both techniques do suffer from the need to adjust the
sary. The blades may therefore be exact geometrical boundary layer control device as operating conditions are
representations of those in the turbomachine. Low speed changed. It is often implied that the boundary layer profile
annular cascades, which are operated at scale Reynolds num- should be as near to an equilibrium profile as is possible
bers, may be distinguished from their high speed counterparts before it enters the blade row, but again, there is little or no
by the fact that the flow within them is essentially evidence that this is the case in the equivalent turbomachine.
incompressible. Many are of large scale, offering improved Since equilibrium only returns many boundary layer thick-
spatial resolution, lower capital and running costs, and ease of nesses after a disturbance, the requirement of an equilibrium
use. profile means that the length of the inlet section effectively
The scale of all annular cascade rigs is largely constrained by sets an upper limit to the thickness of the inlet boundary layers
similar considerations to those that apply to linear cascades, that can be achieved. Vortex generators, rather than trip-
notably by the mass flow-pressurerise characteristic of the air wires, appear to be more suitable for generating large
supply and the values which are required to establish the cor- boundary layer thicknesses (Ref. 3.1-19) although trip wires
rect operating conditions in the cascade. retain the advantage of simplicityand they can be easily inter-
changed to alter the thickness of a boundary layer.
3.1.4 Inlet and Exit Flow Conditions Many turbomachine blade rows do not operate with axial
The ability to establish and investigate the effects of the inflow. Thus, it may be necessary to use some form of flow
appropriate radial distributions is often quoted as the most guidance to produce the correct swirl angle distribution at
important reason for building an annular rather than a linear inlet to the blade row. Inlet guide vanes may be inserted into
cascade. Therefore, great emphasis is often placed upon the the annular space ahead of the blade row under investigation,
need to establish the correct radial distributions of the flow (e.g. Ref. 3.1-20) but time invariant wakes and possibly secon-
variables at inlet to and exit from the blade row under test. dary flow vortices are shed from the guide vanes so that if
The shape and the size of the inlet and exit ducting of annular there is insufficient distance downstream of the guide vanes,
cascades often dominate the particular facility. For this rea- the flow will not be circumferentially uniform at inlet to the
son, the factors which influencethe nature of the inlet and exit test section.In the case of supersonic inflow to the test section,
flow are introduced prior to a discussion of the cascades the guide vanes’ trailing shock waves may also affect the cas-
themselves. cade under test and any bow shocks of the cascade may
interferewith the guide vanes and be reflected back to the cas-
3.1.4.1 Inlet Conditions cade.
The majority of annular turbine cascade tests are conducted Bolcs (Ref. 3.1-21) describes an axial cascade test facility
with blades that are designed for axial inflow. These designs which uses a radial inlet in order to circumvent the problems
are the simplest to test since axial flow tends to occur naturally of wake and shock wave interference. The facility is shown in
upstream of the cascades. Under these circumstances it is Figure 3.1-1. The inlet guide vanes are placed in the radial
only necessary to create the correct radial profiles of total portion of the inlet duct, where the velocity (and therefore
pressure, total temperature and turbulence and to remove any Mach number) is much less than at inlet to the test section.
residual swirl. The flow in the guide vanes is always subsonic and any shock
In all of the annular cascade tests known to the authors, the wave interference effects are avoided. The flow is then accel-
stagnation pressure is usually constant outside of the hub and erated as it progresses towards the test section so that
casing boundary layers. A series of gauzes, like those used to although the wakes may still be identified, the total pressure at
simulate inlet distortion in compressor tests, could be used to cascade inlet is effectively axisymmetric. Boundary layer
create a radially varying inlet stagnation pressure profile but bleeds are provided ahead of the test section in order to limit
the authors are unaware of such an experiment. Information the effects of boundary layer growth.By appropriate setting of
on radial temperature distributions is also rarely available. It the guide vanes, subsonic or supersonic inflow can be
is even less common to attempt a simulation of the actual dis- achieved at inlet to the test section. In the case of the latter
tributions so that in the majority of annular cascade tests, the condition, the sonic throat upstream of the cascade is self
stagnation temperature is uniform. For aerodynamic mea- starting, its location depending upon the inlet conditions.
surements it is usual for the inlet temperature to be nominally Even when the mainstream flow is nominally axial at inlet to
ambient but rigs that have been designed for.heat transfer the blade row under investigation,the absence of an upstream
measurements may operate at significantly higher pressures rotor may significantlyaffect the performance of the test cas-
and temperatures (e.g. Refs. 3.1-11 and 3.1-13). cade. It is well known, for example, that the unsteadiness
A survey of the open literature concerned with turbine cas- produced as a result of blade row interactions can affect the
cade testing, would suggest that at inlet, the hub and casing blade loading (e.g. Ref. 3.1-22), the loss (e.g. Ref. 3.1-23) and
boundary layers are turbulent and have a displacement thick- the heat transfer (e.g. Ref. 3.1-24), though the importance of
ness equal to about 1 percent of the blade chord (e.g. Refs. the effects of wake-boundary layer interactions can now be
3.1-14 and 3.1-15). However, thereis little or no evidence that assessed (Ref. 3.1-25). Two very simple techniques have been
these values correspond to those of the equivalent turboma- used to simulate the wakes shed from an upstream rotor. A
chine. Fortunately, the thickness of the endwall boundary spoked wheel (e.g. Ref. 3.1-26) driven by, say, an electric
75
i n l e t valves
o u t e r and inner s e t t l i n g chambers
o u t e r and inner p r e s w i r l vanes
aerodynamic probes up- and downstream of'
t h e t e s t cascade
9 c y l i n d r i c a l o p t i c f o r Schlieren v i s u a l i s a -
tion
10 s t a t i c w a l l pressure tappings
11 o u t l e t .valve
# suction
Fig. 3.1-1 Annular test rig for investigationof steady and unsteady transonic flows (Bolcs)
motor, may be placed ahead of the cascade. Alternatively, 'a bars for the simulationof unsteady inflow will also tend to cre-
lightly cambered, virtually unloaded blade row may be placed ate skew (Ref. 3.1-32).
ahead of the test section (Ref. 3.1-20). A turbulence grid may
also be placed upstream of the cascade in order to increase 3.1.4.2 Exit Conditions
the levels of background disturbances (e.g. Ref. 3.1-27). The Many annular cascade tests (e.g. Refs. 3.1-35 to 3.1-40) have
effects of potential interactions, whose strength is usually been conducted with nothing more than a constant annulus
greater at higher Mach numbers (Ref. 3.1-28), cannot be downstream of the cascade which subsequently exhausts into
simulated. a constant pressure plenum. Several investigators have drawn
attention to the consequences of the imposition of a constant
The change of frame of reference which occurs as the flow pressure exit boundary condition upon the flow through a
leaves a rotating blade row results in a skewing of the hub and, cascade. Concern is usually expressed since there is a tend-
when a shroud is present, the casing boundary layers. The ency for the hub endwall layer to separate as the pressure rises
effects of endwall skew upon the secondary flow of a blade towards the exit. In effect, this pressure rise occurs because
row are significant and are well documented (e.g. Refs. 3.1-29 the conservation equations for mass, momentum and energy
and 3.1-30). Even so, few annular cascades have the ability to ensure that the static pressure at mid-span at the cascade exit
create a skewed inlet boundary layer. Rotating hubs can be plane is similar to that in the downstream plenum while the
used to generate the inlet skew (Ref. 3.1-31) in the same way presence of exit swirl means that the hub static pressure at the
that moving belts are used in linear cascade studies but long cascade exit plane is lower than that at mid-span. The possib-
surface distances are required in order to establish the correct ility of flow separation will increase as the hub-tip radius ratio
conditions and the additional complexity that such an decreases and the exit swirl increases (Refs. 3.1-41 and
arrangement requires cannot be ignored. Some investigators 3.1-42). The problem is encountered more often in turbine
have added uniform roughness or fences to moving belts (e.g. rather than compressor cascades. The avoidance of a separa-
Ref. 3.1-33) to shorten the surface length needed to establish tion should not, however, be regarded as absolutely necessary
the correct conditions. An upstream rotor carrying rotating though it does represent good practice. It may be sufficient
76
that separation happens downstreamof the exit measurement has a direct impact on the time available to establish the cor-
plane, although a large region of reversed flow might create rect operating conditions and then to acquire the data.
unsteadiness or so much blockage that the radial equilibrium Nevertheless, short duration blowdown cascades may be of
condition is disturbed ahead of the separation. In the case of considerable size. The major difficulties associated with the
zero exit swirl, a radially constant exit static pressure is, of design and operation of a transient rig arise due to the much
course, consistent with the actual flow. greater complexity of the control systems and the need for fast
The majority of annular turbine cascades are used to investig- response instrumentation and rapid traverse systems. Con-
ate high pressure turbine nozzle blades. In practice, there are sideration must also be given to the tremendous aerodynamic
few instances of separation after turbine cascades when the forces that occur during the operation of large scale high
hub-tip radius ratio is of the order of 0.8 or higher (e.g. Ref. speed cascades.
3.1-42).In the case of low hub-tip ratio cascades, this is Continuous flow facilities must be employed when steady
usually not the case. Spurr (Ref. 3.1-43)describes a method of flow conditions are required over sustained periods of time.
establishing the correct radial flow distributions at exit from a The mass flow rate and pressures in these facilities are simply
low hub-tip ratio cascade. He successfully used a perforated determined by the setting of valves and compressor speed.
plate to simulate the rotor of the last stage of a steam turbine, Fast response instrumentation is only required for specialized
the radial variation of the open area ratio of the plate and, applications and area-traversingis more easily accomplished.
therefore, the static pressure drop across the plate being The power required by a contiyous flow high speed system is
chosen according to simple fluid mechanic principles. Squire essentially governed by the need to provide the choking mass
et al. (Ref. 3.1-44)applied the same technique at exit from a flow rate of the cascade at the highest pressure ratio that is
low hub-tip ratio transonic turbine cascade. It would seem required. The required pressure ratio may be reduced by the
that in many cases, it is unnecessary to employ a rotor in order inclusion of a diffuser between the blade row and the down-
to simulatethe correct boundary conditions at exit from a sta- stream plenum. If the pressure rise-mass flowcharacteristic of
tionary cascade. the compressor is not vertical, it may then also be possible to
test a cascade of larger scale.
3.1.5 High Speed Annular Cascades
3.1.5.1 High Speed Test Facilities The power requirements of continuous flow facilities are such
In principle, a high speed annular cascade permits the testing that the cascades under test are often smaller than those that
of blade geometries in flow conditions that closely match could be tested in blowdown facilities. However, since pres-
those in actual machines. The design of high speed annular sure transducers can be remote from the pneumatic probes
cascade rigs is largely constrained by the relationshipbetween and settling times are relatively unimportant, the difference of
the size of the cascade and the gas supply required to establish scale only becomes significant for specialized investigations
the correct Reynolds and Mach numbers. such as the traversing of blade surface boundary layers. Of
The cascade under test is usually incorporated into one of two course, it may be possible to test just a segment of an annular
types of test facilities;either a continuous tunnel (e.g. Ref. 3.1- cascade so that larger blades may be used but as yet the issue
10) or a limited duration blowdown arrangement (e.g. Ref. of how to create the correct downstream conditions has not
3.1-12).Both configurations may be part of a closed system been worked out for aerodynamic tests. The dimensions of a
typical low aspect ratio NGV cascade which operates in a
but this is not a fundamentalrequirement.A closed configura-
tion does however offer the possibility of using gases other continuous closed loop facility are given in Table 3.1-1.In
than air to achieve engine conditions at lower power (e.g. Ref. Table 3.1-1
3.1-45)or of varying the Reynolds number by simple adjust- Whittle Laboratory, Closed-Loop Annular Cascade
ment of the mean pressure in the system (e.g. Ref. 3.1-46). Geometry and Typical Flow
Closed systems also facilitate the removal of moisture which
may be necessary if condensation is to be avoided (see Ref.
3.1-46).
The blowdown configuration essentially consists of two ple-
num chambers with an interconnectingduct which houses the
cascade. One of the plenum chambers may be replaced by the
atmosphere. Prior to operation, the plenum or plenums are
charged with gas to specified pressures. In most practical
applications, the upstream plenum will be pressurized whilst
the downstream pressure will be equal to or lower than the
atmospheric value. By controllingthe release of the high pres-
sure air, a constant mass flow and pressure ratio can be
maintained over part of the dischargetime. The back pressure
on the cascade is usually controlled by a throttle in the exit
duct though such a simple arrangement does limit the range of
flow conditions that can be achieved when exhausting to the
atmosphere. Lower exit pressures can be achieved by the use
of ejector pumps (Ref. 3.1-47).The duration of the resulting
transient flow is a function of the size of the cascade and the
pressure vessels. In principle, the scale of a rig is limited only
the available space for the cascade and pressure vessels. In
practice, however, the time taken to raise the plenum to the
required pressure may be an important consideration and this
r
practice, most continuous flow annular cascades will have assembly since each module is of a manageable size and
casing diametus of less than about OSm. weight. The modules are suspended from an overhead track
In summary, short duration blowdown cascade wind-tunnels allowing simple and rapid access to any part of the cascade
provide a method of overcoming the power and thus the size rig. In this particular application an overhead system offers
constraints but at considerable expense in terms of insrhu- significant advantages over a conventional floor mounted rail
mentation complexity and cast. The greater measurement arrangement since working space around the rig is limited by
detail that is potentially available from such facilitiescomple- the walls of the enclosing plenum. Overhead suspension can
ments that from the s d e r continuous facilities that can be achieved using a single track if some means of lateral and
more easily provide more wide ranging data in reasonable torsional support is provided when the rig is assembled and
time-SCaIeS. running.For many other applications floormounted railsmay
be more appropriate.
3.1.5.2 The WhinleLobomroryHighSpeedAnnuIarCmcade Air is supplied to the working section via the inlet contraction
The transonic annular cascade at the Whittle Laboratory is which reduces the duct CIOSS Sectional area from 0.43m2 to
described here in order to introduce the main issues that must 0.033mz in the working annulus corresponding to hub and
be confronted when approaching the design ofa continuously casing diameters of 0.357111 and 0.412m respectively. Since a
running annular cascade rig. Short duration facilities may recirmlating wind-tunnel arrangement is employed, a non-
share many of these design features and such rigs are dis- uniform and possibly swirling flow exists upstream of the
cussed in more detail elsewhere. inlet. A combination of a honeycomb (as a flow straightener)
The rig makes use of the Whittle Laboratory’s closed circuit, and gauzes (for flow uniformity) is located approximately
variable density. high speed wind-tunnel in which the Mach three inlet pipe diameters upstream of the inlet contraction
number and Reynolds number may be controlled independ- and effectivelyremoves any inlet disturbances.
0 ently (Ref. 3.1-46). Air h m two compressors is supplied to
the teat section via a contraction and a short,p a d e l duct. It
The centre-body of the inlet contraction is supported at the
upstream end of the rigby three slender support shuts of aetu-
exhausts via a radialdiffuser into a large plenum before being foil section (Fig. 3.1-3). Since no significant temperature
returned to the compmors. The compressors may be oper- changes are experienced at the inlet (coolers are included in
ated either individdy or in series or parallel according to the circuit between the compressorsand the rig) radial shuts
the cascade geometry and the requkedflow conditions (Table aresuf6cient.Insomerigsit maybenecessarytoincludetang-
3.1-2). There is no provision for the pressurization of the sys- ential lean to cope with the thermal expansion of the struts.
tem. Therefore, the maximum pressure that can be achieved When complete, the enthe assembly is cantjlevered from the
occurs when atmospheric pressure exists at the inlet to the innerfaceoftheenclosingplenumandallforcesarecarriedat
compression system lko vacuum pumps can be used to eva- thisface. Theinlet casing thereforeincludes asubstantial steel
cuate the system which therefore allows the density in the fabrication which is needed to provide adequate support.
system to be reduced for operation at low Reynolds number.
ForatypicalNGVcascade(Table3.1-1),theexistingarrange- The contoured Liner of the inlet casing and all of the other
mentis sufficient to achieve the densities and pressure rahos modules are machined from aluminium castings. The casing
that are r e q u i d to obtain Reynolds numhers in the range modules are joined by quick release clamping rings (Fig. 3.1-
from well helow 10’ to above lo6 at Mach numbers up to 1.2. 4) so that MYinternal part of the rig can be accessed within
only a few seconds by a single operator. The hub modules,
which rarely need to be disassembled, are interconnectedby
bolted flanges. Each ofthe hub modules is hollow for ease of
Table 3.1-2 insnumentation and to reduce the overall weight of the rig.
Whittle LaboratoryTansonic Annular Cascade Air Supply The inlet Contraction leads into a parallel section which, when
present, can be used to control the hub and casing inlet boun-
dary layer parameters. The traverse system and the blade row
itself are incorporated into a single module which is immedi-
ately downsueam of the inlet sections.
For all of the studies carried out to date a simple parallel-
annulus cascade has been tested but the modular
conshuction does allow future studies using flared and con-
toured wall p m e t i e s to be u n d d e n with the
modification or replacement of only a few rig components.
The only major restriction is that imposed by thetraversecas-
ing which limits the maximum diameter. Within this diameter
the wall contour may be revised by the uge of profiled liners or
entirely new components.
In order to achieve supersonic cascade exit conditions using
the Whittle Laboratory’s existing compressom at the required
mass flow rate it is necessary to diffuse the swirling cascade
exit flow before it exhausts into the enclosing plenum. The
meridional geometry of the diffuser is formed by two drcular
Figure 3.1-2 shows the general arrangement of the facility. arcs. Such devices can ereate unsteady flow hut this has not
The entire annular cascade rig is modular so that it may easily been identified in this facility. The diffuser recovers approxi-
be adapted to meet future requirements. Modular design also mately 50% ofthe cascade exit dynamic head when operating
provides ease of access to all parts of the rig and ease of with an entry swirl of 65’ from the axial direction. The hub
Fig. 3.1-2 The Whittle Laboratory transonic annular cascade - general arrangement
Fig. 3.1-3 Inlet spider of Whittle Laboratory transonic annular cascade

79
QUICK RELEASE CLAMP
\
PROBE MOUNT
(rotates with traverse ring)
I
i
CIRCUMFERENTIAL TRAVERSE RING
Fie. 3.1-4 Working Section of Whiffle Laboratory transonic annular cascade

showing circumferential traverse ring and quick release clamps
and casing are rigidly connected at the diffuser exit via axial
inspection of the blade row and provides the potential for
bolts. This not only provides a means of locating concentri- future optical studiesof the flow. However, it should be no!ed
d y the downstream end of the hub and casing but also that the curvaturea of the hub and casing walls add signifi-
conveniently allows the forces acting upon the hub to be cantly to the problems of using optical techniques.Because
transmitted through the casing to the inlet support section. the current blades are mounted on the casing,optical access is
resaicted through the casing. A hub mounted arrangement
To date, the cascades have contained 34blades with chords of may provide greaterflexibilityfor certain optical studies.
approximately 40" and spans of 28mm corresponding to
an aspect ratio of 0.7and a bub/tip ratio of 0.867.The blades The use of a cartridge containing several blades makes
detailed insrmmentationof the cascade relatively straightfor-
are mounted from the casing end only and silicone sealaat is
applied at both the hub and tip of each blade to p m t lea- wardsinceit~becarriedoutonthework~~Thesyst~
kage flows. Four of the blades are mounted in a removable also allows alternative passage instrumentation to be set up
perspex @lexiglass) camidge (Fig. 3.1-5)and a matching per- using a second camidge whilst performing teats on the 6rst
. .
s~exwindowissetintothebub.Theuseof~ersoexaidsvisual cartridge. Since the arrangement of surface pressure tappings
andotherformsofinsr"entationisohensp~ctoaparti-
cularcancade,theuseofinstrumentedcamid~aLsoallowsa
number of cascades to be tested without the need to replace
the casing.
When constant sectionvanes are used it is a relatively straight-
forward procedure to create pressure tappings by drilling
through the blade surface to intersect larger spanwise holes
that have been drilled below the surface f"the hub or tip.
This mahodhas commonly been adopted for the inskumen-
tah0n of blades in high speed hear cascades and is ohen
rather easier to use m annular cascades because the blade
spans are smaller. The technique cannot be uscd when the
blades have complex th" ' ensional surface curvature.If
the blades are cast in resin, it is possible to cast the spanwise
pressure lines into the blade; a procedure that is described in
more detail below.If this is not possible then it may be neces-
sary to cut grooves in the blade surface into which pressure
Nbes are placed. Thegrooves are then 6Uedand the surfaceis
returned to its original shape. The small profile errors that
may be inmduccd by this procedure dictate that the side of
the blade which is opposite to that where the pressure tapp-
ings are located should be machined and that the pressure
Fig. 3.1-5 Blade cartridge of Whiffle Lebomtory transonic tappings therefore should be applied to a blade passage which
annular cascade is bounded by 'good' surfaces in preference to the use of a
singleinstrumentedblade.Infact,it isusualinhighspeedcas- probes are used exclusively in order to “ i z e the effects of
cades to choose passage instrumentation when the small probe stem blockage and the use of a separate traverse
blade size limits the total number of pressure tappings that module would have required an extended probe and would
can be fitted into a single blade. have led to problems associated with probe flexure and vibra-
A three axis traverse me&” enables detailed measure- tion.Such an arrangement would have severely compromised
ments to be made of the flow-field at exit from the cascade measurements with the smallest probes such as those used for
and provides reasonable probe access within the blade pas- measurements within the blade passages boundary layers.
sages. C i e r e n t i a l motion is achieved by rotating a
complete section of the casing relative to the fixedblades (Fig. 3.1.6 Low Speed Annular Cascades
3.1-6).?he rotating ring is driven through a flexible stainless Low speed annular cascades may be distinguishedfrom their
steel cable from a stepper motor via an anti-baddash gearbox. high speed counterpam by the fact that the flow within them
Redid traversing is provided by a stepper motor driven linear is essentially incompressible.Many low speed cascades, such
ball slide. The radialslide is mounted on a rotary table which as the 1.5m dia. rigs at the Whittle Laboratory (e.g. Ref.3.1-
is itself rigidly mounted to the circumferential traverse ring. E),are of large scale. The major advantages of low speed
The traverse system may be controlled by computer or by a cascade testing are: improved spatial resolution; lower costs;
key-pad. Since annular cascade testing often involves a lot of and ease of use. Table 3.1-3 lists typical configurations.
area traversing, computer control of motorized traverse
Table 31-3
Geometry and Flow Condkions of Cambridge Large Scale
I
Low Speed Axial Flow Compressor and Turbine Research
Facllities
3.1.6.1 Low Speed Modetling of High Speed Cascades

I
For true dynamic similarity to exist bemeen the flowin the
cascade model and in the turbomachine, the Reynolds num-
Fig. 3.1-6 Rear view of WhtUe Laboratmy transonic ber and the Mach number must be the same for each and the
annular cascade surface roughness, the freestream turbulence and unsteadi-
ness and the blade row gwmetty must all be scaled correctly.
Inreality,theflowinalowspeed~laseadeisessentiallyincom-
?he modular nature of the rig provides easy access for chang- pressible so that true dynamic similarity cannot be
ing or repositioningprobes. Incorporatingthe traverse system maintained Thus, the use of low speed cascades adds further
and the blade row into a single module, as is the case here, complexity to the modelhng process and a choice must be
placea some limitations upon the geometries of the cascades made as to which parameters are to be regarded as sacrosanct
that can be tested but at the same time it provides consider- and which may be altered to achieve an acceptable compro-
able benefits. Of these, the greatest is undoubtedly the mise.
availabiity of a precision three axis traverse system located Wider (Ref.3.1-48) presents an approach to low speed simi-
immediately downstream of the blade row. Sting-mounted larity modelling for high hubtip ratio multi-stage axial flow
81
compressors which is in use at General Electric. The metho- 3.1.6.2 Low Speed Facilities
dology is very similar to that employed at the Whittle Figure 3.1-7 shows a cross section of one of the two 1.524111(5
Laboratory (e.g. Refs. 3.1-49 and 3.1-50) and elsewhere (e.g. feet) diameter large scale low speed rotating rigs at the Whit-
Ref. 3.1-51) for both low speed compressor and turbine test- tle Laboratory. It is shown configured as a two stage axial flow
ing. turbine with a hub-tip radius ratio of 0.8. These two facilities
At the simplest level, conventional normalizing techniques have previously been used to test isolated compressor rotors
are used to produce an equivalent set of velocity triangles for of low hub-tip ratio (e.g. Ref. 3.1-54), a single stage axial tur-
bine (e.g. Ref. 3.1-15) and compressor (e.g. Ref. 3.1-55) of
the mean gas path and, therefore, an equivalent set of stage
characteristics. The annulus line is chosen to provide the hub-tip ratio 0.7 and as a single stage compressor of hub-tip
equivalent meridional velocity variation, a radial equilibrium ratio 0.8 (e.g. Ref. 3.1-56). The two facilities have an identical
calculation being used to obtain the desired variations. How- casing diameter to those at the United TechnologiesResearch
Centre (e.g. Ref. 3.1-57) and General Electric (e.g. Ref. 3.1-
ever, it should be noted that the pressure and temperature,
and so the density and viscosity, change significantly through 48) two of which are shown in Figures 3.1-8 and 3.1-9
respectively. Though the rigs at the Whittle Laboratory are
a high speed turbomachine whereas they are nearly constant
at low speed. Since the density variation is negligible at low different in detail to the other large scale facilities mentioned
above, they share a number of common features and it is use-
speed, the annulus height will vary much less than at high
speed. Quite often, turbomachines are designed for approxi- ful to examine these in some detail.
mately constant axial velocity. Given the cost of manufacture Each rig operates as an open loop wind-tunnel and therefore
of casings, it is therefore common to find that most general has a large inlet bellmouth drawing air from the atmosphere
purpose low speed annular test rigs have constant hub and into the working section. The open loop configuration, with
casing radii in the working section, though when the axial the mass flow control device and suction fan (if any) after the
velocity is not constant at high speed, it is necessary to accept working section ensures that axisymmetric, low turbulence
that the velocity triangles will not be identically matched in flow can be obtained at entry to the blade row under investiga-
the model rig. tion. The added complicationsof a closed loop circuit cannot
be easilyjustified in the case of low speed rigs of this type. The
The Reynolds number characterizes the effects of viscosity in throttling device and fan, where fitted, are placed downstream
the flow, whether in terms of .loss generation or blockage. of the test section. The throttle may be a simple plate fitted to
Therefore, the operating Reynolds number of a low speed the rear of the facility,a conical plug valve, adjustable vanes or
cascade is usually set equal to that found in the turbomachine. other axisymmetric device. When the suction fan is of the cen-
Along with the Mach number, this is the most important of the trifugal variety, experience gained at the Whittle Laboratory
non-dimensional parameter groups or similarity parameters. suggests that these are best operated without a volute or other
Because the density and viscosity are nearly constant at low flow collection system if sufficient power is available. If this is
speed and the velocity ratios are similar to those at high speed, so, the throttle must be placed ahead of the fan.
it is impossible, for example, to match the inlet and exit Rey-
nolds numbers of a blade row and an appropriate (constant) In general, a combination of filters, honeycomb and turbul-
value of density and viscosity must be selected. For the same ence screens is used to remove flow disturbancesat the inlet to
reasons, it is not generally possible to simultaneously match the bellmouth. The facilities at the Whittle Laboratory are
the Reynolds numbers of the rotor and stator of a low speed mounted relatively close to the ground. The above arrange-
rig to those of the high speed machine. ment effectively removes the distortion caused by the ground
vortex so that at inlet to the working section, the resulting cir-
Because the low speed cascades are of larger scale but oper- cumferential variation of the dynamic head is less than one
ate at the correct Reynolds number, the flow velocities and so percent when the hub-tip radius ratio is 0.8. The facilities at
the stresses are reduced so that the cascade may be of a lighter UTRC are horizontal, like those at the Whittle Laboratory
and simpler construction. The power requirements are also whereas at General Electric, the facilities are mounted verti-
substantially less. For the same reasons, probe stems may be cally and will not suffer from the effects of inlet distortion due
of smaller dimension so that there are fewer problems asso- to ground proximity.
ciated with blockage (Ref. 3.1-44). Since the flow velocity is
everywhere much less than at sonic conditions the effects of A bullet is inserted into the bellmouth of each of the rigs in
probe blockage are further reduced. Probe sensing heads, order to provide a parallel annulus upstream of the working
though often no smaller than their high speed counterparts, section. The contraction ratio between the inlet and the test
are also relatively smaller by virtue of the increased diameter section is of the order of 1O:l. The hub may be supported by
of the low speed cascade. Thus, the low speed investigations inlet guide vanes when these are in use but otherwise must be
can be more detailed. It is the authors’ opinion that the supported by streamlined struts. The rig shown in Figure
improved spatial resolution alone justifies the use of large 3.1-7 uses three pairs of struts of 25mm chord to support the
scale rigs in fundamental investigations of turbomachinery weight and the axial thrust of the inlet section of the hub. The
flows. length of the inlet sections may be varied by inserting or rem-
oving hub and casing rings so that different boundary layer
In deciding upon the shape of the low speed blade profiles, it thicknesses can be achieved. Alternatively, trip wires may be
is usual to design for the same normalized velocity distribu- fitted to the annulus or hub (e.g. Ref. 3.1-15).
tion. Under certain conditions, the Prandtl-Glauert
transformation rule (Ref. 3.1-52) or similar (e.g. Ref. 3.1-53) In each of these large scale facilities, the inlet section leads
may be applied but in general these are far too restrictive, even into a parallel working section. At low speed, a parallel annu-
when the flow may be regarded as two dimensional. A boun- lus leads to a constant mean axial velocity through the
dary layer analysis, though limited in accuracy, should be used working section. In general, this will not be correct but it does
to ascertain that equivalent boundary layer behaviour is offer a large degree of flexibility and cost saving since the
achieved in terms of separation, re-attachment and transition. same casings, hubs and traverse systems can be used with
82
Radlal transverse
E
E
m
N
In
Circumferentlal tronsve
h E rollers and kmfe edge
E
s!
In
Fig. 3.1-7 Large-scale 2-stage low speed axial turbine at Whittle Laboratory
EXHAUST
MEASUREMENT
YDRAULIC MOTOR
I '1.''
Fig. 3.1-8 Large-scale 1;-stage low speed axial turbine at UTRC

83
I Flow Direction
.<' ..
-1 \\ f
I
\ylniet
L---
Screen
Bellmouth
Plane 0.1 Bellmouth Flow
Measurement Plane
Plane 0.5 IGV Inlet
Plane 1.0 Rotor 1 Inlet

Plane 1.5 Rotor 1 Exit
Plane 2.0
Plane 4.5
Plane 5.0
Laboratorv Floor
Fig. 3.1-9 Large-scale 4-stage low speed axial compressor (Wisler)

0
many different blade geometries. The flexibility is enhanced taken through the wall of the mould. After the epoxy resin has
by using a modular construction technique for the hub and begun to harden around the twine, the moulding process is
casing. Each stator blade row, for example, may have its own completed. When the resin is fully cured and the blade has
casing and hub ring and may be fitted to the rig as an assembly. been released from the mould, the Nylon twine is removed
On some of the facilities, the casing rings and inlet section are simply by pulling on a free end. Approximately forty-eight
mounted on wheeled camages. Access to the air path is then pressure lines are set into each blade. When the blade shapes
easily achieved by sliding the components along the axis of the are particularly complex, it is often necessary to cast pneu-
rig. Alternative access to the gas path is achieved by using matic tubes directly into the surface (Ref. 3.1-49).
blade cartridges similar to those in use on the high speed facil-
ity (Fig. 3.1-5). Probe traversing systems vary significantlyfrom rig to rig. It is
The blades may be made from a glass reinforced epoxy resin beyond the scope of this Section to review the various meth-
cast from a mould or machined from aluminium alloy. Each ods used to move the probes radially or to yaw them. The
blade is usually mounted on a cylindrical spigot so as to per- techniques used to traverse circumferentially do, however,
mit easy adjustment of the stagger angle and throat areas. The merit a description. Typically, probes need to be traversed
casting technique is preferred at the Whittle Laboratory (Ref. over two or three pitches in the circumferential direction.
3.1-58), partly because of cost but also because a significant Movement of this magnitude is conventionally achieved using
number of static pressure lines can be cast into the blades. a geared mechanism or a mechanism similar to a rack and pin-
These are made by laying Nylon twine which has been dipped ion arrangement. An alternative approach is employed on the
in epoxy resin onto the gelcoat. At least one end of the twine is rig shown in Figure 3.1-7. Here, a toothed belt is wrapped
04
around part of the traverse ring and then passed over a motor- sealed slots are cut into the casings along which the probes
ized pulley being being attached again to the ring as detailed traverse in the circumferential direction. An alternative
in Figure 3.1-10. Similar arrangements are now employed on method of circumferential traversing is used at GE. This is
several rigs at the Whittle Laboratory. also employed by Sieverding et al. (Ref. 3.1-59) and the
arrangement is shown in Figure 3.1-11. Here, the stator row is
moved circumferentially while the probe is traversed radially.
For very different reasons to those given above, Bindon (Ref.
Tension pulley 3.1-60) uses a rapidly rotating casing to provide relative
I
,Drive pulley motion between a stationary cascade and the casing in a study
of tip leakage flows. Whatever the traversing arrangement,
computers should be employed to control the motorized
mechanisms.
In summary, low speed annular cascades are usually of large
scale offering improved spatial resolution, lower costs and
ease of use. At the same time, true dynamic similarity with
high speed counterparts cannot be maintained and additional
compromises are required.
3.1.7 Experimental Considerations

The instrumentation fitted to a cascade may be divided into
the basic instrumentation which is necessary to establish the
Fig. 3.1-10 Circumferential traverse system operating conditions and that used for the investigation in
hand. It is beyond the scope of this Section to review the full
range of instrumentation that can be used in annular cascades.
Much of the instrumentation is common to linear cascade
In the facility shown in Figure 3.1-7, 20mm wide rings are testing. Here, we only consider those aspects which signifi-
mounted between the blade rows and supported by adjustable cantly affect the design of the facility.
wheels. On a more recently constructed facility (Ref. 3.1-49)
The basic instrumentation should include all that is necessary
and on the rotating traverse gear shown in Figure 3.1-7, the
to commission the cascade under investigation. The inlet
ball-races have been replaced by commercially available
boundary layer parameters for the hub and the casing must be
V-notched wheels which locate on a knife-edge on the peri-
determined. It is within this inlet section that the mass flow
phery of the traverse ring to aid axial location. At UTRC,
measurements are usually made. In practice, this requires a
TO A X I A L GIFFUSER
--
Fig. 3.1-11 Test section of annular cascade (Sieverdinget al.)

85
series of radial traverses at different locations around the olane traverse probes which have their sensing heads
annulus in order to check the axisymmetry of the flow field mounted directly on the end of the stem. In a cascade, the abs-
and to determine the mean inlet quantities and the thicknesses ence of a downstream rotor provides the opportunity to
of the hub and casing boundary layers. However, once these locate the radial portion of an L-shaped probe stem suffi-
measurements have been made for a range of flow conditions ciently far downstream that the effects of its blockage are
and compared to at least two reference pressures, it is possible minimized. In the case of high speed cascades, this facility is
to use the reference pressures for this measurement. These particularly important.
reference pressures might be a casing static pressure and a
mean radius total pressure. At inlet, the pressure tappings The use of an L-shape probe introduces new problems which
should be located sufficientlyfar upstream of the cascade that must be addressed at the design stage. An appropriate aper-
the data are not affected by the upstream pressure field of the ture must be provided in the casing, perhaps in the form of a
blade row and not too close to any boundary layer control slot. In the transonic cascade at the Whittle Laboratory, the
devices. slot in the casing is filled by a matching piece on the traverse
gear to which the probe is mounted. The axial location of the
A flattened pitot tube is often used to traverse the inlet boun- traverse plane must be considered since the rigidity of the
dary layer flow, with the static pressures being provided by probe may be reduced. In addition, if the probe faces in any
adjacent surface tappings. The probe shown in Figure 3.1-11 direction other than along the rig axis then it must be curved
has two pitot tubes, one to traverse near to the hub and the to match the annulus if it is to be traversed over the entire
other near to the casing. The rake on the probes, which is to height of the annulus. The curvature of the probe depends
enable the tip to touch the appropriate surface, can lead to upon the pitch-wise angle to which it is set and is unique to
inaccuracies in the measurement of the stagnation pressure that geometry. The forward projection also prohibits its rota-
e'
unless it is first calibrated. The rake may also lead to a deflec-
tion of the probe tip which will change according to the flow
conditions so it is essential that the exact position of the probe
tion about the radial axis without changing the axial location.
Thus, the probe angle must be set to the mean yaw angle of the
flow and the probe must be calibrated over a much greater
tip relative to the wall is known. Since both the inlet section range of yaw angle than would be necessaryfor a simple radial
and the boundary layer probe are usually manufactured from probe. Even when set to the mean yaw angle of the flow, inci-
electrically conducting materials a reliable method to esta- dent flow angles onto the probe may be in excess of 30
blish a datum position is to traverse the probe towards the degrees and many researchers would question the reliability
wall until electrical contact is established. This technique of data acquired under such circumstances.
requires that the probe is electrically insulated from the test
rig. This should be considered essential for all probes. 3.1.8 Concluding Remarks
The pitot traverse and static pressure measurements should The design of an annular cascade rig and the methods that are
be repeated at a number of circumferential locations around adopted for its control and instrumentation must be carefully
the inlet of the annulus to ensure that the flow is axisymmetric. planned according to the primary objective of the research
In addition, a yaw probe must be used to check the flow direc- that is envisaged. There are always compromises to be made.
tion. It is much harder to establish the periodicity of the exit Speed, size, complexity, power supply and instrumentation
flow field, but an indication can be obtained from casing and must all be considered together with the nature of the mea-
hub static pressure tappings and by using surface flow visuali- surements that are required. It is not coincidental that rig
zation. configurations and testing methods vary almost as widely as
The mass flow rate is usually derived from inlet data and this do the flows that are to be investigated and it is therefore
uniquely determines the flow in a low speed cascade of fixed impossible to present a definitive set of design rules for annu-
lar cascades. However, it is hoped that this Section provides
geometry. In high speed cascades, where the flow can be
an insight into the many factors that must be considered.
choked, the mass flow rate can be independent of the pressure
ratio and so it is also necessary to monitor a static pressure
which is representative of the operating Mach number. It 3.2 ROTATING ANNULAR CASCADES
should be noted here that a minimum of instrumentation J. Amecke and E Kost
should be used to monitor the flow conditions once a more 3.2.1 Introduction
thorough investigation has been undertaken to establish the For the experimental and theoretical developmentof bladings
validity of the measurements. for axial flow turbomachines the simulation of the flow field
by a straight cascade is a very helpful tool. But in a test facility
Although many of the problems associated with high speed the number of blades is limited and therefore it is very difficult
annular cascades are shared with their low speed counter- to provide the required periodical flow field. Especially in the
parts, such problems tend to be exaggerated at high speed. transonic and supersonic flow regime this is a serious pro-
Not least is the problem of probe design. The high power blem.
requirement of high speed continuous flow annular cascades
has in practice limited casing diameters to less than about 0.5 Consequently an annular cascade is the test set-up assigned to
metres. Since high speed testing is most applicable to H.P. overcome this deficiency,because it is closed in itself. It repre-
blade rows where aspect ratios tend to be small, the resulting sents the flow in one axisymmetrical stream surface of the
blades may typically have both spans and chords of less than turbomachine (Ref. 3.2-1) and therefore, in contrast to a
30mm (e.g. Ref. 3.1-10). Probes must be correspondingly bucket ring, the following features are typical:
small if resolution and blockage effects are to remain within Constant cross section of the blades (prismatic)
acceptable limits. However, small probes can lead to pro- 0 Straight blades (untwisted)
blems both structurally and in terms of response. 0 Large hub-to-tip ratio ( > O B ) .
In a turbomachine, the close proximity of the rotating and sta- In general the duplicated stream surface is cylindrical,but fac-
tionary blade rows means that it is difficult to avoid using exit ilities with conical walls are also utilised.
86
It is the intention to simulate by the annular cascade only the 3.2.2 Upstream Flow Conditions
relative flow field of the turbomachine. Therefore the curva- The rotating annular cascade requires a homogeneous inlet
ture of the stream surface is not duplicated. Consequently the flow parallel to the axis of the rotor. This can be achieved by a
annular cascade is defined by the same parameters as the standard wind-tunnel installation assembled from settling
straight cascade: chamber, flow straightener, screen(s) and an optimised short
Profile shape contraction. The latter is very important in order to provide a
Pitch-chord ratio minimal boundary layer thickness at the inlet of the cascade.
Stagger angle. In contrast to this, the fixed annular cascade facility demands
In general a well designed wind-tunnel provides a parallel and a pre-swirl generator or an inlet guide vane either adjustable
non-rotating flow field upstream of the model. This permits for the range of inlet angles according to the envisaged
the investigation of inlet guide vanes (inlet flow angle = operating range. The following effects of this installation must
90") without any additional measures. But a non-axial inlet be accounted for:
flow has to be established either by a pre-swirl generator (=
jixed annular cascade) or by rotation of the cascade (= rotat- Wake Disturbances
ing annular cascade). The distance from the cascade, the number and the circum-
ferential location of the guide vanes must be arranged in a
In both cases the same relative flow field is simulated, but the way that the wake disturbances do not have any influence
absolute flow vectors are basically different. This becomes on the cascade measurements.Therefore a reasonable dis-
evident by the velocity triangles (Fig. 3.2-1). tance upstream of the cascade should be kept. An
increasing of the path length from theinlet guide vane to the
cascade unavoidably increases the boundary layer dis-
placement thickness too. 0
Radial Equilibrium
The balance of the radial pressure gradient and the centri-
fugal force in a rotating flow causes an increase of the static
pressure from the inner to the outer wall. Therefore, with
constant total pressure, the Mach number will decrease and
the flow angle will increase from the inner wall towards the
outer.
Stability
A rotating flow field between coaxial cylindrical surfaces is
instable against axisymmetrical disturbances. These dis-
turbances generate the well-known Taylor ring vortices. In
a helical flow field - as considered here -the destabilising
effect is forced by the radial gradient of the axial velocity
(Ref. 3.2-2).
_..a
-../ For a fixed annular cascade with a choked turbine cascade

the stability criterion has been investigated in (Ref. 3.2-3)
for transonic turbine cascades. The result suggests that the
range of instability can well coincide with the range of prac-
0
tical important turbomachinery parameters.
All disturbing effects listed above are functions of the
required inlet angle (turning of inlet flow). This leads to the
conclusion that the fixed annular cascade can be a useful test
Relative Absolute Circumferential set-up if no or only small turning of the inlet flow is requested.
-
- --- ___ The application of a rotating annular cascade test set-up is
- -........-
Turbomachine
-........-
Rotating annular cascade mandatory if high turning of the inlet flow is required.
Fixed annular cascade
On the other hand it must be pointed out that the technical
expenditure to run a facility for rotating cascades is a severe
penalty for this type of tests because an additional rotor
Fig. 3.2-1 Comparison of the velocity triangles including a high accuracy speed control system is necessary,
besides the complete wind-tunnel installation.
Measurements at the blade surface in a rotating annular cas-
An annular cascade test section can operate as part of an cade are much more complicated than in a fixed annular
arbitrary whd-tunnd circuit.This may be a closed or an open cascade and special devices for data transmission from the
loop, a continuous or intermittent system respectively. But the rotating to the fixed frame have to be provided for this.
available measuring time has fundamental influence on the The power output of the rotor is determined by the momen-
required measuring equipment. In general it may be con- tum equation. Considering the non-rotating inlet flow field in
cluded that the expenditure in instrumenting increases with the rotating cascade facility,the work per unit flow rate can be
decreasing testing time. calculated from the circumferentialvelocity of the rotor U and
87
the circumferential component of the absolute outlet velocity the rotational speed and the radius in general are different
vzu: from that in the corresponding turbomachine.
AW=-U ' V2, As a consequence of the rotation the centrifugal force and the
(3.2.1) Coriolis force have to be taken into account. For constant
rotational speed about a fixed axis (angular velocity vectors)
the following relation between the total change of the absolute
This equation shows that the rotor must be braked ( A W >O) velocity vector $and the relative velocity vector 'w holds (Ref.
if U and v2, have opposite directions and it must be driven 3.2-4):
( AW <0) if U and v2" have the same direction.
The transformation into the relative frame yields the loading _ - -D e
D?
+6x ($ x 7) 2 + - 6x G
coefficient as a function of the relative outlet velocity w2 and Dt Dt
the relative outlet angle f12 (Fig. 3.2-2): (3.2.3)
-A -
W w2 The second term on the right hand side represents the centri-
- - (1 + - - c o s p 2 )
U* U (3.2.2) fugal acceleration, induced by the rotation of the relative
frame. It is a vector perpendicular to the axis:
1.5 2.50
2.25 - U*
acenlrifugal --r (3.2.4)
1 .o 2.00 ,
1.75 \
3" The third term on the right hand side is known as Coriolis
.5 1.50 >-
t acceleration:
1.25 5
w
n
.o 1.00 =- a,,. = 2.-.
w
a
.75
I-
2 oriolis r (3.2.5)
\ 3
5 -.5 SO
c
Z .25
w For the investigation of cylindrical stream surfaces the radial
:-1.0
w
LL
.oo component of the velocity w, can be neglected and with the
0 circumferential component of the velocity w, only this equa-
0
tion becomes:
$
- -1.5
0
-x U W"
s a, . .= 2 . - . w "
oriolis
r
= 2.-.
U
acentrifugal
(3.2.6)
-2.0
Under this condition the Coriolis acceleration is also a vector

-2.5
perpendicular to the axis.
The effect concerning the flow field under consideration is
-3.0 two-fold (Ref. 3.2-1):
1. Boundary Layer
-3.5 In a rotating flow field the centrifugal force and the Coriolis
0 30 60 90 '120 150 180
OUTLET ANGLE 8 2
force are balanced by the radial pressure distribution. But
in the boundary layer this balance is disturbed by the pres-
Figure 3.2-2 Loading coefficient of a rotating annular sure distribution impressed by the suKounding flow field.
cascade In a guide vane consequently the particles in the boundary
AW AW layer follow the gradient of the pressure field, because their
(- >0: Braking -
u2 <0: Driving) absolute velocity is very small. This entails an increase of
U*
the boundary layer in the hub region.
Braking of the rotor is required for turbine rotor cascades On the other hand in a rotor a drifting of the boundary layer
towards the blade tip is observed. This is generated by the
designed for a high degree of reaction (typically R > O S ) and
in general for rotor cascades of axial flow compressors. fact that the relative velocity of the particles is small and the
guiding motion is dominated by the absolute velocity.
These effects of the centrifugal acceleration and the Corio-
3.2.3 Secondary Flow lis acceleration increase with the length of the blade.
In a linear cascade the secondary flow is mainly caused by the Therefore it may be concluded that the above-mentioned
blade-to-blade pressure gradient acting on low momentum effects are only small in a facility for rotating cascades with
fluid. Against that in a rotating annular cascade the secondary a large hub-to-tip ratio as defined.
flow is governed by the radial pressure gradient induced by In a rotating cascade the motion of the blade row relative to
the rotation. the annulus wall boundary layer leads to skewed endwall
But it must be remembered that a facility for rotating annular boundary layers thereby also influencing secondary flow
cascades simulates only the relative flow field and therefore development.
88
2. Potential Flow has to be proved prior to the tests.

In general a potential function is defined by the continuity
The transformation of the flow parameters from the absolute
equation and the condition of non-rotation:
frame into the relative frame is obtained with the aid of the cir-
div?= 0 rot;= 0 (3.2.7) cumferential velocity at mid-height of the blades.
The velocities of the absolute and the relative frame are Because of the large hub-to-tip ratio.- required to ensure a
related by: two-dimensional flow field - specific measuring problems
;=;+;iX'; (3.2.8) appear at the downstream end of the cascade. Therefore spe-
cial evaluation methods have been developed by Rhomberg
Hence (Ref. 3.2-7) and H. Meyer (Ref. 3.2-8) at ABB, Baden
div?=div$ rot;= rot w +2 "w (3.2.9) respectively, by Ludwieg and his co-workers J. Meyer (Ref.
The continuity equation is satisfied for the absolute and the 3.2-9 and 3.2-10) and Heinemann (Ref. 3.2-11) at DLR, Got-
relative frame. But if the absoluteflow is non-rotational, the tingen. The latter conducted a systematic treatment of the
relative flow has a constant rotation of -2 o. Conse- - area.
Based on wake measurementsin different planes downstream
quently the relative flow is no potential flow.
of a straight cascade J. Meyer calculated the increasing of the
Special attention must be paid to minimise the tip leakage entropy. About two pitches in the axial direction downstream
flow because in an annular cascade with a large hub-to-tip of the cascade the flow is nearly uniform and at least 90% of
ratio this secondary flow can seriously affect the proper simu- the losses have been developed. Consequently for a fixed sta-
lation of a two-dimensional flow field. tionary probe this should be the minimum distance of the
If a rotating annular cascade is investigated with conical flow measuring plane.
surfaces it is necessary to re-examine the secondary flow, On the other hand at this distance the displacement thickness
since the conical flow through the blading of an axial turbine is of the boundary layer at the outer and inner wall of the casing
similar to the flow past an aircraft wing with sweep and dihe- cannot be disregarded. Assuming that the boundary layers
dral. This type of flow leads to a significant amount of stream from both walls have not grown together and therefore an
surface twist; so any assumption of an axisymmetrical stream unaffected flow field remains in the centre, Heinemann has
surface is seriously violated (Ref. 3.2-5 and Ref. 3.2-6). worked out an evaluation method for rotating annular cas-
cades. He presumed furthermore that radial components of
3.2.4 Measuring the Evaluation of the Cascade Data the velocity vectors and heat transfer at the bIades can be neg-
The flow field upstream of the cascade is defined by the mea- lected.
sured values in the absolute frame according to Table 3.2-1.
In general the definition of the homogeneous flow field down-
The inlet angle in the absolute frame is a, = 90' as stipulated. stream of a cascade requires four independent values. But
But the validity of this stipulation in the individual test facility only the properties according to Table 3.2-2 are measured.
Table 3.2-1
Upstream measured properties
Measured property Technique Location

Total temperature ,T
, Probe Settling chamber
Total pressure , ,p Probe Settling chamber
Static pressure p, Wall taps Inlet duct
Circumferential velocity U Counter for RPM Rotor shaft
Table 3.2-2
Downstreammeasured properties
Measured property
power^
Mass flow m
Total pressure po,
I 11;: 1
Technique
Venturi nozzle
Location
Rotor shaft
Wind tunnel circuit
Center of outlet duct
1
J. Meyer I H.-J. Heinemann (DLR Gottingen):
Tota I temperature ,T Probe Center of outlet duct
Total pressure po, Probe Center of outlet duct
89
Under the stated assumptions, these values should be unaf- All other values such as Mach number, velocity and loss coef-
fected by the boundary layer development. ficient can be calculated from by standard relations.
The two lacking values will be provided by the momentum The reading of the total temperature must be corrected for the
equation (Euler's turbine equation) and the continuity equa- recovery factor and the total pressure with respect to the bow
tion. This system of equations will be solved for a perfect gas shock (supersonic flow,only). Both corrections are functions
with constant specific heat defined by the gas constant R and of the actual Mach number obtained from static pressure
the ratio of specific heats K. measured at the wall.
In the ABB method the outlet total temperature is calculated The ABB method delivers average properties for the total
from shaft power and mass flow: blade height - not only for the mid-section as intended. Its
accuracy is limited by parasitical losses as bearing friction and
K - 1 1 P ventilation at shaft and disc.
To, = Ton - -* - * - The DLR method presents under the stated assumptions
K R m (3.2.10)
information about the mid section of the blade only. But as
analysed by Heinemann (Ref. 3.2-11) and Braunling (Ref.
In the DLR method the outlet total temperature is measured 3.2-12) this evaluation method can be hampered by the res-
directly. On this basis the circumferential component of the tricted measuring accuracy. This applies especially to the
absolute outlet velocity v2" can be calculated with the help of temperature, because the difference through the rotating cas-
the momentum equation: cade is in general small, and consequently, even small
measuring errors entail large discrepancies in the downstream
K R flow values.
V2" = - * - . (Tovl- To,) (3.2.11)
K - 1 U With respect to power requirements the cross section of most
test sections is relatively small and the blockage of the flow
The procedure becomes more comprehensible if the total field by the heads of the probes downstream of the cascade
temperature and the total pressure are based on the axial must be observed carefully. In continuous working facilities
velocity component vZa. Consequently we obtain by applica- this problem can be handled by the installation of minimised
tion of the energy equation: single purpose probes for angle, total pressure and total tem-
perature in different circumferential positions of one plane
downstream of the cascade. During the test the actuator of the
angle probe is adjusted to the flow direction by balancing the
pressure difference on the wedge and the other probe actua-
tors are turned parallel.
3.2.5 Test Facilities

With the aid of the relation for isentropic change of state we It must be remembered that in general the rotating annular
get from this equation: cascade is more a testing procedure than a specimen of facilit-
ies. More or less every test facility for axial turbomachine
stages may be utilised for this purpose.
Po,. - Po, . (3) (3.2.13) In the following some facilities are listed, which have been
applied for rotating annular cascade tests.
0 For the continuity equation holds with the densities p1 and p2

ABB (BBC) Baden, Switzerland
This test rig became first known for the test procedure consid-
of the inlet and outlet flow respectively: ered in this section (Ref. 3.2-7 and Ref. 3.2-8). It is installed in
a closed circuit and powered by a compressor for continuous
operation. The rotor speed is controlled by a brake, which
also allows measuring of the torque. Therefore only cascades
with power output according to Figure 3.2-2 can be investi-
Inserted into the relation of the dimensionless mass flow rate gated.
based on the stagnation point values we obtain:
The characteristic data are (Fig. 3.2-3):
Mean diameter D, = 350 mm
Blade height h - 30 mm
Hub-to-tip ratio D,/D, = 0.842
Rotational speed %ax
-
- 16 000 RPM
Circumferentialvelocity U, - 293 m/s
By applying the equation of state of the perfect gas, the equa-

tion for the velocity of sound and the isentropic relations we
finally receive:
(3.2.16)
90
91
The pressure ratio of the compressor allows for high super- compressor research. The main purpose was the visualisation
sonic downstream Mach numbers (Ma, >1.7). The data of the flow field upstream and inside of the cascade; therefore
acquisition and evaluation method is developed for measur- the hub-to-tip ratio is rather high. A stator is installed down-
ing the homogeneous upstream and downstream flow stream of the rotor.
properties only. The circuit is closed and Freon 114is used for testing in order
A special feature is a shadowgraph system installed down- to keep the power low. The rotor is driven and adjusted to the
stream of the cascade. desired speed by an electric motor. An additional compressor
The facility was installed and applied mainly for the investiga- is not installed in the circuit.
tion of the last blade rows of low-pressure steam turbines. The pressure level in the circuit is adjustable. The tempera-
ture is set to a constant value by a heat exchanger.Therefore it
DLR Gottingen, Germany is possible to adjust the Mach number and the Reynolds
The DLR wind-tunnel for rotating cascades (Ref. 3.2-10 and number in a limited range independent of each other.
Ref. 3.2-11)was in particular designed and built for the testing The characteristic data are:
of annular cascades for turbines and compressors in a wide
range of parameters. Nevertheless an essential design goal Mean diameter D, = 455 mm
was to create a facility which is as versatile as possible. Blade height h -
- 10 mm
Hub-to-tip ratio Dh/Dt = 0.957
The test section is part of a closed circuit (Fig. 3.2-4) for con- Rotational speed n = 3 000 ...6 500 RPM
tinuous operation with adjustable pressure and temperature Circumferentialvelocity U - 71 ... 155 m/s
level for independent variation of Mach and Reynolds num- Inlet pressure Po1
-
- 10...20 kPa
ber. Inlet temperature To1 = 320 K
The circuit is powered by a 4-stage radial compressor with a Blade Reynolds number Re = 106
variable speed electric motor and equipped with an addi- Freon mass flow "ax - 1 kg/s
tional bypass valve. This renders possible an operation in the Driving power Pmax
- 40 kW
complete performance graph including the surge range and Because Freon 114 has a low velocity of sound (130m/s) only
the critical speed range of the compressor and quick unload- low rotational speed is required.
ing of the test wheel for the emergency case. Downstream of
the compressor a water cooled heat exchanger is installed in The flow visualisation is obtained by means of a laser con-
order to maintain a constant temperature in the system. trolled Schlieren system using polished inner walls of two
blade channels and the upstream part of the rotor for beam
The pressure in the system is controlled by a combination of a reflection (Ref. 3.2-15). A special optical system with a spher-
vacuum pump and a high pressure air feed. In general the sys- ico-cylindrical lens adapted to the visualisation window and
tem pressure is below ambient pressure. The humidity of the the blade-carrying hub was designed for this test facility.
air leaking into the tunnel is removed by a bypass dryer instal-
lation. The inlet flow conditions are evaluated from the total pres-
sure and temperature upstream of the cascade and the static
The speed-control of the test rotor is provided by an electric pressure measured at the casing wall. Probe measurements
motodgenerator, which allows driving and braking in both are not performed with respect to the narrow flow channel.
directions of turning. The test section is shown in Figure 3.2-
5. The flow enters the test section via the settling chamber. NAL Bangalore, India
The annular flow channel is shaped from exchangeable liner This facility (Ref. 3.2-16) is designed for a similar application
rings. to the ABB and the DLR installations.
The characteristic data are: The rotational speed is controlled by a dynamometer. This
Mean diameter D, = 512 mm allows the testing of cascades with power output only.
Blade height h 36 mm The characteristic data are:
Hub-to-tip ratio Dh/Dt = 0.869
Rotational speed %ax
-
- 14 500 RPM Mean diameter D, = 320 mm
- Blade height h - 26 mm
Circumferential velocity U, - 402 m/s
Power of the test rotor P, -
- 500 kW Hub-to-tip ratio Dh/D, = 0.850
- n = 15 000 ...26000 RPM
Inlet pressure Po1 - 10 ...150 kPa Rotational speed
251 ...436 m/s
Inlet temperature To1 = 295 ...335 K Circumferential velocity U =
Flow rate Vmax

-
- 15.5 m3/s Inlet pressure Po1 - 100 ... 700 kPa
Compressor pressure ratio n = 6 Inlet temperature
Flow rate
Tol
Vmax =
= 300 ... 570 K
1.2 m3/s
During the tests, the casing above the rotor is provided with an Dynamometer load '
m
a
,
- 1200 kW
abradable material in order to achieve minimal tip clearance.
The local values of total pressure, total temperature, static
The circuit was originally designed for the operation with air
pressure and flow angle upstream and downstream of the cas-
and Freon 12 in order to extend the application range and to
cade are measured by radial traversing of combined wedge
reduce the power requirements.But with respect to pollution
probes. For determining the flow angle the probe is adjusted
considerations all equipment for Freon 12 operation has to the flow direction by balancing the pressure difference at
recently been removed.
the wedge.
ONERA Chatillon, France Special attention was paid to the measurement of the static
This rotating annular cascade facility (FA20) (Ref. 3.2-13 and pressures near the wall by correcting the wedge probe read-
Ref. 3.2-14) was designed with regard to supersonic axial ings for wall proximity effects.
92
HP-System ( , 7 O B A R )
MAIN COMPRESSOR
( 1 -STAGE RADIAL 1
P :1000 KW
VACUUM PUMP
Fig. 3.2-4 Circuit of the DLR wind-tunnel for rotating cascades
Fig. 3.2-5 Test section of the DLR wind-tunnel for rotating cascades
93
3.2.6 Concluding Remarks 3.1-7 Hill JM,Lewis RI

Comparative experiments in a straight cascade and a rotating Experimental Investigations of Strongly Swept Tur-
annular cascade facility have been presented by Heinemann bine Cascades with Low Speed Flow, ASME J.
(Ref. 3.2-11) for two typical turbine cascades. The conclusion Mechanical Engineering Science, Vol 16, No 1,
is a reasonable agreement of the results under the stipulation 1974.
that the location of the measurement in the rotating annular 3.1-8 Camus J-J, Denton JD, Soulis JV,Scrivener CTJ
cascade has to be defined carefully in accordance with the An Experimental and ComputationalStudy of Tmns-
guide lines established (Ref. 3.2-9 and Ref. 3.2-11). onic Three-Dimensional Flow in a Turbine Cascade,
Another very thorough comparison by Paulon et al. (Ref. 3.2- ASME J. Eng. for Gas Turbines and Power, Vol106,
13) comprises tests with a supersonic compressor blading in a 1984.
straight cascade facility as well as in either a fixed and a 3.1-9 Gostelow JP
rotating annular cascade facility. The comparison of the Cascade Aerodynamics,Pergamon, pp 77-81,1984.
results is restricted by some differences in the test set-up and
the instrumentation. Finally, it must be concluded that the 3.1-10 Dominy RG
results integrate each other well but up to now no type of test- The Whittle Laboratory Transonic Annular Cas-
ing is able to replace all others completely. cade Wind Tunnel 9th Symp. on Measuring
Techniques for Transonic and Supersonic Flow in
The above compilation was mainly concentrated on annular Cascades and Turbomachines,Oxford, UK,1988.
cascade facilities and test procedures simulating cylindrical
stream surfaces.But the application of non-cylindrical stream 3.1-11 Dunn MG
surfaces has already been indicated. As an example the simu- Phase and Time-ResolvedMeasurementsof Unsteady
lation of conical stream surfaces in the DLR facility by Heat Transfer and Pressure in a Full-Stage Rotating
Braunling (Ref. 3.2-12) must be mentioned. But it must be Turbine,ASME paper 89-GT-135, 1989.
stated that these tests raised several additional problems con- 3.1-12 Epstein AH, Guenette GR, Norton RJG
ceming the proper interpretation of the results. The MIT Blowdown Turbine Facility, ASME Paper
It was already mentioned above that a rotating annular cas- NO84-GT-116,1985.
cade may be a useful tool for the investigation of the 3.1-13 Brooks AJ, Colboume DE, Wedlake ET, Jones TV,
interaction between centrifugal forces (includhg Coriolis for- Oldfield MLG, Schultz DL, Loftus PJ
ces) and secondary flows. For this the combination of a The Isentropic Light Piston Annular Cascade Facility
rotating annular cascade with an adjustable pre-swirl genera- at RAE Pyestock,AGARD CP 390, Paper 31,1985.
tor should be the ideal test installation,because it would allow
3.1-14 Hodson HF',Dominy RG
the independent adjustment of the absolute flow field and the
Three-DimensionalFlow in a Low Pressure Turbine
relative flow field or - in other words - the investigation of a
Cascade at its Design Condition,A S M E J. Turboma-
flow field in the relative frame exposed to different radial for-
chinery, pp 177-185, April 1987.
ces.
3.1-15 Hunter IH
REFERENCES Endwall Boundary Layer Flows and Losses in an
3.1-1 Langston U, Nice ML, Hooper RM Axial Turbine Stage, ASME J. of Engineering for
Three Dimensional Flow within a Turbine Cascade Power, Vol 104, pp 184-193,1982.
Passage, ASME J. of Engineering for Power, Vol99, 3.1-16 Hamson S
NO 1, pp 21-28,1977. Secondary Loss Generation in a Linear Cascade of
3.1-2 Yaras M, Sjolander SA High-Turning Turbine Blades, ASME paper 89-GT-
Development of the Tip-Leakage Flow Downstream 47,1989.
of a Planar Cascade of TurbineBlades: VorticityField, 3.1-17 Ilyas M, Norbury JF
ASME Paper No 89-GT-55,1989. Effect of Axial Velocity Variation on the Subsonic
3.1-3 Kiock R, Lehthaus F, Baines NC, Sieverding CH Flow through a CompressorCascade, I Mech E Conf.
The TransonicFlow through a Plane Turbine Cascade Pub1 3, p 276,1973.
as Measured in Four European Wind Tunnels, 3.1-18 Schlichting H
ASME J. Engineering for Gas Turbines and Power, Boundary-Layer Theory, McGraw-Hill, 7th Edition,
Vol 108,1986. pp 536-542. Translated by Dr J Kestin.
3.1-4 Hoheisel H, Kiock R,Lichtfuss HJ,Fottner L 3.1-19 Hunter I, Cumpsty NA
Influence of Free-Stream Turbulenceand Blade Pres- Casing Wall Boundary-Layer Development through
sure Gradient on Boundary Layer and Loss an Isolated Compressor Rotor, Trans ASME J. of
Behaviour of Turbine Cascades, ASME Paper No Engineering for Power, Vol 104, pp 805-817,1982.
86-GT-234,1986. 3.1-20 Gallus HE, Grollius H, Lambert2 J
3.1-5 Lewis RI, Fisher EH, Saviolakis A The Influence of Blade Number Ratio and Blade Row
Analysis of Mixed-Flow Rotor Cascades, ARC R & Spacing on Axial Flow Compressor Stator Dynamic
M 3703,1971. Load and Stage Sound Pressure Level, ASME paper
3.1-6 Breugelmans FAH, Carels Y, Demuth M 81-GT- 165,1981.
Influence of Dihedml on the Secondary Flow in a 3.1-21 Bolcs A
Two-Dimensional Compressor Cascade, ASME J. of A Test Facility for the Investigation of Steady and
Engineering for Gas Turbines and Power, Vol 106, Unsteady Transonic Flows in Annular Cascades,
pp 578-584,1984. ASME Paper No 83-GT-34,1983.
94
3.1-22 Moses HL,OBrien WF Jnr Turbine Stator Vanes, AGARD CP 214, Secondary
The Chord-Wise Pressure Distribution on a Rotat- Flows in Turbomachines,1977.
ing Axial Flow Compressor Blade, Presented at 2nd
Int. Symp. on Air Breathing Engines, Sheffield, 3.1-36 Williamson RG, Moustapha SH
March 1974. Annular Cascade Testing of Turbine Nozzles at High
3.1-23 Lopatitskii AO, et al. Exit Mach Numbers, ASME J. of Fluids Engineer-
Energy Losses in the Transient State of a Incident ing, Vol 108, No 3, pp 313-320,1986.
Flow on the Moving Blades of Turbine Stages, Ener- 3.1-37 Govardhan M, Venkatrayulu N, Prithvi Rai D
gomashinastroenie, Vol 15, August 1969. Secondary Losses in a Large DeflectionAnnular Tur-
3.1-24 Doorly DJ, Oldfield MLG bine Cascade: Effect of the Entry Boundary Layer
Simulation of Wake Passing in a Stationary Turbine Thickness,ASME Paper No 86-GT-171,1986.
Rotor Cascade,J. Propulsion and Power, Vol 1,No 4, 3.1-38 Sjolander SA
pp 316-318,1985. The Endwall Boundary Layer in an Annular Cascade
3.1-25 Hodson HP of Turbine Nozzle Guide Vanes,Carleton University,
Boundary Layer and Loss Measurements on the TR MEIA 75-4, Dec 1975.
Rotor of an Axial Flow Turbine, ASME J. of Engi- 3.1-39 Dominy RG, Harding SC
neering for Gas Turbines and Power, Vol 106, April An Investigation of Secondary Flows in Nozzle
1984. Guide Vanes, Paper No. 7, AGARD CP 469, Secon-
3.1-26 Yurinskiy VT, Shestachenko IY dary Flows in Turbomachines, 1989.
Losses in an Impulse Turbine Cascade in an Unsteady
Flow, Fluid Mechanics - Soviet Research, Vol3, No
1,Jan-Feb 1974.
3.1-40 Boletis E, Sieverding CH
Experimental Study of the Flow Field behind an
Annular Turbine Nozzle Guide Vane with and with-
a
3.1-27 Baines WD,Peterson EG out Downstream Rotor, ASME Paper No 84-GT-15,
A n Investigation of Flow through Screens, Trans. 1984.
ASME 73, p 467,1951. 3.1-41 Bammert K, Klaukens H
3.1-28 Greitzer EM Nabentotuasserhinter Leitradern von axialen Strom-
Thermodynamicsand Fluid Mechanics of Turboma- ungsmaschinen, Ingenieur-Archiv, Band Xvn, pp
chinery, NATO Advanced Study Institute, Turkey, 367-390,1949.
Chapter 8, Sept 17-28,1984. 3.1-42 Williamson RG, Moustapha SH, Huot JP
3.1-29 Camck HB The Effect of a Downstream Rotor on the Measured
Secondary Flows and Losses in Tbrbine Cascades Performance of a Transonic Turbine Nozzle, ASME
with Inlet Skew AGARD CP 214, Secondary Flows J. of Turbomachinery, Vol 108, No 2, pp 269-274,
in Turbomachines,Paper No 9,1977. 1986.
3.1-30 Boletis E, Sieverding CH, Van Hove W 3.1-43 S~UXTA
Effects of a Skewed Inlet Endwall Boundary Layer A Computational and Experimental Study of Fully
on the 3-D Flow Field in an Annular Turbine Cas- 3 0 Transonic Flow in Turbomachinery,PhD Thesis,
cade, Paper No 16,AGARD CP 315, Vkcous Effects University of Southampton, March 1980.
in Turbomachines, 1983. 3.1-44 Squire LC
3.1-31 Bindon JP
The Effect of Hub Inlet Boundary Layer Skewing on
the Endwall Shear Flow in an Annular Turbine Cas-
Effects of Probe Supports on Measurements in Steam
Turbines,ASME Paper No 86-GT-213,1986.
3.1-45 Epstein AH, Guenette GR, Norton RJG
a
cade, ASME Paper No 79-GT-13,1979. The MIT Blowdown Turbine Facility, ASME Paper
3.1-32 Schultz HD, Gallus HE NO86-GT-116,1986.
Experimental Investigation of the Influence of Rotor 3.1-46 Gostelow JP, Watson PJ
Wakes on the Development of the Profile Boundary A Closed Circuit kriable DensityAir Supplyfor Tur-
Layer and the Performance of an Annular Compres- bomachinery Research,ASME Paper No 76-GT-62,
sor Cascade, Paper No 21, AGARD CP 468, 1976.
Unsteady Aerodynamic Phenomena in Turboma- 3.1-47 Baines NC, Oldfield MLG, Jones TV,Schultz DL,
chines, 1989. King PI, Daniels LC
3.1-33 Haas JE A Short-Duration Blowdown Tunnel for Aerody-
Analytical and Experimental Investigation of Stator namic Studies on Gas TurbineBlading,ASME Paper
Endwall Contouring in a Small Axial-Flow Turbine, NO82-GT-312,1982.
NASA Technical paper 2023, AVRADCOM Tech- 3.1-48 Wisler DC t
nical Report 82-C-4,1982. Loss Reduction in Axial-Flow Compressors through
3.1-34 Yamamoto A, Janagi R Low-Speed Model Testing, ASME Paper No 84-GT- I
Production and Development of Secondary Flows 184,1984.

and Losses within a Three-DimensionalTurbineSta- 3.1-49 Huntsman I, Hodson HP, H ill SH
tor Cascade,ASME paper 85-GT-217,1985.
The Design and Testing of a Radial Flow Turbinefor
3.1-35 Goldman LJ, McLallin KL Aerodynamic Research, ASME Paper No 91-GT- I
Effect of Endwall Cooling on Secondary Flow in 220,1991.
95
3.1-50 Inoue M, Cumpsty NA 3.2-4 Vavra MH

Experimental Study of Centrifugal Impeller Dis- Aero-Thermodynamicsand Flow in Turbomachines,
charge Flow in Vanelessand Vaned Diffusers, ASME John Wiley & Sons, New York, London, 1960.
J. Gas Turbines and Power, Vol 106, pp 445-467,
1984. 3.2-5 Potts I
3.1-51 Joslyn HD, Brasz JJ, Dring RP The Importance of S-1 Stream Surface Twist in the
Centrifugal Compressor Impeller Aerodynamics (an Analysis of Inviscid Flow through Swept Linear Tur-
Experimental Investigation), ASME paper No 90- bine Cascades, Proceedings of IMechE Conference:
GT-128,1990. Turbomachinery - Efficiency Prediction and
Improvement, pp 231-248, Cambridge, Sept. 1-3,
3.1-52 Glauert H 1987.
A Theory of Thin Aerofoils, ARC R&M 910,1974.
3.2-6 Kost F
3.1-53 Lakomy C Three Dimensional Transonic Flow Measurements in
The Calculation of the Flow of the Compressor Cas- an Axial Turbine with Conical Walls, ASME Paper
cade at High Subsonic Speeds, Strojnicky Casopis 92-GT-61,1992.
Xvn, cl, 1965.
3.2-7 Rhomberg F
3.1-54 Deverson EC, Marsh H, Oxford JTB Investigations into Rotating Blade Cascades for
The Performance of an Aerofoil in Stationary and Transonic Flow, Brown Boveri, Rev. 51, pp 762-773,
Rotating Cascades, Cambridge University Internal 1964.
Report, 1967.
3.2-8 Meyer H
a Dong Y, Gallimore SJ, Hodson HP
3*1-55 Three-Dimensional Flows and Loss Reduction in
Axial Compressors, ASME Paper No 86-GT-193,
Transonic Flow in the Last Rotor Blade Row of a
Low-Pressure Steam Turbine, Proceedings of
IMechE Conference: Internal Aerodynamics (Tur-
1986. bomachinery), Cambridge, pp 83-94, July 19-21,
3.1-56 McDougall NM, Cumpsty NA, Hynes TP 1967.
Stall Inception in Axial Compressors, ASME J. of 3.2-9 Meyer J
Turbomachinery, Vol 112, pp 116-125,Jan 1990. Theoreticaland Experimental Investigationsof Flow
3.1-57 Dring RP, Blair MF, Joslyn HD, Power GD, Verdon Downstream of Two-Dimensional Transonic Turbine
JM Cascades, ASME Paper 72-GT-43,1972.
The Effects of Inlet Turbulence and Rotor Stator 3.2-10 Meyer J
Interactions on the Aerodynamics and Heat Transfer Untersuchungder Stromung hinter ebenen transson-
of a Large-scale Rotating Turbine Model, Vol I - ischen Turbinenschaufelgittern, BWK 25, pp
Final Report, NASA CR 4079, UTRC-R86- 240-247,1973.
956480-1,1986.
3.2-11 Heinemann H-J
3.1-58 Gregory-Smith DG, Marsh H Ein neues Me& und Auswerteverfahren fiir
The Manufacture of Glass Fibre Rotor Blades with Untersuchungen an rotierenden Ringgittern, VDI-
Pressure Tappings, Aero J. Vol 75, pp 213-215, Forschungsheft 594, pp 1-31,1979.
March 1971.
3.2-12 Braunling W
3.1-59 Sieverding CH, Van Hove W, Boletis E Untersuchungen zum Einji’ufl der Konizitat auf die
a Experimental Study of the Three Dimensional Flow
Field in an Annular Turbine Nozzle Guidevane,
ASME Paper No 83-GT-120,1983.
Kenn werte rotierender Turbinen-Ringgitter im
transsonischen Geschwindigkeitsbereich, VDI-
Forschungsheft 627, pp 1-56,1985.
3.1-60 Morphis G, Bindon JP 3.2-13 Paulon J, Reboux J, Sovrano R
The Effects of Relative Motion, Blade Edge Radius Comparison of Test Results Obtained on Plane and
and Gap Size on the Blade Tip Pressure Distribution Annular, Fixed or Rotating Supersonic Blade Cas-
in an Annular Turbine Cascade with Clearance, cades,J. Eng. for Power 97, pp 245-253,1975.
ASME Paper No 88-GT-256,1988.
3.2-14 Fabri J, Paulon J
Unsteady Phenomena in Turbomachines, as
Revealed by Visualizations and Measurements,
AGARD - CP 177, pp 7/11-12,1976.
3.2-15 Philbert M, Fertin G
3.2-1 Scholz N SchlierenSystemsfor Flow Visualizationin Axial and
Aerodynamik der Schaufelgitter, Verlag G. Braun, Radial Flow Compressors,J. Eng. for Power 97, pp
Karlsruhe, 1965. 254-260,1975.
3.2-2 Ludwieg H 3.2-16 Murthy MVA, Soundranayagam S,Murthy SJK
Stabilitat der Stromung in einem zylindrischen Experimental Determination of the Relative Pressure
Ringraum, Z. Flugwiss. 8, pp 135-140,1960. Field at the Tip of a High Speed Turbine Rotor Blade
3.2-3 Amecke J Row, Proceedings of First International Symposium
Stabilitat der spiraligen Zustromung eines sperrenden on Experimental and Computational Aerothermod-
Ringgitters, DLR Internal Report IB 222-91 A 03, ynamics of Internal Flows, Beijing, China, July 8-12,
1991. 1990.
96
Chapter 4
Particular Aspects of Hot Cascades
C.T.J. Scrivener
4.1 INTRODUCTION When evaluating metal temperatures a choice has to be made

Most major manufacturers of gas turbines have hot cascade whether to carry out the evaluation in a static rig or in a full
facilities of some form. It is not very useful to describe each engine or demonstrator. The engine environment produces a
individual facility in detail since they are generally dedicated complex flow structure within the aerofoil passage resulting
to proprietary work. Instead this section describes those par- from turbulence generated by the combustion chamber in the
ticular aspects and characteristics of hot cascades which combustion process, nozzleguide vane wakes, rotation for the
differentiate them from more conventional cascade facilities. rotor blade and passage secondary flows. The engine effects
Hot cascade facilities are generally more complex, more that are difficult to reproduce in conventional cascade tests
expensive and more difficult to operate than conventional include the absolute turbulence level, the wakes from
cascades. Because of the operating environment they can upstream aerofoils, the full gas temperature distribution and
involve operational compromises to achieve valid and useful temperature level and unsteadiness or turbulence. However,
measurements. It is on these particular aspects that this sec- there is a significant role and usefulness for hot cascades
tion will concentrate. because of their relative simplicity compared with hot engine
or demonstrator measurements.
4.2 OBJECTIVES Of course, static cascades are unable to reproduce the effects
The objectives of hot cascade testing are generally the valida- of rotation and upstream guide vane wakes which are present
tion of hot turbine component thermal designs. This is aimed for a rotor blade. Consequently it is more easy to interpret the
at establishing that the metal temperatures in the turbine result from hot static cascade for nozzle guide vanes than for
blade and vane components in the engine will be in line with rotor blades. However, providing this is appreciated then use-
original design intention. Other typical objectives of hot cas- ful investigation can be camed out on rotor cascades.
cade rigs are the investigation of a particular problem seen in
an engine development or in a technology demonstrator pro-
gram. They can also be used in the demonstration of
modifications to the cooling configuration of a component in 4.3 GENERAL,ARRANGEMENTS AND OPERATING
order to improve the design achievement in terms of metal CONDITIONS
temperatures or life or in optimisation of the cooling standard A diagram showing the complexity of a typical hot cascade
to permit higher operating temperatures. facility is shown in Figure 4.3-1. Air is supplied from com-
I --- - COL0 SUPPLY LINE

I - I I
TESTFACILITY TEST UNIT -1
Fig. 4.3-1 High pressure, high temperature blade cooling test facility
97
Fig. 4.3-2 Cooling experimental rigs and techniques airflow capability of HP/HT facility
pressors and heaters are also used to increase the inlet gas adequate level of inlet turbulence; a measurement by inlet
temperature.For a typical facility the supply system will allow radial traverse for gas temperature and pressure, a mounting
independent control over the temperature and pressure of the system for the cascades which permits easy removal and out-
mainstream and the cooling air feeds. This particular facility let ducting which does not impose restrictions on the outlet
as shown has two legs within the test cell, one for rotor blade gas flow direction. Three typical facility arrangements will be
test units and the other for nozzle guide vane test units. This is described below. For the first as shown in Figure 4.4-1 water
generally convenient because of the significant differences cooled inlet and exit ducting is employed where both gas pres-
between gas deflection required in rotor and vane cascades. sure and temperature loads are camed on the same set of
Each leg features a combustion chamber for heating for high ducting, a five passage cascade is employed to achieve satis-
temperature testing, but it can be shown that useful testing can factory aerodynamic conditions although for very large scale
be achieved with non vitiated (non-burning) air at moderate components fewer passages can be used provided the calibra-
inlet temperatures as supplied by the basic compressors plus tion test establishes that the aerofoil velocity distributions are
heaters. Typical air flow capability, gas pressures and temper- sufficiently representative.
atures and mass flows are shown in Figure 4.3-2 for both
mainstream gas and cooling air. An altemativefacility and test unit arrangement is as shown in
Figure 4.4-2. This is more complex in concept but is able to
There is a fundamental requirement to satisfy similarity achieve higher ultimate operating conditions. Typical inlet
between the engine conditionsand the hot cascade test condi- pressures that can be achieved on such a facility are SOOpsi,
tions. This requires the achievement of representative Mach inlet temperatures 1800K, mass flows up to 7/8kgs per sec-
numbers, Reynolds numbers, gas-to-coolant temperature ond and a pressure ratio across the cascade greater than 2. In
ratios and pressure ratios. Furthermore the correct ratios of this arrangement the whole cascade is encased within a pres-
gas-to-coolant conductivity and gas-to-metal conductivity
0 are required to fully represent engine conditions.
sure vessel which takes the full pressure load. The inlet and
exit gas flow is contained in water cooled ducting which only
As has been implied above for selected test cases this can be carries the temperature thermal load. The gas pressures
achieved at more modest operating conditions within the hot across the ducting are balanced by raising the pressure in the
cascade facility. pressure vessel.
It is important to establish that aerodynamic conditions The pressure vessel is mounted on a bulkhead and is a single
within the cascade are representative of design operating con- skin uncooled cylinder flanged at the inlet for bolting to the
ditions. This can conveniently be established through an facility preheater and mixer casing. It is closed at the exhaust
aerodynamic calibration test. If camed out prior to any cool- end by a removable hemispherical cover plate. The cascade
ing test, this enables the cascade pressure ratios to be set to working section or cassette is shown in Figure 4.4-3. It con-
obtain a good match between measured and predicted aero- sists of inner and outer support segments bolted to the outlet
foil Mach number distributions on the centre (instrumented) flange. Six nozzle guide vanes are located in the working sec-
aerofoils.At this time the cold flow conditions in the inlet duct tion using engine hardware and held in position by a mounting
can be checked by traverse to establish uniformity of inlet block. The ends of the cassettes are formed by end plates in
total pressure and temperature. the shape of a half nozzle guide vane, to preserve the shape of
the end passages, and with simulated platforms. Instrumenta-
4.4 DESIGN APPROACHES FOR HOT CASCADES tion and cooling air feed pipes from the cascade pass through
There are a number of altemative approaches in the design of the outlet flange and pressure vessel in mechanical seals and
hot cascades. These have to balance practicality, cost effec- terminate at connector plates. Provision is made in the inlet
tiveness and the simulation of the engine flowfield and for radial traverse planes together with wall static pressures
operating conditions to achieve a useful and flexible rig with- on the end-wall and the side-wall. The advantage of this form
out prohibitive expense. Typical requirements are a mixer or of construction is that the pressure loads are carried in the
mixer plate at inlet to achieve temperature uniformity and an large pressure vessel and the high temperature thermal loads
98
Fig. 4.4-1 Test unit - high pressure, high temperature nozzle guide vane cooling test unit
2 Qaa Stream Tlca-4 Statlca 1-1
4 1
Fig. 4.4-2 Schematic of the HPIHT test section

99
are carried in the interior water cooled ducting. A further of the nozzle guide vane cannot be understood in detail. In
advantage shown in Figure 4.4-3is that the outlet ducting pro- other words detailed film cooling effectivenessdata cannot be
vides less constraint to the exit gas flow direction. This has the extracted from the measurements because the operative inlet
potential for providing an improved quality of airflow and temperature condition for each streamline cannot be esta-
better passage repeatability in the cascade. blished. However, such a facility can achieve very valuable
A typical facility of this kind would require cooling air sup- work in NGV cooling assessment and cooling system
plied through multiple feeds, around 10, each line requiring development. There is an added advantage in cost effective-
heating and lagging to maintain cooling air temperature. The ness in that combustor and nozzle guide vane thermal
cooling pipes would be connected to the top of the test unit evaluation can be carried out simultaneously. An additional
and then piped to common root and tip air boxes. It can be factor is that the air supply to the combustor must be able to
seen from Figure 4.4-3that the inlet duct and flare are thin produce adequate high pressure and temperature conditions
walled fabrications which are water cooled and pressure bal- for the testing to be representative.
anced against the mainstream gas by high pressure water.
Within the inlet flare a mixer plate or turbulence grid can be 4.5 MEASUREMENT TECHNIQUES
bolted to permit the achievement of uniform temperature TemperatureMeasurements
profiles and adequate turbulence level at cascade inlet. A The primary measurement made in hot cascades is of metal
prime advantage of the above facilities is that cooling design temperatures. The most common form of this instrumenta-
evaluation can be carried out with accurately known inlet tion is through some form of thermocouples. These have the
temperature and pressure, because these flat inlet profiles can advantages that temperatures can be measured over a range
be employed when required. of operating conditions. They are an accurate measurement
There is a third style of facility that is useful and productive of surface temperature, and the data can be recorded through
for hot cascade testing but for which this benefit does not automatic computerised data logging systems. They do have
apply. In this set-up an annular ring cascade of nozzle guide the consequent disadvantage that significant time and effort is
vanes is mounted behind a fully annular combustion chamber. required to install thermocouples. Further factors are that the
Simultaneous testing of combustion chamber and nozzle accuracy of the measurement can be dependent upon the
guide vane thermal performance can be achieved in such a quality of the installation of the thermocouple. Furthermore,
facility. It has the advantage of producing fully representative the definition of the data that can be obtained is controlled by
profiles of temperature and pressure both radially and cir- the number of thermocouple measurements that can be
cumferentially at exit to the combustor and inlet to the nozzle obtained. In principle, surface film thermocouple measure-
guide vane cascade. This provides a full simulation of engine ments can be used as an altemative to the embedded
conditions and thus a full evaluation of NGV cooling per- thermocouple. A typical cascade would require 80-100 ther-
formance. The fundamental disadvantage is that, because the mocouple measurements from a set of aerofoils to provide
inlet temperature profile is not flat, the thermal performance high quality data definition.
Fig. 4.4-3 View of the HP/HT working section

100
It must be recorded that alternative measurements techniques cooling and leakage flows to be obtained. It can be carried out
for surface temperature are available, these include tempera- both in the mainstream and in the cooling air flow. Main-
ture sensitive paints or thermal paints, or optical pyrometry. stream flow visualisation is best carried out with outlet
Given sufficient development such methods can be used satis- ducting removed and conventional methods using coloured
factorily, however they do have a number of advantages and or fluorescent particles in a liquid suspension. Typical exam-
disadvantages. Optical pyrometry provides a very good mea- ples could be Dayglo powder in diesel oil, or Magnesium
sure of the surface temperature distribution and can provide Oxide powder in Butyl Phylate. Both these examples are
quick testing. However there are many detail issues of calibra- aimed at low temperature application.
tion and measurement which need to be explored carefully. Cooling air flow can be similarly traced (for film cooling or
Thermal melt paint techniques also have the advantage of leakage flow). A typical method would use a fluorescent dye
providing full surface measurement. However they are less in an acetone solution. Many alternatives can be chosen
flexible in not being able to accommodate a range of operat- according to individual preference and circumstances.
ing conditions in one cascade test. Repainting or coating is
required. Some of the advantages and disadvantages of these 4.6 TYPICAL MEASUREMENTS
techniques are shown in Figure 4.5-1. A typical comparison between a high pressure, high tempera-
ture rig test, and engine measurements for an HP NGV are
Pressure Instrumentation shown in Figure 4.6-1. Precise agreement has not been
A typical cascade will require around 50-60 pressure tapp- obtained, but this is thoroughly explicable. The rig measure-
ings to measure the pressure distribution around the aerofoil ments were obtained with a flat inlet temperature profile, and
and to establish, through a calibration test, that the velocity so the surface effectivenessis precisely known. In contrast the
distributions around the aerofoil are fully representative. This engine measurements were obtained with a representative
is particularly important when a cascade with limited number
of passages is employed. The periodicity of the cascade can be
significantlydegraded in these circumstances. For such a cas-
radial and circumferential temperature distribution. The
quoted effectiveness values were based on a single gas tem-
perature value. However hotter or cooler gas could have
a
cade an aerodynamic calibration test can prove very valuable. reached some portions of the aerofoil, producing an apparent
reduction or increase in surface effectiveness.
Instrumentationof Cooling Air Supply Figure 4.6-2 shows rig test results from progressively blanking
Each cooling air feed system must have adequate instrumen- film cooling rows. By this means the film cooling contribution
tation for static pressure and temperature to define coolant of individual rows can be established. However the changes in
supply conditions. This will require one or two static pres- operating conditions for the remaining rows must be correctly
sures and thermocouples in the root and tip feed galleries. allowed for in the analysis.Such testing can provide a valuable
Furthermore these will need supplementing by multiple static diagnostic technique in design analysis. However, it is worth
pressure tappings and thermocoupies close to the cooling air repeating that data of this form can be most reliably obtained
entry points at hub and tip of the cascade. It is very important by testing with uniform inlet gas temperature distributions.
to establish a good understanding of cooling air supply condi- This task is further aided if modest inlet temperatures are
tions to the cascade. If this is not done then cooling employed whilst retaining fully representative temperature
performance can be significantlymisunderstood. ratios and other non-dimensional parameters.
Flow Visualisation 4.7 MEASUREMENT OF EXTERNAL HEAT

It can prove very useful to employ flow visualisation tech- TRANSFER COEFFICIENT
niques in hot cascade rigs with full engine hardware. This Whilst they are not ideal for the purpose, it has been shown
enables a better understanding of actual flow conditions, and that valid measurements of aerofoil external heat transfer
a
THIN FILM GAUGES OPTICAL PYROMETRY THERMAL MELTS
TECHNIQUE LAID ON BASE COAT SURFACE RADIATION AEROFOIL COATED WITH METAL
[ELECTRICAL [cg. SILVER)
INSULATION)
ON AEROFOIL
ADVANTAGES QUICKER/EASIER TO WHOLE SURFACE ACCURATE MELT LINE
INSTALL cf. T/C's COVERED
USE ON THIN WALL QUICK PREPARATION GIVES FLOW VISUALISATION
AEROFOILS
DISADVANTAGES LIMITED EXPERIENCE MANY DIFFICULTIES LIMITED CHANGE POINTS
TO BE SURMOUNTED
I BASE COAT IS A
THERMAL INSULATION
HIGH GAS TEMPERATURE
REQUIRED
Fig. 4.5-1 Design validation of cooled turbine components metal temperatures by other techniques
101
EFFECTIVENESS
(E)
ENGINE THERMAL PAINT TEST
/
/
0
I I I I
100 80 60 40 20 LEADING 20 40 60 80 100

EDGE
SUCTION SURFACE PRESSURE SURFACE
% SURFACE LENGTH
Fig. 4.6-1 Comparison of rig and engine cooling effectiveness data
Removing Films
+ Test 1
+--
Test 2
- - - Test 3
m ....)(... Test 4
m
W
z --I+- Test 5
w
>
H
I-
U
W
LL
LL
W
a
z
H
II
I
I
-I I
I
I
0
I
0 I
I
U I
I
I
I
I I I I I I 1 I I I 1
-100 -80 -60 -40 -20 0 20 40 60 80 100

SUCTION SURFACE %PERIMETER PRESSURE SURFACE
Fig. 4.6-2 HP NGV cooling effectiveness in rig

102
coefficient can be obtained from a continuously running high REFERENCES

pressurehigh temperature cascade facility. An example of
this sort of work is shown in References 4.7-1 and 4.7-2. In 4.7-1 Barry B, Forest AE, White AJ
that experiment a non film cooled nozzle guide vane aerofoil Measurement of Heat Transfer Coefficients on Gas
was made in thin-walled form for HP/HT cascade testing. Turbine Components; Part I; Description, Analysis
Testing was camed out varying the temperature of the internal and Experimental Rrification of a Techniquefor use
cooling air. Two techniques were employed, the first using sin- in Hostile Environments, ASME Paper No. 82-GT-
usoidal variation W A ) and the second using a pseudo 174,1982.
random variation (PRBN). Both are adapted from standard
control system techniques. In each case the variation of exter-
nal wall temperature was measured. Mathematical analysis
has shown that if the amplitude ratio and phase relationship 4.7-2 Beacock RJ,Horton FG, Kirker TJ, White AJ
are measured then the external heat transfer coefficients can Measurements of Heat Transfer Coefficients on Gas
be deddced. The measurements made in References 4.7-1 Turbine Components; Part IZ: Applications of the
and 4.7-2 have shown satisfactory agreement with conven- Technique Described in Part Z and Comparisons with
tional transient measurements on the same nozzle guide vane Results from a Conventional Measuring Technique
aerofoil. and Prediction, ASME Paper No. 82-GT-175,1982.
e
103
Chapter 5
Transient Cascade Testing
T.V. Jones, M.L.C.Oldfield, R.W.Ainsworth and T. Arts
5.1 BACKGROUND Shock Tunnel was used to measure film cooling in subsonic
flows (Ref. 5-2) and cascade testing using a Shock Tunnel was
Short duration aerod! amic 2 d thermal testing has long
proposed (Ref. 5-3). In the event a new form of tunnel, the
been an established method in several areas. The classical
Isentropic Light Piston Tunnel was developed (Ref. 5-4) and
case is that of high enthalpy flows. In the 1960s when re-entry
later versions were used for cascade testing at Oxford and
aerodynamics were being studied for the first time it was
elsewhere (Refs. 5-5,5-6 and 5-7).In America short duration
necessary to simulate the actual stagnation enthalpy encoun-
facilities were employed at M.I.T. where a Shock Tunnel and
tered during re-entry. This was due to the fact that
Blowdown Tunnel were used (Refs. 5-8 and 5-9). A refined
dissociation, ionisation and excitation have a major influence
rotating compressor, Blowdown Tunnel was reported in 1974
on the flow field and these phenomena were a function of the
(Ref. 5-10) and this led to a similar turbine tunnel (Ref. 5-11) a
absolute stagnation temperature of the flow. During re-entry
decade later. A Shock Tunnel for turbine measurements was
from close earth orbit the specific enthalpy encountered is
also proposed by the Calspan Corporation and measure-
approximately 33MJ/kg whereas that for re-entry from a
ments on a stator reported in 1977 (Ref. 5-12). A large Shock
lunar mission is correspondingly higher at around 6OMJ/kg.
Tunnel facility has recently been commissioned at Calspan
On the other hand the specific enthalpy corresponding to tur- which enables the testing of full scale turbine rotors to be per-
bine entry conditions are over an order of magnitude less at
formed (Ref. 5-13). A history of the development of short
1.8 MJ/kg. The inability to generate continuously re-entry
duration facilities in the Oxford TurbomachineryGroup over
flows led to a range of short duration facilities and also
this period is given in (Ref. 5-14).
spawned a set of instrumentation techniques which are still in
use today in various forms (Ref. 5-1). The range of tunnels As mentioned, instrumentation techniques specific to short
that arose from this period are shown in Figure 5.1 and with a duration facilities were developed and in particular those for
few additions these represent the transient wind-tunnels heat transfer proved extremely successful so as to be pre-
which are available for cascade testing. ferred to steady state measurements (Refs. 5-15 and 5-16).
Other fast response measurements of pressure, for example,
The first use of short duration facilities in the turbomachinery were perfected and these were used to measure pressure dis-
context occurred at the beginning of the 1970s. At Oxford a tributions in a shock tube driven cascade (Ref. 5-17).
SHOCK TUBES
CONTINUOUS TUNNEL
LIMIT
N.O. L. HY PERVELOCI T Y
V K I LONGSHOT -
GUN TUNNELS-
I I I
ILPT
FLOW DURATION
I I~ I~ I I I I I I I
vs 10 Id lo'
ms 1 10 lo2 10'
s 1 10 lo2 io3 10'
Fig. 5-1 Stagnation temperature versus run time for intermittent tunnels (Ref. 5-15).
104
5.2 INTRODUCTION sary. Similarly an ejector pump.could be used to match sub

Severalgeneral papers have been published on the question of atmospheric exit conditions to atmospheric pressure. Con-
short duration testing in this area (Refs. 5-18,5-19 and 5-20). versely the inlet pressure vessel might not be required if
In (Ref. 5-18) the power requirements of continuous testing pressure levels were so low that atmospheric pressure would
are considered and a comparison made between results from suffice for the inlet pressure. In this last case condensation
steady state and short duration cascade testing at Oxford effects could be a problem and it is possible to supply the air
University as discussed in the introduction to the AGARD- from balloons filled with dry air. The,latter avoids the use of a
ograph. fast acting regulator and produces extremely constant pres-
The wider question of short duration turbomachinery testing sure. Thus the model in Figure 5-2 is a generalised version
is discussed in (Ref. 5-19) and the M.I.T. short duration facilit- which can be said to apply to all short duration cascade tun-
ies are taken as typical examples of such wind-tunnels. The nels. Even in the case of transient devices such as shock
subject is reviewed in (Ref. 5-20) and the Isentropic Light Pis- tunnels and Ludweig tubes the result is a volume of gas pro-
ton Tunnel at the Royal Aerospace Establishment at Pyestock cessed to the required conditions, which acts as the reservoir,
is described in detail. A general description of the various and the generalised model still applies.
short duration testing techniques applied to turbomachinery For a given size of facility (i.e., supply pressure vessels and
over many years at Oxford is given in (Ref. 5-14). dump tanks) it can be shown that the product LtR or L2rRis
constant depending on whether the supply or the vacuum
5.3 SCALING tank limits the performance respectively. L is the scale of the
In any cascade testing the questions to be addressed are cascade and zR the run time. Thus by reducing tR the scale of
usually those of geometric scale and the standard require- testing can be significantly increased. This is extremely
ments of Reynolds number and Mach number similarity.Film
coolant flow similarity is also important. In short duration
testing there is the added question of “temporal scaling” as to
important in turbine blading where boundary-layers are very
thin and boundary-layer studies are made easier at a large
scale. The time to establish steady flow obviously limits the
a
whether the steady testing time is sufficient for the pheno- extent to which this course may be pursued and this subject is
mena being studied to take place. This question can be discussed in a later section.
generalised and posed: “Is the flow steadiness sufficient for
There are many characteristic times present in cascade test-
the measurements of a particular phenomenon”. Another
ing. The gross time scales associated with blade bending
requirement rarely considered in steady state aerodynamic
frequency and passing frequency have been discussed in (Ref.
cascade testing is that of the thermal boundary conditions. In
5-19) which is concerned with rotating systems and these are
short duration testing this question can be complicated by the
reproduced in the table below. Further, it may also be shown
fact that the thermal boundary conditions may be changing in
that L may be increased by using lower temperatures or heavy
time.
gases (Ref. 5-11).
A significant advantage of short duration testing is the gain in
Characteristic Times in Turbomachinesfrom (Ref. 5-19)
scale that is possible for a given size of high pressure reservoir
or dump tank. All short duration facilities use a high pressure Flow through time 0.3 ms
reservoir or a large evacuated dump tank or both in order to Rotor-Stator blade passing 0.1-0.2 ms
produce quasi-steadyflow through the cascade for a time, tR, Rotor revolution period 2.0-5.0 ms
as shown in Figure 5-2. The cascade inlet flow may be regu- Rotating stall period 10.0 ms
lated and a downstream sonic throat may be employed to Thermal time constant, thin wall 250 ms
maintain a constant cascade exit pressure. Thermal time constant, thick wall 2000 ms
If the pressure downstreamof such a second throat was above
atmospheric pressure then a dump tank would not be neces-
Blade first bending period 3.0 ms
However, there are other fundamental times which occur in
cascade testing which is the subject of this AGARDograph.
a
For examplethere is the time for a boundary-layerto establish
or the time associated with trailing edge vortex shedding fre-
quencies.For a cylinder of diameter lmm at sonic conditions
CASCADE the Strouhal frequency is approximately 60kHz and thus the
time scales for vortex shedding from trailing edges of this size
will be of order 15 pS. Secondaryflows also have their charac-
teristic times as do separated regions. Even in a “steady”
cascade flow, interaction with the downstream region can also
produce an unsteady flow situation with its corresponding
RE SERVO1R time scale.
DOWNSTREAM
5.4 FLOW ESTABLISHMENT TIME
THROAT If it is assumed that the inviscid external flow field is suddenly
established, there is then the further time for the boundary-
layers and secondary flows, which are viscous dominated, to
/ wt
DUMP
TANK
reach a steady state. Much of the basic study of this process
has taken place in relation to shock tube experiments on flat
plates where the external flowfield is suddenly established by
the passage of a plane shock wave over the plate (Refs. 5-21
Fig. 5-2 Schematic diagram of a representativeshort and 5-22). Physically the nature of the boundary-layer esta-
duration facility. blishment may be considered as the diffusion of vorticity (i.e., I
105
of the velocity gradient) from the surface of the flat plate into by the suffix 2 can be found from the previous expression and
the inviscid flow so as to create the boundary-layer. This this is then the turbulent establishment time. Further turbu-
growth is initially laminar, but, as time progresses the boun- lent growth to the final turbulent boundary-layer thickness is
dary-layer may undergo transition to the turbulent state and then added to this value.
the growth rate will change. The following refers to a flat plate
boundary-layer.
0.37 - Re0:)
The time to establish a laminar boundary-layer is independ- t - t 2 = -
ent of the fluid viscosity and simply depends on the time for P
the flow to reach the relevant point from the leading edge.
This can be deduced from the governingequations (Ref. 5-23) The total time required to establish .a boundary-layer can be
or using dimensional anaylsiswith extended dimensions (Ref. plotted on an'x-t, or Re-t diagram. Figure 5-3 shows a typical
5-24). Thus Re-t diagram for aL = 0.3, a, = 0.6, p = 0.016, Re, = lo5,
Re, = 5 X lo5.As would be expected, the turbulent boun-
dary-layer grows at a considerably faster rate than when
-X - - aconstant = aL laminar. The rates are consistent in magnitude with those
U0t measured by (Ref. 5-22) and modelled by (Ref. 5-28) for the
shock tube situation.
where x is the distance, U, is the freestream velocity and t the A conservative estimate of maximum boundary-layer esta-
establishment time. The constant aL depends on the parame- blishment time should be a factor of at least three times that
ter considered and the establishment criterion. Analysis (Ref. predicted by the above equations.
5-21) and experiment (Ref. 5-25) give a range of values. The Table 5-1 shows estimates of the boundary-layer establish-
0 experimental data of (Ref. 5-22) gives 0.3 whereas analysis
(Ref. 5-25) indicates a value of 0.2 for surface heat flux to be
ment times in air for a large subsonic cascade operating at
ambient temperature and pressure, and a small transonic
within 1% of the steady state value. A value of aL = 0.1 is warm cascade at high pressure, both with Re, = lo6.
therefore taken to give a safe estimate.
Transition takes place at some position along the surface and The boundary-layer starting times estimated here are small
as shown in (Refs. 5-26 and 5-27) turbulent spots form and compared to the run times of Isentropic Light Piston cascades
spot leading and trailing edges travel at 0.9 and 0.5 of the ( > 0.1 s) and Blowdown cascades ( > 1 s), but may be signifi-
freestreain velocity respectively.A safe estimate of the time to -
cant for shock tunnels ( 10 ms). It is interesting to note that,
establish a transitional boundary-layer is thus made by adding in both cases, the boundary-layer establishes itself before a
the spot trailing edge convection time to that required to form three chord length of flow has passed through the cascade.
the upstream laminar boundary-layer. Thus Returning to the question of the required steadiness in the
running time of the facility, recourse may be made to the anal-
1 ysis of (Ref. 5-29) where the effects of sinusoidal freestream
t - tl = -(x - x,) fluctuations on a flat plate laminar boundary-layer are exa-
W O
mined. It is shown that, for slow fluctuations, an in phase
boundary-layer velocity change is produced with the value
where t is the establishment time at x and t, and x1 the esta- determined from the steady state expression and the instan-
blishment time of the laminar boundary-layer at the start of taneous freestream velocity (i.e., a quasi steady situation). In
transition. a,= 0.5 corresponding to the spot trailing edge. addition an out of phase component of velocity is produced
which arises if a quasi-steady assumption is made. The out of
0 Inber
the case of a turbulent boundary-layer the fluid viscosity is
relevant to the establishment time and this is Reynolds num-
dependent. This may be analysed on the assumption that
phase component of surface shear, AT, which is the quantity
most adversely affected is given as
the boundary-layer is always turbulent (to be published) and
results in an equation for the establishment time, t, 1
A t =jo-pAUO6i
2
- X= - P R0.2
Uot 0.37 ex (5.3)
where AU, is the amplitude of freestream velocity fluctua-
tions at frequency w 138 is the boundary-layer displacement
where is a constant between 8 X and 2.4 X lo-'. Re, is thickness. In many transient tunnels such as a blowdown tun-
the Reynolds number based on freestream conditions at a nel there is a continuous fall in pressure and temperature as
position x from the leading edge. The total time to establish well as velocity. The analysis of (Ref. 5-29) may thus be
the boundary-layer must include that for the laminar and adapted to estimate the fractional change in shear stress, At/
transitional periods and these have to be added to the above. to,resulting from a ramp change in freestream velocity,
The results of the calculation can be expressed as a non-
dimensional time, U=U,-bt (5.7)
giving,
t=-t
U:
Y (5.4)
where v is the kinematic viscosity. Thus the time to form a A -t

- x b
- 2.592 --
transitional boundary-layer at the end of transition denoted t 0 U0 U0
106
Table 5-1
Estimates of Boundary Layer Establishment Time (from Fig. 5-3)
Laree Subsonic Cascade Small Transonic Cascade

Chord 300 mm 50 mm
Pressure 10s Pa
Temperature 300 K 450 K
100 m/s 400 m/s
U 1.5 x lo-' m2/s 5.8 x lo6 m2/s
Characteristic time 1.5 ns 36 ps
V
Laminar BL time 0.5 ms 12 ps

(7 = 3.3 x 10s)
Transition BL time 1.7 ms 41 ps

(7 = 1.13 x lo6)
Turbulent BL time to 2.4 ms 58 ps
Re, =
~
lo6
(7 = 1.6 x lo6)
Safe estimate of starting 7.2 ms 170 ps

time t,
~
"Flow length" U, f, 720 mm 69 mm
Thus for a 1% error in surface shear for x = 100" and U, = boundary-layers in this section which will dominate the time
1 0 0 d s the allowed fall in velocity in 100 ms is 0.386 U,. The necessary to set up secondary flows. With nozzle guide vane
above estimate shows how tolerant such quantities are to var- cascades the upstream velocity can be much lower than in the
ying freestream conditions and shows that the quasi steady cascade. Using the curve in Figure 5-3 and the typical cas-
state assumption can be used to infer steady state results from cades given previously, establishment times are given in Table
rapidly varying data. In a blowdown tunnel the pressure falls 5-2.
causing the density and temperature to fall according to the This confirms that the time to set up the boundary-layer flows
isentropic relationships. Pressure changes are transmitted upstream of a cascade working section is the dominant delay
immediately through the boundary-layer but temperature in establishing steady flow in transient tunnels. Of course
changes will have to be transported into the boundary-layer working sections with large contractionsjust upstream of the
with an appropriate response time. blading will reach steady state more quickly.
The case of secondary flow establishment is rather different
5.5 STARTING PROCESS
to that for the boundary-layer. At first sight the diffusion of
vorticity to form this large structure which has a scale equal to The time scales discussed in the previous section referred to
the time taken by the cascade and cascade walls to influence
the passage width would take a very long time. However, the
the flow through viscosity if the flow was everywhere acceler-
vorticity within the secondary flow is convected into the pas-
sage with a velocity comparable with the passage flow velocity ated impulsively to the inviscid ideal flow field. In practice the
from the vorticity in the inlet side-wallboundary-layer. Indeed flow establishes through wave motions and a significant time
the passage vortex can be predicted quite well from a given is required for this to take place. For this description the cas-
cade will be approximated to a one dimensional nozzle with a
inlet velocity profile on the assumption that viscous effects downstream throat which corresponds to the cascade throat
within the passage may be neglected.It is therefore reasonable as shown in Figure 5-4. In practice the valve opening time is
to expect the secondary flow to take the same time to establish usually significant;however, it will be assumed that this opens
as the upstream side-wall boundary-layer. Additional time instantaneously as though it were a bursting diaphragm. In
will be required to form the passage and blade boundary-lay- some instances this is actually the case.
ers beneath the secondary flow and to transport the inlet
vorticity through the passage. The wave motion which ensues is shown in Figure 5-5where it
has been assumed that the dump tank is initially evacuated.
In most cascade tuMelS, there is a parallel walled section When the valve opens the flow immediately assumes sonic
upstream of the working section, and it is the time to establish conditions at the inlet (Fig. 5-5). Unsteady supersonic flow is
107
Time to Steady Boundary Layer Flow

1.8;
U
I
transitional
o*21A,T
OO
0.1
.-
laminar
0.2
- .
,
0.3
. ___
,
0.4
- - .
,
0.5 0.6
~
0.7
- I
0.8
__
I
0.9
.. -
Rex
Fig. 5-3 x-t diagram indicating boundary-layer establishment.
Table 5-2
Cascade Flow Ebtablishnlent Tinles
Large Subsonic Cascade Small Transonic Cascade

Upstream length 1.0 m 0.3 m
Velocity (M = 0.1) 15 m/s 40 m/s
Rex 1.0 x lo6 2 x lo6

1.6 x lo6 2.4 x lo6
Time7 =
U,'
-?
V
I( Establishment time t
I
I 8.7 ms
I
1I Blade BL establishment
IL
time
Time for steady secondary - 110 ms I -9ms
ISafestimates I 27 ms
108
2.0 0 FROM HEAT TRANSFER

f ” SHADOWGRAPHS
T
1.0-
mS
1THROAT2
0- I I I !
Fig. 5-4 One dimensional representationof a cascade inlet
geometry.
I\
I
I
- FLOW
- ASSUMED
DISCONTINUITY
I
DIAPHRAGM THROAT
established downstream of the inlet creating an expansion fan
as shown in the x-t diagram in Figure 5-5.The supersonic flow
at the throat will pass through the throat under the influence Fig. 5-6 Predicted and measured trajectory of the inlet
of attached shocks. With time, the flow Mach number at the starting shock.
throat will decrease to a point where the attached shocks are
not possible and a normal shock is formed just ahead of the
throat. The throat will then become sonic and the normal
of opening of the valve. A sequence of Schlieren photographs
shock will start to move upstream into the unsteady expansion
from a high speed film showing the starting of the Oxford
fan. Thus the flow downstreamof the shock will be subsonic at
Isentropic Light Piston Tunnel (ILFT)is given in Figure 5-8.
approximately the steady state Mach number and the throat
The formation of the shock system and establishment of film
will be choked. The shock moves to the inlet bringing the flow
cooling from a row of holes is clearly seen. It is interesting to
in the nozzle close to the steady state conditions until it
note that even at long times there is significant oscillation of
reaches the inlet. At this point there will be further reflections
the shock system which might not be evident from steady state
back into the nozzle but these are usually small. The major
measurements. Experience has shown that, in “steady state”
starting time is that for the shock to travel from the throat to
cascade flow, unsteady shock and wake systems are usually
the inlet.
present.
5.6 THE SECOND THROAT

In many instances a second throat is employed in order to
L I
keep the cascade exit conditions constant during the tunnel
running time. This enables a tank of modest size to be used to
/ EXPANSION maintain sub-atmospheric pressure downstream of the cas-
SHOCK
FAN.. cade at the expense of having to evacuate this to well below
I
the desired cascade exit pressure. A typical arrangement is
shown in Figure 5-9where a one dimensional approximation
c
is given. The cascade may be subsonic or supersonic at exit
c -
Y whereas the second throat is always sonic in order to isolate
DIAPHRAGM
P the rising pressure in the dump tank from the cascade exit
region. An intermediate cascade exit tank is usually present
downstream of the cascade separating this from the second
Fig. 5-5 Expansion and shock waves in an inlet starting throat. This exit tank may in practice be formed by the down-
process. stream ducting from the cascade.
The required second throat area may be found if it is assumed
that none of the dynamic head in the cascade exit flow is
recovered. The total pressure for the second throat is thus the
A prediction for the trajectory of the shock for a particular cascade exit static pressure. This is not always the case, as will
nozzle is shown in Figure 5-6together with the experimental be seen for the Oxford Blowdown Facility, but for general
points. A series of Schlieren photographs from (Ref. 5-2) purposes this is a fair assumption.Thus the second throat area
showing the motion of the normal shock is also given in Figure AI*, is related to the cascade throat area, A,*, by,
5-7.I t is of interest to note injected coolant entering through a
row of film cooling holes adapting to the various flow regimes
as these change in time.
The above description of the starting process is relevant when
the valve opening time is short compared to wave transit times
from the valve to the throat or cascade. When the valve open-
where Po, and P, are the cascade total and exit pressures and
ing time is much longer than the above shock transit time then
a most complex flow field can exist during the starting phase
Methe cascade exit Mach number.
which will be highly dependent on the geometry of the valve as A further consideration is the volume of the cascade exit tank,
it opens. The cascade flow will therefore pass through a series We. At first sight the requirement is that it should be as large as
of pseudo steady states corresponding to the different states possible in order to produce uniform cascade exit conditions.
108
diaphrasm time from

4 + flows diaphragm
opating
170ps
320us
nops
1.22ms
1I
- lms
t
injection through a double IOW of holes
Fg.5-7 Schlieren photographs of the inlet starting shock in a one-dimensional subsonic nozzle (a) wilhout coolant injection
(b) with coolant injection. Total temperature -1OOOK.
However, the tunnel startingtime may be adverselyaffected if determined by the time for the downstream vacuum tank of
the volume is too large. The time required to establish steady volume W,to rise in pressure so as to unchoke the second
conditions may be found on the assumption that the cascade throat i.e.,
is choked.Except in the final stages of the startingprocess this
willbetrueformostcascadesandintheeventthatthecascade
is designed to run choked then the assumption will always be
true. The pressure history in the exit tank is given by, ,-Y
--
(
- e
( Y + 1 ))L&W/((&)fi%~A;)
(5.12)
(5.10) FOI the runningtime to be great= than ten times the starting
time associated with filling the exit tank,
where r, > 10 re (5.13)
Hence the requirement for the vacuum tank is approximately,
Y+1*
This equationis derived on the assumption that uniform con-
w, > lo(-) W. (5.14)
2
ditions exist in the exit tank and that the gas flow transit time
through the tank is small compared to the rise time rc Thus
the sue of the exit tank has to be small enough such that 1. is
significantly less than the tunnel run time $. T~ will in Nm be W, > 19We (5.15)
110
0.2 - 0. 118. Trailing edge vortex shed as flow starts to move.
-
1.4 2.0 m. Early subsonic flow bends cooling jet onto surface Wake become apparent.
i
5.6 - 6.2 IPS Amval of hot flow shows as striations. Flow nearly sonic.
k
L
--
9.8 - 10.4 m. Trailing edge and suction surface shock wave form as flow continua to accelerate.
11.2 - 11.8 m. Quasi-steady transonic flow established. Shock systems oscillate slightly for rest of run.
- - -
Fig. 5-8 Cascade starting flow in the Oxford ILPT, taken from a high speed Movie Schlieren. The jets from the mw of film
-
cooling holes on the suction are clearly visible. 5000 frames/sec. Blade temperature 288K. Gas temperature 432K.
Isentropic exit Mach number 0.96. Reynold’s number 6.3 X lo5. Interestingly, the Row appears to have become
established in about 12ms, even though the gate valve, in a low-subsonic region, is only part way open. Normally, data would
only be taken alter the gate v a h was fully open at about 40ms.
111
7FAST ACTiNG Thus it may be seen that the second throat and the associated
volumes must be carefully chosen.
5.7 SHORT DURAnON FAclLITIEs

In this sectionthe salient features of the various facilities are
described.
5.7.1 Blowdown Tunmls

+ T ~ ~ < ~ ~ p ~ ; ; ~ ~ ~
Blowdown Facilities have been used for aerodynamic testing
in many areas and running times up to minutes may be
achieved. Here we shall only be concemed with the very short
FIXED PERFORATE duration facilities with runningtimes up to a few seconds i.e.,
PLATE with time sufficient to establish and measure cascade condi-
tiOnS.
(a) MIT Facilifies

The MIT Compressor Blowdown Facility (Ref. 5-10) was
perhaps one of the tint examples (Fig. 5-10). The facility was
in practice used to test compressor rotors at representative
speeds as high as 10,000rpm. However, the facility could just
as well be used for testing static cascades. Indeed the tech-
Fig. 5-9 Schematicdiagram of an adjustable second
0 mroatgeametry.
nique was subsequently used in the MIT B l o w d m 'Ihrbine
Facility (Fig. 5-11) in which annular cascades of nozzle guide
Fig. 5-10 MIT blowdown compressor facilityschematic (Ref. 5-10),

112
COOLINQ AIR
EXAUST TO
ONLET IANK SUPPLV

IOODEQ. K 10 PSI*
zoo Den. K
"'* BOUNDARY
EMLems
e
MAQNElSl&rOR
VALVE R A M SLIP RINQ YWlT

AND
W E E D ENCODER
Fig. 5-11 MIT blowdown turbine facility schematic (Ref. 5-11).
vanes were instrumented and tested upstream of the turbine measurement at a given Mach number and Reynolds number.
rotor (Ref. 5-11). Essentially these facilities consist of a pres-
surised vessel which is suddenly vented through the test
section. In the case of the turbine facility the vessel was cap-
Ofcourse the variation in flow conditions needs to be suffi-
aently slow such that a quasi-steady situation exists as
e
discussed in a previous section.
able of being preheated. The mass flow through the working
A major advantage of the ha facilities was the relatively
section was so high and running times short such that regula-
small volume of the reservoir tank and the vacuum tank.
tion of the supply was not undertaken. Thus the pressure fell Apart from the benefits of size this enabled smallvolumes of
to approximatelyhalfits initial pressure during the combined gas mixtures to be used to match the ratio of specific heats and
starting and running times of 0.1 and 0.4 seconds of the com-
also to adjust the velocity of sound in the reservoir. In the case
pressor and turbine facilities, respectively. w i c a l pressure
of the MfT Blowdown Facilities mixtures of Argon and Freon
records for the two MfT facilities are shown in Figures 5-12 were employed. The use of a gas different to air was also cen-
and 5-13.
tral to the Birmingham Blowdown Facility described below
Second throats were used in both of the hafacilities and the where the working fluid was supercooled steam.
unchoking of this terminated the runningtimes. The pressure
drops during the testing time of approximately 32% for the (b) The Birmingham Blowdown
compressor and 6% for the turbine would result in Reynolds In the expansion of steam through a turbine the expansion
number changes of approximately 24% and 5% respectively. path crosses the Saturation line and the fluid becomes a two
In the experiments reported, the rotor speed was controlled phase mixture consisting of minute drops canied in the
to maintain the blade corrected speed (Mach number) con- vapour. It is thereforenecessary to examine the detailsof such
stant. This aspect will be retumed to in a later section. Thns flows in cascade testing. Unfortunately if saturated steam is
the testing conditions were held effectivelyconstant in a non expanded it does not supercool sufficiently to produce repre-
dimensional sense although there was a small drop in Rey- sentative nucleation until it is supersonic. This is not
nolds number. The latter need not be significant if representative of the turbine where condensation occurs in
insnumentation with a fast response is employed to make a the subsonic region due to the fact that energy is extracted
113
I II
"-
A< I I I t l I I I
.I r. a. IC
.I
Variations of supply lank h), compressor Inlet @z), outiel @3)

and dump lank (pa) pressures wllh iime during blowdown withoul rolor.
Slagnation pressure in oullet annulus @ I ) ) trom traverse h also shown.
Fig. 5-12 Preasure history in the MIT compreasar facility (Ref. 5-10),
2.5 Total Pressure (otm)
2.0
1.5
1.0
.5
0
0 400 800 I200 I600
Test Time(ms)
Fig. 5-13 Pressure history in an MIT turbine facility run (Ref. 5-11 ).
from the flow as work. A method of producing the correet steam is then expanded through the cascade working section.
flow was developed by Bakhtar at Birmingham whereby The facility operated with a pressure of 6.9 bar upstream and
supercooled steam was first pmduced which then was used as vacuum downstream.
the source in a blowdorm tunnel @d 5-30). The source chamber was a large vessel of volume 28m3. The
The Birmingham Blowdown Facility (Ref. 5-31) is shown in supercooled steam was produced by suddenly expanding sat-
Figures 5-14 and 5-15, The supercooled steam is generated urated steam witbin the vessel and stopping the expansion
by venting a c h b e r initially filled with Saturated steam so before rapid condensation occurred. The supercooled
that it goes below the saturation line and becomes super- vapour remained stable within the vessel and the fast acting
cooled. Before condensation takes place the supercooled valve suddenly opened to allow the supercooled steam
114
Fig. 5-15 Photcgraph of the Birmingham steam blowdown facility(Repmduction with fhe kind petmission of the Insfirutbn d
Mechanical Engineers, London)
115
through the cascade. The use of a condenser with a condens- longer running time than those at MIT The running time
ing surface area of 60mZ meant that a large downstream required was a few seconds and this therefore meant that a
vacuum chamber was not required. downstream dump tank would be prohibitively large. Also it
A boiler supplied steam to the main vessel which was lagged was necessary to test at l q e scale in order that detailed tra-
and the vessel was charged with saturated steam. Prior to the versing could be performed and this meant that pressures
test the main vessel was vented to the condenser for a short below atmospheric had to be accommodated. A downstream
period in order that the steam condensed on any foreign parti- ejector pump system was therefore chosen and the complete
cles present which then settled removing nucleation sites. The facility was tested at l/lOth scale (Ref. 5-33)before a full
vessel was then vented again to a predetermined pressure, the scale facility was built (Fig. 5-17).
venting being stopped before the onset of rapid condensation. A schematic diagram of the full scale facility is shown in
The fast acting valve was then immediately opened. The ves- Figure 5-18a where it CM be seen that a large pressure vessel
sel wall was heated by the steam when the vessel was filled and supplies compressed air for both the cascade pressure vessel
required an intemal water “sprinkler”system to retum it to and the ejector system. The flow passes through a large capac-
lower temperatures quickly when required. A further precau- ity, very fast acting, pressure regulator before supplying the
tion necessary was a thin intemal aluminium guard to isolate cascade and the ejectors. This regulator was specifically
the supercooled steam from radiation from the hot walls just designed for the facility and is a crucial component. In order
prior to the run.The thermal expansion of the apparatus was to relieve the requirements on the regulator it only passes a
accommodated using expansion bellows between the vessel fraction of the total mass flow so as to maintain a constant
and the fast acting valve. supply to the cascade and ejectors. The regulator is effectively
a variable aperture in parallel with a gate valve in the main
The fast acting valve was a slide valve with a tapered body flow supply as shown in Figure 5.18a. As the supply pressure
a s h g against tapered seats and this opened in 50111s.
flow underwent a transient phase but was steady within
200ms from the initiation of the run and this was maintained
The falls the regulator aperture automatically opens so as to main-
tain its downstream pressure at a reference pressure fixed
extemally on the regulator. The drop in pressure from the
for at least 500ms. Prior to the run the working section was supply tank takes place across the two gate valves and the
preheated using steam from a bypass line. When optical regulator shown in the schematic diagram. The principle of
access was required the windows could be electrically heated operation of the system can be seen with reference to Figure
to above the saturation temperature hence avoiding conden- 5-18a The valve from the high pressure reservoir is always
sation obscuring the observation. Pressure measurements hlly open being merely a stop valve and when open the down-
were performed with fast response pressure transducers and stream gate valve and the regulator are subject to the tank
these were connected to blade pressure tappings with fluid pressure. Valve mass flow Characteristics are of the form
filled lines. The latter was essential for otherwise the small below if it is assumed that the gas total temperature is constant
bore lines would become partially filled with condensate and throughout the system,
adversely affect the rise time. To avoid vapour locks in the
lines these were purged by a liquid (B.P. calibration fluid was
found to be best). Each pressure transducer was installed in a
m=k,P,f(P,/P,) (5.16)
small manifold with connections to the working section tapp-
ing,a purge liquid header and a vent as shown in Figure 5-16.
With no vapour locks or air bubbles the response time of the where k, is a constant and P, and P, are the pressures
system is very fast and depends on the compliance of the Nb- upstream and downstream of valve respectively.
ing. When operating at low pressures these vapour locks can Thus for constant mass flow through the gate valve and regu-
be problematical although at 0.2 bar there is little problem. At lator in parallel,
lower pressure a greater frequency of purging was required.
Optical measurements of the droplet size were conducted
using light extinction techniques. A force balance to measure
blade forces was installed in this facility.
(5.i7)
ER (5.18)
The pressures Pa, Pbare upstream of these valves in parallel

and k, and k, refer to the constants for the gate valve and
Fig. 5-16 Schematic of liquid filled pressure measurement regulator respectively.
lines in the Birmingham blowdown.
The regulator area opens so as to increase k, as the pressure
Pa falls. For optimum use of the regulator this starts shut and
(c) The Oxford Blowdown Cascade all the mass flow is through the gate valve. As the run conti-
In order to facilitate the measurement of aerodynamic pro- nues the regulator progressively opens and compensates for
perties, using boundary layer and wake traverses for example, the drop in P, hence keeping Pbconstant. If the valves are run
the Oxford Blowdown Facility (Ref. 5-32)was devised with a above the criticalpressure then the regulator should be able to
116
WORKING CATCHER PSTAGE

SECTION
*
BAFFLES
+THROTTLE
CASCADE
AIR
FEE-+
5 200 psig EJECTOR AIR SUPPLY
WIRE MESH
SCREEN I AIR SUPPLY
LOO psig
LOO psig m x
Fig. 5-17 Phdograph of the 1/10 scale version of the Oxford blowdown (Ref. 5-33).
pass a mass flow equal to the design value times the fractional The two stage ejectors are described in (Ref.5-33) and (Ref.
drop in reservoir pressure. Thus the larger the drop in reser- 5-34) and together were capable of pumping the cascade exit
voir pressure allowed, the larger the regulator capacity is down to 10 Wa. In this case most of the mass flow from the
quired. reservoir passed through the ejectors and there was little cas-
cade mass flow. The mass flow through the various
The pressurehistoryforatypicalrunis showninFigure5-19.
components of the system as a function of plenum chamber
The reservoir line pressure falls continuously whereas the
pressure (cascade plenum pressure) is shown in Figure 5-20.
regulated pressure acquires its regulated value after approxi-
For plenum pressures corresponding to cascade exit pres-
mately 1 second. The cascade plenum pressure takes a little
sures above atmospheric pressure a choke pig. 5-18a) is used
longer to stabilise and steady conditions prevail for approxi-
to raise the exit pressure. The maximum plenum pressure was
mately 3 seconds.
5 bar. The operating regime of the tunnel is shown in Figure
The cascade system is somewhat unconventional in that the 5-21 where it can be seen to encompass Reynolds numbers
exit angle of the cascade is fixed and the inlet angle varies to and Mach numbers of existing and proposed turbine Nozzle
suit different blading. This arrangement is in part due to the Guide Vanes (NGVs) and blading.
ejector system and a "catcher" which recovers some of the As mentioned the aim of this tunnel was to enable detailed
dynamichead of the cascade exit flow. The latter alleviates the loss and boundary-layer traverses to be performed at engine
ejector pumping requirement. Care was taken to ensure that it representative conditions. Recent examples of such measure-
did not adversely affect the cascade exit conditions. ments are given in (Ref. 5-35) and (Ref.
5-36) where a trident
117
KEy !ssure
w Gate valve
H.PAir
Balanced
Regulator
Reservoir
v
I
r G
,h ,€h
I I
Cascade /
Matcher
Fig. 5-18 (a) Schematic diagram of the Oxford blowdown (b) ejectors (c) working section (Ref.5-39).
probe (Ref. 5-36)was used to survey the wake and boundary- the same conditionsfor a cascade with seven passages. Detail
layer traverses were undertaken. This gave the results in of the static pressure distributions around the leading and
Figure 5-22 showing the contributions of boundaylayer, trailingedgeofthisbladearegiveninFigure5-24andthephi-
shockandmixinglossesasafunctionofe~ttachnumberfor losophy of such testing is discussed in (Ref.5-37).Some early
a turbine blade at an exit Reynolds number of lo6.A further studies in this transient cascade were devoted to verifying the
recent example of the detailed aerodynamic measurements is measurements against steady state facilities and details are
shown in Figure 5-23where boundary-layer profiles around a given in (Ref. 5-38). The identical cascade was passed
large chord blade (300mm)are given. The blade scale was betwean facilities in this exercise and an example of a compar-
made possible by mounting this blade in the centre ofa two ison of measured static pressure distributions is given in
passage cascade and adjusting the end wall shutters to pro- Figure 5-25.None of the slight differences were due to the
duce the conest pressure distributionpreviously measured at transient nature of the blowdown flow.
I18
t
0
2r
Plenum pressure
,
Fig. 5-19 Typical pressure history in the Oxford blowdown (Ref. 5-32).
/- \
\€*Z .\ LI
1
WO
b
iix 300
m chamber hbl pressure IkPal
w
Fig. 5-20 Mass flow division in the Oxford blowdown (Ref. 5-32)
€xt Mach number
Fig. 5 4 Operating Regime, Re-M, for the Oxford blowdown (Fig. 5-32).
119
Fig. 5-22 Result from the Oxford blowdown cascade showing the contributionsto loss by various mechanisms (Ref. 5-35).
9.0
8.0
7.0
L
E
Ql
2' 6.0
(?
$ 5.0
Gi
$ 4.0
e0
U
.$ 3.0
ul
6
2.0
1.0
0.0
0.0 0.2 0.4 0.6 0.8 I .o I. 2
Fig. 5-23 Detailed boundary-layer profiles around a large scale turbine blade measured in the Oxford blowdown at engine
representativeconditions (Ref. 5-37).
120
I .4 I I I I
I *2 _ .........................................
I
L
I I
I
8
E 1.0
z
8
3 0.8
LA
$
0.6
.-
pC
2 0.4
0.2
"..
nn
0.0 0.2 0.4 0.6 0.8 1.0
0.7
0.6
iE 0.5
5
2:
5
-2
0.4
PI
e
0.3
.-I,
4
e
-
: 0.2
VI
0.1
0.0
...
-0.12 -0.08 -0.04 0.00 0.04 0.39 0. I2
0.6: SlirfJce Lucdtiuii. S/S, lSuctiuri surface is 'YC)
. ,5
D 0.a !i.....................................................................................................................
-
2,
m
U
2
f
-
9
PI
0.58
\
PI
a
2 0.1
t
.-I,
U
9
G
0.54
0.52
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Surface Distance / Trailing Edge Diameter (Suction surface is +vel
Fig. 5-24 Detailed measurements around a large scale turbine scale turbine blade (a) surface Mach number over pressure
and suction surfaces (b) leading edge detail (c) trailing edge detail. (Ref. 5-37).
12l
l2 t
a 02 OL OK 08 IO
BLADE COORDINATE r/c
o 02 OL 06 a8 10
BLADE COORDINATE x h
F!g. 5-25 A comparison of resub on Me same cascade in continuous and short duration facilities (Ref. 5-34), (a) facilities
RG-VKI, 60-DLR Gotlingen, BS-DLR Braunschweig, OX-Oxford University (transient) (b) and (c) Mach number disbibution
comparisons.
122
5.7.2 The Isentropic Light Piston licnnel (ILPT) RESERVOIR NOZZLE

The concept of the Isentropic Light Piston b e l arose from THROAT DIAPHRAGM
the requirements for the heat transfer testing of turbine blad-
ing. In this case the non dimensional quantities required for VALVE 7
RESERVOIR \,,”\ OR VALVE
\ ~~~~~’ THROAT
simulation a x Reynolds Number, Mach Number and Gas to
WallTemperatureRatio.ThelastqUantityiS1.5foragas tem-
pera” of 1800K and a wall temperature of 12OOK.When
testing with models at m m teInperaNre this means that gas
temperatures of only approximately 450K are required and
these may be achieved using simple isentropic compression. HIGH PRESURE PUMP TUBE DUMP
RESERVOIR TANK
This thereforeled to the development of the Isentropic Light
Piston b e l .
The operation of the tunnel may be likened to that of a bicycle Fig. 5-26 Schematic diagram of the Isentropic Light Piston
pump where a piston compresses and heats the test gas and Tunnel, ILPT
this is suddenly allowed out through the working section by
opening a fast acting valve. The running time is determined by
the time hetween opening the valve and the piston reaching the temperahlre of the Wt gas before this exhausts in the
the end of the tube.In practice the piston is driven down the usual manner. The ILPTwas developed at the University of
tube by air from a high pressure reservoir (Fig. 5-26) and Oxford, UK (Ref. 5-4),and versions of the ILPT have been
steady conditions maintained during the run by making the commissioned in several countries, notably at the von W-
volume flow rate into the tube behind the piston equal to that min Institute, B N S S ~ Sand the DRA Royal Aerospace
leaving through the working section.The piston tunnel is said
to be operating under ”matched” conditions when this situa-
Establishment, Pyestock, UK,and are used to test cascades.
The Oxford and RAE tunnels have annular working sections
0
tion applies. If the mass flow from the high pressure reservoir and that at VKI is a linear cascade. However, at Ma a large
is constant then the pressure rises linearly during the com- anndar cascade ILPThas just been completed. Examples of
pression stage. The LPT is in essence a blowdown tunnel in these facilities are shown in Figures 5-27, 5-28, 5-29 and
which some of the stored e n e w is used to compress and raise 5-30.
Fig. 5-27 Photograph of the Oxford ILPT linear cascade.

Fig. 5-26 Photograph of the Oxford ILPT annular cascade and mtOr facility.
Fig. 5-29 Photcgraph of the RAE, ILPT annular cascade facility.

124
Fig. 5-30 Photograph of the VKI, ILPT new annular cascade facility
Before dealing with the theory of the tunnel it is instructive to isentropically but it is only necessary that conditionsare adia-
consider the pressure variation in the general case where a batic as far as the inflow is concerned.
volume has inflow, o u ~ o wand the volume, V, itselfis varying netheory for the ~m is giMn in 5-4) and (Ref, 5-7)
with time, t, as shown in Figure 5-31. The general equation and is outlined below. me for matchg is immedi-
governing the variation of the pressure, P, is, ately apparent from the general equation above if the piston is
assumed to be massless and hence pressure is still uniform
dP throughout the fixed piston tube volume.
dt V V dt (5.19)
dP dV
In this case - = 0 and ~= 0 hence
dt dt
PW, = PIWl (5.20)
dV
Similarly during compression - = 0 and WO= 0 thus
dt
Fig. 5-31 Generalised model ofinflow, outflow and (5.21)

boundary movement in a volume.
In practice the tunnel operates as though the piston was semi-

Losses are ignored and it is assumed that the process is suffi- bly massless and the matching condition above is a valid
ciently slow such that there is pressure equilibrium in the criterion.However, a small pressure fluctuationis present due
volume. WO and \k, are the volume flow rates at outlet and to the finitepiston mass and estimates for this are given below.
inlet, and PI is the pressure in the inlet reservoir.The inlet flow Using the general equation to determine the pressure varia-
is defined at inlet reservoir conditions and the outlet volume tions on both sides of the piston (Fig. 5-26), and the equation
flow rate is de6ned at conditions within the volume, y is the of motion for the piston, of mass M; the period, T, and ampli-
ratio ofspeciiic heats. It is assumed that the outtlow occurs tude of the pressure fluctuations,AP,may be found as:-
125
The suffix 0 refers to initial tube conditions and a is the veloc-

ity of sound.
(5.22)
The resulting running time becomes therefore,
VA and V, are the volumes ahead of and behind the piston of

area A. P is the matching pressure and W the total volume of
the tube equal to the sum of VA and V,.
(5.23)
i.e.
where tRUN is the running time of the EFT.The piston oscil-
lation frequency increases as the piston approaches the end of
the tube during the run as V, decreases and this is observed in (5.28)
practice. The variation of velocity and pressure is shown
schematically in Figure 5-32.
where is the run time if no compression took place.
It is possible to compensate for the initial deficit in the veloc-
ity of the piston when the nozzle opens and this is performed
by allowing an excess mass flow rate into the tube during the
compression stage. The mass flow is suddenly reduced to the
“matching” mass flow when the valve opens. The gas velocity
at the high pressure reservoir end of the tube will equal the
matching value when the nozzle opens. However, the piston
velocity will be a fraction of the latter, equal to the ratio of the
distance from the nozzle end of the tube to the tube length.
The fraction is determined by the volume compression and
hence temperature ratio required. Thus the mass flow neces-
sary for compensation is simply
t.NOZZ LE OPENS
;.t (3” (5.29)
- MASS LESS PISTON ---- FINITE PISTON MASS times the matching value. This can be a significant increase in
mass flow requirements.
The pressure fluctuations may be reduced by reducing the
Fig. 5-32 Variation of the piston velocity with time in the
piston mass. However, even for a massless piston oscillations
ILPT and the pressure history.
will occur due to the mass of gas within the tube. The effective
mass of the piston due to the air in the tube is approximately
given by half the total mass of air. Thus it is not necessary to
The running time may be found from the compression build extremely light pistons.
required for the desired operating temperature T.The volume A further source of pressure fluctuations can arise if a signifi-
after compression is given by cant volume, AV, exists upstream of the working section
+?)”
nozzle throat (Fig. 5-26). This causes a sudden drop in pres-
sure ahead of the piston when the nozzle fast-acting valve is
opened. The fall in pressure AP, ,is given by
(5.24)
and the flow rate through a choked nozzle of area A* and stag- (5.30)
nation velocity of sound 5 is
As can be seen from Figure 5-32 this oscillation will be 90’
out of phase with the oscillation due to piston mass, U,.The
(y resultant pressure oscillation thus becomes,
(5.25)
where AP =
r ( (+)2 + (-=) );
AP,
P (5.31)
The analysis above indicates how pressure fluctuations may

(5.26) be estimated and the tunnel designed to minimise these
126
effects. In general they have only a small effect on tunnel employed to measure heat transfer to gas turbine blading at
operation and the discussion above is included to indicate one MHI using a linear cascade (Ref. 5-41). The use of a super-
of the considerations which are peculiar to short duration fac- sonic expansion with subsequent normal shock to generate
ilities. the test gas was developed by Calspan and is described below
The working section inlet valve may be opened slowly in (Ref. 5-12). In China a shock tunnel has recently been
order to reduce shock loading of the model in the working described whereby the gas is expanded from the reflected
section for example. When such opening times are signifi- shock region through a contraction into a linear cascade (Ref.
cantly longer than the piston period of oscillation, pressure 5-42).
fluctuations are reduced. Another possible source of pressure The shock tube is a facility usually used to produce high
fluctuations are those due to acoustic waves which originate enthalpy flows typical of re-entry and has been well docu-
from the opening of the nozzle. Such oscillations are usually mented (Ref. 5-1). Hence it is not proposed to describe the
extremely small. A more detailed description of the theory is general operation of the tunnel other than to say that ,in a
given in (Ref. 5-4). shock tunnel a primary shock is produced by bursting a
diaphragm situated within a tube dividing a high pressure
5.7.3 The Shock Tunnel driver gas from the test gas in the channel. The shock travels
The shock tunnel was the first transient wind-tunnel to be down the tube and reflects off a closed end producing a stag-
used for testing under simulated turbine conditions. Heat nant gas which has thus been processed twice by the shock
transfer in the presence of film cooling was presented in (Refs. wave. Thus the gas pressure and the temperature are
5-2, 5-8 and 5-9). This work was extended to fully annular increased. The gas processed in this manner is expanded
cascades at Calspan (Ref. 5-12) and complete rotating stages through a nozzle into the test section when a subsidiary diaph-
are now routinely tested in the facilities at this establishment. ragm is burst by the shock reflection. These conditions are
The description in this section is therefore largely based on only sustained for a short period until the expansion wave
the latter. A linear cascade operating from a shock tunnel was produced by the original diaphragm bursting arrives at the
proposed by (Ref. 5-3) and such a facility was built at General nozzle after reflection from the opposite end of the tube. Test
Electric Co., Schenectady (Ref. 5-17) and flat plate studies times are typically lOmS for tubes of overall length 10 metres.
representative of turbine conditions were reported in (Ref. An x-t diagram showing the shock and expansion waves'
5-39). motion is given in Figure 5-33.
A shock tunnel was used by Mitsubishi Heavy Industries at In the Calspan Shock Tunnels a unique technique is used to
their Nagasaki Technical Institute in 1975 to measure pres- produce the correct stagnation pressure for the annular cas-
sure and flow in a transonic cascade typical of low pressure cade. This involves expanding the flow supersonically in a
steam turbines (Ref. 5-40). Subsequently a shock tunnel was nozzle to high Mach number and then reducing the flow to
EVACUATED DUMP TANK

9' ID x 33' LONG
NOZZLE
DRIVEN 10.5 CONE
DRIVER 18" ID x 60' LONG \
FLOW-CONTROL
NOZZLE
,
'' REFLECTED
RAREFACTION
-X 0 *X
DISTANCE
Fig. 5-33 x-t diagram for a shock tunnel and the shock tunnel arrangement. I
127
subsonic conditions through a stationary normi I( develope It 0: rd @$ 5-45). Encapsi ed liquid crystals
formed across the entrance to a tube which represents the are a convenient me& of measuring surface temperatures.
inlet section to the cascade. This is shown in Figure 5-34. The They are applied to the model surface as a thin layer of cap-
advantage of the system is that the shock tube may be oper- sules approximately 2 0 p thick which produce a colour play
ated at a constant high pressure, producing a large mass of as a function of the surface temperature. The heat transfer
processed gas. The cascade total pressure is then chosen by coefficient may be inferred from the transient thermal
varying the position, i.e., the Mach number, within the pri- response of the model which is found from the liquid crystal
mary n o d e at which the flow is reduced to subsonic colour change. The detail of the method whereby the heat
conditions by the normal shock. The exit pressure of the cas- transfer coefficient is determined is considered in a following
cade is determined by an adjustable downstream throat. The section and here details of the tunnel are given. The liquid
total enthalpy, of course, is unaltered. crystal method also gives a very good overall visual indication
The facility at Calspan has been employed for many years of the surface flowfield as revealed by the surface heat trans-
measuring pressure and heat transfer in a range of cascades fer.
(Refs. 5-12 and 5-43). Rotor stages have also been tested in In such testing it is necessary to have a significanttemperature
the facility where the rotor is brought up to speed before the difference between the freestream gas and the model. For
shock tunnel is fired (Ref. 5-44). A new larger shock tunnel relatively low mass flow rates it is perfectlyfeasible to heat the
has recently been commissioned which can test full scale first flow electrically, for example, and suddenly pass this through
stage rotors for most modern gas turbines (Ref. 5-13). This is the model. The latter technique is employed at Oxford to test
shown in Figure 5-35 and employs a new fast acting valve to the heat transfer characteristics of blade and vane internal
open the nozzle. cooling passages (Ref. 5-46) where a 40kW heater is used.
The range of possible stagnation temperatures is wide However, cascade flows are usually of a much higher mass
although, with room temperature blading, only modest tem- flow and the heating requirements become rather high. In this
case it is easier to preheat or precool the cascade itself. A sim-
peratures are required in order to simulate gas to wall
temperature ratios as described in a previous section. Typical ple Oxford cascade is shown in Figure 5-36 where this
conditions within the original Calspan shock tube were 5 54K, preheating is achieved by surrounding the cascade with a
73 bar, and the NGV inlet total pressure was 6.8 bar. The new water bath. In this manner a uniform model surface tempera-
larger shock tunnel is of diameter 0.46111 and has driver and ture can be maintained prior to the nm.Although it is not
channel lengths of 12.2111 and 18.3111 respectively. The dump essential to have the initial temperature uniform it is much
tank is 2.75111 diameter and 10.4111 long. w i c a l driver pres- more convenient to analyse the data if this is so. By use of mul-
sures are 75 bar and running times between 35 and 40ms. tiple liquid crystals initial model temperature non uniformity
Mass flow rates of 248kg/s are achieved and wheel diameters can be taken into account (Ref. 5-47).
of lm may be accommodated. The running time of the m e 1 is determined by the time
required to initiate the flow. In the case of the liquid crystal
5.7.4 Liquid Clystnl Cascades - Cold Heat Transfer Tunnel measurement it has to be assumed that there is essentially a
Short duration heat transfer testing of cascades has recently srep function in heat transfer c d c k n t over the model in
been undereken using transient liquid crystal techniques order to analyse the results. The reason for this is that the
Fig. 5-34 Schematic diagram of the Calspan shock tunnel, annular Cascade working section.
Fig. 5-35 Photograph of the new Calspan shock tunnel.
Fig. 5-36 The Oxford pmtotype cold heat tranr er tunnel.

129
measurement gives a value of surface temperature (or several working d o n to the supply pressure. In the caseof a linear
values)atagiventime(ortimes)andtheformofthehistoryof cascade, slide valves are commonly used (Refs.5-5.5-6 and
the heat transfer ooef6cient has to be h m although its 5-45). Suchvalvesare possible as theyopen in the direction of
value is to be determined. This would not be the case if sur- the blade span which is usually relatively short. exam-
face temperature was continuouslymonitored where it would ples of slide valve construction are shown in Figure 5-37a‘0’
only be necessary for the flow to be at its steady condition at rings seal the valve and FTFE or Nylon bearing surfaces are
one instance in the run. Thus for the liquid crystal tunnel the used to reduce friction.A Weage shaped valve plate can also
running time must exceed the starting time by an order of be used which ensures sealing when closed but which reduces
magnitude. For large cascades starting times may be of order friction in the opening phase (Fii 5-37).
looms and hence run times of order seconds are necessary. Actuation of the valve is u d y performed using pneumatic
This was the situationfor the cascade in Figure 5-36. cylmders which pull the valve plate with a c o ~ e d rod q In
In situationswhere the starting pmceas is unduly long and p the case of the ILPT,for example, there is the added m@e-
.
there is a d a e n t length of steady runningtime the cold heat ment that the valve should open at a certain time determined
transfer tunnel technique may still be employed by using by the pressure history within the pump tube. The use of
model injection for example. Such a situationdoes exist in the detonators triggered by the output from pressure traapducers
Oxford B l o w d m Cascade facility where no emphasis was measuring the pump lube pressure enables this to be
placed on sudden startiag but rather on adequate steady state achieved. As shown in Figure 5-37athe pneumatic actuating
test time for aerodynamic mwurements (seeFig. 5-19). This cylinder is pressurised but is restrained by a hollow plastic
facility is presently being modified to indude preheating or tube loaded close to its buddingpoint The detonator is situ-
precooling of test blades which are suddenly exposed to the ated within the tube and this desmys the tube on Sring
flow when steady conditions are achieved. Thus a class of allowing the pneumatic drive to accelerate the valve plate
short duration, cold heat “fer c~scadeswill become avail- without the delay apsoCiated with W g the pneumatic cylin-
able for transient heat “der measurement. der. %idly, opening times of approximately20msfrom the
initiating signal can be achieved for a 229” X 76.2” inlet
5.8 FASIACTINGVALVES channel. Related techniques can employ a method whereby
An essential component of most short duration facilitiesis a the initiating device actually moves the valve and actuator so
fast acting valve wbich is interposed between the cascade as to uncover gas ports which enable the pneumatic actuation
f
Cailng
‘Nylbn laced
qllda
I To 5-K)
air cylinders
slm
Fig. 5-37 (a).Slldevalve constructionfor the oxford ILPT, (b) Birminghamslide valve (Ref. 5-31 ). (ReprOduCtiOfl with Me
kind permission of the Insrimkm of Medranical Engineers, London)
to then come into play. This initial acceleration can be pro- cade technology when going to an annular cascade by
duced by electromagnetic or explosive devices. employing a rotating gate valve with multiple ports, Figure
The valves for annular cascades are rather different due to the 5-38.The small rotation is achieved using a detonator
large aperture which has to be opened. Both the annular cas- destroying a plastic tube as before. The disadvantage with the
cade at RAE and that at MIT use the high pressure of the above system is that care has to be taken to smooth the flow
reservoir to drive the valve. Thus there is no requirement for a downstream from the ports (Ref. 5-48)
separate pneumatic actuator. These facilities differ however
in that the h4IT cascade valve can be triggered when desired The MIT turbine blowdown plug valve is of annular form as
whereas the RAE, ILITvalve has to be opened at a particular shown in Figure 5-39and in the closed state is held shut by the
time in the compression phase of the tunnel run. The ILPT at reservoir pressure. This seals at points L and K in Figure 5-39
Oxford on the other hand has retained the original linear cas- and when unseated at K by a small rearward movement the
- R A D I A L SVOKES
TRANSDUCER
I
Fig. 5-38 Rotating gate valve for the Oxford ILPT mmr.
Fig. 5-39 MIT annular plug valve.

pnssureforadrivesitbadctowardsthereservoiruntilsealL facility "the valve design and manufacture developed into a
disengages.Subsequentlyseal L mengagesand captures and major engineering effort" and computer simulation of the
wmplpsses gas which may be used to decelerate the valve. valve movement and gas flowwithin it was required.
Controlled leaks on this system allows adjustment of the
deceleration. Opening times of SOms were reported. A It is therefore worthwhile explaining the basis of design of
mechanical shock absorbing system was used in addition to such large valves with reference to that on the RAE ILPT
arrest the valve. The MlT wmpressor blowdown facility annular cascade tunnel. In this caw the valve was a "true" plug
employed a large diaphragmwhich was ~pturedby explosive valve which opened a large 3SOmm diameter apenure in
saips. These saips of "Datasheet" were detonated from the ZOms and had the added consmint that it had to open at a
centxe of the diaphragm so this "petalled" in equal segments gimpressureleveland hencetimeinthewmpressionphaPe.
(Ref. 5-10). As the authors state when describing the turbine The valve is shown in Figure 5-40, havinga mass of 43kg and
Far1 acltng
plvp ".\IC
Pump lube
II2m diamclcr)
I ..
Fig. 5-40 RAE plug valve schematic and photograph.

132
a required movement of225m.Theplugwas closed bypres- The suffixes refer to the regions in Figure 5-40. i refers to the
surising the valve to a value above that achieved in the piston initial d u e . a is the velocity of sound and C(x) the venting
tub. To activate the valve a seal ring on the pilot seal was area incorporating a discharge c d c i e n t .
pneumatically driven forward a small distance by switching a The variation of pressure within the plug valve was deter-
fasrmagneticvalve(D.C.supplieshavetobeusedforthispur- mined according to the general equation for rate of changeof
pose as A.C. cycle times are significaut). ' h e plug pressure pressure given in Section 5.7.2, giving this in terms of the plug
dropped as gas escaped into the rear volume of the valve and velocity U as
the piston tube pressure then caused the plug valve to move
backwards. This in Nm opened the area of the venting aper-
ture as the aperture stop was carried on a rod attached to the
plug. Thus very rapid venting of the valve occurred and the (5.33)
plug accelerated quickly. The subsequent requirement was to
decelerate the valve smoothly and bring it to rest without
undue impact on the cascade assembly This was achieved by
capturing a controlled amount of gas within the plug using a
"thimble" on the rod already mentioned which blocked the
ventingaperturefortherequiredtime.Thevalvecametorest A is the cross-sectional area ofthe plug. The venting area was
just before its stop, when the thimble has just passed through a prescribed functionof plug position, x, and hence these equ-
the venting aperture, dumpingthe gas which caused the decel- ations were solved together with the equation of motion of the
eration and allowing the plug to seat on its stop at low velocity. plug. Use was also made of the energy equation for the system
Without this last venting the plug would bounce back into the which expressed the overall gain in energy of the plug and gas
mainthroat in terms of work done by the gas within the piston tube. Typi-
The motion of the plug was predicted by first calculating the cal predictions and experimental results for the plug motion
gas flow rate, W, through the venting orifice according to the and pressures within this are given in Figure 5-41. Frictional
pressure within the plug and that in the rear volume. effects are small in such large devices.
The valve was therefore. adjusted to have apertures according
to the predictions for a given operating pressure within the
piston tube. When operating at different pressure levels, dif-
ferent apertures would be necessary in order for the valve to
operate correctly. This is evident, for the gas flow time con-
5 Time
-theory
+x expt
S Tlm
Fig. 5-41 Theoretical and experimental results for the RAE plug valve motion
133
stant would be independent of pressure level whereas the ture compressor gas is fed through channels in the blade so as
mechanical time constant would decrease as pressures to emerge onto the blade surface thus giving a cold protective
increased for the plug mass was constant.In order to avoid the gas film.Therefore in most heat transfer testing the filmcool-
inconvenience of such adjustments the gas used within the ing must be modelled. The injection system for this coolant is
plug valve was a mixture of air and helium. At low pressure hence an important aspect of the experimentalapparatus and
levels air was used and at the highest pressure pure helium is described below. In simulating such coolant flows it is
was employed. A factor of 7.5 in pressure level could be necessary to obtain the correct density ratio between coolant
accommodated with this technique. The valve thus executed and freestream gases. Since in the model tests the gas temper-
the same form of motion at each pressure level, the time being ature is approximately 450K this might imply in some
shorter at the higher pressure levels. In terms of non-dimen- instances providing a coolant gas at temperatures below
sional parameters this meant that the quantities ZOOK. This may be avoided if a heavy gas such as carbon diox-
ide is employed as the coolant and such an expedient has been
PD3 at fdlyjustifiedexperimentally(Ref.5-49)asfarasheat transfer
--and-
Ma2 D (5.35) measurement is concerned.
The actual design and construction of the injection mechan-
ism is worthy of comment. Figure 5-43 shows a schematic
were the same in all cases where P is the pressure level, M the diagram of the apparatus. It is not possible to start injecting
plugmassandDthediameter.Pressureratiosintheplugvalve the coolant prior to the run for this may be at a different tem-
were held constant. perature to the blade and produce a pre-run non equilibrium
Bursting diaphragms are conventionally used in shock tun- temperature.It is essential to start the coolant flowsas the run
nels to open the working section nozzle as well as to initiate starts and this is accomplished by opening the constant pres-
the primary shock. Diaphragm particles can be a nuisance suresupplyvalveinFigure5-43,Thisfillstheintemalpassage
and a perforated plate was used to collect them in (Ref. 5-42). to a constant pressure as the run starts. Immediately after
The running time in the large Calspan shock tunnel was very filling the calibrated, choked ofice, the supply valve opens
long (40111s)and a fast acting valve was employed to open the and the constant pressure line is sealed off.This C O M ~ Cthe ~ ~
n o d e leading to the annular cascade and rotor. In addition blade with a constant mass flowsupplythrough a choked, cali-
the valve was required to close &er the hot running time in brated, orifice. In this manner the blade intemal passage is
order to prevent turbine ovenpeed and also to stop debris rapidly filled so as to prevent ingestion of free stream gas and
from the main diaphragm from entering the working section. yet during the run the coolant mass flow is set at the desired
The valve took the form of a centrebcdy upstream of the level. The actual pressure established within the cooling pas-
throat, the centrebody containing a cylindrical valve (Fig. sage is measured continuously during the run in addition to
5-42).The flow was therefore around the centrebody and the coolant temperature.
then radially inwards through the cylindrical valve before
passing axially through the throat. Opening time was 3ms and 5.10 SIMULATIONOF BLADE PASSING IN CASCADES
closing time 6ms. Flow unsteadiness exerts considerable influence on the aero-
dynamic and heat transfer performance of turbomachinery
5.9 FlLM COOLING AND INJECTION SYSTEMS blades. A significant contribution to this is caused by the
In modem gas turbines the high pressure turbine stage is blades passing through the unsteady wake and shock wave
inevitably filmcooled. In this process relatively low tempera-
e DRIVEN TUBE OPENING PORT

I
VALVE SLEEVE
I
/
CLOSING PORT CLOSING CHOCK
Fig. 5-42 The fast acting valve utilised in the Calspan shock tunnel.
134
CONSTANT PBESSUUE
U.P. SUPPLY FOR

CHoKgD ORIFICE
CONSTAUT W S lLov
INITIAL VALVE
SOLENOID
CHOKED ORIFICE
IIWLTEWCOOLER JACKET
SWITCE OVER VALVX
CARRYING CHOKED
ORIFICE
Fig. 5-43 Film cooling injection system described in section (Ref. 5-9).
fields shed by upstream blade rows. It is possible to simulate include all those fluctuating phenomena present in continu-
these effectsin a twoditnensionalcascade tunnel by moving ous facilities and does not refer to uniform flow alone. Thus,
suitably sized bars upstream of a stationary cascade.In exper- as in many cases, very high frquency capability is necessary
iments at Oxford (Refs. 5-68 and 5-69) the effects of wake to make relevant measurements in even a wntinuous facility,
passing have been studied using thin film gauges. The bars the instrumentationcan be essentially the same. Details of a
were brought up to speed in a vacuum and then the short wide range of these unsteady measurements are given in
duration EFT was fued,the flow passing through the bars (Refs. 5-16, 5-50 and 5-51). Hence in this section we shall
before the cascade. Again this demonstrates the advantage of consider those measurementswhich are particularly suited to
the short duration facility in terms of the power required to short duration cascades, such as heat transfer, and also dis-
spin the bars. The experimental arrangement is shown in cuss the requirements and limitations for conventional
Figure 5-63. The boundary-layer can be clearly identified aerodynamicmeasurements imposed by the finite run time.
passing from a laminar to a turbulent state and back to a lami-
nar state as the wake passes through the passage. The effects
of shock w a w on the blade heat transfer were also studied 5.11.1 Pressum Meesuremeots
(Ref. 5-70) by running the bars at high Mach number. The Those measurements which are meant to correspond ta the
complex shock motion could be seen within the passage using time average in a continuous facility can usually be effected
Schlieren photography and could be identified from the heat quite readily in most short duration facilities so long as the
transfer traces in Figure 5-64. basic transducer has sufficientlyfast response. Also there
should be no extreme transientsin the starting process.
The success of these unsteady measurements depends crit-
The values of rise time, r, obtainable in practice are deter-
ically on the availability of high frequency pressure and heat
transfer insuumentation.A typical blade passing frequency of
mined by the length of the pressure tubing, restrictions at
entry or elsewhere, and the volume at the transducer end of
5kHZ requires a bandwidth of at least 50kH.7,if line detail is
the tubing. The latter has a large effect on the rise time and
to be resolved. The heat &er processing system described
care is usually taken to minimise this. Sensym pressure trans-
in (Ref.5-59) has a sufficient WkHz bandwidth.
ducers (Ref. 5-52) are a reasonable cost solution to theselow
Geneally, the signal-to-noise ratio in cascade experiments frequency applications. These are shown in Figure 5-44 and
with rotating bar simulators is higher than that obtained on when connected to the model by 300mm of lmm I.D. hard
fully rotating experiments, where slip rings are used, and so it plastic tubing give rise times of approximately lOms if the
is not so necessary to mort to phase-locked averaging to transducer volume on the inlet port is filled to minimise the
extract the periodic signal components. volume. For multiple pressure measurements the Scanivalve
ZOC system has found use in short duration cascade facilities
5.11 MEASUREMENTS (Ref. 5-53). This system has a block of miniature transducers
Measurements in short duration cascades are essentially the which can be connected to the tappings, each transducer
same as those made in their steady state counterparts.Indeed, being scanned at high frequency and calibrated during each
the main aim of the short duration cascade is to produce a scan. A typical nozzle guide vane, with end wall film cooling,
steady state measurement. The term steady state is meant to instrumented for static pressure is shown in Figure 5-45.
a
REGULATOR
Fig. 5-44 Sensym pressure transducer schematic diagram.
Fig. 5-45 Typical pressure tapped NGV used in me RAE ILPT.

136
In the application above, such measurementsdo not show any tory at the transducer which would approach the extemal
sigdftcent oscillation due to the viscous damping within the value with each reflection as shown in Figure 5-46. This
narrow pressure tubmg. However,if the tube is made short in characteristic wave transit time gives a characteristic fre-
order to decrease the response time then the transducer pres- quency for such pressure tapping geometries. It can be seen
sure can fluctuate. In the extreme limit if an instantaneous that the period of oscillation of the pressure signal cone
pressure rise took place at the tapping a compression wave or sponds to four wave transit timec and hence the signal
shock wave of this magnitude would travel down the tube oscillation frequency is
(Fig. 5-46). The shock would reflect offthetransducer and the
f = -
1
4a (5.36)
i.e., the fundamental frequency for acoustic resonance in an

open ended pipe. Thus the operatingfrequency limit for such
tappinggeomeuies mustbesignilicantlybelow this resonance
frequencywhichcenbeoftheorderof10kEkfortubelengths
of lo”. This frequency response may be adequate for mea-
suring the mean pressure and pressures on traversing probes
in transient tunnels but does have severe litations for
detecting the high frequency phenomena present in rotating
>
stages. In the latter case surface mounted transducers may be
TRANSDUCER
used (Ref. 5-54), as is discussed in the next section. Miniature
transducers such as those manufactured by Kulite or
Endevco, for example, can be mounted in a short tapping.
a
They are also suitable for mounting in wake traversing probes
such as the three hole probe shown in Figure 5-47.
Fig. 5-46 Wave motion and pressure in a Short tapping.
5.11.2 Fast Response Pressure Measurementsin cerades
Pressure measurements in cascades generally fall into one of
two categories, essentially determinedby thefrequency band-
instantaneous value initially measured would be far greater width of the required data In the first case, measurements
than that at the tapping (in the acoustic limit it would be twice may be requiredat a relatively low bandwidth ( <400Hz),and
the tapping value). The reflected shock would return to the here there are many possibilities for using a combination of a
tapping aperture and then, not being matched to the extemal pneumatic pressure tapping linked to a sensoror bank of sen-
pressure, would reflect as an erpansion wave back into the sors some distance from the measurement point (maybe
tapping. The process would repeat itself giving a pressure his- several metres distant). This technology has been discussed m
Fig. 5-47 Three hole traveming pmbe mounted in the RAE ILp7.
137
theprevious section. In the second me, where possible wake Temperature effects are more usually accounted for by using
interaction effects are being studied, a higher bandwidth is compensathg electronic circuitry anached to both the sensor
reqired (>loo&), and this cao best be provided by the diaphragm and to the exterior of the transducer. The most
mounting of miniature semiconductor transducers on or recent work in miniaturisingsensingelements has enabled the
close to the surface where the pressure measurement is compensation to be carried out in software, post measure-
required. This might be the surface of a blade in cascade, or ment. Here, only the silicon diaphragm and its supporting
the surface of a fast traversing probe, Figure 5-47. sensor need be attached to the cascade blade or aerOaynamic
Semiconductor pressure transducers have been used for probe surface, and a much greater spatial density of measure-
some considerable time for these purposes, and (Refs. 5-54 ment pants is then achievable.
and 5-55) trace some of the significant developments. There The technology of pressure measurements is still a rapidly
are essentially threemainissues which affect the use of semi- developing one.At the research level the use of piezoelectric
conductor devices for these purposes. In principle, the device film may permit "whole field" measurements to be made, and
usually consists of a silicon diaphragm with a Wheatstone's there is no doubt that many improvements to both time mean
bridge network of resistors etched onto its surface, Figure and fast response measurements will be made during the cur-
5-48. Pressure 18 applied, so that the diaphragm ddects, and rent decade.
the resistivebridge, being strain sensitive, produces an output
voltage proportional to the applied pressure. The threemain 5.11.3 High Speed Pmbe Traversing
issues which have to be addressed anz (i) the inherent sensi- The problems associated with traversing probes across cas-
tivity of the transducer to temperature as well as pressure; (U) cade flows in short duration tunnels are considered in this
the sensitivity of the device to strain not assodated with the section. Such traversing is necessary for loss measurements
applied pressure; (iii) the stifhess and mass of the diaphragm, and other flow diagnostics. For these it will often be necessary
which affectthe frequency responseofthe sensor. to measure p M e s of total and static pressures, flow angle
and total temperature. For non two-dimensional cascades
area traversing may also be nquind.
Stnln gauge clamnra In tunnels wheretheflowisessentiallycontinuous,theinvesti-
gator is not too much concemed with the frequency and
transient response of the instrumentation when moving the
probe from one instrumentation pomt to the next. The data
are often effectively time averaged at each measurement
point
.75 mm s f l h (100 slkz)
In short duration cascade tunnels the limited testing time
available makes consideration of instrumentationfrequency
and transient response imperative, The main issues arising
I
f"this are now considered.
Speed of m r s i n g mechanism
Obviously, the desire to " k the number of measure-
ment points per tunnel run suggests that high traverse speeds
should be used. This is limited in practice by a number of fac-
tors:-
a) The response time of fhepmWtmnsduceruscd.It is
necessary to give suf6cient time for the muwrement
value tosettleat eachmeasurcmentpoint.E.g.aprobe
with a l O O H z bandwidth will trike -7ms to settle to
1%. With measurement points 0.2mm apart, this
limits the traverse speed to 28"/sec, probably
inadequate for a 300ms run time In cascade flows,
downstream shock waves do give rise to sharp pres-
Fig. 5-48 Typical silicon preawre Sensor &sign. sure discontinuities,and it is desirable to be able to
resolve them to at leapt the dimensions of the probe
head,typically lmm for a simple pitot For wake ua-
versa the spatial discrimination repuired is less. At
the measurement plane, wakea have a typical width of
The design of solsing elements has improved much over the 35 to 70% of the cascade pitch (typically 6 0 " ) , and
last decade, and the susceptibility of the sclls01 to base strain sopermittheuseofeitherasloweruansducerorafas-
effects has been much reduced by cardul design ofthe strain ter traverse
isolatingpiUars on which the silicon diaphmgm sits. Sicethe
na~frequencyoftypicalminiaturesmsors(like.theKulite b) nyMmic effects on the mcnnued variables. The
XCQ or LQ series) is 300-400kHz,and they are normally motion of the probe will change the m e a s d pro-
lightly damped, the frequency response is often assumed flat perties:
to 250%. For measurements at higher frequencies, it is i) The measured flow angle 0 will be changed
necessary to measure. the frequency response. This is often by the probe velocity v transverse to the flow
achieved using a shock tube to provide a step input to the sys- velocity U
tem, the step response being monitored.
138
Choice ofpmbe and tmnsducer

(5.37) Most of the probe geometries used in continuousNMek (e.g.
trident, wedge, 3-hole and 5-hole) can be adapted for use in
-- - short duration tunnels. The previous sectioncovers the use of

Consequently, for U lOOm/s, a limit of 88 pressure transducers for unsteady flow measurements, and
0.1' gives a max. v 170mm/sec. this is also relevant to probes used for traversing. Miniature
U) The measured djnamic head q P,5a, --
Pwc for incompressibleflow,will be m a d
semiconductor pressure transducers (Kulite,Endewo etc) are
commonly used.
by the streamwise velocity U: The design is usually a compromise between the desire for a
small probe tip, and that for fast mponse. The fastest
response probes have the transducers (e.g. semiconductor
pressure gauges) mounted directly on the probe tip, at the cost
(5.38)
of a larger probe size. The installation of the transducers
within the nose of the probe is, nevertheless, limited by two
-
For a 1%error this limits U to 500mm/s for U
10Ods.
effecb:
i) As indicated pwiously, most pressure transducers
iii) The measured totnl tempemme will be are quite d t i w to temperaNre chauges and fluctu-
mcdifiedbyu: ations; careful compensation M comction is
therefore required over a rather large temperature
range (e.g. ambient to 550K in an ILPT NM~).The
transient nature of the flow can also result in transient
(5.39) the& stresses on the transducer diaphragm as the
probe heats up uneyenly during the tunnel NIL This
For a 1K error, in air(Cp- 1010JhgK)at U

lOOm/S, this limits U to I O d S e c .
- can cause transient offsets not easily compensated or
c o d .
U) The volume occupied by the (even miniaturized)
Before deciding on a traversing speed, all these limits should transducer withinthe probe head incresses the aero-
be considered, especially when measuring regions such as dynamic blockage. The effect becomea of major
wakes,where velocities may be considerably below that ofthe importance in transonic flow regime9.
mainfreesaeamnow. In probes where the pressure transducers are remote, every
effortshould be made to mount them close to the probe head,
Tmwrsing Mechanism and to COMect them with short lengthsof wide bore tubin& in
The high traversing speeds used in short duration tunnels order to minimise the response time of the probetransducer
have led to the almost universal use of computer controlled combination Effortsshould also be made to "isc the
automatictraversingsystemsusingavarietyofpropulsivesys- dead-air volume in front of the transducer.
m S : Thetimeresponseoftheprobeshouldbech~usingafast
a) Pnemksystems, used both at Oxford (Ref. 5-56) rising pressure step, from a device sucb as shown in Figure
and VKI (Ref. 5-57), exhibit a number of definite 5-49. An example of such tests carried out at VKI are shown
advantages such as simplicity, flexibility, repeatability in Figures 5-50 and 5-51, using Kulite and Sensym naosdu-
cers @reviously National SemiCOndUCtM) connected to the
and a wide range of operation.The speed of traversing
canbemadealmostconstant,bymeansofchoMori- calibrator with metal tubes of length 1 and diameter The +.
fice.~in the feeds to and from the actuating cylinder. It results show that the Se", which has a lager dead-air
is, however nefessary to measure the position: time volume. has a slower response,and that the tubes reduce the
trajeaoryoftheprobeduringtherunu~geither ana- frequency responses to values far below that of the naked
logue (linear potentiometer or LVDT) or digital transducers (typically lOOlrHz and lOlrHz for the Kulite and
encoding. Sensym respectively). The typical oscillatoryresponseis quite
undamped and can give rise to misleading spurious spectra in
b) Stepper motor drives are amactive for blowdown the transducer signals.
tunnels, e.g. (Ref. 5-32), where traverse times of over a
second are available. In principle, the position of the S.11.4 Heat Treader Me;lauements
probe can be controlled in time to l O 0 p accuracy by Heat transfer measurements are used in cascadesin two ways:
suitablechoice of stepper motor and gearing.In prac- firstly to measure the heat transfer characteristicsof the pro-
tice, the relatively high probe velocities mean that the files concerned; and secondly to infer aerodynamic
stepper motors are running at near their maximum properties of the flowfields. In both cases a transient facility
speed, and it is prudent to use encoders to k k for can have advanrages in that it is ofteneasier to measure heat
missed steps. Dedicated single chip microprocessors transfer in short duration flows. Hot wire and hot filmprobes
can be used to control the individual axes, with trajec- i.e., electrically heated probes are used widely to measure
tories downloaded from a host computer. The flow characteristi~incontiouousfacilitiesandcanbeusedin
stepping action can give rise to annoying vibrations transient fadties in a similar manner. However, there are
which interfere with the probe signals. sometimes difticulties assodated with tuning the sssociated
electronics in constant temperature operation. This arises
c) D.C. servo-motordrives, tightly controlled by feed- from the factthat the adjustments to optimise a constant tem-
back, offer a smooth running alternative to stepper perature anemometer are dependent on the flow over the hot
motors, at a cost of more development effort. wire which is absent until the run starts in a transient facility.
139
PRESSURE PUL E TESTING APPARATUS
ATMOSPHERIC
r p
Ap iPz -Pq
p1 I:'
-
Fig. 5-49 Apparatus used at VKI to generate a fast pressure step for traneducer calibratbn (1 small volume reservoir: V,,
- -
2 large volume reservoir V,, 3 - solenoid, 4 connection to pressure transducer, 5 - valve, 6 -flange, 7 spring). -
TUBE 1 i 155 mrn

Ol", i 1 mrn
AP = Z O O m Hg h 6 v
X : 2m I I GRAD
Y ' 1 V I GRAD
lbl
Fig. 5-50 Typical transducer outputs using the calibration apparatus of Fig. 5-49,
140
15 I i i 1 I I
tl lmsl
T lms I
10
0 I I I I I I
50 100 150 I l m
Fig. 5-51 The response time of typical transducers as a function of the IengIhof tubing connecting this to the pressure
tepping.
Surface thin film heat transfer g a u p on the other hand do (a) Thin Film H e a Tmnsfer Gauges
not rely on electrical heating and are essentially used to moni- The surface heat flux q is related to the surface temperature
tor a model surface temperature during transient heating. history by,
Thus this form of heat transfer measurement is i d d y suited
to short duration testing. The basic analysis whereby the sur- (5.40)
face temperature history of the model is related to the heat
transfer rate is given in (Refs. 5-15and 5-16).The analysis is
also the same as that used in inferring heat transfer from sur- where pck is the material thermal product and s the Laplace
face temperature measured with liquid crystals. The transform variable.
difference between the u8e of liquid crystals and thin film
gauges is that the latter measures the temperature contiou- ?he inverse Laplace transform of the above equation may be
found by taking the convolution of the inverse of & and T
ously whereas the former usually only pmvidm a
measurement of certain temperature levels during the run.
Modem image analysis techniques,however. may allow some
giving. e
continuous measurements to be made with liquid crystals.
,"(t - TI- (5.41)
Many methods are available to perform thisintegrd and some
arereViewedin~S.5-15and5-58).Thesingularityat t - r
has to be taken into account
These gauges have been discussed extensively in the literature
(Ref. 5-15)and only a brief account of recent models is
included here. Complex blade and vane models can be
machined from Coming Machinable Glass Ceramic. The
platinum thin film gauges are then hand painted onto the
model surface using c o m m e d y available paints and fired
Fig. 5 5 2 The semiinfinite conduction model of the at approximately 600% in order to pmdum the films. An
transient heat transfer measurement. exampleisshowninFigure5-53.Acons~tcurrentispassed
through the platinum film in order to monitor the resistance
change. The voltage output may be recorded dvectly or used
In the analysis of the thin film gauge it is assumed that the as input to an electricalanalogue of the one dimensional con-
duration of the test is so short that one dimensionalunsteady duction equation which gives an output correspondingto the
conduction into the gauge material can be assumed. In such instantaneousheat transfer rate. A typical circuit is shown in
transient SituatiOriS Laplace transform techniques are Figure 5-54 and a frequency response of 1OOkHz can be
employed and are denoted by overbars in the following. expected for the gauge with this circuit (Ref. 5-59).
141
I
Fg. 5-53 A tvplcel ceramic. nozzle guide vane model instrumented wNh thin film heat transfer gauges.
Constani Current Source
Fig. 5-54 The h$h frequency heat transfer analogue circuit f” (Ref. 5-59).
142
A recent development of this technology has been the two may be either e l d c a l l y supplied via the gauge itself or by
layer gauges used in rotating systems. Due to the inadequate external radiant heating. By suitable processing of the output
strength of the ceramic blades, insulating layers have been both the enamel and the metal properties may be found. A
placed on metal blades. In the case of the Oxford gauge (Ref. typical gauge is shown in Figure 5-56.
5-60) the layer was enamel whereas in the ha gauge (Ref.
5-61) a thin sheet of Kapton was glued onto the blade.
A very brief description is given here of these developments.
The Oxford gauge (Ref. 5-60) is shown schematically in
Figure 5-55. A thin film resistance gauge is painted on top of
Fig. 5-55 Schematicdiagram of the Oxford thin film multi-

layered thin film heat transfer gauge.
Fig. 5-56 The Oxford two layer, enamel;thin film heat

the enamel covered blade and fired in the usual manner. At transfer gauge (Ref. 5-60).
short times the gauge responds in the semi-inkite mode. At
longer times the two layer equations given in (Refs.5-15 and
5-16) and developed in (Ref. 5-62) have to be employed and
the equation for the Laplace trausform of heat fluxbecomes, TEMPERATURE
SENSORS
(5.42) Fig. 5-57 Schematic diagram of the MIT two-lawr

. gauge
- -
employing a thin polyimideinwlator sheet.
a
The MIT gauge is shown schematically in Figure 5-57. The
thinfilmresistancethermometerisserpentineinfonnsoasto
increase the 6lm mistance. Contacts are fed through the
1is the thickness of the enamel layer and a its thermal diffusiv- Kapton sheet so that all four contacts for upper and lower
ity. Suffixe8 1 and 2 refer to the enamel and the metal filmsare on the upper surfam. N i i is deposited to form the
respectively. In the case of a two layered gauge it is not practi- thinfilmbyD.C.sputteringandgoldtagsaredepositedonthe
cal to use an analogue circuit However, a similar circuit to output leads. The thin film gauge typically has a resistance of
that for a semi-in6nite case may be profitably used in order to 500 ohms. Calibration is performed by radiant heating using
boost the high frequency components of t h e n ” tempera- an argon ion laser supplying 30kW/m2 to the film. The ther-
ture signal before it is pgssed to the rest of the data acquisition mal properties of the Kapton film are found from the short
system. The reason for this is that the high frequency time response for a step functionin surface heat transfer rate,
components of the heat transfer rate produce lower surface s,as,
temperature fluctuationsthan those at lower frequency Con-
sequently it is best to correct this inequality, at least in part, 2qh
before passing the signal on to an amplifier chain with con- qt)= (5.44)
stant gain at all frequencies. In the caseof the Oxford gauge
when installed on a Nrbine rotor this signal adjustment is i.e., parabolic in time. In practice an old technique described
employed before passing through the slip ring. Full details are in(Ref.5-15)isused toavoidthenecessityodetemriningthe
given in (Ref. 5-60). CaKbration of this gauge may be accom- laser power. A liquid drop of known thermal properties is
plisbed by pulse heating and observingthe output This pulse placed over the gauge and again subjected to the same laser
143
pulse. In this case some of the power conducts transiently into (b) Liquid Crystal Methods
the liquid and consequently the film output is smaller. A com- A review of liquid crystal methods in heat transfer and aero-
parison of the signal with and without the liquid enables the dynamic testing is given in (Refs. 5-63 and 5-64). Liquid
Kapton sheet thermal properties to be found with respect to crystal materials are made from large rod-like molecules
the liquid properties. The characteristic time of the heat pulse which form a mesophase exhibiting liquid properties but in
to penetrate the Kapton layer is typically 1.6ms and at times which there is a long range order or pattern of distribution of
greater than 20 times this value, steady state conditions may the molecules. An example of these materials is the nematic
prevail. Thus from the difference between the two sensors on liquid crystal which is found in liquid crystal electronic dis-
either side of the sheet at large times the necessary informa- plays. A development is the chiral nematic liquid crystal
tion about the thickness and thermal conductivity may be which forms a cholesteric mesophase and it is this phase
found. which exhibits a colour play on reflecting white light (Ref.
5-63). The reflected wavelength is dependent on the material
The Kapton film had 20 such gauges mounted on it and this temperature and hence this property may be used to measure
was wrapped around the turbine blade and glued in position. temperature.
Obviously this does impose limitationson the curvature of the
surface being instrumented. Typical output signals from the In practice the liquid is encapsulated in extremely small
turbine environment are shown in Figure 5-58.The upper gelatin spheres with a typical diameter of 1Opm. Such encap-
and lower temperature signals are given and the inferred heat sulation protects the liquid crystal material from other
transfer rate from the signals is also shown. It can be appre- influences such as mechanical forces and also enables more
ciated that for such a gauge the signal processing to obtain the than one material to be employed at one time by mixing cap-
final heat transfer rate is somewhat complex. However the sules. The mechanism whereby colours are produced is
information about flow phenomena present make this effort somewhat similar to B r a g reflection from solid crystal layers
0 well worthwhile. which selectively reflect a certain wavelength due to interfer-
ence effects although this analogy does not fully explain the
polarisation of light which also occurs. As the temperature
varies, the spacing between the molecular layers changes and
hence the selectively reflected colour also changes. The
colour display temperatures and temperature range over
which the visible spectrum is selectively reflected may be
altered by choice of a suitable material.
In practice it has been found preferable to use a liquid crystal
which passes through the visible spectrum over a narrow tem-
perature range of order 1'C. By observing one particular
colour the temperature may be determined to an accuracy of
f 0.2"C.The capsules containing the narrow band liquid crys-
Bottom Sensor - tals are sprayed or screen printed onto the model surface so as
to give a monolayer of capsules using a water based binder
such as polyvinyl alcohol.From the early days at Oxford it has
30 1 1 1 1 1 1 I I I I been usual to use mixtures of capsules (Ref. 5-46) and two
0 IO0 200 300 4 00 500 components for example are sometimes used. Thus with this
Time ( m a )
system of narrow band liquid crystals, two isotherms will pass
over the model surface as its temperature rises during the
0 1 transient test. The advantages of this will now be discussed.
4
160
For a step function of hot flow the surface temperature his-
tory, T,is given by,
N
- T-- Ti
4 120 - - 1 - e'* erfc(p) (5.45)
5 Tg - Ti
-
x
-
L
a
ul
= 80 where Tgand Ti are the gas and model initial temperatures
?
l- respectively,
c
0
I (5.46)
40
0
8.0
I I
8.2
4
' I
8.4
I
Blode Possing
I
8.6
I I I
8.8
I '
''O
where his the local heat transfer coefficient.Thus if Tg,
are known then h can be found from a single measure-
Ti and
ment of T at a time t. It is often advantageous to use a mixture

of two liquid crystals of narrow bandwidth to obtain tempera-
Time ( m s )
tures TI and T, at time t, and t,. The ratio of the values of
Fig. 5-58 Typical results from the MIT gauge of Fig. 5-57 are the same over the model and hence as at any point his con-
(a) output signal, (b) computed heat flux from (a), (Ref. 5- stant then the ratio of times t,/t, is also the same over the
model. This test gives a very convenientand rigorous check on
144
the experimental results which may be performed immedi- short duration facilities the heat transfer rates are usually
ately the experiment is completed. In certain instances a higher and impose intolerable normal temperature gradients.
greater number of liquid components are employed (Ref. In this case different systems have to be employed and the chi-
5-46). ral nematic materials can only be used qualitatively.The detail
In many cases it is advantageous to preheat or cool the model of the molecular thermal response of the liquid crystal in its
(Refs. 5-45and 5-65)and this in turn means that uniformity cholesteric state has been examined by BoMett at Oxford
of the model initial temperature may be difficultto attain. The (Ref. 5-67).Image processing of the video signals from such
use of two liquid crystals does enable the local initial model testing may enhance both accuracy and productivity.
temperature to be found in addition to the heat transfer coef-
ficient as demonstrated in (Ref. 5-65).In certain cases the (c) Examples of Heat Transfer Results
determination of h and Ti from the two equations found from Heat transfer distributions around a nozzle guide vane at mid
the valus for the two crystals is not very well conditioned and span for different Reynolds numbers as measured by thin film
use may be made of the ratio tl/t2 to refine the result. As can gauges are shown in Figure 5-59. By using several vanes or
be imagined when a preheated model is employed a certain blades in a cascade, and instrumenting these with thin film
time must be allowed for temperature gradients normal to the gauges at a different location, a map of surface heat transfer
model surface to relax before the test. However, even when may also be constructed.An example is shown in Figure 5-60
this is so there may be temperature differences from widely where a fitting routine to produce contours of Nusselt number
spaced locations on the model. It would be expected that this has been used.
variation would be smooth and hence that the variation of the The high frequency capability of the thin film gauge may be
ratio t1/t2 should also vary smoothly. Examination and cor- used to detect boundary-layer transition on blades. An exam-
rection of data to ensure a smooth variation of t,/t2 gives ple of the changingform of the heat transfer signal through the
improved accuracy in the heat transfer coefficient especially transition is shown in Figure 5-61from (Ref. 5-68)and the
in a region where the latter has large variations. This tech- signal is similar to that for a hot film response as would be
nique is described fully in (Ref. 5-47). expected.
Throughout the above discussionit has been assumed that the The usefulness of liquid crystal methods in measuring global
liquid crystal responds immediately to the surface tempera- heat transfer distributions is demonstrated in Figure 5-62
ture and this had to be demonstrated. The thermal response from (Ref. 5-45)where the cascade endwall heat transfer
time of encapsulated liquid crystal as a surface thermometer coefficient distribution is shown for a nozzle guide vane. The
was first measured at Oxford (Ref. 5-66)and typically a value passage and horseshoe, leading edge, vortices are clearly
of 3ms may be assumed.This is certainly adequatefor experi- seen. Similar heat transfer coefficient distributions on pres-
ments of duration of the order of seconds, however, in very sure and suction surface are also shown.
Fig. 5-59 Distributions of heat transfer around a nozzle guide vane as measured using thin film gauges in the Oxford ILPT
linear cascade for different flow conditions of Reynolds number and turbulence level.
145
4200
3800
3400
3000
2600
2200
1800
1100
Fig. 5-60 Contours of heat-transfer coefficient measured using thin film gauges on a nozzle guide vane in the annular
cascade of the RAE, ILPT.
! I ! 1 ' 1 I ! 10
l&?.
0. I .
I .
0. 1 , ZA
Fig. 5-61 Transition as detected using thin film heat transfer gauges on the suction surface of a blade in the Oxford ILPT.
146
--
0.9 0% 0.7 06 0.5 0.4 0.3 0.2 0.1

-
X
P
Pressure surface
09 0.8 0.7 0.6 0.5 04

-
X
S
Suction surface
Fig. 5-62 Contours of heat transfer coefficientson a vane and end walls using liquid crystal techniques in a cold heat
transfer tunnel. From (Ref. 5-45).
5.11.5 Thin Wire Probes and again enable temperatures to be measured in a passive
Thin wire probes may be used in transient testing. Wire manner.
diameters of low and less have a sufficiently fast thermal
response for mean flow properties to be determined and in A survey of the passive methods of recovery temperature I
many instances fluctuatingparameters may also be measured. measurement is given in (Ref. 5-16). Hot wire techniqueshave
Both passive and active use of the thin wire may be employed, been extensivelyreported in the literature and a review of this
for example, a thin platinum wire can be used to sense the field is given in (Refs. 5-71, 5-72 and 5-73) and no further
freestream recovery temperature using a small sensing cur- description of these will be given except to point out the dif-
rent to determine its resistance and also the same wire can be ferences when operating in a transient tunnel. Constant
heated as a constant temperature anemometer to determine current operation of a hot-wire may be difficult as the flow
the flow velocity. The thin wires can also be made from dissi- conditions will be quite different during the run to that prior
milar materials with a thermocouple junction at the centre to or after the run.Thus it may be necessary to switch the wire
147
B
WAKE-PASSING SIMULATION
IN FIXED
CASCADE OF TURBINE BLADES.
Rotor Cascade Section X X ’
Fig. 5-63 Schematic diagram of the rotating bar simulation of the wake passing experiment of (Refs. 5-68 and 5-69).
Fig. 5-64 Typical variations of heat transfer with time from the experiments in Fig. 5-63.
148
on only during the run in order to avoid overheating the wire. state and allowed to cool. The model should then appear
Constant temperature operation obviously eliminates this colourless. The flow is then suddenly started and the time
problem. However, it is sometimes difficult to set up the taken to exhibit colour at a given point is measured. Calibra-
adjustment of the constant temperature control frequency tion of this time allows a quantitive measurement to be made.
response and gain prior to the run. Adjustment is normally Typical calibration curves are shown Figure 5-65for materi-
performed in the flow being tested in order to optimise the als of different viscosity.
system. Thus a hot wire system adjusted to be optimum prior
to the run might well go into oscillation during the run.Wire
probe vibrations can often be introduced in the starting pro-
cesses in the testing and also wire integrity may be a problem.
Particles present in the flow which result from diaphragm
bursting can also destroy hot wires.
Aspirating hot wire probes also have their place in transient
testing. In this case the hot wire is located in a very small pitot
tube either upstream or downstream of a sonic throat within
the tube. The downstream region is evacuated using a vacuum
line. Thus for a given upstream total pressure and tempera-
ture the signal from the hot wire will be independent of
freestream velocity and dependent on gas constitution for
0
example. This was the basis for the fast response concentra- 200 400 600 800 1000
tion probe of (Ref. 5-74)and similar systems are currently Time From Flow Onset/s
available commercially. Another form of the aspirating probe

is that of (Ref. 5-75)where two hot wires set at different over- Fig. 5-65 A typical calibration of shear stress and texture
heat values are employed. Local wire Mach number is thus change time for a cholesteric liquid crystal.
fixed as in the concentration probe and the measured Nusselt
number is dependent on total pressure and total temperature.
The use of two overheat values thus enables both total pres-
sure and temperature to be determined. The above aspirating
probes are useful in cascade testing for the concentration If the material is in the G texture, shearing will produce colour
probe may be used when film cooling is present and the change. Liquid crystals insensitive to temperature changes are
double wire probe is of importance in measuring losses down- employed such that colour change can be solely attributed to
stream from cascades. the rate of shear of the material. At present this technique has
been used in a qualitative mode and is subject to rather large
5.11.6 Shear Stress Using Liquid Crystals variations resulting from viewing and illumination angles.
Shear ‘stress may be measured in the conventional manner Nevertheless,this does prove to give a good indication of lam-
(Preston tube, hot film) in a short duration facility. However, inar to turbulent boundary-layer transition in flight tests.
one property of unencapsulated liquid crystals does lend itself Fluctuating shear stress patterns have been observed using
to the short duration measurement. Unencapsulated liquid this technique (Ref. 5-77).
crystals have been used to indicate shear stress level in flight A new method using twisted nematic liquid crystals has also
tests in America for some years (Ref. 5-76).The mechanism been developed for quantitativemeasurements of shear stress
whereby the effect occurs has been studied at Oxford Uni- in transient facilities-and this is described in (Ref. 5-63).
versity (Ref. 5-67)and two distinct processes take place
dependent on the manner in which the liquid crystals are 5.11.7 Flow Visualisation and Optical Methods
applied to the surface and also on the surface state. Usually if Short duration cascade facilities c-an usually make use of all
the material is sprayed in a solvent onto a surface and allowed those optical methods applicable to continuous cascade faci-
to dry it has a diffuse scattering appearance in the Focal Conic lities if the illumination is sufficient to record the desired
(FC) texture. If this is then sheared the flow aligns into a situation. In the case of shadowgraph and Schlieren the light
colourful or Grandjean (G) texture. In both textures the liquid sources are usually of extremely short duration in order to
crystal is actually in the same phase but the FC texture is more “freeze”the events and hence these methods readily transfer
fragmented than the G texture. to transient cascade testing. Holographic interferometry has
also been used in the Oxford ILPT (Ref. 5-78)and an exam-
Once the material has the G texture, further shearingwill pro- ple shown in Figure 5-66 where the inferred blade surface
duce a change in colour. This process is reversible as it will Mach number is compared with that measured by surface
assume the original colour when shearing is stopped. On the static pressure.
other hand the FC to G texture change is irreversible. The lat-
ter may be used to measure shear stress in a transient Surface flow visualisation may be performed in many ways
wind-tunnel test by measuring the time taken for the colour but the use of surface oil dots has been useful in the shorter
play to appear. duration facilities such as the ILPT. Titanium dioxide is mixed
with vacuum oil and small dots placed on the blade. These
The FC to G texture change occurs sooner in a region of dots are swept along the surface according to the local shear,
higher shear as the material flow aligns after a certain strain the amount of material in the dots determining the length of
has taken place. Thus the time taken depends on the viscosity the trace. Fluorescent dyes in oil are also employed in a simi-
of the material. A suitable material is chosen and sprayed lar manner. Important considerations in the interpretation of
onto a black model so as to be in the FC texture. If this is not any such visualisation are the tunnel starting and stopping
the case then the material can be heated into the isotropic transients. These may well constitute a significant time com-
Mach
Fig. 5-56 The Oxford two layer, enamel; thin Rim heat transfer gauge (Ref. 5-80).
p a r d to the run time of the tunnel and the shear stress in these and include anti-aliasing low-pass filters. Experience shows
cau signilicautly distort the surface flow p i e . that these should be set with a break point well below the
Nyquist limit of halfthe sampling rate.
Particle Image Veloeimetq (PlV) is a technique which has
been applied to the measurement of flowfields in transient Q p i d ~ a b a n d w i d t h o flOOHzisusedforthe300samples/s
facilities (Ref. 5-79).In this method the flow was seeded with channels and 1SOlrHZfor the SOOk samples/s channels.
small particles and then illuminated with two laser pulses 2. Slow AID conversion channels. Generally, it is preferable
(- l h ) separated by a short duration (- lms). The reflected to have stand-alone systems containing A/Ds and memory
light is " I e d photoppbically and the result analysed to which are conmUed by the computer and download data
6nd the m o w n " of the particles and hence the relevant after the run
component of the flow velocity.This technique was used in a
W e n t annular cascade where a wbole flowfield was 3.Fast AID conversion is best achieved by the use of Tran-
mapped with a resolution of 0.5mm to a velocity accuracy of sient Recorders (e.g., Datalab Multi-trap system) with fast
4%. Conventional h e r Doppler Ane"etq (LDA) may datalinLstothecomputer.In~thcaseslZbitAIDsampling,
alsobeemployedintransientfacilitiesalthoughtheadvantage giving a resolution of 1/(4096)of full scale, will suffice.
of PIV is that the complete flowfield may be mapped whereas 4. Interface between the AID and the computer the IEEE
with the LDA, measurements at a point are obtained, (i.e., HPIB) parallel interface is almost standard, although
Ethernet should ofkr advantages as it becomes more avail-
5.12 DAU ACQmSrlTION AND PROCESSING able.
To reap the full bene6ts d a t e d with short duration wind-
tunnels, it is necesary to use sophisticated, computer based 5. The computer should be dedicated to the tunnel. Modem
data acquisition and processing systems. Generally, there is microcomputers, such as the IBM PC compatible "80368"
not sufficienttime to modify tunnel conditions during a run, and"80486"systemsnowoffer~cientpoweratlowprices.
and so most instrumentation systems will concentrate on It is recommended that safety critical tunnel monitoring and
acquiringlargeamountsofmulti-channeldataduringtherun, data acquisitionbe carried out on separate d m m p u t e r s .
for post-processing immediately afterwards (Ref. 5-50), 6.Data storage. AU raw data should be stored, either on 1.44
m c a l data requi"nts would be: Mbyte floppy disks, or on other mass storage systems such as
tape streamers or optical writ-dy disks.
Si-four Channels of "slow",300 samples/second/channel
data to capture steady flow measurements. 7.As far as possible data should be p d using commer-
daUy available Somuare,such as "MATLAB" (Ref. 5-80).
Thiny-two Channels of fast (5OOk samples/see/channel)
taken in a burst mode (- 4K points/channel) to capture This softwareshould have a "Macro" language to automate
unsteady phenomena such as turbulent spots, transition, processing, as well as allowing interactive manipulation of
simulated wake passing etc. data in graphical form.Modern techuiqua of spectral analy-
sis and correlation by means of the Fast F d e r Transform
A typical system would comprise. algorithm are essential, together with windowing and
1. Amplifiers and signal conditioners near the transducers. filtering utilities. It is essential that those U- the padrage
These amplify the signal to a level high enough forthe AIDS, understaad the pitfalls inherent in such techniques.
150
The system should be realised With the minimum of in-house 5-12 Dunn MG and Stoddard FJ
hardware and soliware, as commerciallyavailablesystems are Application of Shock-Tube Technologyto the Measm-
more likely to survive the inevitable changes of computer ment of H e a t - T e r Rate M Gas Turbine
every five years or so. Components, in Shock Tube and Shock Wave
Research,Proc. of 11th Int Symp. on Shock Tubes and
5.13 CONCLUSIONS Waves, Eh., Ahlbom, Hertzberg A and Russell D,
Transient cascade testing has been recognised as a worth- Seattle, 1977.
while means of determining the aerodynamic characteristics
of cascade blading for over a decade. The development of 5-13 Dunn MG, Bennett W, Delany Rand Rao K
high frequency insbumentation and the advent of fast data Inwsrigarion of U n s t d y Flow through a Tmnsonic
acquisition systems have now established the transient meth- Turbine Stage: Pan II - DatdPdiction Comparison
ods as routine. This review has chronicled this development for Tw-Awmgedand Phe-RrsolvedRwsure Data,
and has provided a reference for the techniques available. AIAAPapex90,2409,0rlando,1990.Alsodetailsin
Such methods are still advancing and, as indicated in the text, WRDC-TR-89-2027, Operating Point Verilication
rotating turbine stages are now being tested in a similar man- for a Large Shock Tunnel Test Facility.
ner. 5-14 JonesTV
REFERENCES Gas Turbine Studies at Oxfonf, I%PI!%i: ASME,
5-1 Lukasiewicz J Paper 88-GT-112, Amsterdam, 1988.
Experimental Metho& of Hypersonics, Vol. 3, Gas 5-15 SchultzDLand JonesW
Dynamics, Ed. Wegenex PP, Marcel Dekker, New Heat TmnsferMeasmments in Shon-DwarionHyper-
5-2
York, 1973.
Jones TV and Schula DL
Fdm Cooling Studies in Subsonic Flows using a Shock
sonic Facilities, AGARD, AG165.1973.
5-16 JonesTV
Heat Tmqfw, Skin Friction Total Tempemnue and
a
Tunnel, Proc. 8th Int Shock Tube Symposnnn, Eds., Concenmation Measurements; in Mcasmmerus of
GaydonAGandOwenPQChapmanandHall,1971. Unsteady Fluid Dynamic Phenomena, Ed. Richards
5-3 Jones TV and SchultzDL BE, Mdiraw-Hill, pp 63-102.1977,
A n ~ ~ m e n t o f t h e ~ s p e TurbineBlade
c~of Cml- 5-17 Kercher DM, Sheer Jr RE and So RMC
ing m Shock Tubes and Shock ~ n n e l r ,Oxford Shon Dumtion Heat Tmnsfer Studies at High Free-
University EngineeringLaboratory Report No.1099/ Stream Tempemwes, ASME, Paper 82-GT-129,
70,1970. 1982.
5-4 JonesTV, SchultzDL and Hendley AD 5-18 Schdtz DL, Ainsworth RW and Scrivener Cl'J
On the Flow in an Isennopic Light Piston " e l , UK, The Role of Shon Dunuion Tating in Gas Turbine
MOD (Proc Exec), Aero. Res. Counc., R and M No. Research, P m . 8th Jnt Symp. on Air Breatbing
3731,1973. Engines, Paper No 87-7031, Cincinatti, 1987.
5-5 SchultzDL, Jones TV, Oldfield MLG and Daniels LC 5-19 Epstein AH
A New Tmm'ent Casmde Facility of the Measurement Shon Dumtion Testing for Tiubomachinery Research
of Heat T m e e r Rates, AGARD Symp. on High Tem- and Dewlopment, Keynote Paper, 2nd Int. Symp. on
perature Problems in Gas Turbines,CP299, Paper 31, Transport Phenomena, Dynamics and Design of
Anksl.a, 1977. RotatingMachinery,Honolulu, 1988.
5-6 Consigny H, Ville J-P and Richards BE
Shon Duration Mmurements of Heating to Cooled
5-20 Brooks AT and Jones T V
The Use of shm Dwarion Facilities for Testing Gas
a
Gas TurbineSwfoces,ASMEPaper78-GT-22,1978. Turbine Blading, Pmc. 1989 European Propdsion
5-7 Brooks AJ,Colhoume DE. Wedlake ET, Jones TV, Forum on Modern Techniques and Developments in
Oldfield MLG, SchultzDL and LoftusPJ Engine and Component Tesiing, Royal Aero. Soc.,
The Isempic Light f i t o n h l at RAE &stock, Paper 10, Bath, 1989.
AGARD CP390, Paper 31,1985. 5-21 CookWJ
5-8 Louis JF, Goulions GN, Demirjian AM, Topping RF Experimental Study of the Shock-Induced Unsteady
and Wiedhope IM Laminar Boundaty Imycr on a Flat-Plate, Phys. of
Shon Dumtion Studies of Turbine Heat Tmnsfer and Fluids, Vol. 17, No. 5, pp 900-902,May. 1974.
Film Cooling Efectiwms, ASME Paper 74-GT-131, 5-22 Davies WR and Bemstein L
Zurich, 1974. Heat Tmmfer and Tmnsition to Turbulence in the
5-9 Louis JF Shock-Induct?dBoundaryyl*rycron a Semi-Infinite F l a ~
Systematic Study of Heat Tmnsfer and Film Cooling Plate,J.FluidMech.,Vol.36,Panl,pp87-112,1969.
Effectiwness, AGARD CP 299,Paper 28,1977. 5-23 YomgAD
5-10 Kerrebrock JL, Epstein AH, Haines DM and Boundary Layers, BSPProfessional Books, 1989.
Tompkins WT 5-24 de St Q. Isaacson E and de St 0. Isaacson M
i%e MIT Blowdown C o m p w r Facile, ASME, Dimensional M e t M in EngineeringandPhysics:Ref-
Paper 74-GT-47,1974. erence Sets and theirPossibilitiesofExtenswn, Arnold,
5-11 Epstein AH, Guenette GR and Norton RIG 1975.
The MIT Blowdown Turbine Facduy, ASME, Paper 5-25 Cook WJ and Chapman GT
84-GT-116,1984. Chamcte&ia of the Unsteady S h o c k - I h e d Lami-
151
nar Boundnry-Layer on a Flat Plate, Phys. Fluids Vol. 5-39 Pedrosa ACF and Nagamatsu HT
15,No. 12,pp 2129-2139,1972, Convective Heat Transfer on Flat Plate at Very High
5-26 Wygnaski L, Solokov M and Friedman D Tempemture and Pressure Gradient9ASME, Paper
On Turbulent Spot in a Luminar Boundary Layeq J. 83-GT-113,1983.
F l ~ Mech.,
d Vol.78, Part 4, pp 78549,1976. 5-40 Kuramoto Y, Nagayama T and Masamichi M
A Study on the Transonic Turbine Cascade by Shock
5-27 Schubauer GB and Klebanoff PS Tunnel, Mitsubishi Heavy Industries Nagasaki Tech-
Contributions on the Mechanics of Boundary Layer nical Institute, Report MTB101, July 1975.
Tmnsirion,National Bureau of Standards Report No.
1289,1956. 5-41 Mitsubishi Heavy Industries - Private Communica-
tion, 1990.
5-28 Sidov VK,Zuev VI and Eighenson EB
Numerical Modelling of Non-Isothermal Compressible 5-42 Jingmei L, Rumeng Z, Jinming H and Suqing D
Turbulent Boundary Layers on a Flat Plate in a Shock Application of Shock Tunnel at "re and Heat
Tube? Anglo-Soviet Seminar on Mathematical Trader Experiments on Turbine Blade, Current Top-
Modelling of Heat and Mass Transfer, Eds. Jackson ics in Shock Waves, 17th Int. Symp. on Shock Waves
ID,Byme JE and Cotton MA, Univ. of Manchester, and Shock Tubes, Lehigh University, Bethlehem,
UK,April 1990. 1989,Am.Inst.Phys. CP208,1990.
5-29 Lighthill,MJ . 5-43 DunnMG
Turbine H a t Flux Measurement: Influence of Slot
The Response of Laminar Skin Friction and Heat
Transfer to Flucfu5IWns in the Smam EIOcity, Proc. Iyection on Vane Trailing Edge Heat TMNer and
Roy Soc., Vol.224 A, pp 1-23, June 1954. Influence of Rotor on Vane Hear Tmnsfe5 ASME, 84-
GT-175,1984.
5-30 Bakhtar F and Heaton HV
On the Potential of BlowDown Conditions in Studies 5-44 DunnMG
of Nucleating and Two-Phase Flows of Steam, Proc. Phase and Time-Reso[wd Measurements of Unsteady
8th Int. Heat Transfer Conference, San Fransisco, pp Heat Transfer and Pressure in a Full-Stage Rotating
2277-2282,1986. Turbine,ASME, Paper 89-GT-135, Ontario, 1989.
5-31 Bakhtar F, WebbRA, Shojaee-FardMH and Siraj MA 5-45 Harvey NW, Wang Z, Ireland PT and Jones TV
A n Expetitnental Facility for Studies of Nucleating and Detailed Heat Transfer Measurements in Nozzle Guide
WetSteam Flows in TurbineBlading, Proc. Inst. Mech. Vane Passages in Linear and Annular cascade^ in the
Engrs. London, Turbomachinery: Latest Develop Presence of Secondary Flows, AGARD, Roc. of Conf.
ments in a Changmg Scene, Paper C423/003, March on Secondary Flows in Turbines, CP469, Paper 24,
1991. 1990.
5-32 Baines NC, Oldfield MLG, Jones TV, Schultz DL, 5-46 Ireland PT and Jones TV
King PI and Daniels U3 Detailed Measuremenrs of Heat Tmnsfr on and
A Shon-Dumtion Blowdown Tunnelfor Aerodynamic around a Pedestal in Fully Developed Channel Flow,
Studies on Gas Turbine Blading, ASME 82-GT-312, Proc. 8th Jnt Heat Transfer Symp. San Fransisco, Vol.
1982. 3, pp 9754'980,1986.
5-33 Schultz DL, Jones TV, Oldfield MLG and Daniels LC 5-47 Ireland PT and Jones TV
Boundary Layer Studies on Turbine Blading -Report Note on Double C v i a l Method of Measuring Heat
on SERC Feasibility Study GR/A/29685, March Tmnsfer C a i e n t , Oxford University Engineering
1978. ScienceLaboratory, Rept No.1710/87,1987.
5-34 Baines NC, King PI, Oldfield MLG and Daniels LC 5-48 Ainsworth RW, SchultzDL, Danes MRD, Forth CJP,
The Use of Tandem Ejector Pumps in an Intermittent Hilditch MA, Oldfield MLG and Sheard AG
Blowdown Tunnel,ASME, 84-GT-226,1984. A Tmnsient Facility for the Study of the Themfluid-
Dynamics of a Full Stage Turbine under Engine
5-35 Mee DJ, Beines NC and Oldfield MLG and Dickens Representative Conditions. ASME, Paper, Paper 88-
TE GT-144,1988.
A n Examination of the Contributions to Loss on a
Tmnsonic Turbine Blade in Cascade,ASME, 90-GT- 5-49 Teekaram AJ,Forth CJP and Jones T V
264,1990. The Use of Foreign Gas to Simulare the Effect of Dens-
ify Ratio in Film Cooling, ASME, J. of
5-36 Mee DJ, Baines NC and OldfieldMLG Thrbomachinery, Vol. ID,pp 57-62.1989.
Detailed Boundary Luyer Measurements on a Tmns-
onic TurbineBlade,ASME, 90-GT-263,1990. 5-50 Oldfield MLG
Experimental Techniques in UnsteadyFlows, Lectures
5-37 MeeDJ in the von Kirmh Institute, BNSSelS, Course on
Large Scale Turbine Cascade Testing at Engine Mach Unsteady Aerodynamics,April 1988.
and ReynoIds Numbers, Experiments in Flu&, (to he
published). 5-51 Sieverding CH (Editor)
Proc. of Symposium on Measurement Techniques in
5-38 Kiock R, Lchthaus F, Baines NC and Sieverding CH Transonicand Supersonic Flows in Cascades and Tur-
The Tmnsonic Flow through a Plane Turbine Cascade
bomachiner,Brussels,VK.I., 1990.
as Measured in Four Eumpean Wind Tunnels,ASME,
J. Eng. for Gas Turbines and Power, Vol. 108, pp 227- 5-52 Sensor Technics, Sensym Pressure Sensor Handbook,
284.April1986. 1991/92.
152
5-53 Scanivalve Corp.,San Diego, ZOC Hiscan 2000 Pres- T and Jones TV
5-66 Ireland l
sure Measurement System, 1990. nte Response Time of Liquid Crystal Surface Ther.
5-54 Ahworth RW, Dietz AJ and Nunn TA momefers, J. Phys.E., Vol.20, pp 1195-1199,1987.
The Use ofSemi-conductor Sensorsfor Bhde Swface 5-67 BonnettP
Pressm Measurement in a Model &bine Stage, J. of Applications of Liquid Crystals in Aemdynamic Test-
Engineering for Gas Wbmes and Power, Vol. 113, p ing, DPhil Thesis, Dept of Engineering Science,
159,Apd 1991.
University of Oxford, 1989.
5-55 Ainswonh RW, Dietz AJ and Allen JL
Investigating the Performance of Mininhve Semi-con- 5-68 Dwrly DJ, Oldfield MLG and Scrivener CTJ
ductor plesrm Tmnsducersfor use in Fast Response Wake Passing in a Turbine Rotor Cascade, AGARD
Aemdynamic Probes,Roc.of the 10th Symposium on CP 390,1985.
Measuring Techniques for T " I i c and Supersonic
Flow in Cascades and Wbomachiues, Brussels, 5-69 Dwrly DJ and Oldfield MLG
1990. Simulation of Wake Passing in a Stationary Turbine
Rotor Curcade, ASME J. of Propulsion and Power,
5-56 Oldfield MLG,Schultz DL and Nicholson J H Vol. 1,No. 4, p 316,1985.
Loss Measurements Using a Fast Tmverse in an ILPT
Tmnsienf-Cascade,Symposium on "Measuring Tech- 5-70 Doorly DJ and Oldfield MLG
niques in Transonic and Supersonif Flows in Simulation ofthe Efects of Shock Wave Passing on a
Cascades and lbrbomachines, Lyon, France, Oct. Turbine Rotor, ASME J. Eng. for Gas lbrbines and
Power, Vol. 107, p 998,1985.
5-57
1981.
Sieverding CH,Arts T, Pasteels M H and Klinger P
Tmnsonic Cascade Performance M e a w " e n t s Using
5-71 LomasCG
Fundamentals of Hot Wire Anemomezry, Cambridge
e
a High Speed Probe Tmwrsing Mechanism in a Shon University Press, 1986.
Dumtion wind Tunnel,Symposium on "Measurement 5-72 Comte-Bellow G
Techniques in Transonic and Supersonic Flows in Hot Wire and Hot Film Anemometers, in Memure-
Cascades and lbrbomachiues", Oxford, UK,1988. ments of Unsteady Fluid Dynamic Phenomena, Ed.
5-58 Oldfield MLG,Jones TV and Schultz DL Richards, BE, M C ~ Hill,
W pp 123-162,1977.
On-Line Computer for Tmnsient-Turbine Cascade
Inslnuneiuation, IEEE. Trans, AES, 14, p 738,1978. 5-73 Comte-Bellow G
Hot WireA n m w r n e ~Ann.
, Rev. Fluid Mach.,Vol. 8,
5-59 Oldfield MLG,Burd HJ and Doe NG pp 209-231,1976.
Design of Wdeband A M I o Cixuits,
~ 16th S y m p
sium of ICHMT, Hemisphere, 1982. 5-74 Brown GL and Rebello M R
A Small Fast Response Pmbe to Measure Composition
5-60 Ainsworth RW, AUen E,D a n a MR, Forth U P , in a Binary Gar M i " , A I M . Vol. 10, No. 5,
Hilditch MA, Oldfield MLG and Sheard AG p 649,1972.
Dewlopments in Instrumentation and Processing for
Tmnsicnt Heat Tmnsfer Measurements in a Full-Stage 5-75 Ng, WF and Ep~teinAH
Model Turbine,ASME J. lbrbomachinery, Vol IU,pp Unsteady Losses in T m o n i c Compnsrors. ASME J.
20-27.1989. Eng. Gas 'liubines and Power, Vol. 107, pp 345-352,
5-61 Epstein AH, Guevette GR, Norton RJG and
Yuzhang G
1985.
5-76 Holmes BJ and Gall PD
e
High Frequency Response Heat F l w Gauges, Rev. Sci. Liquid C~~stals
for High Altitude in Flight Boundary
Insh~~.,
57(4), pp 639-649.1986, Layer Flow Vilisntion, AIAA Gea Aviation Tech-
nology Cod., Paper 86-2592, Anaheim,1986.
5-62 Doorly JE and Oldtield MLG
The Theory of Adwnced H a t Tmnsfer Gauges, Int. J 5-77 RedaDC
Heat Transfer,Vol. 30, No. 6,1987. Obsemtion of Dynamic Stall Using Liquid Crystal
AIAA J., Vol. 29, NO. 2, pp 308-310.1991.
Cmtirg~,
5-63 Jones TV
The Use of Liquid Crystnlr in Aerodynamic and Heat 5-78 Oldfield MLG,Bryaaston-Cross PJ, Nicholson JH
Tmnsfer Testing, Proc. 4th Int. Symp. on Transport and Scrivener CTJ
Phenomena in Heat and Mass Transfer, KN Paper 7, A Smdy of Passage Flow through a Cascade of Turbine
pp 1199,Elsener, Sydney, 1991. Blndes wing Image-Plane Hologmphy, AGARD CP
399, Paper 33,1986.
5-64 Jones TV
Recent Developments in Tmnsient Heat T e r Mea- 5-79 Bryanston-Cross PJ, Towers CE,Judge T, H a r a s g m
surement, Proc. 9th Eurotherm Seminar, pp 73-80, SP,Hopwood ST and Towers DP
Entropie, Bochum, 1989. The Application ofhnicre Image velocimeny (Pw) in
a Shon Dumtion TmnsonicAnnular Turbine Cascade,
5-65 Jones TV and Hippensteele SA ASME Paper 91-GT-221, Orlando, 1991.
High Resolunon Heat Tmnsfer Coefficient Maps
Applicable to Compound CurveSurfaces UsingLiquid 5-60 The Math Works Inc., Matlab for 80386-Based MS
Crystals In a T m i e n t Tunnel, ASME HTD-Vol. 71, DOS Personal Computers, Users Guide, 21 Eliot
pp 1-9,1987. Street, South Natick, MA 01760, USA., 1989.
153
Chapter 6
Unsteady Flow in Cascades: Flutter & Forced Response
Daniel H.Buffum and Sanfordk t e r
6.1 INTRODUCTION function, with frequency close to a blade resonant natural fre-
Aerodynamically induced vibrations are generally classified quency, acts on the blades in a given row. Such forcing
as either flutter or forced VibratIOII. For flutter, the unsteady functions are generated at multiples of the engine rotational
aerodynamic forces acting on a blade depend only on blade frequency and arise from a variety of sources including aero-
motion, with the energy to sustain the motion supplied by the dynamic disturbances resulting from the presence of
air. For forced response, the aerodynamic forces that excite upstream and downsheam struts, stator vanes and rotor
the blade motion are independent of the motion In both blades, and disturbances due to inlet flow nonunifomities,
cases,thevibratorymotiouofthebladecanleadtofatiguefai- rotating stall patterns and compressor surge. A primary
lure. source of resonant forced vibration is the aerodynamic inter-
There are fundamental differences between the flutter and action between adjacent blade rows. The two principal types
forced vibration characteristics of turbomachinery blading of such interactionsare potential flow or staticpressure inter-
action and wake interaction. Potential flow interactions result
e
and those of aircraft wings. These are assomted with (Ref.
6-1) the relatively high density of the blade and (Ref. 6-2) the
neighboring rotor blades. As a result, blade vibration fre-
from the variations in the velocity potential or pressure field
assodated with the blades of a given row and their effect on
quencies tend to be higher than thaw of wings, with the the blades of a neighboring row moving at a different rota-
unsteady aerodynamics acting on a given blade depending tional speed. This type of interaction is of serious concern
intimately on the aerodynamic interaction between this blade when the axial spacings between neighboring blade rows are
and its neighbors. Also, because of the large disparity small or flow Mach numbers are high.Wake interaction is the
between the densities of the blade and of air, the mode and effect on .the flow through downstream blade rows of the
natural frequency of blade motion are essentially independ- wakes shed by one or more upstream rows and persists over
ent of aerodynamic effects. considerable distances.
6.1.1 Flutter
Under certain conditions, a blade row operating in a uniform t
I
flow field can enter into a self-excited oscillation known as
flutter. The blade motion is sustained by the extraction of SUPERSONIC STALL
[FLUTTER
energy from the flow during each vibratory cycle, with the
flutter frequency correspondinggenerally to one of the lower
blade or coupled blade-disk nafural frequencies. ’Ihis results
in high stressesin the blading and leads to very short term high Q
cycle fatigue failure. Compressor blade failure due to flutter
has been recognized as a major problem almost from the
5
Lu
U
SUPERSONIC
beginning of the development of axial flow compressors. In 3

U)
modern engines flutter can be encountered over a wide range U)
Lu
of operating conditions and imposes significant constraints U
a
on the design of fan and compressor stages. ’nubine blades,
particularly high aspea ratio aft stages of low pressure power
turbines, are also susceptible to flutter, although to a far lesser
extent than compressor blades. In addition, with the perform-
ance trend toward higher flow velocities that lead to shocks in
turbine blade passages, flutter may also become an increas- INCIDENCE
ingly important concern in the turbine stages of gas turbine FLUTTER
engines. MASS FLOW RATE
Boundaries for the most common types of flutter are depicted
schematically on a compressor map, Figure 6-1. The signific- Fig. 6-1 Principal types of flutter
ance of the steady flow field on flutter is clearly indicated by
the names of the individual flutter boundaries. Because of
their frequency of occurrence, subsonic/hansonic positive
incidence flutter and classical supersonic flutter have gener- The rotor speeds at which significant forced vibrations may
ally been regarded as the two most important typesof flutter. occur are predicted with frequency-speed diagram, Figure
Hence, they have received the most attention. 6-2. These display the natural frequency of each blade vibra-
tionmodeand theforcingfunctionfrequenciesasfunctionsof
6.1.2 Forced Vibration rotor speed. The intersections of these curves indicate the
Destructive forced vibrations can occur in fan, compressor, possibility of high vibratory stress. It is not possible to elimin-
and turbine blading when a periodic aerodynamic forcing atethesourcesofforcedvibrationfromtheoperatingrangeof
154
a NrbOInaChine. As a result, the number of vibratory fatigue ing engine environment.However,this environment is h d e
failures attributable to forced vibration problems may e d with regard to both instrumentation and isolation of
those due to flutter. Over the lifetime of the engine, forced parameters. To circumvent these problems, experimental
vibration problems can often prove to be more significant in investigators have turned to specialized facilities in which the
terms of cost and repair time than flutter. However, since essential features of the flow are preserved, but the environ-
forced vibration problems are not as crucial to an engine ment and condtions for scquiring meaningfulmeasurements
development program, they have not received the same are much more favorable. The particular experiment objec-
degne of attention. tives then specify the appropriate facility.
To investigate and quantify the unsteady aerodynamics of
flutter and forced response, experiment objectives may be
A IDLE DESIGN
categorized as: (1)analysis validation and M o n ; (2) flut-
ter data bank; and (3) concept investigations.
6.2.1 U p i s Validation and Direction

Thedevelopment ofanalyticallybasedflu~erorfo~vibra-
tion prediction design systems is directly dependent on the
availability of models which accurately predict the unsteady
aerodynamicsin the various flow regimes. Thus,detailed data
are necessary to expe.rimentally determine the range of valid-
ity of existing unsteady flow analyses and to direct the
development of advanced models. These data must be
obtained from experiments which model the fundamental
aerodynamic phenomena inherent in flow-induced turboma-
Chinery vibrations. Thus, the parameters which must be
modeled characterize both the steady and the unsteady flow
fields.
- The basic steady aerodynamics data of interest are the inlet
and exit flow field and the detailed blade row characteristies
ROTOR SPEED (RPM) including the surface pressure distributions and regions of
flow aeparation.These data are acquiredwith standard steady
Fig. 6-2 diagram
C~~~pb ell state instrumentation.
For forced response, the unsteady a d y n a m i c forcingfunc-
tion must be experimentally modeled, with the resulting
6.1.3 Gust and Motion Induced Unstendy Aemdylumics gust-induced unsteady aerodynamic blade IOW response
Theunsteadyaerodynamicsofflutterandforcedresponseare measured. Note that in most forced response unsteady aero-
closely related.Their classification is delineated by the rela- dynamic experiments, only the gust-induced, not the
tionship between the blade motion and the unsteady motion-induced, unsteady aerodynamics are investigated.
aerodynamics acting on the blading, as previously noted. Thus the unsteady data quantify the aerodynamic forcing
Namely, forced vibrations occur when the unsteady aerody- function, the harmonic surface pressure and pressure differ-
namic forcingfunctionsactingon the blading are independent ence distribunons and their phase relation with respect to the
of the blade motion. In c o n a t , flutter or self-induced aerodynamic forcing fimction, the unsteady aerodynamic M
osciuations occur when the unsteady aerodynamics acting on and moment, and the unsteady characteristics of the separ-
the blading are functions of the displacement, velocity or ated flow regions. Dynamic ' ' tation used to acquire
acceleration of the blading and these forces feedenergy into these data include hot-wire probes, LDVs, hot 6lm gages, and
the system. This leads to the use of the following WO quantit- mimature high respnse pressure transducers.
ies: the gust and the motion-induced unsteady aerodynamics, For flutter, the airfnls must be harmonidy oscillated at a
also termed the aerodynamic gust and the aerodynamic frequency such that the reduced f q u e n c y 1s of merest, i.e.,
damping, respectively. The gust induced unsteady aerody- thereduced~uencymustbclargeenoughthattheunsteady
namics are d e h e d herein as the unsteady aerodynamics of a aerodynamics is unsteady, not quasi-steady. The unsteady
blade row which are independent of the blade motion and are data quantifythe harmonic surfacepressure and pressure dif-
generated by upstream or downstream phased periodic dis- ference distributionsand their phase relations with respect to
turbances. The motion-induced unsteady aerodynamics the mohon of the bods, the unsteady aerodynamic lift,
detine the blade IOW unsteady aerodynamics due to har- moment, or aerodynamic damping and the unsteady charac-
monic, phased motion of the blades. Thus, it is the teristics of the separated flow regions. Dynamic
motion-induced unsteady aerodynamics which drive flutter. instrumentationused to acquire these data include hot-wire
For forced response, both the gust and the motion-induced probes, strain gages, hot film gages, and miniature high
unsteady aerodynamics come into play, accomplished by the response pressure transducers.
super position of these two fields.
6.2.2 Flutter Bouod.ry Dam Bank
6.2 EXPERIMENTOBJECXIVW The development of an advanced empirical or semi-empirical
From the dehitions of the flutter and forced vibration p m flutter design system requires an extensive and systematidly
blems, it is readily apparent that the experiments and obtained flutter boundary data bank, with operating condi-
measurements needed are most properly made in an operat- tions and geometries extended beyond curreat experience
levels. The parameters whch must be considered in the data Table ei
baak include the flutter mode, the reduced frequency, the Flutter and forced reaponse reducedfrequencies
blade incidence angle 01loading, the inlet Mach number, the
blade and cascadegeometry, the surge and chokemargins,the
air density ratio,and the temperature. FLUTTER TYPE REDUCED
(Mode) FREQUENCY, k
The primary data of interest quantify the effect of these para-
meters on the flutter boundary characcwisticsincluding: the Classical Supersonic (Tonion) 0.7-1.3
aerodynamic opera- conditionsat the nutter boundary, the Supe~nicStall (Bending) 0.2-0.5
flutta frequency, and the interblade phase angle. These data
SubsonicflransonicPositive Incidcna 0.4-1.6
are acquiredby means of standard insbumentation,mcluding
blade mounted &gages. (Tonion)
Negative Incidence (Bending) 0.3-0.5
6.2.3 Concept Investigatiom
New or advanced concepts fornutter stability or aerodynamk FORCED RESPONSE REDUCED
d y forced mponse enhancement must be experimentally MODE FREQUENCY, k
verified before they can be utilized in an engine design.
Depending on the pariicular concept,the necessary experi- First Bending 0.3-1.2
ments may involve the validation of math models, or the Fin[ Torrion 0.8-4.0
measurement of flutter boundary or forced response charac-
teristics. Second Bending 1.0-6.0
Second Tasron 1.6-10.0
6.3 EXPERIMENTDESIGN Chordwire Bending 2.0-20.0
In the design of cascade experiments directed at the acquisi-
tion of valid and useful unsteady aerodynamic nutter or
forced mponse data, the driving turbomachine phenomena
must be considered, including the key unsteady parameters For forced response, the interblade phak angle is dependent
inherent in hubomachines. on both the number of periodic disturbances per mlution
and the number of blades in the row. As the number of dis-
6.3.1 Ullstendyp.nmeters tonedregions,vanes,andbladesarevariedinnubomachines,
The unsteady aerodynamics of nutter and forced response in the interblade phase angle can range over almost any value.
turhmachines are characterized by a number of nondirnen- Note, however,that in contrast to the flutter of a given blade
sionsl parameters, including the Mach number, the blade row where N interblade phase angles are possible, them is
.
loading, etc. Tko of the primary unsteady Uow parameters only one forced response phase angle for a given blade row
whichmustbeconsideredwhendesigninganexperimentcon- and aerodynamic forcing function.
cemed with flutter or forced response are the reduced
frequency and the interblade phase angle.
For example, to analyze the unsteady aerodynamics gener-
ated by wakes from periodic flow obstructions upsh" of
When a turbomachine blade row encounters a flutter boun- the given blade row, it is assumedthat the wake has beenFou-
dary or a forced vibration resonance, the blade response rier decompwd, with only one component considered. Thc
involves an harmonic oscillation at a frequency (D. The classic wake spacing and the relative velocity of motion be"the
sssumptionthatallairfoilsintherowhavethesamefrequency obstructions and the blade row specifies thefrequency, w, and
and damping implies that a phased relatiou between the the interblade phase angle to the blade row.
motion of the airfoils exists. The phasing between rotor air-
foils is classically defined as the interblade phase angle, a.
The unsteady aerodynamics acting on the blading, for either
flutter or aerodynamidy forced response, are functions of
the reduced frequency, k. ILfpical values of nutter and forced
response reduced frequenaes in turbomachines are given in
Table 6.1. 6.3.2 U M W Flow ~ Generation
Analytical and computational models for the prediction of
WC rotor blade flutter uaneady aerodynamics and the motion-
k =- induced component of forced response typicdly sssume
V harmonic, constant interblade phase angle oscillation of cas-
cadedairfoils.Experimentsforthevalidationanddirectionof
where (D is the oscihtoxy frequency, C denotes the blade these models may be performedwith the same assumptions.
chord, and Vis the relative now velocity. In these, the ability to achiexe realistic values of the reduced
For both flutter and forced response, the blade vibrations are frequency at controlled interblade phase angles are major
characterized by a constant interblade phase angle, a. For obstacles. In particular, the following reguirements must be
nutter, the possible interblade phase angles depend on the considered.
number of blades in the rotor. with an infinite number of inter- 0 Tbe harmonic torsion and translation mode ~ ~ ~ i l h t i
blade phase angles possible. must be at a frequency such that the reduced frequency is
equal to that of nutter or forced response for the blade row
2n n
anuM= j = 0,...,N-1 being experimentallymodeled. This is t y p i d y a high fre-
"Vlrblndcr quency of osdation because of the high now velocitiesof
interest.
156
An essentially two-dimensional motion of the airfoilsin the

proper mode must be attained.
GRAPHITEEPOXY AIRFOIL
0 The torsion and W a t i o n mode drive systems must be
STEEL AIRFOIL
capableof forcingthe airfoils at a speci6ed frequency over a
complete range of interblade phase angle values.
6.3.1.1 Mode Shape and Frequency Requirements

Todemonstratethefirsttwooftheserequirements,asteeland
a graphitcepoxy composite airfoil are i n d g a t e d using
finite element models ofhigh frequency drive systems, Riffel
and Fleeter (Ref.6-1).
The finite element model of a high frequency torsion mode

drive system for double "ion airfoils schematically
depicted in Figure 6-3is considered. Spring bars provide nat-
ural frequency control of the airfoil system, with
moditications of the spxing bars used for minor frequency "0 0 25
adjustments of individualblades. For this example, the airfoil
cbordandspanareboth3.OOin.(7.62cm),withthermnnions SPRING 0AR THICKNESS (tn )
repnsented by a 0.5 in. (1.27 cm) square bar. Boundary con-
ditions at the bearing locations are established mch that the Fig. 5 4 1st torsion frequency vs spring bar thl&ess
airfoil can rotate but not translate.
AIRFOIL,
TOP VIEW
I I
Fig. 6-5 ?anslation mode frequencies vs support stiffness
FRONT VIEW A comparison of the bending mode shapesfor the two airfoils
at approximately 300 Hz is presented in Figure 6-6.The gra-
phite-epoxy composite airfoilis seen to result in nearly rigid
Fig. 5 3 Finite element torsion mode drive system model body translation across the span, whereas the steel airfoil
exhibits spanwise bending. This is significant in that the cas-
cade airfoils must &%it rigid body motion to insure the
The trend of torsional frequency with increasing spring bar two-dimensionality of the experiment.
thickness is established for both the steel and composite air- Thus, as determined by the above described model torsion
foils with this model, Figure 6-4.S i c a n t l y thinner spring and aanslation mode drive systems, airfoils fabricated from
bars are required to achieve a specified system n a n d fre- graphite-epoxy composite material are seen to offer many
quency for the graphite-epoxy composite airfoil than for the significant advantages over steel airfoilsat high frequency.
steel airfoil,thereby substantially reducing the r e q u i d airfoil For high frequency oscillating cascade experiments, airfoils
driving force. have been fabricatedfrom a combination of pre-impregnated
A similar analysis performed for an analogously idealized Kevlar cloth and graphite mat injected with epoxy resinunder
mslation mode drive system has been used to predict the pressure into a booking mold. Cloth fiber orientation was
hanslation mode frequency of the steel and grapbiteepoxy controlled to meet prescribed t o r s i d and bending stress
composite airfoils as a functionof the support stiffness. The requirementswhilemaintainingalowdensityandahighmod-
results, presented in Figure 6-5, show that less stiffness is ulus of elasticity. Hollow steel trunnions were attached to the
again r e q u i d for thegraphite-epoxy composite airfoil. airfoils at midchord. Cloth insertion and epoxy fill into the
GRAPHlTElEPOXY AIRFOIL
-TRUNNION 7- .-.-.-.-.- STEEL AIRFOIL
AlRFOlL-tTRUNNlON--
.-.-.- ._._.-.-.-.- .-.-.

-./.C.
2
a
w c
Y
-I
2U
0 -1.c
Fa. 5 6 Tmslation mode shapes at -3M)Hz
d o n s to provide adequate s t x q t h at critical airfoil- CartaandStHilaire(ReE6-2)usedafourbarlinkagetodrive

"ion interface suess locations. Splines located on the a -de in torsional oscillations.Reduced frequency values
d o n s were used to mount the airfoilsinto either drive sys- in excess of 0.3 have bcen achieved in this cascade for low
tall. speed flows.
The fabricationof the airfoils from graphite-epxy composite A high speed airfoil oscillation mechanism developed at the
material necessitated the use of nonconventional inshumen- NASA Lewis Research Center uses a camshan arrangement
tation techniques. In particular, to maintain the desired to drive the cascaded airfoils in controlled torsional oscilla-
overall composite material properties with no degradation of tions with a constant interblade phase angle. Buffum and
the airfoilsurface contours, provisions for dynamic instru- Fleeter (Ref. 6-3),Figure 6-7.Stainless steel barrel -, each
mentation were embedded in one airfoil during with a six cycle sinusoidalgroove machined into its periphery,
instrumentation. 'This involved molding the dynamic pressure aremountedonacommonrotatingshanwhichisdrivmbyan
transducer lead wires into the airfoilas part of the lay-up and 74.6 k W electric motor. A cam fdower assembly, consisting
molding process. The ends of the lead wires w r e then ofa tiraniumafloycomecthgannwitha stainlesssteelbunon
exposed and the transducers attached. on one end, is joined on the other end to an airfoil trunnion.
Both static and dynamic calibrations are reqmred on the The bunon fits into the cam groove, thus coupling the airfoil
installed dynamic pressure transducers to obtain quantitative to the camshaft.Lubricationfor the cam/foUower assembly is
data In addition, it is neeessBly to account for the accelera- provided by an 8OW-140oil bath. The interblade phase angle
tion sensitivity of the installed transducers, as will be is changed by rotating the c a m to new fmdrelative positions
discussed later in this chapter. on the shaft. Oscillation frequencies in excess of 500 Hertz,
corresponding to a reduced frequency of about 1 fortrans-
onic flow, have been achieved with this system.
6.3.3 DtivenAlrFdls
To address the third of the basic repuirements,namely that the Kobayashi (Refs. 6-4and 6-5)has used a novel approach to
torsion and aanslation mode drive systems must be capable achieve high reduced frequenaesusinga high speed mechani-
of forcing the airfoilsat a specified frequency over a complete cal cam drive system. Freon gas is circulated througha closed
range of interblade phase angle values, two experimental loop annular cascade wind-tunnel. The relatively low sound
techniques have been developed which drive the airfoils at velocity of the Freon allows the desired cascade Mach num-
high frequency and with a specified and controlled interblade ber to be attained with a lower inlet velocity. As a result, the
phase angle: (1) all airfoils oscillating and (2)unsteady aero- desired reduced frequencyis achieved with a lower oscillation
dynamic influence c d c i e n t s . frequency cq Equation 6-1.In practice, the requkd osdlls-
tory frequency is only about 40% of that mpired when air is
6.3.3.1 AllAirfoils Oscillaling the working fluid. For transonic compressor and turbine
The straightfonvardexperimentalapproach is to simulate the geomehies, reduced frequencies in the vicinity of 1.0 have
flutter unsteady aerodynamics by harmonically oscillatingall been reported.
of the airfoilsin the cascade with a constant interblade phase Electromagnetic excitation has been used to osdllste cas-
angle. A centrally-located airfoil is i " e n t e d to measure caded airfoils in both torsion and translation. Fleeter et al.
the fundamentsl quantity of interest, the unsteady airfoil sur- (Refs.6-6 and 6-7),Riffel and Fleeter (Ref. 6-1)and Bolcs
face pressure distribution, or derived quantities such as the and schlafli (Ref. 6-8).In such a system, each airfoilis elasti-
unsteady Me, momens or the aerodynamic damping. Several caUy suspended and electromagnetically excited at the
oscillation mec4anisms have been used,including bar linkage natural frequency of the airfoWsuspension system. Strain
systems, cam arrangements and computer-controlled e l m gages on the spring elements of the suspension system are
tromagnetq. used to determinethe airfoil motion.Computer-wntrol ofthe
SHAFl
SEAL
Fig. 5 7 NASA Lewis high speed airfoil oscillation mechanism
individual airfoil excitation systems allom the interblade cients define the unsteady moment coefficient developed on
phaseangletobefixed, the reference airfoil due to the motion of airfoil n. with the
remaining airfoils stationary
63.3.2 Unnendy Aemdynamic Influence Co&iems
The inherent complexity of driving all of the cascaded airfoils 2-
in simultaneous, controlled oscillations has led to develop
ment of an alternate experimental approach. When the
unsteady disturbances are small, as in a typical flutter stability
0
problem, anunsteady aerodynamicinnuencecoeffiicienttech-
nique may be utilized, Hanamura, Tanaka and Yamaguchi
(Ref.6-9); Davies and Whitehead (Ref. 6-10); Buffum and
Fleeter (Ref. 6-11). In this technique, only one airfoil in the
cascade is d a t e d , with the resulting unsteady quantities
measured on the oscillating airfoil and its stationary neigh-
-
bors. The unsteady aerodynamics of an equivalent cascade
with all airfoils oscillating at any interblade phase angle is then
determined through a vector summation of these Muence
coefficient data. Consider a finite airfoil cascade, F i p r e 6-8, -1
+
-
with 2N 1 airfoils executing constant amplitude harmonic
oscillations with a constant interblade phase angle p. The
resulting unsteady aerodynamic moment acting on the refer-
ence airfoil (Airfoil0)may be expressed as a Fourier series -2
Fig. 5 8 Twodimensional finite cascade model
Sample experimentallydetermined unsteady aerodynamic

where C t are the complex-valued unsteady aerodynamic moment innuence coefficieats for torsion mode oscillations
moment Muence coefficients. Thus these innuence coeffi- are presented in Figure 6-9, Buffum and Fleeter (Ref. 6-11).
159
INFLUENCE OF Nth AIR- Generation of the periodic unsteady flow fields appropriate
FOIL ON REFERENCE AIR- to forced response investigations is difficult in a hear u1s-
FOIL €"
cade, but much more readily achieved in an annular
environment. Both stationary annular cascades and rotating
cascades, as discussed in detail later, are useful for this pur-
pose.
6.4 DATA ACQUISI'TTON AND ANALYSIS
The experiment objectives spec@ the data acquisition and
analysis and facility requirements. In the deaign of cascade
experiments directed at the acquisition of valid and useful
unsteady aerodynamic flutter or forced response data, the
driving turbomachine phenomena must be considered,
-1
including the key unsteady parameters inherent in turboma-
chines. Note that the steady parameters are certainly
significant. However, this chapter is only concerned with
unsteady flow.
-2 Unsteady flowexperiments generate large quantities of data
which require high frequency transmission and recording.
Fig. 6-9 lnituence of nmairfoil on reference ailfoil These data involve a wide band of frequencies,with the ampli-
tude and phase relations between frequency COmpoIIentS of
importance. Hence, high speed, digital data processing is
Also shown are theoreticalinfluence coeffcientsbased on the essential.
classical flat plate analysis of Smith (Ref.6-12), for which the Gostelow (Ref.6-13) described several techniques for the
number of airfoils is infinite. Predicted values of C,(p) are analysis of unsteady turbomachinery data, including the use
obtained for -nspsn and used to evaluate the integral for @ , of time domain signal averaging and fast Fourier eansform
which is obtained by inversion of Equation 6-4. (FFTJbased computer techniques. The unsteady data of
interest for flutter and forced responseare periodic. However,
the general flowfield is comprised of threecomponents: (1) a
steady component, f ;(2) a periodic component, $;and (3) a
random unsteady component,f,. Thus, general unsteady data
can be expressed as follows.
The unsteady aerodynamic moment is seen to be a strong
functionof the unsteady aerodynamics assodated with oscil-
lating the reference airfoil itself and the two adjacent airfoils. f(t) = L + fp + f
The resulting imaginary part of the unsteady aerodynamic
moment obtained via summation of these influence coeffi- The steady component is typically measured indepcndently.
cients is presented in Figure 6-10 along with experimental Thus f is 888umed known and the unsteady data signal, s(t),
data for all airfoils oscillating and summationof the theoreti- may be considered to be composed of the sum of the periodic
cal influence coefficients. and random components.
0 DATA-ALL AIRFOILS OSCILLATING = fp + fl (6.7)

- MTA-INFLUENCE COEFFICIENTS
._....
THEORY (N.41
--- THEORY IN=ml The periodic signal component is determined by a digital
ensemble averaging techniquebased on the time domain sig-
nalenhancement concept. The unsteady signal is sampled and
digitized over a time frame that is greater than the periodic
signal component characteristic time. With the same data
initiation reference, a series of correspondingdigitized signals
is generated by repeating thissignal sampling and digitization
process. The time-vaxiant ensemble average is then deter-
mined by averaging this series of digital data samples.
3 m -120 -80 0 80 120 180

INTERBLADE PHASE ANGLE (DEGREES1
For a sufficiently large number of digital signals in the series,
Fig. 6-10 Imaginary unsteady aercdynamic mar" N >> 1, the ensemble average of the random signal compo-
correlation nent will be zero.
If the time required for signal digitization and updating of the
sums in Equation 6-8 is not small relative to the duration of
the experiment, a frequency domain averaging technique can
be used as an alternativeto the time domain signal averaging
technique. In contlast to the time domain signal averaging
Thus,the periodic component of the time-variant signal rela- technique, the signals covering the duration of the experiment
tive to the data initiation reference is determined by this may be digitized following a single trigger event. Following
ensemble averaging technique. that, each digitized signal is partitioned into contiguous
blockshavinganequalnumber of samples.Ea&blockisFou-
N rier decomposed in turn and referenced to the phase of the
(si) =
1
-
Ni-1
1 fp,, j = 1,2, ...m;N > > 1 (6.10)
airfoil motion signal (which is obtained by Fourier decompo-
sition of the corresponding block of the airfoil motion signal).
Once all of the blocks are decomposed in this manner, an
average for any frequency is obtained by averaging all of the
Figure 6-11 displays a time-variant pressure signal for 1 rotor
block results for that frequency.
revolution and averaged over 25,50,75, 100 and 200 rotor
revolutions. The ensemble averaging significantly reduces the
random flucNatiOns which are superimposed on the periodic 6.5 INSlXUMENTATION
signal with the, pressure signal essentially unchanged when The panicular experiment objective specifies the unsteady
averaged over 75 or more rotor revolutions. instrumentation requirements.
c
w N=100
TIME (msec)
Fig. 6-11 Ensemble averaging of unsteady pressure signals

6.5.1 High Respome Pressure Transducers pressure transducers have been reported by Minkin (Ref.
High response pressure transducers form the measurement 6-14) and Grant and Lanati (Ref.6-15). Lamboume (Ref.
basis of many investigations of fundamental unsteady aerody- 6-16) presents a discussion on measurements of unsteady
namics. The electromechanical configuration of the pressure pressure in which dynamic calibration concerns are noted.
transducer is similar to the basic strain gage in that the sensing These investigations emphasize the necessity of developing
element consists of a four-arm Wheatstone bridge bonded to quantitative techniques to ascertain the installed response
a thin flexible diaphragm, Figure 6-12. These transducers characteristicsof high response pressure transducers.
have several desirable characteristics: (1)frequency response One dynamic calibrationtechnique utilizes a resonant tube to
sufficient to measure at least several harmonics of the oscilla- generate plane acoustic waves to excite the pressure uansdu-
tion frequency; ( 2 ) small dimensions for good spatial cers at distinct frequencies. As depicted in Figure 6-15, the
resolution and minimal flow disturbance;(3) low sensitivity of assembly consists of a long tube with a speaker mounted at
output to transducer acceleration, and (4) invariance of the one end. An instrumented airfoil is inserted in the opposite
transducer dynamic response with change in temperature. In end of the tube, which is open to the atmosphere, with a refer-
addition, a mounting technique which effectively isolates the ence pressure transducer positioned adjacent to the
transducer from airfoil strain is necessary. instrumented airfoil. Amplified sine waves are used to drive
the speaker which in turn creates acoustic waves in the tube
6.5.1.1 Mounring Techniques for excitation of the transducers.
The principal difficulty in using such transducers is assoeiated
with mounting them in the airfoil. Namely, to obtain valid With the transducers in the airfoil, the dynamic response of
unsteady pressure data it is essential that the transducer be each wansducer is experimentally determined. The transdu-
mounted such that the transducer is decoupled from local cers are dynamically calibrated for pressure waves of various
a
s t r m on the airfoil,i.e., that they are isolated from the effects
'.
of nbration or other strain sensitivity Obviously, directly
attaching the transducer to a blade surface is not satisfactory,
frequencies and amplitudes, with the signals digitized and a
Fourier transform analysis performed on the data record.
Each transducer signal is normalized by the reference trans-
since all oscillatory strains in the blade would appear as pres- ducer signal and the amplification factors,A,, and phase shift,
sure outputs.An example of a mounting which decouples the +,are then determined. Figure 6-16 shows a typical response
transducer from the local strain is shown in Figure 6-13, curve for an airfoil transducer connected by a passage to the
Another transducer installation approach is the use of small point of measurement.
cylindrical transducers in cavities machined into the airfoils, To correct for passage effects, the transducer-passage system
Figure 6-14. is modeled as an under-damped second order system. The
amplification factor and phase shift are given by
6.5.1.2 FrequenqResponse
Dynamic calibration of the installed high response pressure
transducer is necessary. This is because the addition of coat-
ings and shields used to protect the transducers from erosion
and damage in the operationalenvironment can detrimentally (6.11)
affect their performance. Also,the pressure sensitive uansdu-
cer diaphragm is often located such that drilled holes, tubes,
or cavities connect it to the desired point of measurement, as
noted above.
-
where r = wez/w,,, we= = excitation frequency, w, - natural
a
Calibration considerations for installation of higb response frequency, and 5 damping factor.
SENSING ELEMENTS (ORIENTED TO

MINIMIZE CENTRIFUGAL FORCE EFFECTS)
0.010 MAX
Fig. 6-12 Dynamic pressure transducer electromechanical configuration

RADIAL DIRECTION . b
Fig. 6-13 Mounting to uncouple transducer from local strain
TRANSDUCER
SENSING
PRESSURE
ORIFICE
t
EPOXY
Fig. 6-14 Reverse dynamic pressure transducer mounting
OUTPUT FROM
TRANSDUCERS
REFERENCE
TRANSDUCER
AUDIO-AMPLIFIER
G
Fig. 6-15 Schematic of dynamic calibration facilily

163
1.25
TAP No. 11
a
p 1.20
0
3 1.15
0
1.10
I!
k
-I
a 1.05
2
1.oo
10
- 8
k?
0
6
3
W
0
w 4
a
8
2
EXCITATION FREQUENCY (Hz)
Fig. 5 1 6 Typical passage transducer response
The solid line in Figure 6-16is determined from thisequation zero mean velocity conditions. Figure 6-17shows an example
using a least squares tit for the value of the natural frequency
of the amplitude response of six coated transducers as a func-
and damping factor. These qmtions with experimentally tion of the acceleration magnitude, Bufhun and Fl&er, (Ref.
determined values for natural freuuencv and damuine factor 6-3). The response is a linear function of the acceleration,
are used to correct the airfoil traniduce;'s signals ;or passage
implying that the acoustic response,which will vary with the
effects. airfoil velocity magnitude, is dominated by the acceleration
When the transducers are mounted flush with the measure- response Significant amplitude variation is apparent among
ment plane, passage correction is not required. For example, the transducers and is probably due to differences in coating
dynamic calibration of flush mounted transducers p r o d u d thicknesses.In this case,the transducer phase angle variation
BBtresP0nseCU~Withle8SthanZdegreesofphsseshihand with frequencywas linear and small.
amplikation factors within 3% of unity for fmpencies l a
than 700 Hz. 6.5.3 L i g b t h b e s
Light probes are a valuable supplementto strain gages for the
6.5.1.3 Smin Isolation measurement of passage and non-synchronous vibration of
Airfoil strain may induce transducer strain, thus resulting in every blade in a rotor, particularly for detecting instabilities
an apparent pressure signal. Potting the transducer in lubber where there is a high degree of circumferentialresponse varia-
@W),as shown in Figure 6-13,has been found to be effec- bility. A constant spatial interval yields a constant output
tive. For example, in tests conducted using an airfoilof 7.62 dection, but flutter, which involves variable spatialintervals,
an chord with blade tip ddectiom of M.05 m no measur- yields a variable output deetion. Casing mounted light pro-
able effect on the transducer output to 0.01 millivolt, which bes are used to measure airfoil tip static and dynamic
corresponds to about 0.07kPa (0.01 psi) was found, Buffum displacements and have also been used to resolve flutter into
and Fleeter (Ref. 6-3). its principal bending and torsional contributions,K u r h and
Dim (Ref. 6-18),Niebading and Pollack (Ref. 6-19),and
6.5.1 Aeeeleration EK& Roth (Ref. 6-20).
During oscillatingcascade eaperimenb, the pressure transdu-
cers are subject to accelerationswhich may produce apparent 6.5.4 Hot-Wires nnd Films
pressure si&. To quan!ify this effect. the instrumented High response hot-wires and films are used to measure
airfoil is oscillated at several frequencies under no-flowor instantaneous flow velocities, with hot films being more dur-
TRANSDUCER advantages of each are considered and their role in defining
the unsteady aerodynamics of flutter and forced response dis-
cussed.
A 4
V 6.6.1 High Speed Rotating Rigs
0 6 High speed rotating rigs, which encompass a range from
single stage rigs to full engines, are the ideal vehicle in which to
establish both the basic aerodynamic and mechanical condi-
tions, as well as the corresponding flutter boundaries and
forced response characteristics, provide the ultimate in
unsteady aerodynamic data as well as the final test of adv-
anced concepts. The inlet and exit flows, the aerodynamic
forcing functions, the reduced frequencies, and interblade
phasing are obviously realistic. However, in this environment,
instrumentation and optical accessibility present major difli-
culties. The high centrifugal fields in which the rotating
airfoils operate limit locations of instrumentation.Concepts
to oscillate airfoils in this environment to obtain aerodynamic
damping information require unique attachment designs.
Requirements for multi-channel rotating, high speed slip-
rings to provide transfer of signals and referencepower across
the stationary to rotating frame of reference introduces
additional complexity. Hence facilities other than high speed 0
rotating cascades are often utilized.
I I I
100 200 300 6.6.2 Linear Cascades
ACCELERATION MAGNITUDE (g) The linear cascade is the most widely used tool in steady and
unsteady aerodynamic turbomachinery research. This is
because of its relative simplicity and flexibility with regard to
Fig. 8-17 Dynamic pressure transducer acceleration the overall facility, the experiment design and hardware,
response instrumentation,and data acquisition, analysis and interpre-
tation. For example, h e a r cascades provide a clean optic&
path for both steady and unsteady flow visualization of the
able but having a decreased frequency response. Hot films airfoil passage flow fields. Also, as there are no centrifugal or
attached directly to an airfoil surface measure skin friction, untwist stresses means that blade mounted instrumentation,
thereby determining the nature of the surfaceflow,i.e., thehot e.g., static pressure taps and high response pressure transdu-
film can be used to identify laminar, turbulent, or separated cers, can be positioned very near to the airfoil leading and
Bow, as well as the transition from one flow regime to another. trailing edges, thereby resulting in the quantification of the
detailed steady and unsteady flow fields.
6.5.5 Strain Cnges
Strain gages are the primary sensor to indicate the actual The objectives of unsteady aerodynamics research in linear
onset of flutter and the occurrences of forced response as well cascades have generally been associated with providing fund-
as the amplitude, frequency and interblade phase angle of the amental oscillating airfoil cascade data appropriate for both
resulting blade vibrations. The strain gage waveform is the flutter and the aerodynamic damping portion of the forced
result of the blade response to the instantaneous flow field. response problem. These data are intended to either validate
Thus, it is indicative of the physical mechanism driving the math models or direct advanced model development and
flutter or the forced response.Strain gages CM also be used to refinement, in contrast to developing a flutter data bank or
measure the unsteady aerodynamic lift, moment and damp- gust-induced unsteady aerodynamics data bank. This is
ing. However, limitations in the application of strain gages because it is extremely difficult to generate proper cascade
currently exist. For example, strain gage data give no measure inlet gust forcing functionsin a linear cascade. However, tech-
of flutter stability prior to the onset of flutter. However, techniques to oscillate cascaded airfoils and acquire data over a
niques to accomplish tbis have been considered, Loiseau, range of interblade phase angles have been developed.
Nicolas, and Maquennehan (Ref. 6-17).
Controlled oscillating linear cascade experiments are the
6.6 CASCADE UNtXEADY AERODYNAMICS most widely used means for establishing the validity and
Each type of research facility currently used to obtain experi- directing refinements of unsteady aerodynamic models.
mental data relevant to the unsteady aerodynamics of flutter These experiments provide increased flexibility as compared
and forced vibration of blading has both advantages and dis- to rotor investigationsin terms of variationsin aeroelasticand
advantages. Experimentalists have taken advantage of the flow conditions as well as experimental objectives. This
uniqueness of each facility and have designed controlled increased flexibility introduces complexities in that the oscil-
experiments to obtain basic data over ranges offered by the lating cascade must simulate the significant rotor Butter
particular facility dependent on the experiment objectives. phenomena. In this regard, a major limitation in obtaining
Four types of experimental facilities are considered herein: oscillating cascade data in a linear cascade is the proper simu-
(1) high speed rotating rigs, (2) linear cascades, (3) stationary lation of the reduced frequency. This limitation is associated
annular cascades, (4) rotating cascades. Advantages and dis- with the spanwise dynamic deflectionsof the airfoils becom-
165
ing non two-dimensional,ie,non-rigid body motions. ALSO, reducedfrequencyatflunercanbemain~nediathecascade.

increasing the reduced frequencyby increasing chord lengths However, because free flutter testing has no provision for
presents difkulties in obtaining required tunnel mass flow maintaining a constant interblade phase angle, the cascade
rates to acbieve proper inlet velocities. Decreasing the inlet and the corresponding rotor will likely have diffemt inter-
velocity to increase reduced frequency results in noncon- blade phasii.
formance to desired experiment objectives. Note that if the
'Ihe free flutter testing of a linear cascade of five airfoilsto
cascade experiment is intended to provide data to validate
determine the supetsonic unstalled flutter boundaries of a
analyses and quantify signiticant unsteady flow features not
analytically modeled, then the cascade geometry, flow condi- rotor was experimentally investigated by Snyder and Com-
merford (Ref.6-21). The data demonstrated non-uniform
tions,and reduced frequency do not have to wrrespond to
airfoil phasing as well as a large factor difference, of the order
th~eofanacNalrotor,althoughitmaybedesirable.Itisonly
necessaryfor the cascade experiment to simulate the assump-
of 2.5, between the fan and the cascade data. This relatively
poor correlation is attributable to the nature of free flutter
tions inherent in the model and for the reduced frequency to
testing in a finitecascade. In particular, the firstairfoil is near
be high enough so that unsteady data, not quasi-steady data,
toanisolated airfoil,andthe~ore,mo~stable thananairfoil
are obtained.
in an infinite cascade. As a result,this first airfoildoes not set
For reasons of wst and geometric fldbility;the use of a sta- up the correct unsteady upstream flow field to the next airfoil
tionary cascade to parametrically investigate flutter in the cascade. As this finite cascade effect progressively dec-
boundaries for a flutter data bank is highly attractive How- reases through the cascade, it might be overcome by utilizing
ever, the relation between the flutter boundary determined large numbers of airfoils.However, this is not typically practi-
from a stationary cascade and that for the Corresponding calinalinearcascade.
rotor must be carefully considered.
0
First, the airfoilcascade must be a valid representation of the
phenomena existing in the rotor blade row,with the unsteady
A number of linear cascade experiments have emphized
analysis validation and direction of fluner unsteady aerody-
namics, making airfoil surface unsteady pressure
data obtained at the high reduced frequency values which measurements in driven cascades over a range of steady flow
correspond to flutter of the rotor. Ranges of reduced fre- Mach numbers. For simultaneousoscillation of the cascaded
quenc~ values of sipifiutnce to rotor flutter are indicated in airfoils,CartaandStHilaire(Refs.6-2and 6-22)havestudied
Table 6-1. It should be noted that data obtained at lower cascades oscillating near stall and the effects of interblade
reduced frequencies and Mach numbers, for example, are still phase angle and incidence angle, all in low speed flows. Carts
of great value in the assessment of modeling assumptions and (Ref. 6-23) also investigated cascade dynamic periodicity in
predictions. the same facility. Fleeter et al. (Refs.6-6 and 6-7) and Riffel
Another consideration for unsteady cascade flows is that and Fleeter p. 6-1) have studied torsion and translation
some airfoilsin the cascade may be either more or less stable mode oscillating cascade aerodynamics in supersonic flow
than others. With a supersonic inlet flowto an airfoilcascade fields.Buffum and Fleeter (Ref.6-3) found that the bounda-
with a large stagger angle, for example,the bow wave from the ries imposed by wind-tunnel walls can have detrimental
second airfoilin the cascade will not intersect the f h t airfoil, effects in subsonic oscillating cascade experiments. Poor
Figure 6-18. Hence, the first airfoilacts as an isolated airfoil, dynamic periodicity was sometimesfound to occur, as a func-
for Mach numbers greater than 1.58, with this "isolated" cas- tion of interblade phase angle, due to pressure disturbances
cade airfoilstable at all frequenciesof oscillation. Thus,in this which interfered with the cascade &er reflection offthe Nn-
case it is unlikely that the finite cascade corresponds to a ne1 walls.
section of an infinite cascade where there is no first blade and
0the aerodynamics and motion of each blade are identical.
Influence coefficient-type investigations include Hanamura,
"&a and Yamaguchi (Ref. 6-9), who found good eorrela-
tion between results obtained by the influence coefficient
technique and data for all airfoils oscillating for flowin a
water channel. Hanamura and Yamaguchi (Ref. 6-24) s u b -
quently obtained intluence coefficient data in subsonic and
hansonic flows.Szechenyi (Ref. 6-25) has applied this tech-
nique to flow fields ranging from subsonic to supersonic.
Buffum and Fleeter (Rd. 6-3) found that wind-tunnel wall
effectsmay also be detrimental to subsonic influence codii-
cient experiments, causing an upstream-travelingwave which
would not exist if the cascade were not enclosed in a duct.
6.6.3 Stationary Annular Cascades

In a stationary annular cascade, a spanwise section of a tur-
bomachine blade row is modeled by stationary airfoil8
arranged in an annulus. Inlet flows ranging from subsonif to
Fig. 6-18 Supersonic flow field for large stagger angle supersonic, depending upon the manner in which flow is
cascade delivered to the cascade, have been generated in stationary
annular cascades.
I h e simplest and most convenient cascade experiment to Stationary annular cascades retain several advantages of the
obtain qualitative rotor flutter boundary data involves the free linear cascade such as optical and insnumentation accessibil-
flutter testing of a two dimensional linear airfoilcascade, with ity, elimination of centrifugal loading on the airfoils, and
the airfoils modeling the tip section of the rotor blading. The control of the velocity field. However, they suffer from some
166
of the same limitations as do linear cascades, for example For example, utilizing the stationaryvane rows of both single
obtaining proper reduced frequencies at realistic velocities and multistage compressors, with the aerodynamic forcing
for experiments directed at analysis validation and direction. function beiig the high reduced frequency WaLeS shed by the
In addition, the flow field may be three-dimensional, not two- upstream rotor blades, the effects of airfoil camber, rotor-sta-
dimensional, unless care is taken in facility and experiment tor axial spacing, and the waveform of the aerodynamic
design. Important advantages of the stationary annular cas- forcingfunctionontbefirstharmonicvanerowresponsehave
cade over a linear cascade include their circumvention of the beeninvestigated,Fleeter,etal.(Refs.6-30and6-31),Capece
periodicity questions and their ability to simulate forced et al. (Ref. 6-32)and Capece and Fleeter (Ref. 6-33).M u s
response forcing function gust unsteady aerodynamics.Also, et al. (Ref. 6-34)urilized five transducers embedded in each
a stationaryannular cascade eliminates many of the problems vane surface to quantify the unsteady lift coefficients corm
associated with using a finite linear cascade to represent a spondiig to the first five harmonics of rotor blade passing
rotor experiencing flutter. Namely, the annular cascade not “cy.
only eliminates the freeflutter interblade phasing and finite To demonstrate the complete data acquisition and analysis
d e considerations, but also offers reduced cost and unsteady aerodynamic technique, the gust response of a two-
increased flexibility as compared to a rotor. Thus, it can be per-revolution wake aerodynamic forcing functionto the first
used to provide a data bank for the development of engine stage rotor blade row of a low speed axial flow compressor is
flutter design criteria. considered, Manwaring and Fleeter (Refs. 6-35 and 6-36).
With regard to unsteady aerodynamics,experiments in annu- Figure 6-19shows an example of the Fourier decomposition
lar cascades have been directed at the analysis validation and of the two-per-revolution wake in terms of the nondmen-
direction of flutter aerodynamics,the development of a flutter s i o d streamwise and transverse harmonic gust components.
The complex unsteady pressure difference magnitude
data bank, and are currently being directed towards forced
response forcing function gust unsteady aerodynamics. For
example, Whitehead et al. (Ref. 6-26),Kobayashi(Ref. 6-27),
response of the first stage rotor blade to the firstfive harmon-
ics of this wake is shown in Figure 6-20for a high steady
0
and Bolcs and Schlafli (Ref. 6-8)instrumented annular cas- 1oadingIevel.Alsoshownarethecorrespondingflatplatepre-
cade blades, in some cases mounted on electrodynamic dictions.
shakers so as to enable flutter measurements. Kobayashi (Ref. For inlet distortions, O’Brien et al. (Ref. 6-37) used six
6-5)utilized an annular cascade facility to experimentally dynamic pressure transducers embedded on each rotor blade
investigate the effect of shock wave motion on transonic flut- surface to measure the unsteady aerodynamic response to a
ter. Jutras, Stallone, and Bankhead (Ref. 6-28)performed a distorted inlet flow field. However, the periodic rotor blade
comprehensive flutter test program at subsonic/transonic row inlet flow field was not measured and thus, the unsteady
Mach numbers in a stationary annular cascade designed to aerodynamic gust forcingfunctionnot quantified. Manwaring
investigate negative incidence choke flutter and positive and Fleeter (Ref. 6-35)investigated the effect of inlet distor-
incidence stall Butter in the Mach number range from 0.2 to tion waveform shape on the unsteady pressure response of a
1.2.Curren~yFleeterandHenderson(Ref.
6-29)areutjlizing first stage rotor row in a multistage compressor. Hardin et al.
a low speed, large scale, annular cascade to investigate the (Ref. 6-38)measured low reduced frequency oscillating air-
forced response gust aerodynamics, with the aerodynamic foil aerodynamics on a rotor of a single stage compressor and
forcing function being the wakes shed by rotating upstream also stated that they performed similar distortion experi-
rods and airfoils. ments, although results were not presented.
6.6.4 Rotating Caseades Aerodynamicallyforcedresponseof centrifugal compressors
Rotating cascades are defined as both the rotating blades and also has a relatively high frequency of occurrence. For exam-
ple, the impeller can be excited by the diffuser, with the
stationary vane rows in single or multiple stage turboma-
chines which have low through flow velocities. As a result, the excitation sourcebeing the steady but non-uniform flow field
generated by the diffuser, induding both vaned and vandess
0
steady loading, in terms of pressure ratio and equivalent
work, is very low. diffusers,and which is traversed by the rotor. Thus, the forc-
ing function to the impeller blades is the downstream
Primary advantages of low speed rotating cascades include potential field, i.e., the potential field of the radial diffuser
optical and instrumentation accessibility, low centrifugal drives the flow induced vibrations of the upstream impeller
loadings on rotating airfoils,and low horsepower required for blades. Other frequent aerodynamic sources of vibrational
drive considerations.In rotatingcascades,both upstream and response of centrifugal compressors include strong reverse
dowhstreamunsteadyaerodynamicforcingfunctiongustscan flowin the impeller, rotatjng stall, surge, and other flow inter-
be realistically simulated, with the exception of shocks from actions, Ellison et al. (Ref. 6-39),Haupt et al., (Refs. 6-40and
supersonic exit turbine nozzles. Specially fabricated screeus, 6-41),Bryan and Fleeter, (Ref. 6-42).
rods, and vane and blade rows can yield variations in gust
amplitudes, reduced frequenciesand interblade phasing. Var- Minimal quantitative data cumtly exists de6ning centrifugal
iations in the number of rotor and stator blades can be used to compressor unsteady aerodynamic excitation mechanisms.
investigate interblade phasing and reduced frequency effects This is due to both the dominance of the axial compressor for
on either the blading of rotor or stator airfoils. Spanwise aircraft applications and also the extremely complex flow
effects present challengesin measurement not encountered in fields.Namely, the flow in the impeller is highly complex, with
two-dimensionallinear and annular cascades. The major dis- the impeller interactions with the diffuser, ie., the vaneless
advantage of rotating cascades is their low Mach number and/or vaned diffuser, also play a significant role. As a result,
range. there are very few facilities for fundamentalcentrifugal com-
With regard to unsteady aerodynamics, experiments in low pressor unsteady aerodynamics resea& with primarily high
speed rotating cascades have been directed at the analysis val- speed test rie which are not necessarily suitableto direct the
idation and direction of forced response pmt aerodynamics. development or veritj unsteady aerodynamic models utilized.
167
FREQUENCY (Hz)
Fig. 6-19 Fourier decomposition of transversa and streamwise velacitbs for a WO-per-revolution wake
8
0
6
z
!
!
Y
lL
e
8W
38.5%
4
W
I+
U
W
Y
Y
a
W
U
3
U)
U)
W
E 10
>
$ z
U) t 5
6.0%CHORD
3 s o
1 2 3 4 5 HARMONlCNo.
0.3 0.6 0.9 1.2 1.5 REDUCEDFREOUENCY
Fig. 6-20 Blade unsteady pressure difference response to the first tnre harmonicsof the two-per-revolution wake
1BB
6.7 SUMMARY 6-5 Kobayashi H

Although engine testa offerthe true mechanical and aerody- Efict of Shock Wave Movement on Aerodynamic
namic environment, the difficulty encountered in Instability ofAnnular Cascade Oscillating in Transonic
insmudenting an engine and also the difficultyin isolating the Flow, ASME Paper 88-GT-187,1988.
fundamental parameters involved have made cascade enperi- 6-6 Fleeter S, McClure RB, Sinnet GT and Holtman RL.
ments customary for the study of forcedvibration and flutter. Supersonic Inlet-Torsional Cascade Flutter, AIAA J.
A number of experimental techniques are available for the Aircran,Vol. 12,No. 8,pp. 664669,August 1975.
study d flutter and forced vibration phenomena and the
fundamental components of these problems, the gust and 6-7 Fleeter S , Novick AS, Riftel RE and CaruthersIE
motion-induced unsteady aerodynamics.The chosen experi- A n Experimental Determination of the UnsteadyAer-
mental technique will largely depend upon the experiment odynamics in a cONmh?d Oscillnring Cascade,
objectives,forwhichthreecategonesareiden~ed:(l)analy- ASME J. of Enginsring for Power, pp. 88-96, Janu-
sis validation and direction;(2) flutter boundary data b a , ary 1977.
and (3) concept investigations. With the present emphasis on 6-8 Bolcs A and schlatliD
development of advanced analytical tools, analysis validation Experimental Investigation of the Tzme-Dependent
and direction experiments are at the forefront. Flow in a Kbmting Annular Cascade Operating in the
Avalidexperhntwillmodelthe trueunsteadyaerodynamic Tmnsonic Flow Regime, Proceedings of the Fourth
environment with appropriate steady and unsteady flow con- International Symposium on Unsteady Aero-
ditions.~eMaehnumberandsteadyloadinghaveimportant dynamics and Aemehticity of ~ b o m a c h i n e sand
influences on the unsteady behavior, and reduced frequency Propellers,Gallus HE and Servaty S,editors,pp. 245-
and interblade phase angle are primary unsteady flow para- 269,1988.
meters. In oscillating experiments, the ability to achieve
realistic values of reduced frequency is a major obstacle.
6-9 HanamuraY,%nakaHandYamaguchiK
A Simplijied Method to Measure UnsteadyForces Act-
a
High frequency responseinstrumentation,in theform of pres- ing on the WbmtingBlades in Cascade,Bulletin of the
sure transducers, strain gages, light probes and hot-wires, is JSME, Vol. 23, NO.180, pp. 880-887,1980.
essential. Digital data processing is also essential for dealing 6-10 Davies MRD and Whtehead DS
with the inevitably large quantitiesof data.Since the unsteady Unsteady Aerodynamic M u r s ~ r e m e nin~ a Tmnsonic
aerodynamic excitations are typicaUy harmonic, Fourier anal- Annular Cascade, Unsteady Aerodynamics of llu-
ysis is used to extract information on the various harmonics bomachines and Propellers, Cambridge University
from the time-depdent signals. Engineering Department, Cambridge, pp. 487-502,
Among the typea of cascade facilities in use, linear cascades 1984.
are most attractivefor the study of motion-induced unsteady 6-11 Buffum DH and Fleeter S
aerodynamics due to their relative simplicity and flexibility, Oscillating CascadeAerodynamics by an ExperimeNal
good instrumentation access and precise control of the flow Influence Coqpienl Technique, AIAA J. of Propul-
field. However,the difficultyin simulating the periodic excita- sion and Power, Vol. 6, No. 5, pp. 612-620,
tions &a'" 'c of forced vibration makes the stationaxy September-october 1990.
annular cascade and the mta!ing cascade the facilities of
6-12 Smith SN
choice for studying the unsteady aerodynamics due to peri-
odic forcing functions. In addition, due to their inherent
Discrete Frequency Sound Genemion in Axial Flow
Tiubomachines, Cambndge University Report
periodicity, annular cascadesare often used to obtain motion-
CUEDfA-Turbo/TR29,1988.
induced unsteady aerodynamic data.
6-13 Gostelow JF'
A New A p p m h to the Experimental Study of Tur
a
bomachinev Flow Phenomena, J. of Engineering for
Power, January 1977.
6-14 MinkinHL
REFERENCES Performance of Some Miniature Pressure Tmrwducers
6-1 Riffel RE and Fleeter S Subjected to High Rotational S p e d and Centripetal
Experimental Modeling of UnstalledSupersonic Tiub- Accelemtion, ISA AS1 76317,1976.
ofan Flutter, AZAA J. Aircraft, Vol. 18, No. 9, pp.
6-15 Grant HP and Lanati GA
718-724,September1981.
Instrumemtion for Measuring Dynamic Pressure on
6-2 Carta FO and S t Hilaire A 0 Rotating Compressor Blades, NASA CR-159466,
Experimentally Determined Stability Pammeters of a 1978.
Subsonic Cascade Oscillating Near Stall, ASME J. of
6-16 Lamboume NC
Engineering for Power, Vol. 100, pp. 111-120, January
Experimental Techniqw in Unsteady Aerodynamics,
1978.
AGARD Paper 679, Special Course on Unsteady
6-3 Buffum DH and Fleeter S Aerodynamics, 1980.
Aerodynamics ofa Linear Oscillating Cmade, NASA
6-17 Loiseau H, Nicolas J and Maquennehan B
'IM 103250, August 1990.
Measurement and Prediction of the Aerodynamic
6-4 KobayashiH Damping qfC o m p r e s m r B W , Aeroelasticity of ntr-
Unsteady Aerodynamic Chamcte&ics of Annular bomachines, Proceedings of the 2nd Intemanonal
Cascade Oscillating in Tmnsonic Flow. Bulletin of Symposium held in Lausanne, Editor: Suter P, Juris-
JSME,Vol. 29,No. 256,pp. 3303-3312,1986. Verlag-Xurich, pp. 399-408,1981,
169
6-18 KUrkW A and ECUS J neering for Power, Vol. 100, No. 4, pp. 664-675,
Synthesis of Blade Flutter Vibnuing Response Using October 1978.
Stationary Transducers, ASME Paper 78-GT-160, 6-31 Fleeter S , Capece VR,Manwarhg SR and Fleeter S
1978. Unnendy Blade Row Intenactions in a Multi-Stage
6-19 Neiberding WC and Polla& JL Compressor,AJAA J. of Pmpulsion and Power, Vol. 2,
Optiear Detection of Blade Fluter, ASME Paper 77- NO.2, pp. 168-174, March-April 1986.
m-66 andNASATM 73573,March 1977. 6-32 Capece VR and Fleeter S
6-20 RothH The UnsmdyAemdynamics ofa First Stage YoneRow,
Vibmtion Mmmmem on Turbomachine Rotor Experiments in Fluids, Vol. 4, NO.2, pp. 72-78,1986.
Blades With O p f h ~ l l WASME
~ ~ , Joint Fluids Engi- 6-33 CapeceVR and Fleeter S
neering and Gas "bine Conference, pp. 215-224, Unsteady Aerodynamic Intemtions in a Multi-Stage
New Orleans, 1980. Compmor, ASME J. of mrbomachinery,Vol. 109,
6-21 Snyder LE and Commerford GL NO.3, pp. 420-428, July 1987.
Supersonic UnstaUed Flutter in Fan Rotors: Analylical 6-34 Gallus HE, Lambertz J and Wallmann T
and Experimental Results, J. of Engineeringfor Power, Blade Row Intemction in an Axial Flow Subsonic
Vol. 96,NO.4, pp. 379-386,1974. Compmor Stage,J. of Engineering for Power, Vol.
6-22 CartaFOandStHilaireAO 102, pp. 169-177, January 1980.
Efici of Interblade Phase Angle and Inciuknce Angle 6-35 Manwaring SR and Fleeter S
on CascadePitchingStability,ASME J.ofEngineering Inlet Distortion Genemted Periodic Aemdymmic
for Power. Vol. 102, pp. 391-396, April 1980. Rotor Response, ASME Paper 89-GT-299, ASME
@6-23 CartaFO International Gas Turbine Conference, June 1989.
Unsteady Aemdynamics and Gapwive Periodiciiy of 6-36 Manwaring SR and Fleeter S
Oscilkuing Cascaakd AirfoiLv, ASME J. of Engineer- Forcing Function Efim on Rotor Periodic Aemdy-
ingforPower,Vol. 105, pp. 565-574, July 1983. namk Reqonse, ASME Paper 9O-GT-109 ASME
6-24 Hanamura Y and Yamaguchi K International Gas Turbine Conference, June 1990.
An Experimental Investigation on Aemdynamic Inter- 6-37 OBrien WF,cousinsW T and Sexten M R
blade Imemtions of a Vibmting Cascade in Tmnsonic Unncody h m Mensurements and Data A M W
Flow,JSME International Journal, Vol. 30, No. 270, Tmhniques in Axial Flow Compmors, ASME Joint
pp. 1919-1927.1987, Fluids Engineering and Gas %bine Conference, pp.
6-25 SzechenyiE 195-202, New Orleans, 1980.
Fan Blade Fluner -Single Blade Instability or Blade to 6-38 Hardin LW, Carta FO and Verdon JM
Blade Coupling?,ASME Paper 85-GT-216.1985. Unsteady A~rodyMmicMeasmments on a Rotating
6-26 Whitehead DS, Watson PJ, Nagashima T and Grant C o m p m r Blade Row at Low Mach Number, J. of
RI Turbomachinery. Vol. 109, pp. 499-507, October
A n Experiment to. Measm C0C;Bicienrr for Aiifoils 1987.
Oscillating in Cascade, Revue Franpise de Micha- 6-39 Ellison FF, Miere CE and Partridge JM
Nque, Num610 Speual, pp. 123,1976. Vane Vibmtion in Radial Flow lkbochaqws, hth-
6-27 K o l ~ p h i H tion of Mechanical Engineers, 1978.
Measuremenr of Unsteady Pressure Distortion on 6-40 Haupt U and Rautenherg M
Oscillating Airfoils in a Turbine A d a r Cascade, Blade Vimtion Measurements on Centr@gal Com-
International Gas lbrbine Congress, Tokyo, 1983. pressor by Means of Te&mehy and Hologmphic
6-28 J u t m RR,Stallone UT and Bankhead HR Intefemmehy, ASME J. of Engineering for Power,
Experimental Investigation of Fluter in Mid-Stage Vol. 106, No. 1, pp. 70-78.1984.
CompnuorDesip, AIAA Paper 80-0786.1980. 6-41 Haupt U, Jin DF and Rautenberg M
6-29 Fleeter S and Henderson G On theMechanism ofDangerow Blade Vimtion Due
NASA Lewis Researeh Center Grant NAG 3-656, io Blade Intenactions on CE-I Compmors,
3-0 Unsteady A e m i p m i c s and Aemehficity of ASME Paper 89-GT-291,1989.
Adwnced Turboprops. 6-42 Bryan WB and Fleeter S
6-30 Fleeter S, Jay RL and Bennett WA Flow Induced Forced Response of a Radial Cascade
Rotor Genemted Unsteady Aemdynamic Including h f l e and Incidence Eficts, AIAA Paper
Response of a Compmsor Stator, ASME J. of Engi- 90-2352, July 1990.
170
Chapter 7
Three-Dimensional Laser Anemometry in Annular Cascades
LJ. Coldman
SUMMARY The main advantage of the LFA system is that the Doppler
This chapter first presents a brief overview of the different shined frequencyis independent of the observation direction.
laser anemometer (TA) techniques available for the design of ThisallowstheuseoflargeaperturecoUectingopticsresulting
a tlueesomponent measurement system. Then, representa- in better signal-to-noise ratios. A schematic of a simple LFA
tive LA systems that have been used to measure all systemisshowninFigure7-1.Thelaserbeamisfirstsplitinto
three-components of velocity in annular geometries are twoequalintensitybeamsbyabesmsplitter(BS)beforebeing
described. Some of the design considerations unique to the focused by a lens (Ll). A fringe pattern is created in the com-
LA systems are then addressed. Finally, possible LA system mon region or probe volume (TV) of the beams. The velocity
enhancements are presented. component measured by the fringe system is normal to the
laser beams bisector (optical axis) and lies in the plane of the
7.1 U S E R ANEMOMETERTECHNIQUES two beams. However, no information is available as to the
7.1.1 Lnwr Doppler and "e-of-Flight Anemometers
A laser anemometer m e s u m fluid velocity indirectly by
determining the speed of particles moving through the flow
directional sense of the flow. When a frequency shift is intro-
a
duced to one of the incidentbeams (e.g. with a B r a s cell), the
fringe pattern appears to move. Therefore, the detected
field. These particles, which may occur naturally or be intro- Doppler shift increases or decreases, depending on the sense
duced artificially into the flow,must be sufticiently small so of the particle velocity with respect to the moving fringes, and
that they follow the flow closely. If this is the case, the fluid the directional ambiguity is removed.
velocity can then be equated to the measured particle velocity.
The two types of laser anemometers in common use are the
laser Doppler anemometer (LDA) and the time-of-tlight
(TOF) anemometer. Bs Beam splitter
L Lens
The LDA is based on the physical principle that light s a t - M Mirror
tered froma moving particle is Doppler shifted in frequency PH Pinhole
PMT Pholomullipliertube
from that of the incident laser light Measuring this Doppler 0 0 Greenbeam
shifted frequency allows the particle velcdty to be calculated. Laser beams 7
LDA theory has been covered extensively in the literature
(e.g. Refs. 7-1 to 7-4) and these references can be consulted
forfurtherdetails.TheTOF(ortwo-spotortwo-f~)anem-
ometer is based on measuringthe time it takes a particle to
cross two discrete, but closely spaced, laser beams. Knowing
the separation of the two beams allows the partide velocity to
be calculated. The principal advantage of the TOF anemome
ter results from the laser power being concentrated into a
smaller measurement r@on producing a larger signal-to-
noise ratio. This permits measurements to be made closer to
surfaces that are normal to the laser beam direction.In addi- Fig. 7-1 Schematic of single-channel fringe laser
tion,signals can be obtained from smaller particlesthat better anemometer system (backscatter configuration)
track the flow. The main disadvantages are that the M F
method gives a lower data rate and is unusable in highly turb-
ulent flows. Schodl (Ref. 7-5) has reviewed the available
information on time-of-flight measurement systems. The backscattered light from the seed particles passing
Although the M F technique has been used in turbomacbin- throughtheW(shownasdashedlinesinFi~7-l)iscoU~
ery mearch, it has not been widely applied to throughthefocusinglens(Ll).Whilemorescatteredlightcan
three-component velocity measurements in these machines. be collected in the forward d i d o n than in the backward
Therefore, further discussion of this type of anemometer is direction, and would be preferred, this is generally not poss-
delayed until Section 7.3. ible in complex annular geometries. After reflection by a
mirror (M), the scattered light is focused by another lens (LZ)
7.1.2 Sie-Component Systems onto a photomultiplier tube (I'm. A pinhole pH)is placed
The mathematical analysis of the Doppler shifted frequency in front of the PMT to "izethe extraneous light incident
equation by Boutier et al. (Ref. 7-6) led to their classification on the detector. The signal coming from the LFA can be pro-
of the four types of single-component laser anemometers: cessed by correlators, frequency trackers, spectrum
specoometer, Type la; reference beam, me lb; one beam, analyzers, counters, or frequency domain processors; coun-
ters are,however, the most widely used.
m e IC;and fringe, Type 1d. Of these, the most widely used
system today is the vpe Id or laser fringe anemometer (LFA). Another singlecomponent laser anemometer technique that
17l
has found use in special circumstances is the spectrometer 7.1.3 -0-Component System
method (Boutier's Type la), which includes both Fabry-Perot 'bo-component laser anemometer systems can be built from
interferometers (FPI) and Michelson interferometers. The the aforementioned singlecomponent or single-channel
principal advantages of the spectrometer method are that the measurement techniques. A singlechannel is used herein to
velocity component measured is parallel to the optical axis (as refer to the use of an optical train with a single PMT in the LA
opposed to normal in the LFA system) and that very large vel- system. Thus, two- or three-channel systems would have two
ocities can be measured easily. A schematic of a or three PMTs, respectively.
single-component FPI system is shown in Figure 7-2; it is 'bo-component LA measurements are commonly made with
similar to the LFA system discussed previously (Fig. 7-1). The either a single-channel or a two-channel fringe system. In the
basic difference is the addition (in the collection optics)ofthe single-channel system the laser beams (and resulting fringe
FPI, which consists of two partially transmitting mirrors fac- pattem) are rotated about the optical axis allowing the mea-
ing each other. These mirrors can be either flat or spherical. surement of velocity components in the plane normal to the
Spherical mirrors (shownin Fig. 7-2) are often used because optical axis. Measuring two different components of velocity
allows the velocity magnitude and flow direction to be calcu-
lated in this plane.
6S Beam spliaer A simple two-channel fringe system, shown schematically in
L Lens Figure 7-3, is essentially a combination of two single-channel
M Minor fringe systems that use two different colors. For argon-ion
PH Pinhde lasers operating in multiline mode, the green and blue lines
PMT Photomullipliirtube
oG Grwnbeam are usually used since they are the most powerful ones avail-
a Laser beams -j able. 'bo-channel, two-color fringe systems can use either
three or four beams, but four beam systems we more com-
mon.AsshowninFigure7-3,inthefourbeamsystem.thetwo
green (G) beams are arbitrarily in the horizontal plane while
the two blue (B) beams are in the vertical plane. The optical
axis is in the middle of the four beams. In a three beam system,
one beam is made up of overlapping a green and a blue beam.
-,i
a-PH The optical axis is usually the overlapping green-blue beam.
In addition to using Werent colors to separate the two chan-
nels, different polarizations or different frequencies can also
be used,however, these techniques are not as widely used.
Because the measurements of each channel can be forced to
be coincident in time, correlated turbulence information can
be determined in addition to the velocity information.
7.1.4 Three-Component Systems

Three-component k r anemometer measurement systems
can be built from the previously discussed single-channel and
two-channel techniques.Many of these LA systems have been
reviewed by Boutier (Ref. 7-6), Dancey (Ref. 7-7), and Mey-
en (Ref. 7-8); hoyever, the systems generally include only
a Fig. 7-2 Schematicof single-channel Fabry-Per01
measurements made in freejets or for flows in simple wind-
tunnels. Only more recently have the three-component LA
techniques been successfully applied to the flow passages of
interferometerlaser anemometer system (backscatter turbomachinery. As far as can be determined, few experi-
configuration) ments appear in the open literature where three-component
LA measurements have been obtained within the blade pas-
sages of complex machines. Of these experiments,fewer were
performed in annular cascades. However, since the major
the interferometeris less sensitive to mirror alignmentfor this emphasis of this chapter is on the descriptionof three-compo-
configuration. If the spacing of the spherical mirrors is equal nent LA measurement systems, all experiments performed in
to their common radii of curvature, the FPI is referred to as a annular geometries have been included because the LA tech-
confocal interferometer. The mirrors of the FPI form an opti- niques used therein would work equally well in annular
cal cavity in which successive reflections of the incident cascades.
monochromatic light create multiple beam circular interfer-
ence fringes (rings). The FPI can be used in a scanning mode The three-component LDA techniques that have been suc-
by placing a PH aperture at the entrance of the FPI (as in Fig. cessfully employed in annular geometries can be categorized
7-2) to restrict the transmission to the central fringe. The as follows: single-channel fringe system, Type 3a; two-channel
wavelength of the transmitted light from the FPI is directly fringe system, Type 3b; three-channel fringe system, v p e 3c;
related to the mirror separation. Therefore, different wave- and mixed configuration system, 3d. In the fringe sys-
lengths (or frequencies) can be transmitted and measured by tems (Types 3a, 3b, and 3c), the basic idea is to make three
adjusting the mirror spacing using a piezoelectric element. In independent noncoplanar component measurements, allow-
this mode of operation, the FPI can be considered as a scanning the velocity vector to be calculated. The method of
ing optical spectrum analyzer. accomplishing this is described in Section 7.2. However, the
l72
angles between the different velocity component measure- The four different threecomponent LDA systems just dis-
ments are generally much less than 90' because of the cussed are listed in Table 7-1 for future reference.
geometric constraints ofthe facility hardware.
7.2 THREE-COMPONENTLASER DOPPLER
?hisbas a negative impact on the accuracy ofdetermining the ANEMOMETERMEASUREMENTSYSTEMS
on-axis velodty component (along the he-of-sigh ot
f the This section presenis a descnption ofthe different types of
laser) and is discussed in Section 7.4. laser Doppler anemometer systems that have been used to
ob- three-component velocity measurements m annular
In the mixed configuration system, a spectrometer channel geometries.Also discussed are the methods of optical access,
(Boutier's la) is used in conjunction with a fringe system flow seeding, and probe volume positiorung. Fringe system
(Emutier's me Id) to measure all three components with a designs are presented first followed by the mixd-co&ura-
single optical axis or viewing direction. Small on-axiscompo- tion system. The main features of these facilities and laser
nent measurements are ditlicult to make using the anemometersystemaresummarized inTabIe7-2.Recentlya
spectrometer technique because of the presence of large fiber optics JDA p m k has been used for three-component
amounts of light scattered offwalls normal to the optical axis. measurements;it is described at the end of the section.
Bs Beamsplii
DESDichroic beam splittsr
DP Dispemkmpism
IF Intetferenosfilter
L Lens
M Mim
PH Pinhole
PMT Photomultiplier tube
.B Bluebeam
0 0 Greenbeam
GB GmandMuebeam
Fig. 7-3 Schematic of lwo-channel. two-color, four-beam fringe I s e r anemometer system (backscatter configuration)
Tab0 7-1
Three-component 18881anemometer systems
Desiunation Method
Single-channel fringe system with fringe rotation and variable

optical axis orientation
Two-channel, two-color fringe system with variable optkal axis

orientation
Thncchannel, three-color fringe system with two optical axes at

fixed orientation
Type Sd Sigle-channel fringe system with fringe rotation combined with a

single-channel Fabry-Perot interferometer with a single optical axis
or viewinw direction
173
Table 7-2
Three-component laser anemometer experiments
I University
Purdue
Strathclyde
University
Cincinnati
University
Virginia
Polytechnic
Institute
NASA Lewis
beareh
Center
Facility:
Machine Compressor Pnmp Turbine Compressor Turbine
Flow Axial' Mixed flow Radial Axid Axial
Blading Stator Rotor Stator Rotor Stator
Mphine ahe Luge Mdcl Actual Large ACtd
Laser ayatem:
Design (Table 7.1) Type Sa Type s b Type sc Type Sc Type Sd
L- power, w 1 a 5 4 4
Robt volume &e 100 pm d i m . BO pm dim. 50 pm d i m . 65 pm d i m . 1W pm d i m .
x I" x 1.4 mm x 0.6 mm x 1.3 mm xlmm
Optical access:
Window type Cwed Curved Flat Flat Cwed
Window material Plexiglsl Acrylic PldglM PldglM GI-
seeding:
Material Ropylene glycol Oil Propylene glycol S o p Fluorescent dye
Putisle sise, pm d i m . 0.6 2.0 2.0
Traversing syntem:
Typs of motion, no. 5 4 s S 6
Stage .ccur.cy, p m so to so0 2w 15 50
7.2.1 Single-Chnnnel Fringe System

For a single-channel system m p e 3 4 Table 7-1). three non-
I \
coplanar velocity component measurements are needed. Frmt r m , Rear
Generally these measurements are obtained by changing the
orientation of the optical axis with respect to the cascade. A
system of this type has been used hy Stauter and Fleeter (Refs.
7-9 to 7-11) in a stationary subsonic annular cascade of com-
pressor blades. Details of the Purduc Annular Cascade
facility design are given in (Ref. 7- 12).7he facility and the LA
system are shown in Figures 7-4 and 7-5, respectively.
The laser anemometer system employed a 4-W argon-ion
~aseroperatinginasingleLinemodeata514.5-nmwavelength
(green h e ) prcdumg a mmmum of 2.3 W of power. The
backscaner optical design included beam expansion (to dcc-
rease probe volume size) and frequencyshihjng. A field stop
was utilized in the LA system to provide good spatial resolu-
tion for the receivingoptics. This allowed measurements to be
made to within 3.5mm (2.2 percent of span) of the cascade
endwalls. The probe volume was about lmm in length and
less than l o o p in diameter and contained 30 hinges. Signal
detection and processing were done with a photomultiplier
tube and a counter-type processor.
Blade
Optical access to the test section was provided by a curved identification J
Id-mm-thick Plexiglas window covering 4o'of circumferen-
tial arc. The five airfoils centered on the window were
~ ~ t i l ~ ~from
e r ethed hub to provide maximum optical access
for the laser.
The flow was seeded upstream of the bellmouth with 0 . 6 - p -
diameter propylene glycol particles. A specially designed
seed injector (Ref. 7-11) provided particles with the same
velocity as the inlet flow and with minimal injector wakes.
cambered
airfoils (36)
J
,,
- Inlet wpmn
st"
..
The traversing system (Fig. 7-5) had four types of motion:

axial, vertical, tilting (corresponding to tangential motion), Fig. 7-4 Schematicof Purdue annular cascade facility
174
take into account the optical deviations caused by the curved

window.
The axial velocity component was measured directly by
orienting the beam plane parallel to the cascade axis. The
tangential and radial components were determined by placing
the beam plane perpendicular to the cascade axis and by
obtaining two separate measurements each with the optical
axis at different angles to the radial direetion.The o p t i ”
angles (largest values consistentwith the hardware geomeuy)
were computed automatically, depending on the desired
probe volume location.At each measurement location, 3000
data points were acquired within typically 30 to 45 sec.
1.2.2 n ” e l Fringe System

For a two-channel, hvo-color fringe system mpe 3b,Table
7- I), measurements are needed at only two orientationsof the
optical axis. Since two components are measured for each
optical axis position, one redundant measurement is
obtained. A system of this type has been used by Carey and
Fraser(Refs. 7-13to7-18)inamcdelmixed-flowpump.Ref-
traversim I
erences 7-13to 7-16 present results for the rotahg blading
(impeller) while (Refs.7-17and 7-18)give data for the vane a
less region downstream of the rotor. Detailsof the Strathclyde
University facility design are given in (Ref. 7-19).A plan view
of the facility and the LA system is shown in Figure 7-6.A
Fig. 7-5 Schematic of traverse system of Purdue annular cross sectional view of the pump is shown in Figure 7-7.
cascade facility The laser anemometer consisted of a two-channel, two-color,
three-beam, color separation system with confocal back scat-
and radial (along the optical axis). Axial motion was accom- ter collection optics. A 2-W argon-ion laser provided green
plished by sliding the positioning superstructure along the and blue lines.An achromatic focusing lens produced a probe
substructurewith a leadscrew. The vertical and tilting motions volume of 80 pm in diameter and 1.4mmin length.The probe
were obtained from two independently controlled leadscrews volume contained orthogonal sets of green and blue fringes
that supported both ends of the optic table. Radial motion with spacings of about 4.7 and 4.2 pm, respectively. A Bra=
was accomplished by moving the focusing lens with a fourth cell was used to frequency shift the combined green-blue
leadscrew. These leadscrews were driven by stepper motors beam.The effective velocities of the green and blue fringes
and had a positioning resolution of O.O064mm,except in the could be independently varied using separate photomulti-
axial direction where the resolution was 0.025mm. A fifth plier signal mixers.
type of motion (liinge rotation) was provided by rotating the Access for the laser beams was provided by two large win-
optical train of the laser anemometer. A computer was used to dows that covered the region from the rotor inlet to the stator
control moving the traversing system to the proper locations exit. The windows were molded from 2.5-mm-thick acrylic
to obtain measurements at a specified probe volume position. sheet to conform to the rotor and stator casing profiles. Both
In addition, the computer allowed corrections to be made to moldings were done on forms in a furnace and the windows
Bwster fan Dynamometer

/
Fig. 7-6 Plan view of mixed-flowpump facility at Strathclyde University

175
Stator - I
Fig. 7-7 Sectional view of model mixed-flaw pump at

Strathclyde University
then polished to give good optical clarity. The rotor window
acovered appmximately 20' of circumferential arc while the

stator window covered 90' of arc. In addition, the stator win-
dow was designed to be easily removable so that the inner
surfaces of both windows could be cleaned periodically.
The flow was seeded by injection of 2-pn-diameter oil drop- Fig. 7-8 Relationship between model and traverse
lets produced by a blast atomizer and was injected just coordinate systems and measured velocity components at
downstream of the honeycomb flow straighteners (Fig. 7-6). Strathclyde University
When the seed particles passed though the probe volume the
signals from the green and blue channels were measured by a
pair of counter-type processors. The digital outputs from the by Hathaway et al. (Ref. 7-20). 'Ibis system is used to obtain
counters were sent to a computer-conmlled data acquisition measurements in a low-speedcentrirugal compressor.
system. Information obtained f"an optical encoder on the
rotor shaft enabled the computer to calculate the relative shaft 1.2.3 Three-chplmel Fringe System
position associated with each measurement. Whenever fre- Three-channel fringe systems 3c, Table 7-1) are essen-
quency measurements were made on both channels tially a combination of a single-channel fringe system (as in
simultaneouslyfrom a single particle, the computer recorded Fig. 7-1) and a two-dmnnel, two-color fringe system (as in
the two frequencies together with the angular position of the Fig. 7-3). A third laser color (i.e., violet for an argon-ion laser)
shaft at the time of the measurement The rotor blade pitch is used in the single-channel system which is set at an angle to
was divided into 40 equiangular sectors (or windows) and the the t w o d a m e l fringe system. Three velocity wmponent
data was sorted into these sectors according to the rotor posi- measurements can be made with only one orientation of the
a '
tion. The average number of t w o d a m e l observations
obtamedineachsectorwas 1000,withamini"ofappmxi-
mately 400.
laser optics and these can be either non-simultaneous or
simultaneous measurements.
7.2.3.1 Non-simulraneour Measmmenn
The laser anemometer optical system was mounted on a tra- A threeshannel, threecolor fringe system was used by
versing table that allowed movement in a horizontal plane in Emglu and 'Ihbakoff (Ref. 7-21) for measurements in the
two perpendicular directions;one direction being parallel to guide vanes of a radialintlow turbine at the University of Cin-
the pump shaft axis or z-direction (see Fig. 7-8). Vertical cinnati. For this same turbine con6gurati011, data were
movement of the table along its support structure was also obtained in the vaneless region by Lakshminarawnhs ' et al.
possible. The optical axis of theIA was always horizontal and (Ref. 7-22) and in the turbine scroll by Malak et al. (Ref.
normal to the vertical plane of the pump shaft axis. Therefore, 7-23). For all these tests,the rotor was replaced by a body of
displacing the optical axis from the horizontal plane of the revolutioninstalled to provide a smooth continuousflow path
pump shaft by a distance h (see Fig. 7-8) was equivalent to d o w n s " of the nozzles.Coincidence of the thnevelocity
inclining the optical axis to the radial direction of the pump by component measurements was not used because of the long
an angle 5 Measurements taken with the optical axis at two experimental times (low data rates) needed to obtain simul-
different vertical displacements allowed all three components taneous data. Therefore, data were collected independently
of velocity to be calculated. In addition, tilting the optical axis from the three channels.
from the radial direction was used to minimize the size of the The laser anemometer system shown in Figure 7-9 used a
region not accessible to the laser beams because of the shad- 5-Wargon-ionlaserasitslightsource.Thelaserlightwassplit
owing caused by the blade twist. Also, the position of the into its three strongest components (pen, blue, and violet)
window in the rotor casing could be adjusted, depending on by a dispersion prism. Polarization of each beamwas adjusted
the vertical position of the optical axis, by mta!jng the model with a polarization rotator, and each beam was split into two
pump casing about the shaIi axis. equal intensity componentsby beamsplitters. Frequency shif-
Another LDA system of this type has recently been reported ters were used between the beam splitters and the coUeCting
Pressure
Orificemeter,
. regulating
valve
High-pressure
air supply
1 Optid system lable
2 5 W argon laser
3 Polarizatiin rcdabrs. beam splitters. Eragg cdls
for green snd blue beam. and purpb team
phdodaedor
4 PolMzation rdslor, beam splitter for purpb beam.
z a t i n i L
valve and 2 pholodaecton tor green and blue beamp
5 Two 3.75 beam expanders and 480 mm lenses
6 Thm frequency shifler dawnmixers
7 Three phdunultpliim
8 Dataprocesson
9 IBM PC/AT mnputsr
10 Radialturbine
11 Flowadapter
12 *ing chamber
13 AIomizar
Fig. 7-9 Schematic d radial inflow turbine stator test rig and LDV system at University of Cincinnati
optics to reduce fringe bias and to determine flow direction. and 3mm or 23.6 percent span from the back endwall. It was
The blue and green beams were used in an optical train in the possible to measure as close as 0.5" from the blade surface
axialdirectionoftheturbineaodmeasuredthehorizontaland when the blade geometry did not shadow the measuring
vertical components of velocity The violet beam passed region.
through a second optical train at a 30' angle to the axial direc- As seen in Figure 7-10, one side of the scroll's square cross
tion and measured a non-orthogonal component of velocity seaion lines up with the nozzle sidewall. Thisflat wall, which
(Fig. 7-9). The orthogonal on-axis velocity component was was made of 6.45-mm-thick Plexiglas, allawed optical access
calculated from a transformation relation between the mea- for the laser measurements.
sured blue and violet components (Ret 7-21). W Oseparate
beam collimators, one on the blue-green train and the other An atomizer was used to seed the Bow with 2-pm diameter
on the violet train, were used before the beamsplitters to propylene glycol particles which were injected into the bot-
insure that the focused beam pairs intersected at the beam tom of the settliig chamber upstream of the radial turbine
waists. All six beams crossed at a wmmon measuring point (see Fig. 7-9). The signals from the laser anemometer were
resulting in a probe volume of about 5 0 p in diameter by processed by threecounter-type signalproces.sors. Data from
0.6mm in length and containing about 19 fringes.Scattered the signal processors were transferred to a computer for fur-
light from particles passing through the probe volume was ther processing and data storage (Ref. 7-21).
collected in the off-axis backscatter mode,toreduce the effect The laser anemometer system was mounted on a milling
of wall flare. In addition, the shroud pack) endwall was machine table which could be traversed 254mm in the axial,
coated with flat black paint to reduce retlections from the 457.2mm in the transverse, and 558.8mm in the vertical
laser beams. The closest that measurements wuld be made directions.The accuracy of the traversing system was 25 pm
was lmm or 7.9 percent span from the fmnt Plexiglas endwall in all three directions.
177
Polytechnic Institute's Axial Flow Research Compressor

Facility. Although the compressor could be operated with two
stages, for these tests a single-stage configuration was used.
Inlet The laser anemometer system m p e 3c, Table 7-1), shown in
+ Figure 7-12, used a 4-W argon-ion laser. Since the LA optical
system was similar to the University of Cincinnati system
described previously, it will not be discussed further herein.
However, a somewhat larger angle of separation of the two
optical channels (approximately40' as compared to 30') was
employed in this system.
A 4 Cross seckm A 4
Fig. 7-10 Schematicof experimental turbine configuration

at University of Cincinnati
a7.2.3.2 Simultaneous Measurements

A threechannel, three-color fringe system, similar to that
described previously, bas been used to make simultaneous
LA measurements by Chesnakas and Dancey (Refs. 7-24 and
7-25) in a low-speed, axial-flow research compressor rotor
blade passage. Figure 7-11 shows a plan view of the Virginia
Fig. 7-12 Low-speed axial flow compressor and LDA
system at Virginia Polytechnic Institute and State University
Drive motor - The position of the compressor shaft was encoded, with one
shaft revolution being divided into 480 windows, so that each
of the 24 rotor blade passages contained 20 measurement
zones. Data were recorded only when a window was enabled.
At midspan each window was 2.5mm long. The relatively
large window size was chosen to maximise the data rate at the
expense of spatial resolution. In addition, the measurements
were averaged over all blade passages to increase the data
a rate.
This method of taking data allowed the filter senings for each
window location to be adjusted depending on its velocity level
(which can vary widely in going from the blade pressure to the
suction side of the passage).
Optical access to the rotor blade passage was provided by a
6.4-mm-thick, flat, uncoated Plexiglas window. The use of a
flat window caused a maximum deviation of 0.8mm from the
contour of the cylindrical casing. The window dimensions
allowed optical access from 17mm upstream to l2mm down-
stream of the rotor. Access in the spanwise direction was
limited by the window size, the angle of separation of the two
optical axes, and the tilt of the optic table to values from
Discharge valve -, 45percent span out to the blade tip. Measurements near the
hub endwall could, therefore, not be made with this optical
setup.
The flowwas seeded with sugar particles, generally less than 2
pm in diameter. The selection of this seed material satisfied
the criteria that it be benign,since the flow from the compres-
sor dumped directly into the room. The sugar seed was
generated by supplying an aerosol seeder with a solution of
Fig. 7-11 Plan view of axial flow research compressor sugar and water. Evaporation of the water resulted in spheri-
facilityat Virginia Polytechnic Institute and State University cal sugar particles; this was verified by scanning electron
microscope photographs of the seed collected on glass slides. In addition to the velocity measurements,Dancey (Ref, 7-26)
A solution of 11 parts water to 1 part granular sugar by presented the complete Reynolds stress tensor measurements
volume was found to produce the best particle size distribu- obtained in this machine.
tion. The seed was introduced locally in the flow, about
1 5 0 ” upstream of the compressorbellmouth.This resulted A similar LDA system was recently used by Seelhorst et al.
in a maximum data rate of about 5OO/sec when collecting (Ref.7-27) to obtain measurements in afan turbine engine.
data in a single encoder window and averaging over all blades.
The seed injection position was chosen to minimize flow dis-
turbances at the measurement location while maintaining 7.2.4 Mixed configuRtioo system
some control over the seed trajectory. Evaporation of the In the mixed configuration system CrLpe 3d, Table 7-1). a
water to assure solid sugarparticles at the probe volume loca- spectrometer channel (Boutier’s Type la) was used in con-
tion was tested by running with pure water. A disappearance junction with a fringe system (Boutier’s Id) to measure
of the LA signal conkned that all the water evaporated all three components with a single optical axis or viewing
beforereachingthe measurement location, direction For a single-channel fringe system, rotation ofthe
fringe pattern (needed for the third velocity component mea-
The laser system was mounted on an optical table which was surement) prevents thistechnique from making simultanmus
attached to a mechanical traversing mechanism (see Fig. measurements.
7-12). With this device, the optical table could be moved
parallel and perpendicular to the compressor a d s (x- and A system ofthis type has been used by Seasholtz and Gold-
r-directions). Manually powered machine screws attached to man (Refs. 7-28 and 7-29) in a stationary annular cascade of
a dial index allowed positioning to within 0.05mm.Rotation core turbine vanes at the NASA Lewis Research Center. Fea-
of the rotor allowed for a third probe volume motion within sibility tests and preliminarythree-componentmeasurements
the blade passage. In addition, the optical table could be usmg the spectrometer technique are reported in (Refs. 7-30
rotated about the r-axis of the compressor.To maximize opti- and 7-31). The cascade design is described in (Ref. 7-32). A
calaccess to the blade passage, the table was tilted at an angle schematic of the cascade and LA positioning system is shown
of 45’25’, approximatelythe stagger angle of the blades. in Figure 7-13.
Annular
cascade 7
Fig. 7-13 Schematic cross-sectional view of core turbine stator cascade and laser positioning system at NASA Lewis
Research Center
179
With the mixed configuration system, the transverse (axial- the dichroic beamsplitter was passed through a narrow band
tangential plane) velocity components were measured with a laser line filter (LLF) to remove any residual fluorescence.
single-channel fringe system using a fluorescent aerosol seed Then it was focused through another 100-p-diameter PH
technique (Ref. 7-33). The radial component was measured before entering the confocal Fabry-Perot interferometer
with a scanning confocal Fabry-Perot interferometer using (CFPI). The CFPI had a free spectral range of 3 GHz, a maxi-
the light elastically scattered by the seed particles (i.e., not the mum transmission of about 10 percent, and was scanned with
fluorescence). a linear ramp generator. Because of this scanning, velocity
A schematic of the optical layout of the laser anemometer is versus time information was not available. The light exiting
shown in Figure 7-14. The 4-W argon-ion laser was equipped the CFPI was detected by photomultiplier PMT2 (RCA
with a temperature-controlled etalon and had a maximum 8850).Photon counting electronics provided a digital count
output power of 0.8W at a 514.5-nm wavelength (green line). rate corresponding to the light intensity passed by the CFPI.
The ethalon provided a single laser frequency that was neces- Additional factors to be considered when using a CFPI for
sary for the Fabry-Perot portion of the measurement system. laser anemometry are discussed in (Ref. 7-30).
Lenses L1 and L2 (see Fig. 7-14) functioned as mode-match- A Bragg cell was included in the optics (Fig. 7-14) to generate
ing lenses to position the beam waists at the focal plane of lens a reference signal offset from the laser frequency by 400
L3. The beam divider (constructed from two appropriately MHz. This signal allowed the CFPI to be scanned over a fre-
coated fused silica plates) split the single beam into two equal quency range less than its 3-GHz free spectral range,
intensity parallel beams. The divider was mounted in a motor- providing faster data acquisition.Further details of the optical
driven rotary mount so that the orientation of the fringes design can be found in (Refs. 7-28 and 7-29).
could be set at any desired angle. It was necessary to place the laser and the optics in a wooden
box covered with a layer of acoustic shielding to reduce the
vibration of the laser. Any laser vibration would causejitter in
the laser frequency, making the Fabry-Perot measurements
tProbe Dichroic impossible. (The acoustic noise level near the cascade was
IF. . . LArrL beam 105 dB.)
Optical access for the laser beams was provided by cutouts in
the outer vane ring and in the cascade outer housing located
downstream of the vane row (Fig. 7-13). In the window region
the vanes were machined to the vane tip radius to permit the
window to fit flush with the tip endwall. Both windows were
made from 3.175-mm-thick window glass. These windows
were formed into a cylindrical shape that matched the tip
radius by sagging them in a vacuum furnace onto a machined
graphite form (Ref. 7-32). The windows at the vane row and at
the downstream location covered about 39' and 56"in the cir-
1
'1
I Laser
t A
L1
cumferential direction, respectively, and were each 4cm high.
The flow was seeded with a fluorescent dye aerosol (Ref.
7-33). A liquid dye solution (rhodamine 6G in a mixture of
benzyl alcohol and ethylene glycol) was atomized with a com-
Fig. 7-14 Optical layout of system at NASA Lewis mercial aerosol generator. The seed particles (1.0- to 1.5-pm
Research Center diam.) were injected through a 6-mm-diameter tube into the
flow at the entrance of the bellmouth. The fluorescent dye
technique allowed measurements to be made by the fringe
The two parallel beams from the beam divider were focused part of the anemometer system close to the hub, vanes, and
by lens L3 after being reflected by mirror M5. Since this mir- windows. Fringe measurements could be made to about lmm
ror was mounted on a motor-driven goniometer stage, the of the hub (about 2.5 percent of span), while Fabry-Perot
optical axis could be positioned along the radial direction in measurements could only be made to about 3mm of the hub
the stator cascade. This allowed a direct measurement of the (about .7.5 percent of span). In addition, the FPI required a
radial velocity component with the Fabry-Perot interferome- higher seeding rate (typically 10 kHz) to obtain high quality
ter. signals than was necessary for the fringe system.
The probe volume was about 100 pin diameter and lmm in A three-axis positioning system with a 1 2 - p positioning
length with about eight fringes. Light scattered from particles accuracy and 1 - presolution was used to move the laser and
passing through the probe volume was collimated by lens L3. optics. The positioning system was controlled by a remotely
The collected light was then split with a dichroic beam splitter, located minicomputer. The beam divider and the goniometer
which reflected the green light and passed the longer wave- mounted mirror were also controlled by this system.
length fluorescence orange light. With the fringe system, seven independent components of the
After passing through a long wavelength pass filter LWF (to transverse velocity were measured at lo" intervals centered
remove any residual green light), the fluorescent beam was about the expected flow direction by rotation of the fringe
focused through a 100-p-diameter PH onto a photomulti- pattern. This allowed not only the axial and tangential velocity
plier tube PMTl (RCA 4526). The signal from PMTl was components to be determined by least-squares techniques
processed by a counter-type processor, using four fringe (Ref. 7-31) but also provided an estimate of their statistical
crossings, to provide velocity components transverse to the accuracy. For the Fabry-Perot system, radial velocity mea-
optical axis (in axial-tangential plane). The light reflected by surements were made at each of the seven orientation angles
180
used for the fringe system transverse component measure- The probe was designed to measure speeds up to 150mhec
ments. The standard deviation of these seven measurements using a IO-mW helium-neon laser operating at a 632.8-nm
provided an estimate of the statistical accuracy of the aver- wavelength for the light source. A 80-~unprobe volume was
aged radial velocity component. situated 12mmfrom the front face of the probe. A fringe spac-
With the mixed configuration LA system, on-axis (radial) ing of about 5 pgave 15 fringes in the probe volume. Further
velocity component measurements using the CFPI were only details of the probe design can be found in (Refs. 7-34 or
made (Refs. 7-28 and 7-29) in regions of the s-wall hub con- 7-35).
tour. For this situation, the optical axis was about 45” to the Initial tests of the instrument were made in a one-dimensional
mirror-like-finished s-wall, and the light scattered from this flow of air in a square duct. Measurements were made at six
surface tended to be reflected away from the collection optics angular positions (only three required) of the optical axis, and
back-scatter direction. In areas upstream and downstream of the consistency of the calculated velocity was found to be very
the s-wall, where the hub was cylindrical,CFPI measurements good. These tests also indicated that the laser produced insuf-
were not possible. Here, the optical axis was normal to the ficient light intensity in the probe volume, resulting in
endwall and large amounts of light were collected, resulting in measurements limited to low flow velocities. A 30-mW laser
an inability to extract the small on-axis signal from the noise. diode was considered as a light source to overcome this pro-
The mirrorlike finish was probably a detriment on these cylin- blem.
drical surfaces, since the light was reflected mainly in the Recently Ahmed et al. (Refs. 7-36 and 7-37) reported three-
backscatter direction. This is supported by the fact that inter- dimensional measurements made with this fiber optics probe
ferometer measurements were previously made in this in a small high-speed centrifugal compressor impeller.
cascade with a cylindrical hub configuration(Ref. 7-30) when
the endwall was not polished.
7.3 THREE-COMPONENTTIME-OF-FLIGHT
7.2.5 Fiber Optic Fringe System MEASUREMENT SYSTEM
A miniature three-component fringe-type laser anemometer Schodl and Forster (Refs. 7-38 and 7-39) have developed a
instrument based on fiber optics was developed by Ahmed et multicolor fiber optic probe based on the time-of-flight laser
al. (Ref. 7-34 and 7-35). This probe was intended for measur- anemometer technique for three-dimensional flow measure-
ing the boundary-layer in centrifugal compressors, and the ments. In this probe (see Fig. 7-16) the light from an argon-ion
design was based on a laser Doppler system using a digital laser operated in multicolor mode was coupled to a
correlator for signal processing. To obtain the three-dimen- single-mode fiber and guided to the optical head. The laser
sional measuring capability, the optical axis was designed to beam passed through a Wallaston prism, which produced two
be at an angle (about 22.5’) to the mechanical axis of the slightly divergingbeams of orthogonal polarization. From the
machine, as shown in Figure 7-15. Three-component mea- inner part of the collimator lens the two parallel, multicolor
surements can be obtained by positioning the laser beams at beams of different polarizations were imaged into the probe
three different angles by rotating the probe about its mechani- volume. Because of the chromatic aberration of the lens LS1
cal axis. The flow vector can then be resolved from these three each color was focused with a slight displacement along the
measured velocity components. optical d s (Fig. 7- 16).This three-dimensional measurement
Measurement
volume
QQ 98 Q
Section A-A
1 Mirror 10Front cover
2 Beamsplitter 11Receiver lens module
3 Receiver lens 12Main body
4 Window 13Rear cover
5 Transmitter lens 14Transmitter module
6 Pinhole 15Bearing plate
7 Single-mode fiber T Transmitter window
8 Multimode fiber R Receiving window
9 Cone
Fig. 7-15 Sketch showing general arrangement of three-dimensional microlaser anemometer

181
II .
I I
LS27 rPP2 r P P 1
. .
0
0
B Blue beam
G Green beam
1 Perpendicular polarization
l e II
PP
LS
Parallel polarization
Wollaston polarizing prism
Collimator lens
L1 Fiber collection lens
SM-F Single-mode fiber
MM-F Multimode fiber
MV Measuring volume
Fig. 7-16 Beam path of three-dimensional multicolor L2F velocimeter
system consists basically of two two-dimensional time-of- and an improved signal-to-noise ratio. In addition, another
flight systems. The first system (system A) measured the TOF fiber optic probe was described, and it was used to make
between beams 4 and Gl and the second system (system B) three-dimensional measurements in the same turbofan. This
measured the TOF between beams Gl and Bb Data acquisi- probe design uses two separate TOF systems inclined at
tion was obtained simultaneouslyfrom each system using two angles of approximately f6' to the optical axis of the probe.
sets of electronics. The measurement frequency (or data rate) 7.4 LASER ANEMOMETER DESIGN
of each system depends on the flow angle with respect to the CONSIDERATIONS
two beams and is a maximum for flow going through the cen- 7.4.1 Laser Anemometer System
ters of the two beams. For a given flow direction, the One of the first design decisions to make when measuring the
measurement rate cannot be a maximum for both systems A fluid velocity is which type of laser anemometer system, laser
eand B. Therefore, the ratio of the measurement frequency of
these two systems (i.e., (A - B)/(A -tB)) was found to be a
sensitive measure of the axial (or on-axis)velocity component
Doppler FDA) or time-of-flight (TOF), to use. The advan-
tages and disadvantagesof these two systems are discussed in
Section 7.1.1.Basically, a TOF system is superior for near-wall
(Refs. 7-38 and 7-39). It would seem that this technique for measurements while an LDA system is better for high turbu-
measuring the on-axis component would be quite sensitive to lent flows. Usually the choice will be dictated by the
the turbulence level (similar to the standard two-dimensional experimental hardware, test conditions, and expected flow
TOF systems) and to the uniformity of the seeding rate. regime under investigation. Comparing the two measurement
Preliminary tests using this laser anemometer probe were techniques by Schachenmann and Tropea (Ref. 7-41) and by
made on a small free jet. The laser anemometer had a beam Wemet and Oberle (Ref. 7-42) shows significant differences,
waist diameter of 8 pn and an axial center displacement of thus making this design considerationextremely important.
0.2mm for the green and blue beams and a separation of the A compromise between an LDA and a TOF system is a laser
different polarized beams of 0.2mm. Measurements were anemometer that uses elongated spots instead of circular
made on the nozzle axis (one nozzle diameter downstream of spots. These elongated spots cause the flow acceptance angle
the exit) at different orientations of the instrument axis to the to be more in line with the fringe system and, thus, measuring
nozzle axis. The experimental results showed good agreement highly turbulent flow near walls becomes feasible. Systems of
with the theoretical predictions. It was estimated that the this type have been described by Wemet and Oberle (Ref.
probe was capable of determining the flow vector to an angu- 7-42) who use four (two pairs of partially overlapping) ellipti-
lar resolution of about 1'. cal spots and by Boutier (Ref. 7-43) who uses two dashes.
Recently, Forster and Schodl (Ref. 7-40) presented measure- Promising developments in the design of miniature three-
ments made in a turbine guide vane and in a propfan using a dimensional laser anemometer probes using fiber optics also
modified version of this TOF system. In this new design, the make these systems candidates for turbomachinery research.
transmitting part of the velocimeter did not use fiber optics. These probes are particularly useful in situations constrained
This resulted in a higher light intensity in the probe volume to use small windows, such as in centrifugal compressors.
182
7.4.2 Cascade (possibly by local melting) when the probe volume is close to
The laser anemometry techniques discussed previously can the window surface.The very fine scribe lines observed on the
be applied equally well to turbine or compressor components. plastic prevented further high quality signals from being
Generally, LA measurements are easier to make in axial obtained in these window areas (Ref. 7-32). It is recom-
machines than in radial machines. This is the result of the mended, therefore, that glass be used for the window material
blade heights being smaller in the radial machines and the fact whenever possible.
that the probe volume is more often near the endwall surfaces.
Therefore, more diffuse light is scattered off these surfaces 7.4.5 Flow Seeding
and its collection by the photodetector results in lower signal- Particles that occur naturally in the air are, generally, not of
to-noise ratios. As the probe volume moves away from the proper size nor of sufficient concentrationto be useful for LA
endwall surface and less light is scattered off the wall, the measurements. Therefore, seed particles are added to the
measurements become easier and of better quality. For simi- flow and must be small enough to accurately track the flow
lar reasons, LA measurements are generally easier to make in changes that occur in passing through the machine. A particle
large-scale'turbomachinery than in actual size hardware. dynamics calculation (Ref. 7-46) should be made before the
Measurements using laser anemometry can also be made in experiment to determine the size of the seed needed. Since
both stationary and rotating cascades. If the blade row is sta- commercially available seed generators produce a distribu-
tionary, the probe volume must move in all three dimensions, tion of particle sizes, it is also important to estimate the actual
if the flow field is to be mapped completely. For a rotating size of the seed particles detected by a specific LA system.
blade row, one type of motion is already provided for by the The laser system parameters that affect this detected size are
rotation. Therefore, to completely map the flow field in this discussed in (Ref. 7-32).
situation, only axial and radial motion of the probe volume
would be necessary. However, additional information is 7.4.6 StatisticalAccuracy
needed to provide the probe volume circumferentialposition For three-component fringe LA measurement systems, the
relative to the rotating blading for each laser measurement. statistical accuracy of measuring the on-axis component is
Strazisar (Ref. 7-44) has reviewed the different methods for generally less than that for the transverse components. Orloff
determining the rotor rotational position using a once-per- and Snyder (Ref. 7-47) detefmined that the accuracy of the
revolution or a once-per-blade timing signal. on-axis velocity component is strongly dependent on the
included angle between the two optical axes or optical orient-
An interestingtechnique for obtaining measurementsin rotat-
ations used in the LA system; the smaller the angle, the more
ing machinery, described by Fagan and Selbach (Ref. 7-45),
inaccurate the calculation of this component. For turboma-
uses a TOF system in conjunction with an image derotator.
chinery applications, this included angle is limited by
This system allows measurements to be made in both the
geometric constraints to values typically between 30' and 40'.
absolute and relative frame within an axial-flow fan. However,
These small angles result in the statistical accuracy of the cal-
optical access along the axis of the machine is necessary.
culated on-axis component being significantly less than that
7.4.3 Optical Elements for the transverse components. By contrast, the spectrometer
To obtain the best LA measurements possible, the optics technique measures the on-axis component directly and
should employ high-quality, fast (low f-number) lenses. Faster would be expected to provide better statistical accuracy than
optics permits collecting more of the scattered light and, that from the three-component fringe system. This was con-
thereby, increases the signal-to-noise ratio. Faster optics is firmed in the measurements by Goldman and Seasholtz (Ref.
particularly important for turbomachinery applications 7-29) where similar statisticalaccuracies were obtained for all
where backscatter collection is necessary. Lenses of V2.5 or three components.
less are common in advanced system designs. Antireflective Orloff and Snyder (Ref. 7-47) also found that correlated
coatings increaselight transmission efficiencyand should also (simultaneous) measurements are more accurate than
be used on all optical elements, including the optical access uncorrelated (sequential) measurements. The three-channel
window. In addition, better measurements are theoretically fringe system (Type 3c, Table 7-1) described previously is
possible by using higher power argon-ion lasers to increase capable of simultaneous velocity component measurements.
the laser light intensity in the probe volume, thus increasing A system of this type that does not use coincidence to insure
the signal-to-noise ratio. In practice, increasing the laser simultaneous velocity component measurements from all
power might not help if a large amount of flare light is being three channels would seem inappropriate, but the advantages
collected by the PMT. must be weighed against the significantlylower data rates that
occur by forcing simultaneity.
7.4.4 Optical Access
Optical access to the flow field is a requirement for making 7.5 LASER MEASUREMENT SYSTEM
laser anemometer measurements. The window needed to ENHANCEMENTS
provide this access can either be flat or curved. In annular 7.5.1 Optical Improvements
machines, flat windows can lead to flow distortions and are A method of improving the performance of three-color LA
generally not employed. Curved windows must be used with fringe systems, employed by Boutier (Ref. 7-6), involves using
care since they act like lenses in the optical system.Because of a second laser source dedicated to generating the violet laser
this, windows are usually made as thin as is safely possible to line. When a single laser is used for all three colors, the power
prevent the laser beams from uncrossing when passing of the violet line is much less than that for the green or blue
through them. lines. Therefore, using a second laser allows more power for
Generally, windows are made from glass or plastic material. the violet line and results in each of the three fringe patterns
Our experience with plastic windows has not been good. We receiving more equal light energy. Boutier's wind-tunnel
found that the focused laser beams can damage the plastic anemometer system used two 15-W argon-ion lasers as light
183
sources. One laser was operated at 8W using all lines, which because the positioning system must move to adjust the opti-
produced 3-W green and blue lines. The second laser was cal axis for each measured velocity component.
used at maximum power producing a 2.7-W violet line.
Another method of improving the performance of LA sys- 7.6 CONCLUDING REMARKS
tems is the use of beam expansion to reduce the probe volume The laser anemometer systems described in this chapter are
size. This concentratesthe laser power in a smaller region and capable of measuring the three velocity components in annu-
results in better signal-to-noise ratios. The main problem lar geometries in both stationary and rotating blades.
caused by decreasing the probe volume diameter is the poss- Generally these systemsuse fringe-typelaser anemometers in
ible uncrossing of the laser beams in passing through the single-channel, two-channel, or three-channel designs,
curved turbomachinery window. To overcome this difficulty, although time-of-flight anemometers are beginning to be
these systems may require the use of window correction employed. A major difficulty with these systems is the large
optics. The basic idea is to have the window correction optics uncertainty when determining the on-axis component as
“uncross”the laser beams so that the window (which acts like compared to the transverse components. This uncertainty
a lens) can “cross” them again. For cylindrical windows, the results from the geometric constraint of the hardware that
design of window correction optics is described by Wernet limits the angle between the different optical orientations
and Seasholtz (Ref. 7-48). Even for this situation, the position needed for determining the on-axis component. The maxi-
of the correction optics cylindrical lens must move relative to mum angles possible in these systems are between about 30’
the focusing lens depending on how far from the curved win- and 40’. To overcome this difficulty, a scanning confocal
dow the laser probe volume is located. Obviously,this makes Fabry-Perot interferometer was added to a conventional
for a more complicated system.In addition,the glass windows fringe-type anemometer that allows the direct measurement
used in these systems must be of high optical quality since it is of the on-axis velocity component with a backscatter configu-
0 now a more critical optical element. Similar optical distortion
(uncrossing) of the laser beams has also been controlled by
ration. Only a single optical axis orientation is needed for this
system to measure all three velocity components. The advan-
special design of the window outer surface (Ref. 7-49). tages of this system for turbomachinery research are that it
Uncrossing the laser beams can also occur if the optical axis is does not require a large optical access port and that it is cap-
not normal to the window. Therefore, this technique may only able of measuring a relatively small radial velocity component
be useful for the combined fringe and Fabry-Perot interfer- when the transverse component is much larger. The system
ometer system (Type 3d, Table 7-1) since only a single optical also appears capable of obtaining measurementswith smaller
axis orientation (which can be normal to the window) is sized seed particles than are possible for the fringe system.
needed for all the velocity component measurements. However, the Fabry-Perot interferometer does have some dis-
advantages when compared to the fringe system used for the
A laser fringe anemometer system incorporating the window
transverse velocity measurements. These include the follow-
correction technique of (Ref. 7-34) and designed with a probe
ing: (1)inability to make measurements when the optical axis
volume of 50 pm and an f/2.5 focusing lens has been used
of the anemometer was normal to the endwall surface (pro-
recently by the author (Ref. 7-50). Comparing this newly
bably caused by the mirrorlike finish on the hub endwall since
designed system with the original (probe volume of 100 pn
interferometer measurements were previously made when
and an f/4 focusing lens) confirmed the improvements poss-
testing a cylindrical hub endwall that was not polished), (2)
ible with the new technique: the signal-to-noise ratio of the
increased data acquisition time (especially in a rotating cas-
measurements are higher, measurements from smaller sized
cade), (3) higher seeding rate, (4) possible need for an
particles are being obtained, and measurements close to the
acoustic enclosure to protect the laser from high noise levels
endwalls are possible without using the fluorescent dye tech-
(105 dB), and (5) inability to obtain velocity time history
nique. The increased level of complexity makes this LA
0 system more difficult to optically align and operate. Expertise
in optical design may also be necessary to build and fine tune
information.
Promising laser anemometer measurement systems based on
a system of this type. miniature three-dimensional fiber optic probes are also
described. These probe designs utilize both fringe-type and
7.5.2 Enhanced Data Acquisition time-of-flight techniques and should find application in situa-
A single- or two-channel laser anemometer system can tions constrained to the use of small optical access ports.
resolve the velocity vector, if three noncoplanar velocity com- Some laser anemometer design considerations and potential
ponents are measured. As discussed previously, this is problems were also discussed.The choicebetween an LDA or
accomplished by moving the optical axis of the anemometer TOF system is a critical one, since experiments have shown
relative to the hardware while maintaining the probe volume significant measurement differences. As a compromise
position constant.Improvementsin the accuracy of the calcu- between these two systems, one might consider the variants
lated velocity can be obtained by making more than three that have essentially elongated spots. Caution is necessary
velocity component measurements. A similar method of data when using plastic windows because of possible damage to
acquisition for coplanar measurements was described previ- the plastic from the laser beams. The dynamics of the seed
ously for the NASA Lewis Research Center annular cascade particles to faithfully track the flow must also be carefully
laser anemometer. In these tests, seven independent velocity examined. To do this calculation properly, the size of the seed
components were measured, and least-squares techniques particles actually detected in the LA measurement process
were used to obtain the two transverse velocity components must be determined, as aerosol generators produce a distrib-
and estimates of their statistical accuracies. Of course, for the ution of particle sizes. The statistical accuracy of the LA data
present situation, noncoplanar velocity component measure- is not only compromised by the small angles between the dif-
ments must be obtained to determine all three velocity ferent optical orientations used for the component
components. This type of data acquisition takes longer to measurementsbut also by the use of sequential(uncorrelated)
accomplish not only because more data are taken but also data acquisition.The three-channel, three-color fringe system
184
allows coincidence of the velocity component measurements, Pump Operating at Peak Eficiency, J. Fluids Eng., Vol.
and this may result in higher accuracy in the calculated veloc- 107, No. 2, pp. 183-190,June 1985.
ity vector. 7-15 Carey C, Fraser SM and Wilson G
Possible enhancements to the three-component laser anem- Behavior of Air in a Mixed-Flow Rotor Operating at
ometer measurement systems are also suggested. These Off-Design Duty Points, Second International Sympo-
include the following: (1) dual lasers for three-color fringe sium on Laser Anemometry, A. Dybbs and P.A. f i n d ,
systems, (2) smaller diameter laser beams (smaller probe e&., ASME, pp. 27-33,1985.
volume), and (3) multiple noncoplanar velocity component 7-16 Carey C et al.
measurements. Studies of the Flow ofAir in a Model Mixed-Flow Pump
Using Laser Doppler Anemometry, Part 2: Velocity
Measurements within the Impeller, NEL Report No.
REFERENCES 699, National Engineering Laboratory, Glasgow, U.K.,
7- 1 Durst F, Melling A and Whitelaw J H 1985.
Principles and Practice of Laser-DopplerAnemometry, 7-17 Fraser SM and Carey C
2nd ed., Academic Press, 1981. Investigation of LDA of the Fluid Behaviour Down-
7-2 Drain LE stream of a Mixed-Flow Impeller, 2nd International
The Laser-Doppler Technique,Wiley, 1980. Symposium on Applications of Laser Anemometry to
7-3 Durrani TS and Greated CA Fluid Mechanics, Paper 9.1, Instituto Superior Tec-
Laser Systems in Flow Measurements) Plenum Press, nico, Lisbon, Portugal, 1985.
7-4
1977.
WatrasiewiczBM and Rudd MJ
7-18 Carey C et al.
Decay of Rotating Wakesfrom a Mixed-Flow Impeller, 0
Laser Doppler Measurements, Butterworths, 1976. International Conference on Laser Anemometry:
Advances and Application, J.T. Turner and J. Stan-
7-5 Schodl R bury, eds., pp. 247-260, British Hydromechanics
Laser-Two-Focus Velocimetry,Advanced Instrumenta- Research Association, Cranfield, U.K., 1985.
tion for Aero Engine Components, AGARD CP-399,
pp. 7-1 to 7-31,1986. 7-19 Carey C et al.
Studies of the Flow ofAir in a Model Mixed-Flow Pump
7-6 Boutier A, DHumiers Ch and Soulevant D Using Laser Doppler Anemometry, Part I: Research
Three-Dimensional Laser Velocimetry: A Review, Facility and Instrumentation, NEL Report No. 698,
ONERA TP-1984-43,1984. National Engineering Laboratory, Glasgow, U.K.,
7-7 Dancey CL 1985.
A Review of Three-Component Laser Doppler Anem- 7-20 Hathaway MD et al.
ometry, Int. J. Optical Sensors, Vol. 2, No. 5-6, pp. Experimental and Computational Investigation of the
437-469,Oct.-Dec. 1987. NASA Low Speed Centrifugal Compressor Flow Field,
7-8 Meyers JF ASME Paper 92-GT-213,June 1992.
The Elusive Third Component, Second International 7-21 Eroglu H and Tabakoff W
Symposium on Laser Anemometry, A. Dybbs and P.A. LDV Measurements and Investigation of Flow Field
Pfund,eds.,ASME,pp. 247-254,1985. through Radial Turbine Guide Vanes, ASME Paper
Stauter RC and Fleeter S 89-GT-162, June 1989.
7-9
LDA Measurement of the Passage Flow Field in a 3-0 7-22 LakshminarasimhaAN. Tabakoff W and Metwally AM 0
Airfoil Cascade,AIAA Paper 86-0501, Jan. 1986. LDV Measurements and the Flow Analysis in t h e - b e -
7-10 Stauter RC less Region of a Radial Inflow Turbine, ASME Paper
Measurement of the Three-DimensionalAerodynamics 89-GT-157, June 1989.
of an Annular Cascade Airfoil Row, Ph.D. Thesis, Pur- 7-23 Malak MF, Hamed A and Tabakoff W
due University, 1986. Three-Dimensional Flow Field Measurements in a
7-11 Stauter RC and Fleeter S Radial Inflow Turbine Scroll Using LDY ASME Paper
LDA Measurement of the Passage Flow Field in an 86-GT-122, Oct. 1986.
Annular Airfoil Cascade, Third International Sympo- 7-24 Chesnakas CJ and Dancey CL
sium on Laser Anemometry, A. Dybbs, E Ali, and G. Total Wlocity Vector Measurements in an Axial-Flow
Momson, eds., ASME,.pp. 215-224,1987. Compressor Using a 3-Component Laser Doppler
7-12 Stauter RC and Fleeter S Anemometer, AIAA Paper 88-2811, July 1988.
The Design and Instrumentation of the Purdue Annular 7-25 Chesnakas C
Cascade Facility with Initial Data Acquisition and Total Velocity Vector Measurements in an Axial-Flow
Analysis)NASA CR-167861,1982. Compressor Using a 3-Component Laser Doppler
7-13 CareyC Anemometer,Masters Thesis, Dept. Mech. Eng., Virgi-
A Study of the Flow ofAir in the Rotor of a Mixed-Flow nia Polytechnic Institute and State University, 1988.
Pump Using Laser Doppler Anemometry)Ph.D. The- 7-26 Dancey CL
sis, University of Strathclyde,Glasgow, U.K., 1984. Measurement of Second Order Turbulence Statistics in
7-14 Fraser SM, Carey C and Wilson G an Axial-Flow Compressor via 3-Component LDA,
Behavior of Air in the Rotor of a Model Mixed-Flow AlAA Paper 90-2017,July 1990.
185
7-27 Seelhorst KA et al. terfor Three-Dimensional Flow Analysis, AIAA Paper

Time-Dependent 3-Component Laser-Doppler- 88-3034, July 1988.
Anemometer and Simultaneous Position Measure- 7-39 Schodl Rand Forster W
ments in the Flow of an Aircrap Engine, Fourth Interna- A New Multi Colour Laser Two Focus I/elocimeterfor
tional Conference on Laser Anemometry, Advances 3-Dimensional Flow Analysis, 13th International Con-
and Application, A. Dybbs, F. Ali, and B. Ghorashi, gress on Instrumentation in Aerospace Simulation
eds.,ASME, Vol. 2, pp. 425-430,1991. Facilities, pp. 142-151, Gottingen, Germany, Sept.
7-28 Seasholtz RG and Goldman LJ 1989.
Combined Fringe and Fabry-Perot Laser Anemometer 7-40 Forster W, Schodl R and Beversdorff M
for Three Component Velocity Measurements in Development of 3-0 VelocimetersBased on the Laser-
Turbine Stator Cascade Facility, Advanced Instrumen- 2-Focus Technique, Fourth International Conference
tation for Aero Engine Components, AGARD on Laser Anemometry, Advances and Application, A.
CP-399, pp. 13-1 to 13-15, 1986. (Also, NASA TM- Dybbs, F. Ali, and B. Ghorashi, eds., ASME, Vol. 2, pp.
87 322.) 693-699,1991.
7-29 Goldman LJ and Seasholtz RG 7-41 Schachenmann A and Tropea C
Three Component Laser Anemometer Measurements LTVSignal Processing in TurbulentFlows, Proceedings
in an Annular Cascade of Core Turbine Vanes with of International Specialists’ Meeting on The Use of
Contoured End Wall, NASA TP-2846,1988. Computers in Laser Velocimetry, Saint-Louis, France,
1987.
7-30 Seasholtz RG and Goldman LJ
Laser Anemometer Using a Fabry-Perot Interferometer 7-42 Wernet MP and Oberle LG
0 for Measuring Mean Velocity and Turbulence Intensity Laser Anemometry Techniques for Turbine Applica-
tions, ASME Paper 87-GT-241, May-June 1987.
along the 0pticaIAxi.sin Turbomachinety,Engineering
Applications of Laser Velocimetry, H.W. Coleman and (Also, NASA TM-88953.)
P.A. Pfund, eds., ASME, pp. 93-101, 1982. (Also, 7-43 Boutier A
NASA TM-82841.) Laser Velocimetry and Wall Approach Limitations,
7-31 Goldman LJ and Seasholtz RG Second International Symposium on Laser Anem-
ometry, A. Dybbs and P.A. Pfund, eds., ASME, pp.
Comparison of Laser Anemometer Measurements and
Theory in an Annular Turbine Cascade with Experi- 75-82,1985.
mental Accuracy Determined by Parameter Estimation, 7-44 Stazisar AJ
Engineering Applications of Laser Velocimetry, H.W. Laser Fringe Anemometry for Aero Engine Compo-
Coleman and P.A. Pfund, eds., ASME, pp. 83-92, nents, Advanced Instrumentation for Aero Engine
1982. (Also,NASA TM-82860.) Components, AGARD CP-399, pp. 6-1 to 6-32,1986.
7-32 Goldman LJ and Seasholtz RG 7-45 Fagan W and Selbach H
Laser Anemometer Measurements in an Annular Cas- Laser 2 Focus Flow Measurements in the Absolute and
cade of Core Turbine k n e s and Comparison with Rotating Frame inside the Blade Row of an Axial Flow
Theory, NASA TP-2018,1982. Fan, 3rd International Symposium on Applications of
Laser Anemometry to Fluid. Mechanics, Paper 16.1,
7-33 Stevenson WH, dos Santos R and Mettler SC Instituto Superior Tecnico, Lisbon, Portugal, 1986.
Fringe Mode Fluorescence Velocimetry, Applications
of Non-Intrusive Instrumentation in Fluid Flow 7-46 Maxwell BR
Particle Flow in Turbomachinery with Application to
Research, AGARD CP-193, pp. 20-1 to 20-9, 1976.
Laser-Doppler Velocimetry,AIAA J. Vol. 12, No. 10,
7-34 Ahmed NA et al. pp. 1297-1298, Oct. 1974.
A 3 0 Micro-Laser Anemometer for Boundary Layer
Studies, Third International Symposium on Laser 7-47 Orloff KL and Snyder PK
Laser Doppler AnemometerMeasurements Using Non-
Anemometry, A. Dybbs, F. Ali, and G. Morrison, eds.,
orthogonal Velocity Components: Error Estimates,
ASME, pp. 175-178,1987.
Appl. Opt., Vol. 21, NO.2, pp. 339-344, Jan. 15,1982.
7-35 Ahmed NA et al. 7-48 Wernet MP and Seasholtz RG
Miniature Laser Anemometer for 3 0 Measurements, Zoom Lens Compensatorfor a Cylindrical Window in
Meas. Sci. Technol., Vol. 1, pp. 272-276, Mar. 1990. Laser Anemometer Uses, Appl. Opt., Vol. 26, No. 21,
7-36 Ahmed NA and Elder L pp. 4603-4611, NOV.1,1987.
Flow Investigation in a Small High Speed Impeller Pas- 7-49 Williams MC
sage Using Laser Anemometry, ASME Paper Laser VelocimettyStudy of Stator/Rotor Interactions in
90-GT-233, June 1990. a Multi-Stage Gas Turbine Compressor, Advanced
7-37 Ahmed NA et al. Instrumentation for Aero Engine Components,
Fibre Optic Laser Anemometer for Turbomachinery AGARD CP-399, pp 11-1 to 11-10,1986.
Applications, Optics and Lasers in Engineering, Vol. 7-50 Goldman LJ and Seasholtz RG
16, NO.2-3, pp. 193-205,1992. Laser Anemometer Measurements and Computations
7-38 Schodl R and Forster W in an Annular Cascade of High Turning Core Turbine
A Multi Colour Fiber Optic Laser Two Focus Velocime- Vanes,NASA TP-3252,1992.
REPORT DOCUMENTATION PAGE
1. Recipient’s Reference 2. Originator’s Reference 3. Further Reference
1 4. Security Classification
1
I
AGARD-AG-328
I
ISBN 92-835-0717-7
I
I
ofDocument
UNCLASSIFIED/
UNLIMITED
5. Originator Advisory Group for Aerospace Research and Development
North Atlantic Treaty Organization
7 Rue Ancelle. 92200 N e d v sur Seine. France
6. Title
ADVANCED METHODS FOR CASCADE TESTING
7. Presented at
8. Author(s)/Editor(s) 9. Date
Prof. Ch. Hirsch (Editor) August 1993
LO. Authot’s/Editor’s Address 11. Pages

Vrije Universiteit Brussel
194
Dienst Stromingsmechanica
Pleinlaan 2, 1050 Brussel, Belgium.
Consequently, the Propulsion and Energetics Panel decided to initiate an AGARDograph on this
subject in 1988. Since then, several authors prepared chapters for the AGARDograph,
coordinated by the editor. The report includes the chapters Modelling Turbomachinery Flow
Conditions, Linear Cascades, Annular Cascades, Aspects of Hot Cascades, Transient Cascade
Testing, Unsteady Flow in Cascades and 3-D Laser Anemometry in Annular Cascades.
I x
8
8
Q
2
c
3Z
W
do
Y
W
w w
2
4
c
2
4
c
;J
4 2
W
p;'
CQ
w
VI
b
4
c
;J
4
1"
F . 0 . M 2715
DK-2100 coaahuan0
FQ*lnrCsI-
n1
I

AGARDAG328

Uploaded by

Copyright:

Available Formats

AGARDAG328

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AGARDAG328

Uploaded by

Copyright:

Available Formats

AGARD

ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT

Advanced Methods for

ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT

Advanced Methods for

Professor Charles Hirsch

This AGARDograph was prepared at the request of the

- North Atlantic Treaty Organization

Published August 1993

Set and printed by Specialised Printing Services Limited

LECTURE SERIES (LS)

Prof. Ch. Hirsch

Prof. Ch. Hirsch

Recent publications of the Propulsion and Energetics Panel iii

List of Authors viii

Chapter 1 Modelling Turbomachinery Flow Conditions 1

1.2 Modelling Turbomachinery Flows 1

Chapter 2 Linear Cascades 11

Chapter 3 Annular and Rotating Cascades 73

Chapter 4 Particular Aspects of Hot Cascades 96

Chapter 7 Three-Dimensional Laser Anemometry in Annular Cascades 170

Prof. C.H. Sieverding 1.2,1.4,1.5,2.3

Dr R.G. Dominy 3.1

Prof. T.V. Jones, Dr M.L.G. Oldfield and Dr R.W. Ainsworth 5

Turbulence Grid IXgK

Turbulence Grid VIK

Turbulence Grid VIIIK

Fig. 1.4-1 Streamline in meridional plane

Fig. 1.4-2 Twisting of stream surface through blade row

\ Developed Axisymmelric Section

Fig. 1.4-3 Development cross section and projection of an annular cascade

a ber dependence of the base pressure of turbine blades

1.3-1 Gostelow JP bomachinery Research, ASME Paper 76-GT-62.

lboundary layer suction]

' lboundary liyer suction1

Fig. 2.2-2 Cascade test section in a cascade tunnel

Fig. 2.2-3 Control of periodic flow condition (see Ref. 2.2-2)

\ Exit supersonic Supersonic

Axial supersonic inlet: If the axial component of the inlet flow

@ flow direction (0,)

Fig. 2.2-4 Probes used for wake measurements(see Ref. 2.2-3)

Three-dimensionaleffects it is observed that for a given cascade configurationand inlet

Granite optical bench

Fig. 2.2-7 Schematic view of the Schlieren-measurement system

Laser-two-focusvelocimetry A general view of the laser-two-focus system at the cascade

I laser power unit

Profile with mounted sensors

Const. Temp. Anemometer

Low-Pass Filter (96dB.lOOkHz)

Sensor e.g. DlSA 55P61

Const. Temp. Anemometer

2.3 SUBSONIC-CHOKED (TURBINE)CASCADES ers in transonic turbine cascades undergo an overall

layer interactions and local th"

PITCH 3 PITCH 2 PITCH 1

0 .05 .to .PO

Fig. 2.3-6 Probe stem in outlet flowfield of various axial distances

Fig. 2.3-7 Schlieren pholograph of blockage Mecl of probe stem

Fig. 2 .8 Cone pmbe with prismatic shafi in Rowfieldat M - 1.17 (Ref. 2

Fia. 2.3-9 DLR turbine cascade tunnel

Fig. 2.3-13 Interpolationof data