
Aerodynamic Impact of Swirling Combustor Inflow on
Endwall Heat Transfer and the Robustness of the Film

Cooling Design in an Axial Turbine

Vom Fachbereich Maschinenbau

an der Technischen Universität Darmstadt

zur

Erlangung des Grades eines Doktor-Ingenieurs (Dr.-Ing.)

genehmigte

D I S S E R T A T I O N

vorgelegt von

Holger Werschnik, M.Sc.

aus Mainz

Berichterstatter: Prof. Dr.-Ing. Heinz-Peter Schiffer

1. Mitberichterstatter: Prof. Thomas Povey, BA, MA, DPhil

Tag der Einreichung: 27. April 2017

Tag der mündlichen Prüfung: 19. Juli 2017

Darmstadt 2017

D 17


Bibliografische Information der Deutschen Nationalbibliothek: 
Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen 
Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.d-nb.de 
abrufbar. 

 
Creative Commons Lizenz (CC BY 4.0 International), 2017.  
Freies Vervielfältigen und Weiterverbreiten  –  Namensnennung  

 
http://dnb.d-nb.de/


i

Editor’s preface

The series Research Reports by the Institute of Gas Turbines and Aerospace
Propulsion accounts for the advances in research and development in the field
of turbomachinery at Technische Universität Darmstadt. Because of the strong
orientation on applications in this research field, the academic problems reflect
industrial development trends.

The changing political, economic and ecological framework influences the current de-
velopment focus and keeps carrying the turbomachine to the border of technological
feasibility. As a result, it is not unusual for findings to be directly transferred to the
industrial application.

Within this environment, the industry and application oriented research works of
this series originate. They describe current findings of experimental investigations
and numerical simulations which were obtained at the Institute of Gas Turbines and
Aerospace Propulsion at Technische Universität Darmstadt.

Heinz-Peter Schiffer

Darmstadt, 2017


iii

Author’s preface

This dissertation originated during my time as a research assistant at the Institute of
Gas Turbines and Aerospace Propulsion at Technische Universität Darmstadt. First
of all, I would like to express my gratitude to my advisor Prof. Dr.-Ing. Heinz-Peter
Schiffer, head of the Institute, for giving me the opportunity to work on the Large
Scale Turbine Rig. I enjoyed to work on this challenging project and I am especially
thankful about the great responsibility that I have been assigned and the involved
possibility to develop personally and scientifically. I appreciate his tremendous
support and confidence during the course of my time at the institute.

Prof. Thomas Povey, DPhil from the Osney Thermo-Fluids Laboratory at the
University of Oxford I thank for his interest in this thesis and the co-review of it.

I acknowledge the financial support within the framework of ”AG Turbo” by the
Federal Republic of Germany, Ministry for Economic Affairs and Energy, according
to a decision of the German Bundestag (FKZ: 03ET2013K) as well as by Rolls-Royce
Deutschland GmbH and Ansaldo Energia.

Furthermore, I want to thank the LSTR-Turbine-Group, namely Claudius Linker,
Sebastian Schrewe, Alexander Krichbaum, Manuel Wilhelm, Tom Ostrowski, Sebas-
tian Leichtfuss, Johannes Eitenmüller and Leonhard Gresser. They all have their
share in setting up this unique test rig and its capabilities. I enjoyed the great
working athmosphere in the office, the lab or our ”networking events” such as the
turbine group hiking and barbecue events.

I would like to thank the LSTR partners at Rolls Royce Deutschland, Roland Wil-
helm, Knut Lehmann, Christoph Lyko and Thomas Janson and at Alstom/Ansaldo
Marc Henze and Jörg Krückels. Their contribution of technical expertise were a
great asset for my research and I appreciate the support to publish my results on
several turbomachinery conferences and in journal papers.

All the colleagues at the institute I thank for the great working atmosphere during
my time at the GLR. I especially appreciate the support of Marius Schneider and
Manuel Wilhelm for proofreading my dissertation

Not to the last, I want to thank all my students, namely Carlos Sánchez-Utgés,
Marcel Adam, Fari Bakhtiari, Daniel Arenz, David Neubauer, Johannes Stroh,
Christoph Steinhausen, Nam Wahrenberg, Viktor Arne, Janina Herrmann, Sebastian
Wiemer, Niklas Apell, Yannick Fischer, Peter Allard and Daniel Markus. They all
have taken care of a part of the project tasks in their Bachelor-/Master-Thesis or


iv Author’s preface

as a student assistant in the laboratory and in this way helped to finish it successfully.

Besonderer Dank gilt an dieser Stelle zudem meinen langjährigen Freunden Sascha
Hell und Christian von Pyschow, mit denen ich die Doktorandenzeit gemeinsam und
unter anderem bei zahlreichen Mensa-Mittagessen durchlebt habe.

Für die stete Unterstützung und Motivation während meines gesamten Studiums
und der Promotion danke ich meinen Eltern Iris und Herbert sowie meinem Bruder
Nils.

Ich danke meiner Tochter Lea, die mir geholfen hat, die ”komplexen” Problem-
stellungen im richtigen Verhältnis zu betrachten und mich dabei immer mit guter
Laune und ablenkenden Alternativbeschäftigungen versorgt hat. Es especial, quiero
agradecer un monton a mi esposa Galina por el tremendo soporte y el buen animo
qué me dio durante todos los años de mi doctorado.

Holger Werschnik

Darmstadt, 2017


v

Abstract

The development of new gas turbines and aero engines is dedicated to reduce pollutant
emissions in addition to the continuous strive to improve component efficiency and
the consumption of fossil fuels. To foster this trend, new combustion concepts have
come into play such as lean combustion. Whereas the emission of carbon dioxide
can be reduced by lower fuel consumption, the formation of thermal nitrogen oxide
can only be hindered by a leaner fuel-to-air mixture: Lower peak temperatures and
avoiding a stochiometric concentration in the combustion chamber slow the thermal
reaction process responsible for the formation of nitrogen oxides.

Swirl and a recirculation zone are used to stabilize the combustion process and
a redistribution of mass flow towards the endwalls occurs. Additionally, a changed
temperature profile with reduced peak temperature, but increased temperature near
both endwalls due to the reduced injection of dilution air in the combustor approaches
the subsequent turbine stage. Associated, positive and negative incidence, high tur-
bulence intensities and increased thermal load to the endwalls challenge the turbine
design.

To improve the understanding of the complex aerodynamic and aerothermal inter-
action, the aerodynamic impact of combustor swirl on the first vane row of a turbine,
the nozzle guide vane (NGV), is investigated. The experiments are conducted at the
Large Scale Turbine Rig (LSTR) in Darmstadt, which consists of a 1.5-stage axial
turbine that is subject to an engine-representative swirl. A combustor simulator is
used to vary the inflow to the turbine. Further goals of the investigation are to eval-
uate the robustness of its endwall film cooling design and to investigate endwall heat
transfer and film cooling effectiveness experimentally by using infrared thermography
and the auxiliary wall method.

As a reference, axial and low-turbulent inflow to the turbine is investigated. A
variation of the coolant mass flow rate highlights the influence on Nusselt numbers
and film cooling effectiveness as well as the aerodynamic flow field. An increase of
Nusselt numbers by up to 80 % is observed with a concurrent increase of the film
cooling effectiveness by up to 25 %. In a combined analysis a significant heat flux
reduction due to film cooling by 30 % is achieved. A coolant mass flow rate (MFR) of
one yields the greatest benefit. For higher MFR the further improvement of the film
cooling effectiveness is counteracted by the even greater increase in heat transfer.

With applied swirl, the flow field changes significantly. Averaged whirl angles of
15◦ to 20◦ and a mass flow redistribution to the endwalls are detected. The NGV
exit flow exhibits a dominating influence of swirl on pressure losses instead of the
coolant flows as it had been observed for the baseline. For similar settings of the
stage parameters, an increase in Nusselt numbers by up to 40 % is observed. The film
cooling effectiveness is reduced because of the mass flow redistribution. For MFR
greater than 1.5, the increase in Nusselt numbers is less decisive and remains at a
similar level to the reference case. To achieve the same level of film cooling, the


vi Abstract

double amount of coolant air is necessary. These general trends are resolved for two
clocking positions between swirler and vanes, whereby local differences are observed.
The combined analysis of the thermal parameters shows a local increase of endwall
heat flux and a local influence on the coolant injection. The coolant injection is
still beneficial in reducing the heat flux for low injection rates, but the local extent
varies much more. For higher injection rates above 1.5, only sections of the endwall
face an improvement and there is a growing area, where increased heat flux and in
consequence higher thermal load is applied in comparison to the reference.


vii

Kurzzusammenfassung

Neben der Zielsetzung hoher Komponentenwirkungsgrade und niedrigem Brennstof-
fverbrauchs wird bei der Entwicklung neuer stationärer Gasturbinen und Flugtriebw-
erke angestrebt, die Schadstoffemissionen weiter zu verringern. Um diesen Entwick-
lungstrend weiterführen zu können, ist der Einsatz neuer Verbrennungskonzepte wie
der Magerverbrennung notwendig. Während eine Reduktion der Kohlenstoffdioxide-
missionen durch eine Verbrauchsminderung erreicht werden kann, ist für eine Senkung
der Stickoxidemissionen ein mageres Brennstoff-Luft-Gemisch unerlässlich: Die ther-
mische Bildung von Stickoxiden wird durch geringere Maximaltemperaturen sowie
die Vermeidung stöchiometrischer Konzentrationen in der Brennkammer vermindert.
Dieser Verbrennungsprozess wird durch eine starke Drallströmung und die einherge-
hende Ausbildung einer Rezirkulationszone stabilisiert, und eine Umverteilung des
Massenstroms zu den Endwänden tritt auf. Weiterhin entsteht ein flacheres Ein-
trittstemperaturprofil in der Zuströmung der nachfolgenden Turbine durch den re-
duzierten Einsatz von Zumischluft im Brennkammeraustrittssektor. Bei der Ausle-
gung der Turbine müssen aus diesem Grund hohe Turbulenzgrade, Inzidenz sowie
eine erhöhte thermische Belastung der Endwände beachtet werden.

Um das Verständnis für die komplexen aerodynamischen Vorgänge und aerother-
malen Interaktionen zu verbessern, wird die Auswirkung erhöhten Brennkammerdralls
auf die erste Leitschaufelreihe am Large Scale Turbine Rig (LSTR) in Darmstadt
an einer 1,5-stufigen Axialturbine untersucht. Die verdrallte Zuströmung ist dabei
repräsentativ für eine reale Maschine. Ein Brennkammermodul ermöglicht Variatio-
nen in der Zuströmung der Turbine, um auf diese Weise die Robustheit der End-
wandkühlung zu untersuchen. Der Wärmeübergang sowie die Filmkühleffektivität
auf der Endwand werden experimentell mit Hilfe von Infrarotthermographie und der
Hilfswandmethode untersucht.

Dabei wird eine axiale Zuströmung mit niedrigem Turbulenzgrad als Referenz un-
tersucht. Eine Variation des Massenstroms, der zur Endwandkühlung eingeblasen
wird, verdeutlicht den Einfluss auf Nusseltzahlen, Filmkühleffektivitäten sowie das
aerodynamische Strömungsfeld. Die Nusseltzahlen erhöhen sich mit steigendem Küh-
lluftmassenstrom im Mittel um 80 %, während die Filmkühleffektivität nur um 25 %
zunimmt. Mit einer kombinierten Auswertung der beiden Parameter kann die Min-
derung des Wärmestroms bestimmt werden. Dabei zeigt sich bei einer Einblaser-
ate von 1 % des Haupmassenstroms das Optimum für die vorliegende Geometrie.
Bei höheren Einblaseraten wird die eintretende Verbesserung der Kühlung durch die
stärkere Erhöhung des Wärmeübergangskoeffizienten größtenteils aufgezehrt.

Bei aufgeprägtem Brennkammerdrall ändert sich das Strömungsfeld in der Sta-
torreihe deutlich: Mittlere Drallwinkel von 15◦ bis 20◦ und eine Umverteilung des
Massenstroms zu den Endwänden werden festgestellt. Die Abströmung der NGV-
Stufe wird vornehmlich durch die Auswirkung des Dralls auf die Druckverluste charak-
terisiert. Der große Einfluss der Kühllufteinblasung, der für die Referenzkonfiguration


viii Kurzzusammenfassung

festgestellt wurde, tritt nicht mehr auf. Bei gleicher Einstellung der Stufengrößen im
Vergleich zum Referenzfall liegen die Nusseltzahlen um bis zu 40 % höher, wogegen die
Filmkühleffektivität um bis zu 40 % niedriger ist. Bei Einblaseraten über 1,5 % ist die
Steigerung der Nusseltzahlen im Vergleich zur Referenzkonfiguration gering und ein
ähnliches Niveau wird erreicht. Um die gleichen Werte für die Filmkühleffektivität wie
im Fall der axialen Zuströmung zu erreichen, wird jedoch in etwa die doppelte Küh-
lluftmenge benötigt. Diese allgemeinen Feststellungen werden für beide untersuchten
Relativpositionen zwischen Drallerzeuger und Schaufelvorderkante beobachtet, wobei
sich lokale Unterschiede einstellen. Bei der kombinierten Analyse beider Parameter
ergeben sich lokal zudem Gebiete, in denen eine Erhöhung des Wärmestroms stat-
tfindet. Eine Reduktion des Wärmestroms wird für kleine Einblasemengen weiterhin
erreicht, jedoch variiert das lokale Niveau deutlich stärker. Bei großen Einblaseraten
treten hingegen größere Wärmeströme auf als im Referenzfall und die Endwand wird
somit stärker thermisch belastet.


Contents ix

Contents

1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research context and connection to other research projects . . . . . . 4
1.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Theoretical background and literature review 7
2.1 Combustion and emissions . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Combustor-turbine interaction . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Endwall film cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Endwall heat transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Resulting heat flux reduction due to film cooling . . . . . . . . . . . . 27
2.6 Definition of research objectives for the thesis . . . . . . . . . . . . . . 28

3 Experimental setup 31
3.1 The Large Scale Turbine Rig . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Combustor module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3 Operating point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Vane passage nomenclature . . . . . . . . . . . . . . . . . . . . . . . . 41

4 Experimental methods 43
4.1 Infrared thermography - auxiliary wall method . . . . . . . . . . . . . 43
4.2 Gas concentration measurements . . . . . . . . . . . . . . . . . . . . . 50
4.3 Pneumatic measurements . . . . . . . . . . . . . . . . . . . . . . . . . 54

5 Investigation of the reference configuration 59
5.1 Aerodynamic flow field . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.2 Endwall film cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3 Endwall heat transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.4 Resulting Heat Flux Reduction . . . . . . . . . . . . . . . . . . . . . . 80
5.5 Summary of the findings for the reference case . . . . . . . . . . . . . 82

6 Investigation of swirling combustor inflow 85
6.1 Aerodynamic flow field . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2 NGV aerodynamic loading . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3 Endwall film cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.4 Endwall heat transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.5 Heat Flux Reduction due to film cooling . . . . . . . . . . . . . . . . . 115
6.6 Summary of the findings for the swirling inflow case . . . . . . . . . . 117

7 Summary and Outlook 121
7.1 Suggestions for further research . . . . . . . . . . . . . . . . . . . . . . 123


x Contents

Bibliography 127

Nomenclature 141

List of Figures 147

List of Tables 151

A Appendix: Overview 153

B Full measurement results for swirling inflow 155

C Comparison to the measurement uncertainty in comparable studies 163

D Commissioning measurement results 165
D.1 Operating point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
D.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

E Increased turbulence intensity inflow 169
E.1 Aerodynamic flow field . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
E.2 Endwall heat transfer and film cooling . . . . . . . . . . . . . . . . . . 169
E.3 Summary of the findings for the increased turbulence intensity case . . 172

F Endwall streamline detection 173

G Curriculum Vitae 175


Introduction 1

1 Introduction

This chapter illustrates the intention and the background of the thesis and gives an
overview on the research field. Furthermore, the research focus and the importance of
the combustor-turbine interface for engine design are explained. Finally, the research
objective and the outline of the thesis are presented.

1.1 Motivation

Gas turbines are the most common machines for aircraft propulsion and are also
widely deployed in power generation. They are heat engines and convert chemical
energy into mechanical work. In the case of a proplusion device, they generate kinetic
energy as well. Among their advantages is the production of a very large amount of
power with respect to their dimension and weight (Lechner and Seume [86]). Since
the first machines, built by von Ohain and Whittle, the main components have not
changed (c.f. Fig. 1.1):

• The air is compressed in the compressor to a high pressure level to achieve
high efficiency. Large gas turbines usually feature axial flow compressors with
multiple stages. A part of the compressed air is extracted and used for cooling
purposes of the subsequent hot sections.

• In the combustion chamber fuel is injected and combusted, increasing the
gas temperature to a very high level of up to 1900 K.

• The hot gas is expanded within several turbine stages and mechanical power
is generated. This is partly used to drive the compressor, mounted on the
same shaft. In the case of a stationary gas turbine, the remaining shaft power
is converted into electric power and in the case of an aircraft propulsion gas
turbine the air is exhausted with significant kinetic energy to produce thrust.

The design of the components requires specialized knowledge, expertise, design
tools and models. Usually, the aerodynamic and mechanical design process is split
and executed by dedicated groups among gas turbine manufacturers. The design
implies a profound knowledge of the boundary conditions at each station for a good
design. With the uncertainty involved in the complex flow field and the strive to limit
safety margins to only the necessary, an integrated design process is aspired. This is
where combustor-turbine interaction (CTI) research comes into play. The interface
between the two main components (c.f. Fig. 1.1) is crucial in engine design, because
it constitutes the hottest area in the gas path and is therefore subject to very high
thermal load. Knowing the boundary conditions of the turbine inflow with more de-
tail is an asset in improving the performance of the turbine. Due to the complex
flow field uncertainties remain. The robustness of a design to deviations from the
design conditions is a very important aspect. In addition, the trend to use stationary


2 Introduction

Turbine
Combustion

chamber

Compressor Nozzle Guide 
Vane (NGV)

Rotor

LP coolant

HP coolant

RIDN

Combustor-Turbine Interface

Figure 1.1: Modern gas turbine engine. Illustration of the main components and
zoomed view on the combustor-turbine interface, showing the location of
the RIDN coolant injection (adapted from Rolls-Royce [106][107])

gas turbines in a more flexible way to compensate load variations in national power
grids require a more frequent operation at off-design conditions. In this context, an
optimized design requires exact awareness of the incoming flow profile and any devia-
tion can lead to inferior performance or even a lack of functionality. A robust design
on the other hand can encounter deviations and ensure that the system retains its
specifications also under altered conditions. Furthermore, combustor-turbine interac-
tion is inherent in the design process. Not only does the combustor outflow influence
the downstream turbine design. The presence of the turbine determines in the same
manner the design of the combustor, as upstream effects of the turbine on the com-
bustor flow occur. In addition, achievable swirl levels that could assist in improving
the combustion process need to be limited to narrow the complexity of the flow field
in the turbine to a predictable level (Turrell et al. [128]).

The fuel consumption of the gas turbine is a major cost contributor in air travel.
Consequently, in the early years of the development, reducing the fuel consumption
had been the focus of the development. Environmental awareness has led to legislative
requirements regarding emissions and are complimented by the application of noise-
ban rules. Therefore, a goal triangle in the design of gas turbines exist, with the aim
to reduce noise and emissions in addition to the fuel consumption.

With component efficiencies of both combustor and turbine exceeding 90 %, a ma-
jor optimization area of the gas turbine is its secondary air system (Schuler [118]),
responsible for the coolant air supply. In the gas turbine cycle, the use of compres-
sor air for cooling purposes is a significant contribution to the overall losses. Up to
30 % of the compressor air is used for cooling purposes in the combustor and tur-
bine(Bräunling [22]). Thereby up to 10 % is injected to cool the first turbine stage.
Cooling air constitutes a penalty to the engine efficiency because it does not fully
contribute to the thermodynamic cycle. On the other hand, cooling air is required
to control the material temperature of the turbine components. Furthermore, this
is necessary in all operating point of the engine to ensure sufficient component life-


1.1. Motivation 3

time. The gas temperature entering the high pressure turbine (HPT) is well above
the melting temperature of the material of the vanes and blades. Thermal barrier
coatings applied to the surfaces increase the bearable temperatures, but the thermal
load needs to be taken care of by external and internal cooling methods for the en-
gine parts. Efficient cooling techniques and a proper understanding of heat transfer
parameters are necessary for the engine design as a consequence.

The recent development of lean burn combustors challenges turbine design. This
development is driven by legislative requirements as well. For example, the Advisory
Council for Aviation Research and Innovation in Europe (ACARE) to the European
Commission has set a goal of a 90 % reduction of NOx emissions compared to the year
2000 in its Flightpath 2050 vision [2]. Furthermore, the ICAO Committee on Aviation
Environmental Protection (CAEP) emits emission standards that apply to the type
designs of future aircraft. Because air travel has experienced annual growth rates of
5% over the last decades and is forecast to increase with similar numbers in the future,
the development of new technologies is necessary to limit the environmental impact.
IATA envisions an environmentally sustainable growth and has suggested means to
even achieve a reduction of the emissions. Data by Airbus [3] shows that within the
recent 15 years, the fuel consumption and the CO2-emission has been reduced by 34 %
per passenger and trip. However, significantly more advancements are to be made to
achieve the IATA vision.

The application of lean combustion is such a technological breakthrough. However,
less dilution air is used in the combustor and thereby a changed inlet temperature
profile with increased temperature near the endwalls is imposed to the turbine. A
crucial field is the thermal design of the turbine endwall. New design philosophies
have to be developed to cope with the altered boundary conditions imposed by new
combustor outlet flow profiles. This sets the scene for the present investigation.

The hub side endwall cooling of the first stator vane is of special interest to this
thesis. The coolant air is injected at the CTI interface (c.f. zoomed view in Fig. 1.1)
and denoted as RIDN-cooling (Rear Inner Discharge Nozzle). Whereas this coolant in
early engine developments had mainly been used to set the desired radial temperature
profile at turbine inlet by injecting colder air, it is used as for platform cooling nowa-
days. Its ejection characteristics and associated cooling capability largely depend on
the adaptation to the incoming flow. In the context of this thesis, inlet boundary
conditions representative for a lean combustor are investigated. This inflow includes
residual swirl, imposed incidence at the first vane leading edge, an altered radial mass
flow distribution and increased turbulence intensity. The robustness of the cooling
design to these conditions needs to be evaluated. A proper distribution of the cooling
air is crucial for a safe engine operation and sufficient lifetime of all components. For
axial inflow there is large expertise of the factors of influence and various publications
exist among the literature on cooling optimization under laboratory conditions. The
complex flow field at the exit of a lean combustor however requires further investi-
gation. An experimental investigation that models this inflow with a great level of
detail is used to examine the robustness of a cooling design to inflow variations as
specified above.


4 Introduction

1.2 Research context and connection to other research projects

This investigation is part of a comprehensive approach to investigate the interaction
between combustor and turbine within the research framework of ”AG Turbo” during
the course of several years (Werschnik et al. [136]). The content of this thesis is
thereby primarily based on the project AG Turbo 2020 3.2.5 (grant number ”FKZ
03ET2013K”), dedicated to the aerodynamic impact on endwall heat transfer and
film cooling effectiveness on the hub of the NGV. In addition, some results have been
acquired within the projects AG Turbo 3.2.1 A & C (grant numbers ”FKZ 0327719F”
and ”FKZ 0327718R”). The focus in these projects is the aerodynamic effect of the
swirling inflow on the turbine flow field. The measurement period and rig setup has
been shared for the three projects. They are jointly funded by the German ministry
of energy and transport, according to a decision of the German Bundestag, and by
the industry partners Rolls-Royce Deutschland, Ansaldo Energia.

Subsequently, the projects COOREflex 3.1.1 (grant number ”FKZ 03ET7021K”)
and 3.1.2 (grant number ”FKZ 03ET7071O”) continue the investigation with the
focus on rotor hub and tip aerodynamics, film cooling and heat transfer.

Preceding work with respect to combustor-turbine interaction at the institute GLR
include the dissertations of Giller [46], who investigated the aerodynamics of a stator
row in a linear cascade with special regard to the influence on the film cooling ejection
at the leading edge. Schmid [112] numerically investigated three test cases with
increasing complexity: Giller’s linear cascade, the LSTR setup and the Engine-3E
(Klinger et al. [74]). The numerical approach is continued by Schneider and Schiffer
[115] with a parametrized model of combustor exit flow conditions, that is used to
study various turbine design optimizations and their robustness.

All research projects are conducted within the context of the University Technology
Centre (UTC) of Rolls-Royce at Technische Universität Darmstadt with the focus on
the aerothermal interaction of combustor and turbine.


1.3. Outline 5

1.3 Outline

The thesis begins with three basic chapters to familiarize the reader with the topic
and the experimental setup:

Chapter 2: Theoretical background and literature review
An overview of the preceding work and the theoretical background is given.
Relevant findings in the field of combustor-turbine interaction and a literature
review on film cooling and heat transfer is presented. As a conclusion of the
chapter, research objectives for this thesis are defined.

Chapter 3: Experimental setup
In this chapter, the Large Scale Turbine Rig (LSTR) and the experimental
setup used for the present investigation is introduced.

Chapter 4: Experimental methods
The developed and applied measurement methods are illustrated.

The main part with the presentation and discussion of the experimental results is
split into two chapters and followed by a summary:

Chapter 5: Investigation of the reference configuration
This chapter presents the findings for the axial inflow reference case.

Chapter 6: Investigation of swirling combustor inflow
The results for swirling inflow and its associated impact on the aerodynamic
flow field, endwall heat transfer and film cooling effectiveness is investigated.
The findings are presented in comparison to the reference case. The robustness
of the endwall film cooling design to swirling inflow is evaluated based on the
results.

Chapter 7: Summary and Outlook
A brief summary of the main results is given, followed by suggestions for future
research.


Theoretical background and literature review 7

2 Theoretical background and literature review

This chapter contains review on relevant literature regarding the content of this thesis.
Two main aspects are adressed:

• Recent research findings from combustor-turbine interaction studies, both
numerically and experimentally, are presented. The contribution of working
groups to the understanding of the inherent changes to the incoming flow field
towards the turbine and the derivation of design rules for the turbine are em-
phasized.

• A discussion of findings from past endwall heat transfer and film cool-
ing studies. In this way, the theoretical background associated for the present
investigation is illustrated.

2.1 Combustion and emissions

In the combustion chamber, fuel is injected into the flow and energy is added to the
flow. The highly complex combustion process has to fulfill requirements on flame
stability, safe operation and a desired exit temperature profile to maintain the tur-
bine within the operating conditions. Furthermore, the chemical reaction and its
conditions decide about the composition of the exhaust gas. Carbon dioxide (CO2) is
generated during the combustion and the amount increases proportional with the fuel
consumption. In contrast, the formation of other pollutants such as carbon monoxide
(CO), unburnt hydrocarbons (UHC), soot and nitrogen oxides (NOx) largely depend
on the peak temperature, the equivalence ratio Φ and the residence time in the com-
bustor (Hennecke and Wörrlein [58]). Similar to CO2 they can be reduced by a
lower fuel consumption, the specific chemical reactions can be further controlled to
set the total emissions to a minimum. Whereas CO and UHC are fostered by lower
combustion temperatures, the thermal formation of NOx increases rapidly with the
combustion temperature and reaches its peak in the stoichiometric regions, i.e. Φ=1.

Conventional RQL combustion

State-of-the-art combustion chambers operate with the principle of ”Rich Burn - Quick
Quench - Lean Burn” (the so called RQL-combustor). They use a primary zone with
an equivalence ratio greater than one to ensure a stable combustion. The injection
of dilution air brings the flow rapidly to a lean mixture. Nevertheless, the mixture
passes the stoichiometric region and pollutants are formed. Therefore, this concept
limits the reduction of NOx emissions.


8 Theoretical background and literature review

Lean combustion

A solution is offered by the lean-premixed-prevaporized (LPP) combustion principle.
The formation of NOx can be significantly reduced by avoiding the stoichiometric
region completely, by staging below equivalence ratios Φ of about 0.6 within the entire
combustion chamber. To ensure a continuous and safe combustion process, i.e. proper
mixing and flame stabilization, swirl stabilization is used (Huang and Yang [61]). This
causes high turbulence intensity and strong shear regions due to velocity differences
and in this way enhances the mixing of fuel and air. A recirculation zone develops as
consequence of a vortex breakdown and the associated adverse pressure gradient. The
flame is stabilized in the shear region at the boundary of the recirculation zone. The
swirler vortex core proceeds towards the turbine in a precessing motion, leading to an
unsteady and non-uniform pressure distribution at turbine inlet (Jacobi et al. [64]).
About 80% of air enters the combustor already at the swirler nozzle (compared to 20%
in a RQL-combustor (Schiffer [110])) and no dilution ports are used. Usually, effusion
cooling is applied to the combustor liners. This difference has two major consequences
for the turbine inflow: A flatter temperature profile with increased thermal load on the
endwalls and increased residual swirl enters the turbine (Bunker [23]). On contrary,
in a conventional combustor, they are both attenuated by the dilution ports. As a
result, lean combustion poses new challenges to the turbine design. This and other
changes that have connected the design of combustor and turbine very closely have
led to the research field of combustor-turbine interaction.

2.2 Combustor-turbine interaction

The term combustor-turbine interaction refers on one hand the strong interaction be-
tween these two main components of a gas turbine that exists due to the complex flow
field. On the other hand, it also refers the design process, because the components
are usually designed in individual, highly specialized teams. Whereas the overall
efficiency of the machine can be increased by optimizing individual component ef-
ficiencies, a more accurate design process requires an exact knowledge of boundary
conditions at the interfaces. Within an engine development program, these bound-
ary conditions are frequently exchanged, changed and altered, based on the detailed
design-engineering work.

With lean combustion, combustor-turbine interaction becomes very important,
since the boundary conditions at the interface are far away from laboratory con-
ditions as they are assumed and modeled in many test setups and studies. Over the
recent years, the complex flow field has been numerically modeled and experimen-
tally investigated in detail by several research groups. In the following, an overview
on findings with special emphasis on the turbine is given.

Stagnation line shift

The vane stagnation line is altered by the combustor inflow: For inversely oriented
swirl to the vane turning direction and high swirl level, negative incidence occurs near
the hub and positive incidence near the casing (Giller [46]). The leading edge film
cooling flows of the airfoil are therefore influenced which may lead to the situation
that they are washed to the opposing side of the vane as they were designed for.


2.2. Combustor-turbine interaction 9

Furthermore, due to the pressure difference between SS and PS, the stagnation line
shift impresses a span wise pressure gradient at the leading edge. This causes a change
in static pressure ratio across the film cooling holes, leading to outflow variations. Yin
et al. [143] and Griffini et al. [51] illustrate this influence on the shower-head cooling of
a vane. They determine strong local inhomogeneities of film cooling effectiveness and
heat transfer, leading to an increase of the net heat flux in the shower-head region.

Interaction between vane and the inlet pressure distortion

Due to the recirculation zone in the combustor, a 2-dim. total pressure deficit ap-
proaches the vane row. The radial pressure gradient in combination with the incoming
swirl thereby causes the formation of additional vortices in the vane passage as it ap-
proaches the vane leading edge (Illustrated according to Jacobi et al. [64]):

Figure 2.1: Formation of the swirl-/leading edge interaction vortex. The low-pressure
zone triggers the formation of a vortex pair at mid span, that persists
through the vane passage and merges with the passage vortex (Adapted
from Jacobi [64])

Approaching the leading edge, the flow is decelerated and a static pressure gradient
in span wise direction results. This evokes a secondary flow feature (shown by red and
blue arrows in Fig. 2.1) that travels to both sides of the airfoil. It becomes attenuated


10 Theoretical background and literature review

and mitigated depending on the swirl orientation and a resulting vortex pair develops.
In addition, a second span wise static pressure gradient is impressed by the stagnation
line shift at the leading edge, as explained before. Jacobi identifies that the vortex
pair travels according to this pressure gradient; i.e. towards the casing if the swirl is
inversely oriented in relation to the vane turning direction and to the hub endwall if
they have a concordant orientation. Turrell et al. [128] have observed this behavior
in a similar manner for a different swirler geometry. Depending on the extent of
the phenomenon, the vortex merges with the passage vortex, but is in any case an
additional loss term in the stator exit flow field. In addition, Jacobi determines a
decrease in airfoil heat transfer along its trajectory.

Hot-Streaks

Because the use of a lean fuel-to-air mixture alleviates the required number of di-
lution ports or even cut them completely, the recirculation zone and the inherent
flame stabilization mechanisms create a strong non-uniform temperature profile with
discrete hot-streaks at turbine inlet, as Andreini et al. illustrate [4]. The pattern
factor cannot be controlled as well as with conventional combustors. The clocking
of the hot-streak relative to the turbine vanes is of great importance. When it is
directed towards the vane passage, it passes relatively unchanged towards the rotor
row, resulting in a non-uniform temperature profile at the rotor inlet as well. This
can be attenuated by directing the hot-streak towards the vane leading edge. This
has the drawback, that the associated peak temperatures cannot be shielded properly
from reaching the vane surface, as Koupper et al. [80] illustrate. They investigated
a comparison of a passage aligned swirler to a leading edge aligned swirler. Whereas
for the leading edge alignment, the hot streak is split into neighboring passages and
a relatively homogenous temperature distribution is detected at the vane exit, for the
passage alignment, the hot streak travels relatively unchanged through the vane row
and discrete temperature peaks are encountered by the rotor blades.
Beard et al. [17] determine a detrimental effect on performance due to an engine-
realistic inlet temperature distortion. Whereas the NGV’s aerodynamic behavior was
rather unaffected, off-design conditions are observed in the rotor. The observed mass
flow redistribution towards the hub and casing endwalls where the local work contri-
bution is lower resulted in reduced efficiency.
Griffini et al. [51] numerically examine the clocking of the hot streak with respect to
the leading edge for a film-cooled vane. A passage oriented hot spot and swirl ori-
entation is compared to a leading edge oriented position. The leading edge position
was determined more problematic and showed increased airfoil wall temperatures.
Moreover, the shift of the stagnation line influences the coolant mass flow ejection in
the showerhead region.
The location of the hot-streak depends on the design of the combustor and the oper-
ating regime. Lean combustors with a pilot/main flame staging exhibit a hot streak
position downstream of the swirler center in pilot mode and in the interaction zone
in between in main flame mode. This is due to higher shear levels between adjacent
swirlers, which stabilize the flame and therefore the highest temperature is located
here.
To study the propagation of hot streaks, the gas concentration method is used as


2.2. Combustor-turbine interaction 11

well. Cha et al. [29] investigate the combustor outflow profile by introducing CO2
as a tracer through the fuel supply tubes in a gas turbine combustor. Butler [28]
investigated an inlet temperature distortion by the same method, introducing a hot
streak of CO2 upstream of a vane cascade. They determined that the spatial extent
of the hot streak does not increase significantly across the vane row.

Hall et al. [55] illustrate design issues to develop a non-reacting combustor simula-
tor, which is capable of producing a range of swirl and temperature distortion profiles.
They have highlighted the importance of the occurring total pressure and turbulence
distribution. They also have modeled an inlet temperature distortion to achieve an
engine-realistic design.

Migration of the swirl core

A migration of the swirl core in lateral direction (Fig. 2.2) along its trajectory towards
the turbine has been observed. Insinna et al. [62] identify a gyroscopic effect to be
responsible: In their geometry, the mean radius of the combustor increases towards the
vanes due to the inclined hub endwall. Because it is forced upward, the swirler vortex
tube deflects sidewards in response to the change in direction. In the experiment
of Hall et al. [55] for a similar geometry, a lateral movement is observed as well.
Furthermore, a combination of both an annular geometry and an inclined swirler is
studied numerically and experimentally by Bacci et al [10] [11] and Koupper et al.
[80] with a similar finding.

Figure 2.2: Lateral deflection of the swirl core observed by Insinna et al. [62] for
counter-clockwise rotating swirl with respect to the main flow turning
direction

On the other hand, this migration is not observed in investigations with linear,
non-inclined cascades, e.g. Giller [46] and Jacobi et al. [64]. Vagnoli and Verstraete
[130] add that the migration is favored by an annular combustor geometry as well:
Because the volume above the swirler axis is greater than the volume below due to
the curvature of the annulus, more fluid is attracted from the outer section into the
low-pressure swirler core. This results in a momentum imbalance and a migration of
the core in the flow direction of the fluid near the casing endwall occurs.


12 Theoretical background and literature review

The momentum imbalance is thereby already imposed at the combustor inlet by
the swirler itself, as Fig. 2.3 illustrates. The LSTR swirler with three concentric rings
is shown with the associated orientation. The turning angle of the outer ring is very
small and hence the overall imposed swirl of this configuration is counter-clockwise
oriented with respect to the main flow turning direction. Radial and whirl components
have to be evaluated in polar coordinates to calculate the resultant. At any arbitrary
position along the circumferential line at midspan, each area element above the line
introduces a positive whirl component and each area element below a negative whirl
component. Peak whirl components result along the mirror line of the swirler. In
sum, the positive whirl component is greater than the negative component, because
the area above the midspan line is greater. Consequently, a shift of the swirler center
in positive whirl direction results, which is to the left in the figure.

Radius at

swirler center &

annulus midspan

positive whirl component

negative whirl 
component

Figure 2.3: Momentum imbalance imposed by the swirler (Schematic shows the
swirler of the present investigation). For an annular rig geometry, the
mean radius divides the swirler into a larger upper section and a smaller
lower section with associated whirl components and a resulting global shift
of the flow field downstream of the swirler exitin the whirl direction of the
upper section

The migration of the swirl core is relevant to the engine design, because its relative
position to the vane’s leading edge including the hot streak alters the wall tempera-
tures at the turbine. Schneider and Schiffer [115] have shown that a small variation
of 1◦ in the position of the swirler core can drive the hot-streak downward to reach
the endwall and evoke peak temperatures that need to be considered for the cooling
design.

Turbulence intensity

The swirler causes an unsteady distribution of turbulence intensity and length scales.
The influence of moderate inlet turbulence quantities has been widely investigated in


2.2. Combustor-turbine interaction 13

cascade rigs (Giel et al. [45], Arts and Lambert de Rouvroit [9], Cresci et al. [34]). Due
to the highly complex and unsteady flow field, only few studies exist, that take into
account high turbulence intensities as they are observed in the combustor outflow,
such as van Fossen and Bunker [131], Koupper et al. [79] and Bacci et al. [10]. At
turbine inlet, turbulence intensities up to 35 % are encountered (Cha el al. [30] [29]).
The turbulence level depends mostly on the geometry, the swirl level and secondary
flows. The combustion process itself has only inferior contribution to the turbulence
level, but the temperature and velocity increase due to the combustion reduces the
swirl level and the turbulence level. To achieve engine-representative swirling inflow
in a cold flow setup, great care has to be taken. Therefore, the use of real engine
hardware requires an adaptation of the simulator design (Bacci et al. [10]). The range
of turbulence length scales at the combustor exit is increased when effusion cooling
is applied to the combustor liner. Small length scales are generated (Koupper et al.
[79]) near the cooled wall. Accurate modeling of turbulent quantities is key in the
numerical investigation of the interaction between combustor and turbine, as Schmid
[112] illustrates.

Different combinations of swirl orientation have been studied to evaluate their mix-
ing capability and effects on the flow field. Merkle et al. [96] show typical pressure,
velocity and turbulence quantities and their distribution for a co-swirl combustion
chamber, i.e. with all swirlers oriented into the same direction of rotation and counter-
swirl with alternating orientation. Schmid [112] presents a similar study for a linear
cascade.

Numerical approaches to combustor-turbine interaction

Existing numerical approaches to combustor-turbine-interaction studies that take care
of the discussed effects are presented in the following. Schmid et al. [113] numerically
examine a HPT stage with aerodynamic boundary conditions of an aero-engine lean
combustor and find increasing heat transfer levels and decreasing stage efficiency
compared to a low-turbulent, axial inflow. The influence of turbulence intensity is
modeled separately and is determined to be responsible for nearly half of the total
efficiency penalty observed with swirl. Additional losses occur depending on the
swirler orientation and clocking position. The latter aspects are also of importance
for the thermal load in the turbine. Qureshi et al. [104] determine a great impact
of swirl orientation and clocking position on endwall heat transfer as well: Both a
significant increase or a slight decrease in Nusselt numbers is observed, the local
divergence and convergence of wall streamlines and an accumulation or dissipation of
boundary layer fluid to be responsible. This working group in addition investigates
an integrated combustor vane concept (Rosic et al. [108]), showing benefits in terms
of cooling air requirements without an adverse effect on aerodynamic performance,
compared to a standard design.

Numerical studies further include coupled simulation approaches with separate do-
mains for combustor and turbine. They include two solvers that run unsteady and
simultaneously but are coupled in each iteration. Such studies are conducted by
Insinna et al. [62] and Vagnoli [130] [129].

The influence of combustor inflow on leading edge cooling is investigated by Yin
et al. [143]. Experimental and numerical studies of combustor outflow turbulence


14 Theoretical background and literature review

intensity and length scales of an engine representative combustor simulator have been
conducted by Koupper et al. [80] The simulator is dedicated for combustor turbine
interaction studies in a high speed turbine rig. Cha et al. [29] highlight the importance
of the turbulence distribution at turbine inlet for a correct numerical estimation of
turbine aerodynamics and performance. They investigate in addition the migration
of the burner core using a passive CO2 tracer to investigate the flow field numerically
and to identify the NGV upstream effect.

Klapdor [73] investigates the combustor-turbine interaction with the Rolls-Royce
in-house combustor code PRECISE, into which compressibility is included. With the
integrated simulation, the extension of the NGV upstream effect in the combustor is
specified to be one axial chord length and the influence on the combustor flow field
is analyzed in detail.

2.3 Endwall film cooling

Within the present investigation, a film cooling design for a NGV endwall is to be
analyzed. Relevant factors of influence on film cooling effectiveness and the measures
to quantify it are presented in the following section.

Requirement of endwall cooling

The efficiency of gas turbine is optimized by increasing the compressor pressure ratio
to higher values. Following the equation of state, with a higher pressure, the com-
bustor inlet temperature increases. To obtain the same work output the turbine inlet
temperature needs to be increased as well. Over the recent years, temperatures are
raised up to 1900 K. This is beyond operational temperatures of the materials used
in the turbine. Therefore, cooling is required. The use of coolant air is intended at
a minimum, since it constitutes a penalty and does not fully contribute to the cy-
cle. Furthermore, component life increases significantly with a reduction in material
temperature. Hence, efficient cooling techniques and a profound understanding of the
flow mechanisms are required to cope with the design changes imposed by lean pre-
mixed combustion concept. A crucial component area to be looked after with special
care when applying a lean burn combustor is used, is the endwall.

Cooling methods and definition of the film cooling effectiveness

The cooling techniques used in gas turbines to maintain the material temperature
within the limitations of their mechanical strength can be categorized as follows
(Bräunling [22], Baldauf [12]):

• Passive cooling techniques; So called thermal barrier coatings (TBC)
are applied to the surface. They are capable to withstand extreme temperatures
and have a low thermal conductivity to minimize the heat conduction into the
base material.

• Internal Cooling techniques maximize the heat removed from the material
and in this way reduce the material temperature. Techniques applied belong
to the convective cooling methods in internal vane or blade cooling passages


2.3. Endwall film cooling 15

to increase the internal heat transfer rates. A special case is impingement
cooling, such as it is applied in the leading edge region or below the combustor
or turbine casing liner, where very high heat fluxes occur.

• External cooling includes film cooling. This techniques reduces the driving
temperature by introducing a film of colder air through slots or holes on the
surface, resulting in a mixing temperature above the wall, denoted as the film
temperature Tf . Film cooling is the method in focus for this study.

• The combination of internal and external cooling is represented by effu-
sion cooling, whereby a large number of small holes is applied to the surface,
a combustor liner for example. Whereas the primary effect is due to the the
cooling film, the surface temperature is in addition reduced by the extensive
heat conduction in the perforated surface as well.

To quantify the utilization of the available temperature difference between coolant and
main flow for a film cooled wall, the film cooling effectiveness η is defined according
to equation 2.1. The coefficient is usually presented as the adiabatic film cooling
effectiveness and the film temperature is replaced by the adiabatic wall temperature.
This temperature is achieved at the wall with zero heat flux into the wall.

η = T∞ − Tf
T∞ − Tc

= T∞ − TaW

T∞ − Tc
(2.1)

In high speed flow, the compressibility of air needs to be taken into account to
include the influence of the diffusion. According to Schiffer [109], this requires that
in the definition of the film cooling effectiveness in the above equation, the recovery
temperatures of both the main flow and the coolant flow need to be used.

Factors influencing film cooling

The film cooling performance has been improved since its first application and new
manufacturing techniques have allowed to optimize geometries using shaped cooling
hole geometries (Bunker [24]). Numerous studies have identified factors of influence,
whereby the study of Baldauf [12] presents both an extensive review and a wide
experimental parameter study. Friedrichs [41] gives an overview with special regard
on endwall film cooling of axial turbines. Below several parameters of influence are
discussed briefly without making a claim to be complete and further reading to the
topics is provided:

• Blowing ratio M and the momentum ratio I of the coolant injection1:
Thole et al. [125] show that the penetration depth of film cooling jets into the
main flow is dependent on I. According to Baldauf et al. [14], the decisive
parameter in the near hole region is I, whereas the film cooling effectiveness
correlates with M in the downstream region.

• The density ratio DR of coolant and main flow: In the real engine, the
temperature difference between the two air flows results in a DR of about two.

1The definition of these parameters is given in section 3.3


16 Theoretical background and literature review

For studies at ambient conditions this makes it impossible to match both M
and I at the same time, when the DR cannot be matched (Thole et al. [125]).
The use of a foreign gas with a greater density than air as a coolant is a method
to overcome this difference and to replicate both M and I simultaneously in the
experiment.

• Thickness and shape of the boundary layer of the flow approaching the
holes: The coolant jets protrude through a thin boundary layer with respect to
the cooling hole diameter. A thicker boundary layer thereby allows the coolant
jets to remain within the boundary layer and to energize it (Dückershoff [37]).
However, the shear forces responsible for the turning of the jet are greater.
Therefore, a general statement to the influence of the boundary layer thickness
is difficult and other parameters of influence have to be considered at the same
time.

• Turbulence intensity and length scales in the main flow and boundary layer:
Increased turbulence intensity favor lateral and radial mixing, resulting in in-
creased lateral film cooling effectiveness but quicker dilution with downstream
distance according to Baldauf et al. [14]. For film cooling ejection with jet inter-
action, i.e. for narrow hole spacing, the influence of the freestream turbulence
decreases accordingly. Similar findings were observed by Burd et al. [25]

• Favorable or adverse pressure gradients i.e. accelerated or decelerated bound-
ary layers: They are imposed by shape of the airfoil surfaces (Dückershoff [37]).

• Cooling hole interior flow and the flow direction of the incoming flow in the
supply plenum.

• The potential field of downstream components: Thomas [126] illustrates the
upstream effect of the leading edge (LE). It is responsible for local blowing rate
variations of endwall cooling holes upstream of the LE in the case of endwall
cooling. The reduced main flow velocity causes the coolant flow to detach from
the surface, resulting in inefficient cooling.

Furthermore, geometrical parameters of the film cooling configuration are relevant:

• The L/D ratio, i.e. ratio of diameter D to length L of the cooling hole. Small
L/D-ratio holes favor ”jetting” of the coolant farther into the main flow due to
a more inhomogeneous velocity profile at the hole exit, whereas high L/D-ratio
favors the jet attachment to the wall. According to Burd et al. [25] with high
free stream turbulence these differences diminishes.

• Lateral and radial inclination of the cooling hole to the main flow direction:
A lateral inclination influences the vortex system of the coolant jet in a way,
that one of the main vortices is attenuated and the other becomes mitigated.
The new vortex system is less prone to lift off of the surface, according to Wolff
[141]. Baldauf et al. [14] state, that steep radial injection angle promote the
interaction between neighboring coolant jets.


2.3. Endwall film cooling 17

• Various shapes of the cooling hole have been widely studied in the literature. A
common method to reduce the momentum ratio for a coolant hole is fan-shaping.
The coolant hole outlet area is increased in lateral and/or axial direction. This
is beneficial to the cooling effect on the wall, as shown by Bogard and Thole
[21] and Dittmar et al [35][36].

• Lateral spacing to adjacent cooling holes, as identified by Cresci [34], has a
great influence on the performance of a film cooling row. A closer spacing favors
the attachment. A lateral spacing of at least 3D is used in practice to maintain
the structural integrity, according to Kodzwa and Eaton [78].

• Surface curvature: Schwarz et al. [119] showed, that a convex shaped surface
can be beneficial for the downstream film cooling effectiveness, whereby the
effect depends on the blowing ratio M . A pressure gradient is introduced in the
radial direction of the curvature.

• Hole pattern and both lateral and axial spacing to subsequent or preceding
cooling rows: The importance of coolant injection on the NGV inlet flow field
was highlighted by Cresci et al. [34]. In the experiment, a double row of cylin-
drical holes was investigated, highlighting the influence of hole spacing on the
local coolant-to-main flow momentum ratio. A closer spacing helps to prevent
film-lift-off, since the effective bulk flow velocity for the second row is increased
due to the injection from the first row and the effective blowing rate is reduced.
Burdet and Abhari [26] show that a row of cooling holes is significantly influ-
enced by the flow field of its preceding row. A staggered configuration was
determined to have superior performance because their counter-rotating vortex
pairs (CRVPs) combine to a downward influence that keeps the coolant jets
more attached. On contrary, the in-line configuration of cooling holes rows in-
creases the tendency of the trailing cooling hole jet to lift off the surface. The
CRVP between different rows interact in a different manner, depending on dis-
tance and pattern, which can be beneficial or detrimental, in addition to the
effect of the preceding row on the effective bulk velocity (Kodzwa and Eaton
[78]). To evaluate the cooling performance, Kirollos and Povey [70] addressed
the superimposition of subsequent cooling films with an energy-based method.
Similarly, Kneer et al. [76] addressed the superimposition effect as well.

• Surface roughness of the wall and the hole interior is addressed by Schroeder
and Thole [117] and (Lorenz [93]).

Even though all of the above-mentioned parameters have been investigated in detail
isolated from other effects, in the real turbomachine, they appear simultaneously and
interact with each other. Kodzwa and Eaton [78] present existing approaches to
encounter this aspect.

Coolant injection vortex system and near hole flow field

A box with a length of 3 hole diameters both in lateral and in axial direction around
each cooling hole is referred to as the near-hole flow field in film cooling problems.
There exists a characteristic flow field that can be idealized as a jet-in-crossflow with


18 Theoretical background and literature review

an inclination angle. Although the flow field in general has been widely studied
among the literature and an agreement about its flow features exists, their origin is
disputed. Dückershoff [37] has illustrated the vortices that typically develop due to
a coolant injection (Figure 2.4). Studies of Andreopoulos and Rodi [6] and Burdet
et al. [27] support these findings. According to Burdet et al. [27], additional vortices
may develop due to vorticity entering the coolant hole from its plenum.

Figure 2.4: Vortex system due to a coolant jet injection (Dückershoff [37])

The common understanding of the vortex system is summarized by Dückershoff
[37] to consist of four individual vortices, designated Ω1 through Ω4 in Figure 2.4.
In the stagnation region upstream of the jet, similar to an obstruction by a cylinder
in crossflow, the static pressure increases. Downstream, a wake zone with low static
pressure is found. The gradient and the inertial forces of the free stream causes the
coolant jet to bend. Dückershoff [37] argues, that a Von-Karman vortex shedding
occurs (Ω4 in Fig. 2.4), similar to the wake of a cylinder in crossflow. Burdet et al.
[27] on contrary states, that the flow has a rather potential character and the main
mechanism is entrainment of main flow into the low-pressure wake zone. Similarly, due
to the obstruction by the jet, a horseshoe-vortex system forms (Ω3 in Fig. 2.4). The
dominant flow feature is the counter-rotating vortex pair (CRVP, Ω2 in Fig. 2.4) or
kidney-vortices, which form due to the pressure gradient upstream and downstream
of the jet, according to Dückershoff [37] and Fric and Roshko [40]. Burdet et al.
[27] and Moussa et al. [98] however account a vorticity ring, which develops inside
the cooling hole, responsible for the CRVP. They agree that this vortex structure
assists in the entrainment of main flow into the jet’s wake. Leylek and Zerkle [89]
have investigated the flow inside a cooling hole and validated, that a CRVP exists in
the cooling hole already. Lemmon et al. [88] continued the investigation, comparing a
simulation without boundary layers inside the cooling hole to the datum configuration.
Accordingly, decisive for formation of the CRVP in the near-hole flow field is the free
shear layer between jet and mainstream. The main flow is accelerated around the
coolant jet, leading to pressure gradients and a compensating secondary flow (Ω1 in
Fig. 2.4).


2.3. Endwall film cooling 19

(a) Low momentum ratio (b) High momentum ratio

Figure 2.5: Jet-in-crossflow model by Andreopoulos and Rodi [6] for normal injection
and varied momentum ratio

Mixing between the flow occurs around the coolant jet. The momentum ratio of
coolant and main flow is decisive on whether the jet lifts off or remains within the
boundary layer flow. This aspect is also observed by Andreopoulos and Rodi [6]
(Figure 2.5) for a jet-in-crossflow with normal injection. They observe an increased
radius of the curvature of the jet trajectory with increased blowing rate. Kodzwa and
Eaton [78] argue that for low momentum ratio, the flow exiting the cooling hole is
obstructed and hindered by the freestream in the upstream half of the hole. This effect
diminishes with increasing momentum ratio. However, also at very low momentum
ratios, a downstream wake region forms with very low flow velocity, that allows the
main flow to penetrate underneath the jet core. This is experimentally illustrated
by Bernsdorf et al. [19]. They identify, that this main flow air is extracted from the
boundary layer, which therefore becomes locally thinned. This in turn increases heat
transfer. Baldauf et al. [15] demonstrate that this increase is concordant with the
increase in blowing ratio.

Classical view on secondary flows in the vane passage

Because of its great importance for endwall cooling, the flow field in a vane passage
is discussed in detail. The classical view with a typical boundary layer, i.e. an
uncooled endwall is illustrated first. Subsequently, the radial equilibrium equation is
illustrated and the influence of coolant injection on the passage flow field is discussed.
The secondary flows influence the endwall heat transfer in a vane passage as well and
the aspects are discussed with this respect in the next section.

Goldstein and Spores [48] have illustrated the common understanding of the vane
flow field:

Accordingly, two pressure gradients are responsible for the formation of the sec-
ondary flows: The horseshoe vortex (Fig. 2.6, item 1. & 2.) is induced by the radial
stagnation pressure gradient in the incoming boundary layer. It is converted into a


20 Theoretical background and literature review

Figure 2.6: Secondary flows in a vane passage (Goldstein and Spores [48])

static pressure gradient when approaching the leading edge of the vane due to the
flow deceleration. This forces the flow to turn towards the endwall, and thus break-
ing up into two vortex structures. One leg enters the passage along the suction side
(SS) of the vane (Fig. 2.6, item 2.) and the other along the pressure side (PS) of
the vane (Fig. 2.6, item 1.) with opposing rotational direction. The turning of the
vanes induces the second important pressure gradient from the PS to the SS of the
vane. The passage vortex develops (Fig. 2.6, item 3.), where the pressure side leg
of the horseshoe vortex encounters the vane surface at its shoulder, after crossing
the vane passage. The SS leg of the horseshoe vortex remains close to the endwall
up to a certain point, where it lifts off as well and starts to roll around the passage
vortex with opposing rotational direction. Further low-momentum vortices, denoted
corner vortices, develop at the airfoil corners (Fig. 2.6, item 4., 5. & 6.). They are
a result of additional boundary layers in radial direction, which develop due to the
horseshoe vortex’ trajectory at the leading edge. This induces a boundary layer in
radial direction and as a consequence, vortices form in the corners.

Horseshoe and passage vortex cause the formation of a three-dimensional separation
line for the inlet boundary layer. Downstream of the separation line, a highly skewed
new boundary layer develops and is fed by a downwash from the vanes pressure side
(Fig. 2.6, item 8.). Friedrichs [41] states that as a consequence, the area downstream
of the separation line is crucial in endwall film cooling because the cooling air in the
boundary layer as well separates from the endwall. With the downwash from the
PS, airfoil coolant air may be transported towards the endwall and contribute to the


2.3. Endwall film cooling 21

cooling there. Nevertheless, typical gas turbine endwalls usually need to be cooled
with additional passage cooling holes in this region. The benefit of such an injection
at the spot has to be compared to the increased mixing losses because of the higher
velocities. Benton et al. [18] state that a coolant injection upstream of the vane row
is beneficial from this point of view.

Radial equilibrium in the passage flow and influence of coolant injection

As it has been discussed, the passage flow field causes coolant air that is injected
with low momentum is observed to detach from the endwall at the separation line. In
consequence, uncooled areas appear downstream. Responsible is a passage cross flow
from the PS to the SS. This in turn results due to the radial equilibrium with respect
to the turning in the vane passage, as Thomas and Povey [127] illustrates (Fig. 2.7).

Figure 2.7: Mechanism responsible for the passage cross flow according to Thomas
and Povey [127]

The mechanism is driven by the pressure gradient from PS to SS. Because the
static pressure through a fully developed boundary layer is constant, low-velocity fluid
(VA) particles near the endwall are subject to this gradient. The radial equilibrium
according to equation 2.2 requires a smaller radius RA in contrast to a higher-velocity
(VB) fluid particle outside of the boundary layer.

δP

δn
= ρA

V 2
A

RA
= ρB

V 2
B

RB
(2.2)

In analogy, a greater velocity flow near the wall than in the main flow, e.g. caused
by coolant injection through film cooling holes upstream of the NGV-leading edge,
reverses the crossflow direction. This is discussed in the work of Thomas [126] (Fig-
ure 2.8). The study examines a film cooling configuration with a double row of
cylindrical holes on the endwall and varying coolant injection rates. Accordingly, the
classical view on the dominant influence of secondary flows holds only true without


22 Theoretical background and literature review

coolant injection or for low coolant-to-mainflow momentum ratios. For increasing
momentum ratios, the boundary layer holds a surplus of momentum and overcomes
the adverse pressure gradient imposed by the vane pressure field. The mechanism due
to the radial equilibrium is reversed and as a consequence the passage cross flow is
inverted compared to the common direction. In this way, no boundary layer separa-
tion and uncooled region is observed for higher injection rates of 3 % and 6 % because
the near-wall flow migrates from SS to PS of the vane. This explains the observed
cold-streak on the endwall at the TE as well; the coolant is driven towards the PS on
the endwall. The formation of secondary flows can be hindered or mitigated in this
way by the coolant injection. A similar observation determine Colban et al [32][33]
and Knost and Thole [77] for a slot ejection upstream of a vane row.

Figure 2.8: Endwall film cooling contours for varied coolant injection rates
(Thomas [126])

2.4 Endwall heat transfer

The material temperature in the turbine depends on the heat load, i.e. heat flux, im-
posed by the near wall flow field. It depends on driving temperature difference, which
in turn results from the incoming temperature profile, the film cooling performance
and the the local heat transfer situation.

It is predominantly influenced by the shape and size of the boundary layer. Flow
components normal to the wall, as they are imposed by vortices and secondary flows,
drive main flow fluid towards the wall. This reduces the boundary layer thickness and
increases heat transfer.


2.4. Endwall heat transfer 23

Theoretical considerations and definition

The wall-normal heat flux into the flow at any point along a surface is derived with
the local temperature gradient at the wall:

q̇ = λ
δT

δy

∣∣∣∣
y=0

(2.3)

Thereby, λ is the thermal conductivity of the fluid near the wall (y ≈ 0). Due to its
small scale, it is hardly possible to resolve the gradient δT

δy
experimentally (Han and

Goldstein [56]). As a simplification, the thermal transport is treated as a boundary
layer problem. The local heat transfer coefficient h is defined as the local heat flux
normalized by the temperature difference between the wall Tw and a certain reference
temperature Tref.

h = q̇

Tw − Tref
(2.4)

A number of studies use the stagnation temperature of the incoming flow, T∞ as
a reference. However, Moffat [97] states that it is beneficial to use the local adiabatic
wall temperature instead. It is defined as the temperature that the wall would attain,
if it was thermally isolated and no heat flux into the surface would occur. The heat
transfer coefficient that uses this temperature as a reference is called the adiabatic
heat transfer coefficient had. In this form, it is linearly dependent on the temperature
difference between the wall and this adiabatic temperature and only a function of the
fluid properties and the flow characteristics. Experimental approaches to approximate
this temperature by measurement as good as possible apply a very low conductive
material to limit the heat flux into the surface to a minimum and the resulting wall
temperature is measured. A more consequent approach is suggested by Goldstein [47]
and Gritsch et al. [52]. TaW and had are determined by a superposition approach.
This requires to measure at least two combinations of wall temperature and heat flux;
both values are then calculated with a linear regression.

Overview on heat transfer measurement techniques in turbomachinery

A brief overview of measurement approaches to determine local heat transfer coeffi-
cients in turbomachinery is given below. Han and Goldstein [56] used a butt-welded
small-scale thermocouple on a transferable mount to study thermal boundary layer
profiles near a turbine stator endwall. The technique is limited by conduction errors
in thin boundary layers and positioning accuracy of the probe in high flow velocities.

Blair [20] conducted research on local film cooling effectiveness and heat transfer
values in a linear cascade. The study applied two separate experiments, where the
first one used a low conductive material to determine adiabatic wall temperatures.
For the second experiment multiple surface heaters were applied, maintaining the
temperature level elevated, but constant for the whole endwall. To achieve this,
varied heat flux from each heater was required and measured to quantify the local
heat transfer value. Goldstein and Spores [48] used an indirect measurement method:
The heat and mass transfer analogy with the naphthalene sublimation technique
provides a way to acquire representative Stanton numbers. They discovered peak


24 Theoretical background and literature review

heat transfer regions in the NGV endwall region and showed a correlation to the
secondary vortex system in the passage. Moreover, they discovered evidence which
helped in understanding the formation of the pressure side and suction side corner
vortices. Airfoil and endwall heat transfer was also measured by Graziani et al. [49] in
a linear cascade: Surface thermocouples were placed on strip heaters to achieve high
spatial resolution in regions with strong gradients. The study also showed the effect of
the inlet boundary layer thickness: For a thin boundary layer, the three-dimensional
separation line moves further down into the passage. The streamlines downstream of
the separation line were nearly perpendicular to the endwall pressure field, whereas
the heat transfer contours exhibited an angle to it. A low heat transfer region was
discovered near the pressure side. All methods are a challenge for experimental work,
since both the heat flux and the driving temperature gradient have to be measured.
This comes in addition to the limited accessibility in many test rigs.

Giel et al. [44] [43] studied heat transfer on turbine blades and endwalls, using a
constant wall temperature approach. They applied thermochromic liquid crystals on
well-conductive material, which was covered by a low-conductive material. The ad-
vantage of their approach was that conduction into the structure could be neglected
since they measured the local heat flux across the covering material. A similar ex-
perimental technique is used by Laveau [82], who uses a thin layer of Kapton. This
method is suitable for complex 3-dim. shapes and a contoured endwall. Xue et al.
[142] use a transient technique to gather recovery temperatures, film cooling effec-
tiveness and Nusselt numbers in a transonic turbine cascade. They apply a linear
regression and two experiments with different coolant temperatures, but otherwise
constant experimental parameters.

Another approach in a steady setup was used by Nicklas [100]. Series of Infrared
measurements of a film-cooled endwall with multiple constant heat flux settings were
evaluated using the superposition approach and a linear regression. In this way they
determined adiabatic wall temperatures and heat transfer coefficients. An overview
of other direct heat transfer measurement techniques was presented by Kaiser [66].

Endwall heat transfer peak regions

The importance of the vane flow field for endwall film cooling has been discussed
already. The heat transfer in a vane passage as well depends largely on this flow field.
Common areas of peak heat transfer have been determined within previous studies,
summarized by Friedrichs [41]. A typical distribution for Nusselt numbers on a NGV
endwall is shown in Figure 2.9, obtained at the LISA rig at ETH Zürich that has a
comparable setup to the one used for the present work.

The following areas are typically prone to high heat transfer:

• The stagnation region upstream of the vane leading edge A . It is detected
in many investigations, both in cascade and annular rigs (Laveau [82], Panchal
et al. [102], Graziani et al. [49], Goldstein and Spores [48]). In this area, in-
creased heat transfer occurs due to the high wall normal flow transport caused
by the formation of the horseshoe vortex. Main flow is swept down onto the
endwall. The application of fillets limits the extent of the horseshoe vortex and
reduces the peak (Shih and Lin [122]).


2.4. Endwall heat transfer 25

A

B

C

E

F
D

Figure 2.9: Endwall Nusselt numbers in an axial turbine without coolant injection
showing typical peak heat transfer areas (adapted from Laveau [82])

• Following the trajectory of the PS leg of the horseshoe vortex B results in
a heat transfer peak along its trace across the passage visible in the results of
Laveau [82] in Figure 2.9.

• At the suction side shoulder C , both legs of the horseshoe vortex merge
to form the passage vortex (Friedrichs [41]), which then lifts off the wall. A
peak in this area is also identified by Blair [20], Goldstein and Spores [48] and
Graziani et al. [49]. In addition to the vortices, high velocities are regarded
responsible for the heat transfer peak. The location often coincides with the
onset of the corner vortex according to Panchal et al. [102]. The corner vortex
causes increased wall-normal flow transport and therefore as well a peak HTC
value along the suction side corner.

• The velocity increase in the passage throat D and downstream of it is re-
sponsible for peak Nusselt numbers. This is observed by Blair [20], Graziani
et al. [49] and Laveau [82].

• The trailing edge wake E of the vane is subject to downwash from the vane
wake and high velocities. This hot spot is observed by e.g. Goldstein and Spores
[48] and Graziani et al. [49]

• A corner vortex forms in the pressure side corner F and together with the
thin boundary layer in this region causes high heat transfer levels.

Influence of film cooling on heat transfer

Endwall coolant injection through a row of holes increases Nusselt numbers locally
and then diminishes with the downstream distance on the endwall (Baldauf [12]). The
mechanism has been investigated in detail in the literature and it is closely connected
with the near-hole flow field, illustrated before. The wake region downstream of the
coolant injection allows the main flow to penetrate underneath the jet core (Bernsdorf
et al. [19]). This main flow air is drawn from the boundary layer, which therefore


26 Theoretical background and literature review

gets locally thinned. The thinner boundary layer means that heat transfer rates are
increased in the region downstream of the hole. This is shown by Baldauf et al.
[15] [12] [13], who also demonstrate that this increase is concordant with the increase
in blowing ratio. In addition, the local increase of the turbulence intensity, driven
by the mixing and shear processes, results in increased heat transfer. The increase
persists further into the flow direction, which is concordant with the blowing rate as
well. Baldauf [12] presents three typical heat transfer regimes for a coolant injection,
typically found with increasing blowing rate (Figure 2.10). It is notable, that the
phenomena displayed for a single hole have been observed in an experiment with a
row of holes. For blowing rates greater than M=1.7, the jet completely detaches
from the surface (denoted ”penetrating coolant jet”) and the effect of a single hole
becomes amplified by adjacent jets. Low heat transfer rates are found immediately
downstream of the holes and enhanced heat transfer is observed sideways of the jet
due to displacement effects of the ejected coolant. The vortex structure of the CRVP
is responsible for this enhancement. It creates a secondary flow towards the surface
between adjacent jets and a secondary flow away from the surface along the centerline
of the hole and hence a zone of low heat transfer. The resulting striped pattern of low
and high heat transfer can be observed to several multiple diameters downstream of
the cooling hole, before the influence is dissipated and the heat transfer level reaches
that of an uncooled reference.

attached 
coolant jet

transition
situation

penetrating
coolant jet

vortex surface interaction

turbulent wake

downstream wake

midspan low momentum

midspan stagnation

jet displacement

Figure 2.10: Characteristic heat transfer pattern and the origin of the observed con-
tours downstream of a coolant injection in different flow regimes (Baldauf
[12])


2.5. Resulting heat flux reduction due to film cooling 27

2.5 Resulting heat flux reduction due to film cooling

To assess the performance and benefits of a film cooling design to the engine, both
thermal parameters have to be combined. To calculate the material temperature,
the local heat flux needs to be analyzed. This heat flux however is specific to the
achieved absolute wall temperature level, which is a priori unknown. In addition, this
level depends on the local thermal conduction within the material and the convective
heat transfer on the opposing side of the wall, enclosing the internal cooling system.
To study this overall thermal performance experimentally, matching Biot numbers as
well as a realistic internal cooling configuration are required (Chavez et al. [31]).
Sen et al. [120] have proposed the concept of the ”net heat flux reduction” (NHFR) to
evaluate the design according to equation 2.5. This measure has been widely applied
in turbomachinery (e.g. Popović et al. [103], Griffini et al. [50]).

NHFR = 1− q̇W
q̇0

(2.5)

The parameter q̇W denotes the local heat flux for the film cooled design and q̇0 for
the uncooled endwall.
This can be rewritten to

NHFR = 1− hf
h0

(1− ηθ) (2.6)

using the local values of the heat transfer augmentation due to film cooling hf/h0,
the local film cooling effectiveness η and dimensionless temperature ratio θ.

θ = T∞ − Tc
T∞ − TW

(2.7)

As a consequence, the coefficient NHFR is specific to a temperature level, which is
not known a priori for a new design. A method to evaluate a design with absence of the
actual wall temperature level and θ has been suggested by Baldauf [12] (Fig. 2.11). In
the definition of a heat flux reduction Θ accounts for the total thermal transmittance
to the internal cooling system. The heat transfer coefficient hw from the wall to the
back surface internal coolant is approximated with h0, the heat transfer coefficient on
the main flow side. This assumes a 1-dim. conduction through the wall (Baldauf [12])
and that the heat flux is dependent on the external flow field. For the film cooled
case, hw is is approximated with hf , the heat transfer coefficient with film cooling on
the cooled external surface. In this way, the equation2 can be rewritten to:

Θ = 1− hf,ges (TaW − TC)
h0,ges (T∞ − TC) = 1− hf (hw + h0)

h0 (hw + hf ) (1− η) (2.8)

2Baldauf uses TG for the main gas temperature. To remain consistent with the nomenclature
used in this thesis, it is changed to T∞ in all equations of this section


28 Theoretical background and literature review

TG TAW

TC

hfh0

hW hW

q̇0 q̇W

h0ges
h0 hW

hW h0+
------------------- h= f ges

hf hW
hW hf+
-------------------=

Figure 2.11: Heat flux analysis through the film cooled endwall according to Baldauf
[12]

Furthermore it is assumed that invariant internal cooling is given i.e. that the
heat flux is also transported ideally to the internal cooling system. This implies that
hw = h0 and the equation is simplified to

Θ = 1− 2hf/h0

1 + hf/h0
(1− η) (2.9)

As a consequence, only the quotient of hf/h0 and no further assumption of a tem-
perature level is required to assess the total heat flux reduction due to one or several
cooling configurations. This assumption of ideal internal cooling is representative for
a gas turbine. In their design it is intended to achieve a close matching between the
internal cooling design and the imposed external heat load wherever possible.

Approaches that take this goal into account are discussed by Kirollos and
Povey [69] [71] [72]. They investigate the requirements for such an optimized internal
cooling system. For a given internal cooling air temperature they show that a
reverse-flowing coolant to the exterior heat load allows for a significant reduction of
the wall temperature.

2.6 Definition of research objectives for the thesis

With the discussion of relevant literature, open questions and research objectives are
derived that are addressed with this thesis. The presented experimental research
work is dedicated to achieve a comprehensive understanding of the interaction of
combustor and turbine regarding film cooling effectiveness and heat transfer to foster
a robust design. To achieve this, a suitable test specimen is needed. The first vane
endwall has been identified as a crucial area concerning the development of lean burn
combustors and is the focus area of this investigation. In summary, this yields two
research objectives:

• To design a test rig setup that models the complex flow field with great detail.
This includes a representative interface flow field of a lean combustor and a


2.6. Definition of research objectives for the thesis 29

fully cooled vane row, including endwall cooling, airfoil cooling and trailing
edge ejection.

• To develop a measurement method that allows the simultaneous acquisition of
heat transfer and film cooling effectiveness data on the turbine hub endwall with
high spatial resolution.

With this setup and the defined measurement method, a cooling design is investi-
gated for a reference case of axial inflow to understand the aerodynamic & thermal
behavior. A coolant mass flow variation is applied to gain insight into factors of in-
fluence and the benefits and drawbacks of low to high-momentum injection. This is
summarized by three aims:

• Understanding the interaction between the hub side coolant injection and the
flow field in the turbine for a baseline case of axial, low-turbulent inflow.

• An investigation of the impact of varying Rear Inner Discharge Nozzle (RIDN)
coolant mass flow rates.

• The use of a measure to evaluate the combined effect of both parameter sets.
Baldauf [12] presented a suitable method for a representative assessment of the
thermal benefit. This method is applied to the results.

To assess the impact of lean combustion on the turbine, an engine-representative in-
terface flow field is applied to understand the imposed aerodynamic changes. Swirling
inflow in a cold flow setup is used. Two additional aspects are addressed in this way:

• To determine and analyze the changes imposed by swirling inflow on the NGV
flow field and its impact on heat transfer and film cooling.

• An evaluation of the robustness of the film cooling design to swirling inflow and
to derive recommendations for an improved design.

In addition, the results obtained allow their dissemination and enlargement within
following measurement campaigns and research projects in the same working group.
The results acquired with steady measurement techniques presented in this thesis
form the basis for the wider goal to achieve a comprehensive understanding of the
aerothermal impact of swirling inflow on the turbine stage.


Experimental setup 31

3 Experimental setup

This chapter describes the experimental setup used to acquire the results presented
in this thesis. Specific emphasis, aside from the facility description, lies on the illus-
tration of the combustor module for the CTI investigation and the endwall coolant
geometry. An overview on the measurement campaign and the operating point is
given to conclude the chapter.

3.1 The Large Scale Turbine Rig

The LSTR consists of a scaled-up 1.5-stage low Mach number turbine in a closed-loop
configuration based on the aerodynamic design of a high-pressure turbine, scaled to
low Mach-number conditions. It has been set up for the investigations of Linker [90]
and Schrewe [116] to study the impact of purge flow injection. The main flow is
provided by a radial compressor which delivers a mass flow of about 9.5 kg/s. The
secondary airflow is provided by an accessory radial compressor which supplies a mass
flow of up to 0.9 kg/s at a pressure ratio of 1.7. Both mass flows are varied depending
on the coolant injection setting, resulting in a constant rotor inlet mass flow rate.
The rig’s general setup is shown in Fig. 3.1.

Measurement 
section

Primary air
cooler

Primary air
blower

Secondary air
blower

Secondary air
cooler & distributor

Settling
chamber

Generator

Figure 3.1: General setup of the Large Scale Turbine Rig


32 Experimental setup

The mass flows and the rotor speed are adjusted according to the ambient condi-
tions to maintain the Reynolds number of the system constant. A small influence
of changing ambient conditions on the Mach number cannot be avoided at the same
time, but is considered negligible, since the Mach number is low enough for the flow
to be treated as incompressible throughout the turbine stage. The temperature of
secondary and main flow can be adjusted independently using two water driven heat
exchangers. This allows for a temperature difference between secondary and main
flow of up to 20 K.

The test rig features a NGV1 row with 24 vanes, which also feature airfoil film
cooling and a trailing edge slot ejection. The rotor row consists of 36 squealer-tip
blades. An NGV2 row with 34 vanes directs the flow towards the outlet casing.
Krichbaum et al. [81] present a description of the test rig and all capabilities.

Combustor Module Turbine Section
ME00 ME01 ME02 ME03 ME04 ME05

RE1-3RE1-3
RE1-3RE1-3

RIDN Cooling 
Air Flow

NGV1 NGV2Rotor

Swirler

Figure 3.2: Cross section of the test rig, showing the measurement planes (ME) and
RIDN boundary layer planes (RE) as well as a schematic of the RIDN
cooling flow path. In this illustration the swirler module is rotated by 15◦

and is hence is displayed below midspan in the figure. The black arrow
indicates the swirler center axis within the annulus.

The cross section of the test rig is shown in Fig. 3.2. Measurement planes (ME) are
located upstream and downstream of all vane and blade rows. In ME01, the turbine
inlet flow field is determined 0.9 axial chord lengths (Cax) upstream of the NGV’s


3.2. Combustor module 33

LE. The NGV outlet flow field is evaluated in ME02, 0.2 Cax downstream of the TE,
which equals to about one true chord length in the exit flow direction.

Scaling of the vane geometry from engine conditions

The vane geometry used for this investigation has been scaled from a high-speed
turbine rig by the industry partner Rolls-Royce [94]. To achieve a representative
geometry, the distribution of the pressure coefficient cp,vane of the high-speed vane
has been maintained similar in the low-speed design. The flaw of this process is that
using this design rule, a very small leading edge radius results and the airfoil shape is
not similar to the one of the high-speed vane. Because this could cause other design
features, such as the airfoil film cooling and flow features such as the secondary flows
to differ from the engine, a compromise is used. The pressure coefficient is altered
slightly in the leading edge region, such that a larger leading edge radius is achieved.

cp,vane = pt,ME01 − p
pt,ME01 − ps,ME02

(3.1)

3.2 Combustor module

A combustor module was used to vary the inflow to the turbine stage. The flow in
the combustor module is non-reacting and hence isothermal with near ambient tem-
perature. Two configurations were studied: A baseline case of an axial, low-turbulent
inflow without any installations (denoted as AX configuration) and a swirling inflow
configuration, representative for a lean combustor, with two clocking positions relative
to the NGV (denoted SWL/SWP). The goal of the combustor module is to achieve
an engine-representative whirl-angle and pressure profile at turbine inlet.

Annulus geometry near the RIDN injection

The hub endwall in the combustor module is inclined to the turbine axis, shaping a
tapered annulus geometry. This provides an acceleration and a representative inflow
to the turbine stage in the swirling inflow case. In a real engine, the heat release in
the combustor decreases the density by about 60 % and accelerates the flow. This
leads to an attenuation of the initial swirl level in the combustor, since the tangential
components remain constant (Schmid [112]).

The wall inclination φ is 13◦ to the turbine axis downstream of the swirler and,
beginning from a relative axial position of -0.75 Cax upstream of the LE, increases
to φ=39◦. The geometry and angle definitions are specified in Fig. 3.3 and table 3.1,
with the values of wall inclination angle ψ and measurement plane inclination angle
ψ for the three RIDN injection measurement planes RE1-3 and ME01 with respect
to the turbine axis. The angles are also specified for the two injection rows at the
intersection of their centerline axis with the hub wall.

The coolant injection holes are placed with a shallow relative angle of 21◦ to the
surface (row 1) and 25◦ (row 2). Since the curvature of the hub wall begins at the
onset of row 2, the relative angle changes from 21◦ at the windward side to 32◦ at
the lee side of row 2. The position of the measurement planes is specified in axial
direction with respect to the LE, non-dimensionalized with Cax. In addition, the


34 Experimental setup

RE3RE2
RE1

ME01

NGV1

RIDN-
Plenum

x
φ

ψ 1
2xs

Figure 3.3: Main annulus geometry and measurement planes in the vicinity of the
RIDN injection, dimensions specified in table 3.1

distance parallel to the endwall along its curvature xs is listed with respect to the
hole center axis, non-dimensionalized with the hole diameter D.

Swirling Inflow

A swirler insert manufactured from stereo lithography is mounted to a support struc-
ture to generate the swirling inflow (c.f. Fig. 3.6). It was designed by Lohmann [91]
using numerical simulations.

LSTR Swirler Design A characteristic inflow at the turbine inlet with averaged whirl
angles of ± 17◦, representative for a state of the art lean combustor, has been achieved
with the LSTR swirler. It is modeled on the basis of the Engine-3-E -swirler [74] and
consists of three concentric rings with curved blades (Fig. 3.4). Two inner rings with
high airflow turning angles create a strong swirl in counter clockwise direction with
respect to the turbine axis. This high swirl forms a recirculation zone close to the
swirler, which is typical for a lean combustor: In the real engine, this is used to
enhance the mixing of fuel and air and to maintain flame stability. The flow is then
accelerated due to the combustion. In the test rig, this acceleration is achieved by a
tapered main annulus. The acceleration attenuates the whirl angles to engine-realistic
values. The outer blade ring with a low turning angle, counter-rotating to the two
inner rings, is used to adjust the swirl to the desired level at turbine inlet .

12 swirlers are placed within the measurement section. Together with the 24 NGVs
and 36 rotors this yields a CFD friendly domain. Furthermore, the measurements can
be conducted in representative sectors of 30◦, assuming periodicity.


3.2. Combustor module 35

Table 3.1: Measurement planes and geometry near the RIDN injection

Measurement plane φ [◦] ψ [◦] axial pos. [x/Cax] Surface pos. [xs/D]

ME01 13 90 0.9 � 30 ←
RE1 39 48 0.4 � 3.5 ←
RIDN 1 39 150 0.33 � -
RE2 39 45 0.29 � → 2.7
RIDN 2 34 150 0.24 � → 4.5
RE3 23 42 0.19 � ⇒ 2.5

� : upstream NGV LE, measured at hub endwall junction

← : upstream RIDN row 1, measured from the hole centerline

→ : downstream RIDN row 1, measured from the hole centerline

⇒ : downstream RIDN row 2, measured from the hole centerline

Clocking position Two clocking positions of swirler center relative to the NGV lead-
ing edge were studied: Leading Edge clocking (denoted as SWL), where the geometric
center of the swirler hardware was aligned with the leading edge at 50 % span height
and passage clocking (SWP), where it was aligned to the center of a vane passage, i.e.
traversed by half a vane pitch equal to 7.5◦3. Twelve swirler modules were used and
the swirler to vane count is 1:2. As a consequence, every other adjacent vane passage
faces a different inflow condition, denoted as passages A and B (see Figures 3.54).
Hence, the two clocking positions result in four passage data sets for a complete
analysis. For the measurement campaign, both a clean annulus configuration with
axial, low-turbulent inflow (denoted as AX in the following) and a swirling inflow
configuration, as described before, were examined.

Swirl number To describe swirl flows, the non-dimensional swirl number S is defined
as the ratio of the axial flux of tangential momentum to the axial flux of axial mo-
mentum, multiplied with the swirler radius R as characteristic length (Gupta et al.
[53], equation 3.2),

S = Ḋ

lchar İ
=

∫∞
0

[
ρuaxutan + ρu′axu′tan

]
r2 dr

R
∫∞

0

[
ρu2

ax + ρu′ax2 + (p− p∞)
]
r dr.

(3.2)

with the fluctuation part of the turbulent flow u′axu′tan and u′ax2 and the axial thrust
due to the pressure difference with respect to the ambient pressure, (p− p∞). The
definition of the swirl is representative to characterize confined flow such as swirling

3The lateral migration of the core along its trajectory towards the vane leading edge is not
accounted for by this definition.

4The nomenclature of vanes and passages is detailed on Fig. 3.9


36 Experimental setup

Outer Ring

Middle Ring

Inner Ring

Figure 3.4: Triple flow LSTR Swirler design with three concentric blade rings, stream-
wise view. Characteristic flow parameters are listed in table 3.2

Figure 3.5: Illustration of clocking positions and vane passages A & B (Adapted from
Steinhausen [124])

pipe flow. Its use to describe a complex flow field such as in the present setup for
the combustor section is limited due to the additional interaction between several
larger vortical structures (Giller [46]). To allow for a comparison to other studies, the
initial swirl number at the swirler exit plane is presented, derived with a geometrical
estimation according to Gupta et al. [53] for flat vane swirlers:

Sgeom = 2
3

[
1− (Ri/Ro)3

1− (Ri/Ro)2

]
tanφ (3.3)

where φ is the swirler vane angle, Ri the inner and Ro the outer radius of each of the
ring.

To design the LSTR swirler, Lohmann [91] conducted a parameter study, varying
the swirler vane angle settings, axial combustor module length and swirler inclination
with respect to the turbine axis. With the final design, the intent to match an e