Professional Documents
Culture Documents
RTO-TR-044
AC/ 323(AVT-018)TP/ 29
NORTH ATLANTIC TREATY ORGANISATION
RTO/NATO 2002
Single copies of this publication or of a part of it may be made for individual use only. The approval
of the RTA Information Policy Executive is required for more than one copy to be made or an extract
included in another publication. Requests to do so should be sent to the address above.
Report of the RTO Applied Vehicle Technology Panel (AVT) Task Group AVT-018.
Public reporting burder for this collection of information is estibated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing
and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burder to Department of Defense, Washington
Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of
law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
18.
19. NAME OF RESPONSIBLE PERSON
NUMBER RTO Meeting Proceedings from NATO, Bonnie
OF PAGES has CD-ROM
356
lfenster@dtic.mil
19b. TELEPHONE NUMBER
International Area Code
Area Code Telephone Number
703767-9007
DSN
427-9007
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39.18
RTO-TR-044
AC/ 323(AVT-018)TP/ 29
NORTH ATLANTIC TREATY ORGANISATION
Report of the RTO Applied Vehicle Technology Panel (AVT) Task Group AVT-018.
Executive Summary
Simulations have become increasingly important in all aspects of military operations. As modeling
techniques have improved, and as computers have progressed, simulation has assumed an essential role
in planning, executing, and evaluating operations. So too, in the design, manufacturing, and operating
of aircraft turbine engines, accurate performance simulations have become essential. They are being
applied in rapidly expanding ways to reduce acquisition costs, increase system performance, improve
maintenance through improved diagnostics and prognostics, and to improve new system design.
Recognizing the importance of the turbine engine to NATO operations, and the widespread use of
turbine-engine performance simulations, Technical Team AVT-018 was created to provide a manual of
such simulations. These were to range from applications to the latest developments in modeling
technology. Originally focused on operators and applications, the effort has resulted in a manual of
broad scope. There is material for all users of models, ranging from those in original design, to those
who operate and maintain aircraft gas turbine engines. The report is one of the most extensive
available on both the applications and the methodology of engine performance simulations.
Applications of performance models will be of initial interest to many readers and a detailed review of
22 examples is provided. It ranges from preliminary design to in-service support, and includes
educational examples. The document then describes the features of several types of complete engine
models, including the complex and detailed models that are used for the actual design of a new turbine
engine and its components. These models are often of a proprietary nature. All modern performance
models are executed on computer platforms, which often dictate the form or construction of the model,
as well as the method and speed of execution. A chapter describing current computer platforms and
software, and possible future developments is included. Sample executable performance models are
included, an addition made possible by the electronic format of the report. Finally, several recent and
advanced developments in mathematical modeling of components are described. An Appendix
contains the results of a survey of model users, and a Glossary completes the document.
The members of Technical Team AVT-018 hope that the availability of this manual will serve to
increase the use and the value of turbine engine performance simulations in the NATO community, and
encourage progress in performance modeling and related applications.
iii
Synth`ese
La simulation est devenue de plus en plus importante dans tous les aspects des operations militaires.
Avec lamelioration des techniques de modelisation, et levolution de linformatique, la simulation
remplit desormais un role essentiel dans la planification, lexecution et levaluation des operations. De
la meme facon, la simulation de performances precises est desormais indispensable a` la conception, la
fabrication et lexploitation des turbomoteurs. Les simulations sont mises en uvre dans des domaines
sans cesse nouveaux dans le but de reduire les couts dacquisition, daccrotre les performances,
doptimiser la maintenance grace a` un meilleur diagnostic et un meilleur pronostic, et dameliorer la
conception des nouveaux syst`emes.
Etant donne limportance des turbomoteurs pour les operations de lOTAN, ainsi que lemploi
generalise des simulations de performances des turbomoteurs, il a e te decide de creer lequipe
technique AVT-018 en vue de la realisation dun manuel technique pour ces simulations. Louvrage
devait couvrir un domaine allant des applications aux derniers developpements concernant les
technologies de modelisation. Oriente initialement sur les operateurs et les applications, le groupe a
finalement produit un manuel dun champ dapplication tr`es e tendu. Il contient des indications
destinees a` lensemble des utilisateurs de mod`eles, quils soient concepteurs, exploitants ou techniciens
de maintenance de turbomoteurs. Il sagit, en effet, de lun des rapports les plus complets disponibles
sur les applications et la methodologie de la simulation des performances des moteurs.
Les applications des mod`eles de performances interesseront bon nombre de lecteurs non-specialistes et
le manuel presente un examen detaille de 22 exemples. Ils vont de la conception preliminaire a`
lassistance technique en service, avec des exemples didactiques. Le document decrit ensuite les
caracteristiques dun certain nombre de mod`eles de moteurs complets, y compris des mod`eles
complexes et detailles qui sont actuellement utilises pour la conception dun nouveau turbomoteur et
de ses organes. Il sagit dans plusieurs cas de mod`eles de marque deposee. Tous les mod`eles de
performances modernes sont executes sur des plates-formes informatiques, ce qui determine, en
general, la forme ou la construction du mod`ele, ainsi que la methode et la rapidite dexecution. Les
plates-formes informatiques et les logiciels actuels sont examines, ainsi que les developpements futurs
possibles dans un chapitre particulier. Grace au format e lectronique du rapport, les auteurs ont pu y
inclure des exemples de mod`eles de performance executables. Enfin, differents developpements
avances recents dans le domaine de la modelisation mathematique des composants sont decrits. Les
resultats dun sondage dutilisateurs de mod`eles sont donnes en annexe, ainsi quun glossaire des
termes utilises.
Les membres de lequipe technique AVT-018 esp`erent que ce manuel servira a` sensibiliser la
communaute technique de lOTAN a` linteret des simulations des performances des turbomoteurs, et
quil permettra des avancees dans le domaine de la modelisation des performances et des applications
connexes.
iv
Contents
Page
Executive Summary
iii
Synth`ese
iv
vi
vii
Reference
Chapter 1 - Introduction
An overview of current practice and progress in gas turbine modeling and simulation.
Chapter 2 - Applications
Description and examples of applications of engine modeling through the different life cycle
phases.
Chapter 3 - Whole Engine Systems
General considerations, system modeling issues, design and off-design performance modeling, 0,
1, 2, and 3-D modeling, validation and calibration are discussed, and examples given.
Chapter 4 - Computer Platform and Software Implementation
The necessary computer platform and software capabilities are discussed, together with
foreseeable improvements in capability. Several examples of gas turbine simulation systems are
described, some of which are available on the accompanying CD-ROM.
Chapter 5 - Recent Progress
Recent progress in the modeling of both whole engines and of each of the major sub-systems is
described. Performance and simulation of axial compressor performance, turbine sub-systems,
combustor systems, exhaust nozzles and inlet systems, aerodynamics of air systems, and control
systems are all described in detail.
Appendix - Summary Analysis of AVT-18 Survey
vi
vii
Mr Stephane Kammerer
DGA/Centre dessais des propulseurs
91895 Orsay
FRANCE
Olivier Mahias
SNECMA-Centre de Villaroche
Dept AERO-COMBUSTION /YKCN
77 550 Moissy-Cramayel
FRANCE
Maj. A. Kottarakos
Hellenic Air Force
Hellenic Air Force General Staff
C1 Directorate
Mesogeion Ave 237
Holargos 15561
Athens
GREECE
Dr. K. Mathioudakis
Technical University of Athens
Dept. of Mechanical Engineering
Fluids Section, Lab. Thermal Mach.
P.O. Box 64069, 157 10 Athens
(street address: Iroon Polytechniou 9, Athens 15773)
GREECE
Mr. M. J. Broomhead
National Aerospace Laboratory
Dept. VH, P.O. Box 90502
1006BM Amsterdam
THE NETHERLANDS
1-1
Chapter 1
Introduction
1
1
2
3
4
5
6
TABLE OF CONTENTS
Table of Contents...................................................................................................................................................... 1-1
Preamble ................................................................................................................................................................... 1-1
Applications Of Mathematical Engine Models......................................................................................................... 1-1
Introduction to Types of Engine Models .................................................................................................................. 1-2
Computers, Software, and Recent Developments..................................................................................................... 1-4
A Survey of Model Users and Creators.................................................................................................................... 1-4
PREAMBLE
Throughout the years, computer models for the prediction of gas turbine performance and the simulation of the
operational characteristics have evolved into a very wide range of applications.
The present document is addressed to the operators and users in the field, for whom user-friendly, accurate and fast PCbased engine simulation tools are now available. These tools can help one to understand the engine performance
behavior and to identify the causes of possible deficiencies in engine performance in a highly cost-effective manner.
They have also become very useful for efficient mission analysis, the preliminary design studies of engines and their
matching to airframes.
Assuming that applications of performance models will be of initial interest to many readers, Chapter 2 provides a
detailed review of 22 examples of applications of models, ranging from preliminary design to in-service support, and
including educational examples. In the next chapter, the document describes briefly the features of several types of
complete engine models, including the complex and detailed models that are used for the actual design of a new turbine
engine and its components. These models are often of a proprietary nature. All modern performance models are
executed on computer platforms, which often dictate the form or construction of the model, as well as the method and
speed of execution. A chapter describing current platforms and software, and possible future developments is included.
Finally, several recent and advanced developments in mathematical modeling of components are described. The
Appendix contains the results of a survey of model users, and a separate Glossary is also provided.
Engine models are defined here as mathematical descriptions of the physical behavior of a turbine engine. These can
either be paper engines in their design phases or real engines in operation. Engine components, such as the compressor,
can also be separately modeled. During the design integration process, the component characteristics have then to be
matched, for instance with respect to mass flow and shaft speed.
Model Users
Model Creators
Requirements
Response
Requirements
Developers of models
and codes
Researchers
ity
Response
Ma
tur
Op
era
t
De
vel
op
me
nt
En
gin
e
Suppliers of codes
Requirements
tail
De
Manufacturers
Researchers
de
sig
n
Evaluation testers
Development testers
ion
Engine operators
Airframe designers
Response
1-2
bench analyses, where model complexity can still be avoided while providing usable answers. Examples are
performance sensitivity studies, mission analysis and the development of engine control systems.
For the operator, to support specific applications, such as engine diagnostics, the engine manufacturers usually supply
engine-specific codes. For preliminary engine and airframe design studies generic tools are available, also in the public
domain.
For engine maintenance, health monitoring, and diagnostics models the nominal condition of the specific engine is
described by a mathematical model involving the indications of the various sensors (pressures, temperatures, shaft
speeds) in the engine. Deviations from the nominal values indicate the state of health of the engine that can be used for
diagnostics and repair on condition.
A dedicated computer model is then a very cost-effective tool to carry out an appropriate diagnosis to identify degraded
or defective engine components. This application is current practice at major airline companies. Some have developed
their own maintenance-on-condition system with add-ons to the basic health monitoring system offered by the engine
manufacturer. The cost-effectiveness of the maintenance of military engine systems also increasingly benefits from
these developments.
The airframe designers may use generic engine models for mission analysis as part of an aircraft performance model in
the preliminary design phase. In that case engine and airframe sizing and sensitivity analyses are the key words. The
engine designer may also do this exercise to identify requirements for possible new engines.
Other uses of engine models by the airframe designer are:
To check and confirm projected engine performance data provided by the engine manufacturer while the engine is
still in the design and test phase;
To assess installation effects;
To assess engine performance.
Test bed engineers may also profit from a computer model that simulates their engine, and that can be run as a virtual
engine in parallel to the actual engine. Such a virtual engine can also be used as a part of the software for (real time)
aircraft flight simulators. For that application, modeling of the transient behavior (e.g. the response to a slamacceleration) is essential. This applies, in particular, to flight simulators for the latest generation of fighter aircraft with
high agility, in combination with close coupling of engine thrust and aerodynamics.
These real-time engine models are also useful for engine control design and development testing, including a hardware
control system (e.g. a fuel control system) coupled to a software engine.
Another group of users is the evaluation and development testers. The evaluation testers generally represent the interests
of future engine users, including the aircraft designers and manufacturers. The development testers are part of the engine
manufacturers organization. They use the engine models to identify unsatisfactory component performance (like the
health monitoring application mentioned earlier) and to tune the whole-engine performance model made available as a
performance card deck to the potential customers.
The engine manufacturers may use the relatively simple models in the early phase of the design only. In the detailed
design phase highly sophisticated models are used, combined with in-house experience. These models are mostly based
on computational fluid dynamics (CFD) describing detailed flow and combustion phenomena. In the present document
these models will be referred to mostly in a qualitative way.
Researchers also make use of engine and engine component models. In many cases experimental work and theoretical
modeling go hand in hand in that area. The research community may use this document to put their work into a wider
scope. This also holds for trainees and students.
Figure 2 shows the cycle simulations used throughout the engine life cycle. The model users identified in Figure 1 are
now associated with the life cycle elements of the engine and the related activities. In the lower left hand corner rather
simple cycle models, of the type suitable for preliminary engine and airframe studies, are used. In the upper part of
Figure 2 the most sophisticated design tools are employed, constituting the latest state of the art of the engine designers.
In the top right of the figure, the activities of the development and certification testers find their place. Mission and life
analysis and health monitoring are found in the lower right hand corner as activities of the engine operators
A simple way of engine modeling is found in the performance data as provided by the engine manufacturer. These
performance data are generally provided in the form of PC-based card decks from which performance charts or tables
can be extracted. The charts are mostly three-parameter charts, using corrected parameters. These corrections are for
engine inlet pressures and temperatures that are different from standard sea level conditions. An example is a chart
where the corrected net thrust is given as a function of the flight Mach number for a series of engine pressure ratio
values. Limits like the maximum turbine entry temperature or maximum shaft speed are included and form the
operational limits of the engine.
1-3
Simple thermodynamic relations can predict the performance of a generic gas turbine engine. Such a model may be
tuned to a specific engine by using values for the component efficiencies that depend on the operational condition of the
engine. On a test stand, engine component defects may show up when comparing measured pressures and temperatures
with those of the standard engine model. An example is a higher compressor exit temperature than expected for the
measured pressure ratio, indicating degraded compressor efficiency.
1-4
These and other areas of interest to the engine designer lead then to component modeling. The present document will
give attention to component modeling in the context of applications to whole engine performance modeling, and as a
separate subject in a later chapter.
All present-day gas turbine engine performance models are executed on computers, and the capabilities of modern
computers have extended the applications and the range of use of the models. For users wishing to adapt models to their
application, and to delve deeper into the present and future possibilities, a discussion of computers and software is
provided. Because of the electronic format of the document, it has been possible to provide executable examples of
some performance models. For those users wishing to model that which has not yet been modeled, and to examine the
frontiers of the component and engine modeling technology, recent progress in the modeling of both whole engines and
of the major sub-systems is described in the final chapter.
In the formative stages of the Working Group AVT-018 (formerly AGARD WG-29), it was decided to approach the
world community which would potentially have an interest in the type of publication that was emerging as the remit of
the Working Group. This was considered to be useful on three counts:
To identify current modeling practice;
To identify expectations and future needs for performance predictions;
To guide the efforts of the Working Group in their preparation of a worthwhile reference document.
Accordingly, a detailed survey was prepared, and distributed to users and creators of engine models. See Appendix 1.
Recipients included engine and aircraft manufacturers, academic institutions, airlines and supporting industries and
agencies. Information was sought in the following areas:
Scope of modeling activity in the particular organization including details of platforms, languages & i/o;
Model requirements for each application area;
Details on modeling technique;
Interfacing with other models and systems;
Model testing and validation.
The responses varied in detail, form and source (there was a disappointing response from the pure user community), but
were adequate to lead to the following conclusions:
The main modeling activity is firmly centered on aerothermal methods - i.e. modeling the physical processes present
in an engine. There is some use of database modeling techniques.
FORTRAN is still a common language, but there is a trend towards object-oriented and graphical approaches using
C++, Java etc.
The general trend is to move towards workstations and PC systems which offer graphical user interfaces.
Flexible and modular systems are required (engine models are required to run in conjunction with separate
subsystem models). Also, there is a requirement to model components and systems at varying levels of detail.
Increasing multidisciplinary interaction places requirements on efficient interfaces between related models and
codes. The formulation of, and adherence to, recommended practices and standards is therefore important.
The scope of a single methodology for applications is very wide.
There is great reliance on comprehensive and user-friendly program and model documentation.
Some guidance is required on selecting a model type for a particular application.
A full analysis of the survey is provided in Appendix 1. It may be helpful to read this with reference to the original
survey form, which is also provided in the appendix.
2-1
Chapter 2
Applications
1.
2.
3.
2-2
1. INTRODUCTION
As initially mentioned in Chapter 1 of this report, engine mathematical models may have many applications. There are
two groups that use engine models: engineers associated with industry and educators who use models for teaching
purposes. This chapter has been subdivided into two major sections that reflect the different uses and user groups. In
the section for simulations for industrial purposes, the section is organized from an engine life-cycle perspective.
Provided in each of the life cycle phases are synopses (summaries) synopses of published and open literature papers,
from a variety of authors, that provide examples of the different uses and applications of engine and component
simulations. In the section for educational purposes, the information is a compilation of engine models with a brief
description of the intended use and level of complexity of the simulation.
Preliminary Design;
Design and Verification;
Development and Validation;
Post Certification and In-Service-Support.
These phases represent all the aspects of a propulsion systems life; from mission need assessments through to the
eventual retirement from service. Throughout the propulsion systems life modeling and simulation are used to reduce
the time, resources, and risks of the acquisition process and to improve the quality of the systems being acquired and
sustained. The life-cycle phases represent a logical continuum of progress, and all programs tend to go through each of
the phases in order, provided they survive long enough to see each phase. A thorough consideration of any real engine
program must include the technical, managerial and contractual structure of the program.
Three major parties are typically involved in the development of any new propulsion system: the engine manufacturer,
the air vehicle manufacturer, and the user service. The relationships between the parties can vary considerably from
program-to-program, depending on the circumstances. For example, an engine development program for a new engine
type will probably be structured very differently from that for a derivative engine. Another program may involve an
eventual competition between air vehicles, and a competition between engines for an air vehicle. The relationships
between these parties tend to drive the exact location and the nature of the program decision milestones.
Decision milestones tend to be one-way gates, and while a program may backtrack through a gate, it is usually at great
expense. Models are a critical contribution to the milestones because they provide critical packets of information about
engine performance and physical characteristics, for exchange between all parties involved. For example, a cycle deck
might be such a packet exchanged at the completion of the Preliminary Design phase. At this point the air vehicle
manufacturer has baseline engine performance characteristics that will be integrated into the design of the aircraft. The
engine manufacturer uses this cycle deck as a baseline for overall engine performance that must be achieved. The engine
component designers will use this data in the next phase to help determine the eventual configuration of their
components. Models are used to understand the physics of the engine, as well as serve as the basis for configuration
control and establish agreements between the various parties involved in the life of the engine.
A very diverse group of model makers and users works across and within the life cycle phases. In general, the
manufacturers dominate the Preliminary Design through Development and Verification phases, and the using service
dominates the Post Certification or In-Service-Support phase. While the specific needs and expectations of each group of
users are different, there are common considerations. Models must be:
Credible;
User friendly;
Flexible;
Robust.
There is no single master model that does all for everyone. Instead, many different types of models are integrated into
the process of designing, manufacturing, and sustaining aircraft engines. Various types of performance related models
include performance and operability, aerothermal component, control system, and hardware-in-the-loop through manned
flight simulator models, and engine health monitoring and life usage models. The results of these models are integrated
with non-performance models such as: structural, fuel and thermal management, mechanical system and secondary
power system, electrical power system, manufacturing, and cost models.
The Preliminary Design phase is also known as the Concept Exploration phase. This phase typically consists of
competitive, short-term concept studies. The focus of these efforts is to define and evaluate the feasibility of alternative
concepts, and to provide a basis for assessing their relative merits (i.e., advantages and disadvantages, degree of risk,
etc.) at the next milestone decision-point. The analysis of alternatives is used as appropriate to facilitate comparisons of
2-3
alternative concepts. Because this is the time to explore diverse concepts, many iterations need to be run. This in turn
means that the turn around time per iteration needs to be relatively short, typically less than one day. The fidelity of the
models is balanced with the need to run many iterations quickly. The most promising system concepts are defined in
terms of initial, broad objectives for cost, schedule, performance, opportunities for tradeoffs, overall acquisition strategy,
and test and evaluation strategy. This phase may be the shortest, and is typically less than two years.
The Design and Verification phase is also known as the Program Definition and Risk Reduction phase. During this
phase, the program becomes better defined as one or more concepts, design approaches, and technologies are pursued.
Most major parameters become fixed, such as the number of stages of each aerothermal component, and the control
scheme is fixed. A wider range of models is run, including non-performance models and individual models now go into
more detail. Assessments of the advantages and disadvantages of alternative concepts are refined. Prototyping,
demonstrations, and early operational assessments are considered and included as necessary to reduce risk so that
technology, manufacturing and support risks are well in hand before the next decision point. Cost drivers, life cycle cost
estimates, cost-performance trades, interoperability, and acquisition strategy alternatives are key considerations. This
phase is longer than the Preliminary Design phase, and longer model turnaround times are tolerated in order to achieve
increased accuracy.
The Development and Verification phase is also known as Engineering and Manufacturing Development (EMD). The
primary objectives of this phase are to: translate the most promising design approach into a stable, interoperable,
producible, supportable, and cost-effective design; validate the manufacturing or production process; and demonstrate
system capabilities through testing. Low Rate Initial Production (LRIP) typically occurs while the Engineering and
Manufacturing Development phase is still continuing as test results, design fixes, and upgrades are incorporated. The
objective of LRIP is to produce the minimum quantity necessary to:
Provide production configured or representative articles for operational tests;
Establish an initial production base for the system;
Permit an orderly increase in the production rate for the system, sufficient to lead to full-rate production upon
successful completion of operational testing. The model diversity and turn-around times are probably highest in this
phase.
The Post Certification and In Service Support phase includes production, deployment, and operational support. The
objectives of this phase are to achieve an operational capability that satisfies the previously developed mission needs.
This will be the longest phase of all and with derivatives can typically span from 12 to 40 plus years. Deficiencies
encountered in developmental or operational testing should be resolved and fixes verified early in this phase. It is key to
consider that the potential for modifications to the deployed system continues during deployment and throughout
operational support of the propulsion system. Modeling and simulation plays a large role in the implementation of a
propulsion systems life management plan. For example, the Low Cycle Fatigue (LCF) life prediction approach typically
used by aircraft engine manufacturers can be described as consisting of seven tasks: materials characterization, stress
analysis, thermal analysis, missions analysis, life analysis and operating experience. Each of these tasks can utilize a
variety of modeling and simulation tools. Apart from all the above, modeling and simulation plays a significant role in
maintenance practices. This includes health monitoring on-board and in-flight, and diagnostic ground stations. A key
evolving area is prognostics and health monitoring.
Each phase of an engines life is discussed in terms of the use of modeling and simulation. Example synopses
(summaries) synopses with references are presented in each section to give the reader a feel for typical uses of modeling
and simulation in each phase. Within each synopsis there is a format that is generally followed: Each synopsis describes
a specific application, and addresses each of the following:
Modeling Technique;
Potential Benefits;
Cited Example(s);
Limitations of the Modeling Technique Chosen.
3. PRELIMINARY DESIGN
In the initial stages of an engine's gestation, a specification may not be available. Much of preliminary design activity is
concerned with looking at the potential market and working with commercial areas in identifying new opportunities and
customers. This involves keeping a close eye on changing market forces and military strategies, and of course, the
competitor's position. Requirements may be, initially, generic and may be understood in terms of the ability to
competitively fulfil a particular mission. As a concept emerges, the requirements will become more explicit in terms of
the usual constraints (e.g. cost, size, and mass, life, performance, growth capability, maintainability, emissions, stealth,
program risk). The detailed specifications are developed jointly with the potential customer, and should reflect not only
what the customer requires, but also what is technically achievable.
The customer is interested in whole-system performance, and so from the outset there must be an effort to place any new
engine in the context of an aircraft and a mission. Mission analysis involves the analysis of each phase of a mission to
which is associated a particular aircraft configuration and payload. Such analysis can yield fuel burn and flight times that
2-4
can form the basis for economic comparison between alternatives. It can also identify parts of the mission where certain
design parameters are critical e.g. specific fuel consumption (SFC), handling, thrust (for particular maneuver capability),
noise (for civil applications) etc. A mission analysis can also yield throttle movement profiles, which feed into
component life assessment, and the propulsion system Life Cycle Cost (LCC) activities. The interactions between engine
and airframe can significantly affect whole-system performance. An early understanding of such effects can save costs
in later development work.
The cycle selection process starts with consideration of the engine design point. Past experience is used to set a starting
point for an iterative process. Leading cycle parameters, flows, efficiencies & temperatures are chosen, and the
corresponding geometry generated by component areas. Component geometry is allied to component performance, but
overall layout and envelope considerations, together with component interactions will force compromises. The
component teams will redesign and refine assumptions, which will lead to refinement of the performance model so that
the next iteration starts with revised boundary conditions for each component. The aircraft role may also imply
conflicting requirements; the requirements for effective loiter and dash-to-intercept are fundamentally opposed in cycle
definition terms. Computer generation of meaningful exchange rates to trade-off design parameters to achieve a
workable compromise is a prime activity in the preliminary design phase.
Off-design performance modeling may initially use generic component characteristics, suitably scaled to meet the design
point requirements. As component synthesis and rig work refines these assumptions, the performance model is updated
and the iteration continues.
Certain decisions must be taken in the early stages on the inclusion (or exclusion) of some engine features. Performance
models can provide the basis for these decisions. Reference 1 cites the example of offsetting of mission improvements
against mass penalty of a convergent-divergent nozzle. The benefits included a geometric effect on the aircraft drag
characteristics, as well as a thermodynamic benefit of an optimally expanded nozzle at certain flight cases.
Cycle selection primarily addresses the steady-state performance of an engine. The implications on maneuvering
between operating points are wide ranging, impacting on compressor stability, combustion stability, mechanical and
thermal loading etc. which have to be controlled to achieve spec-compliant transient handling. Controls studies in the
preliminary design phase are discussed in detail elsewhere in this section.
3.1.
The conceptual design process of gas turbine engines is complex, involving many engineering disciplines.
Aerodynamics, thermodynamics, heat transfer, materials science, component design and structural analysis are a few of
the fields employed when selecting an appropriate engine configuration. Because of the complexity involved, it is
critical to have a process that reduces engine options without missing the optimum. Various steps are required,
including:
The engine design process is not linear since the steps are interdependent. A number of iterations are usually necessary
in selecting a final engine configuration.
The advent of the computer has made early examination of numerous propulsion characteristics possible. Illustrated in
Figure 1 is an estimate of when various computerized techniques became widely available. In the early years of
computer based analysis, engine selection was based primarily on cycle trade studies and the design engineers
experience. Elements such as installed performance, flow-path, and weight were delayed until the detailed design part of
the overall engine-development process. This could result in the selection of an engine configuration that was not fully
optimized. In the worst case, the selected engine could not satisfy the aircraft requirements, necessitating a costly and
time-consuming redesign. Today, many computerized tools are at the design engineers disposal for considering
component and engine design characteristics, weapon system tradeoffs, and most recently, life cycle cost.
2-5
POTENTIAL BENEFITS
Many different methods exist to integrate various design elements into an overall process. Ideally, designers like to
perform all design steps concurrently in order to minimize the overall time required to conduct a study. However,
several steps must be performed in series since the results of one must feed into the next. Installation and component
design analyses can be performed simultaneously, and the hope is that up-front costs (R&D and Acquisition) can be
integrated into the process at an earlier stage.
CITED EXAMPLES
1.
2.
Stricker, J. M., The Gas Turbine Engine Conceptual Design Process An Integrated Approach, Design Principles
and Methods for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999
Schaffler A., and W. Lauer, Design of a New Fighter Engine The Dream in an Engine Mans Life, Design
Principles and Methods for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999
2-6
Efficient or smart integrated modeling in early design may reduce the first two of the above items.
3.2.
A modern fighter engine has stringent requirements for performance, operability and durability. To meet these
conflicting requirements and to ensure a balanced design, simulations can be exercised during flow-path design, controlmode design and development testing. Use of a simulation helps to ensure problem prevention and reduces development
costs. Representative engine models, which accurately account for off-design and transient effects, can be used early in
the design phase for judicious configuration selection and control-mode design. Favorable component matching is
ensured before hardware fabrication and thus costly mistakes can be prevented. Special flow-path design considerations
in fast-response twin-spool-afterburning-turbofan engines can be analyzed. In addition to flow-path design, simulation
tradeoff studies can be used to optimize the control system to satisfy system requirements. Novel control modes can be
analytically evaluated across the operating spectrum and made practical with appropriate activation criteria that are
readily implemented in digital-control logic. Simulation applications during development and flight-testing include
calculation of hard-to-measure engine parameters using test-data driven transient-engine-models, thus facilitating design
verification.
2-7
Power-balance equation for each rotor with the rotor inertia term. (turbine power = compressor power + parasitic
power + acceleration power)
Continuity equation for each component with transient mass storage term.
An accounting for transient metal heat transfer of each component.
Assurance of static pressure balance at the mixing surface boundary of the duct and core flows.
A multidimensional Newton-Raphson iteration technique was used to simultaneously satisfy all the relationships to
achieve cycle balance at each instantaneous point. The component dynamics, including the moment-of-inertia of rotors
and the heat transfer characteristics used in the equations were obtained from the manufacturers design technology
groups. In addition to the above relationships, the model had to be modified to accurately model off-nominal variable
geometry effects, and the turbo-machinery performance and compressor stall line were also adjusted for the deviation of
transient clearance from that obtained during steady state operation.
POTENTIAL BENEFITS
Engine transient simulations which properly account for non-equilibrium conditions can provide system analysis and
tradeoff studies for judicious configuration selection and control-mode design, thereby effectively preventing operational
problems and reducing costs. System design and optimization can be particularly challenging for an afterburning
turbofan fighter engine with fast rotor response and augmentor transient requirements. Under transient conditions, the
engine components can operate in far off-design conditions. To prevent any aerodynamic matching problems and
adverse component interactions, engine configurations are thoroughly evaluated using transient simulations before
hardware fabrication commitment. This ensures timely identification of the required flow-path modifications and
compensatory control actions.
CITED EXAMPLE
3.
Khalid, S. J., Role of Dynamic Simulation in Fighter Engine Design and Development, Journal of Propulsion and
Power, Vol. 8, No. 1, January-February 1992, pp. 219-226.
One example in the cited reference deals with bypass duct and fan deceleration stall margin. The fast deceleration rates
of a fighter engine make the decrease in fan flow lag the decrease in compressor flow with a resulting increase in
transient bypass ratio. The bypass duct pressure loss is a non-linear function of the bypass duct entrance Mach number,
which increases with increasing bypass ratio. This non-linearity causes large increases in duct pressure loss during the
bypass ratio excursion of a deceleration.
There is approximately a one-to-one relationship between duct pressure loss increase and fan stall margin loss.
Illustrated in Figure 2 is the simulated deceleration fan operating lines, stall line, and steady state operating lines with
low duct loss representations evaluate on the F100-PW-229 engine model. It should be noted that the deceleration
operating line with the initially proposed high blockage duct rises significantly above the steady-state operating line
(13%) in spite of opening the exhaust nozzle area. With a high blockage duct, a bigger exhaust area during deceleration
further increases the duct Mach number, causing an increase in pressure loss. However, when the duct was streamlined
to increase the effective flow area, the simulation showed a deceleration operating line rising only 2% above the steady
state level. A low duct loss also increases the effectiveness of exhaust nozzle action in increasing fan stall margin. It
should be noted that the effect of duct pressure loss characteristic on the deceleration operating line is more pronounced
than on the steady-state operating line due to the higher duct-corrected-flow to fan-corrected-flow relationship during
deceleration.
2-8
3.3.
A particular engine type or cycle is selected on the basis of its ability to produce the required steady-state installed thrust
and SFC for its particular application. Clearly, other engine attributes must be considered, (mass, noise, price,
dimensions etc.) but the performance aspects are fundamental. Thrust requirements typically exist at a number of flight
points, and perhaps for a range of ambient temperatures, and at various ratings (e.g. maximum dry power, idle,
maximum afterburner). The agility requirements of an aircraft may place requirements on the transient times between
ratings. Retention of all aspects of performance over the life of the engine may also be specified.
It is possible to define an engine operating point in terms of any gas-path parameter, or combination of parameters. Each
combination may have different implications in terms of how the engine operation will be affected by external
influences such as customer bleed, power-extraction and inlet airflow distortion. Stator outlet temperature (SOT) is
traditionally used in early stages to define an operating level in view of its relevance to technology level. However, gas
temperatures such as SOT (apart from being difficult to measure) are not closely related to thrust, and so are not
necessarily ideal parameters upon which to base a control scheme. It is clearly important to consider, early in the
preliminary concept definition phase, how specification criteria can be achieved i.e. how the engine must be controlled
to achieve the functionality required by the customer.
REPRESENTATION
An engine model must model the physical processes to a degree, to allow the derivation of meaningful steady-state
sensitivities. Engine models used for controls investigations are not necessarily the same as those used for performance
studies, although there is a trend towards the wider use of cycle-match models for all functional design work. As the
cycle selection process is focused on the steady-state performance of the engine, dynamic modeling is not a prime
consideration. However, the main dynamics associated with an engine are generic, and once the methods are established,
they require engine-specific data such as shaft inertia and gas-path geometry. Therefore, a 0-D dynamic representation of
up to 30Hz is easily established. High-order models such as these are not necessarily required for initial control-loop
selection. However some candidate schemes, which appear viable under steady-state conditions, may become less so
when gas-dynamics are considered. This is especially true for variable-cycle engines where power level (or operating
point) is being dictated by variable geometry. In a conventional gas turbine, the dominant dynamic associated with
power level is shaft inertia; if the power level can be changed at a constant shaft speed, then other dynamic terms dictate
the design of the control loop. Gas dynamics are also relevant when considering the finer details of control-system
implementation e.g. actuator response.
2-9
Thermodynamic models are usually confined to the normal operating range of the engine. However, a model that is
capable of running down to zero speed (or steady windmilling speed) is needed for exploration of starting strategies.
Overspeed control is fundamental to engine integrity, and so a model should be capable of running to the conditions
arising from system failures. Similarly, control of the engine to recover or avoid compressor stall or surge requires the
post-stall behavior to be modeled. It could be argued that these aspects are not a prime part of control studies in the
preliminary design phase, and so just the normal operating range may be adequate. However, if one thinks in terms of
capability acquisition, generic full range models should be available so that control strategies (or part-strategies) are onthe-shelf ready for maturation on forthcoming projects.
All likely sensor stations must be modeled. This requires geometry assumptions for flow areas for static pressures.
Detailed gas temperature (2D) profiles are not required at this stage, but may become relevant as a design matures. The
modeling of the sensors themselves is not directly relevant to the choice of control-loops, their own dynamics may
influence the detailed design of a control-loop but ought not to impact on the overall control concept.
COMPATIBILITY
Controller design tools and methods are commonly based on linear methods. A linear representation of the sensitivities
across the dynamic frequency spectrum can be obtained through manipulation of the engine model using a process
known as linearization. This is explained in the cited reference, using a simple worked example based on the shaft and
volume dynamics of a single spool turbojet modeled using iterative methods.
VERSATILITY
A model should be able to generate steady-state points. This can be achieved by iteration (time fixed) or by stabilization
over time. Cycle-match models have an advantage in their ability to run to a specified level of an output quantity, not
just in steady-state mode but also in dynamic simulation mode. A typical thrust transient profile can be specified,
perhaps in conjunction with a fan working-line constraint. Consequently, the requisite control input profile is generated
(e.g. fuel and final nozzle area), which can be useful in determining the level of controller complexity required, or for
investigating the potential to simplify or relax control constraints during transients.
POTENTIAL BENEFITS
Early consideration of control-system issues can lead to a better final product at reduced cost. Product specifications can
be developed with potential (realistic) control schemes in mind; promises of unrealistic levels of performance may thus
be avoided. For example, the sensor set should be chosen alongside control-law definition activities. Late consideration
of control-laws may lead to a position with a sensor or a control-law incompatibility. Even if soluble, this can cost
money and program time. Exposure of engine handling qualities to potential pilots using real-time whole engine-system
simulation can provide early identification of control-system shortcomings, which can be rectified at more cheaply if
identified in the preliminary design phase.
CITED EXAMPLE
4.
Horobin, M., Cycle-match models used in functional engine design - an overview, Design Principles and Methods
for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999
2-10
Engine deterioration and engine to engine scatter also apply to inputs other than fuel flow, which suggests closed-loop
control of all inputs. The drive for simplicity, which may lead to lower cost and greater reliability, suggests a
combination of open and closed-loop implementations. However, with the increasing complexity of engines (e.g.
variable-cycle designs) and the potential rewards for tighter control of the engine operating point (e.g. life management,
fleet uniformity etc.), closed-loop control of all engine inputs is becoming more desirable.
Control-loop interaction is a hazard when multiple closed-loops are used. Ideally, each control-loop should be isolated.
That is, the input should only have an effect on its own feedback parameter. Realistically, this cannot be the case for a
gas-turbine engine where an increase in fuel flow has a direct effect on many parameters, some of which may be used as
feedback terms for variable geometry (say). There are mathematical techniques that can reduce interaction. However
(natural) interaction should be minimized by careful selection of control parameters. It is here that there may have to be
a compromise between ideal performance requirements, and the feasibility of implementing certain combinations of
control-loops in a robust fashion, with adequate stability margins.
The engine must be controlled within its safe operating range - critical parameters must be measured and demands
overridden if necessary. The control laws must also be designed to be able to accommodate possible system failures.
Control of a demanded transient could involve relaxation of some control-loops; open loop scheduling from an
appropriate engine parameter might control some inputs. Transient constraints (limits) are typically associated with
compressor and burner stability, although, depending on the specific case, appropriate shaping of an acceleration or
deceleration can have an impact on engine life.
It is trades such as these that can be explored in the early design stages, using suitable engine models to establish
exchange rates and sensitivities. An understanding of the engine thermodynamic cycle helps to identify a short-list of
candidate schemes for control. However there are black box techniques that can identify viable schemes.
3.4.
In gas turbine performance simulations the following question often arises: What is the best thermodynamic cycle
design point? This is an optimization task, which can be attacked in two ways. One can analyze a series of parameter
variations and pick the best solution, or one can employ numerical optimization algorithms that produce a single cycle
that fulfills all constraints. The conventional parameter study builds strongly on the engineering judgement and gives
useful information over a range of parameter selections. However, when values for more than a few variables have to be
determined within several constraints, numerical optimization routines can help to find the mathematical optimum faster
and more accurately.
The traditional way to select the thermodynamic cycle of a new gas turbine employs extensive parameter variations. For
a complex engine with many design variables this is a time consuming task. One looks for the optimum solution in a
certain respect. Instead of screening a wide range of potential solutions for the design variables with systematic
parameter variations it is also possible to do an automatic search for the optimum engine design with the help of
numerical optimization routines. A numerical optimization algorithm will only find the optimum of the mathematical
model, rather than the true optimum. If the result of an optimization run is an exotic cycle, it usually hints to a
deficiency in the model. In such cases, most probably a design constraint has been overlooked when defining the
problem. In addition, it is always of interest to know about the neighborhood of the optimum solution. From a
parametric study, limited to the region of interest it becomes obvious which design variables and constraints have the
biggest impact on the result. One of the advantages of numerical optimization is that the region where parameter studies
should be performed is significantly reduced.
2-11
is to be minimized. For a fighter engine it might be that the specific thrust should be maximized. One can also think of a
weighted combination of these parameters. When values for more than a few variables have to be determined while
several constraints exist, then numerical optimization routines can help to find the mathematical optimum, the minimum
or maximum of the figure of merit, faster and more accurately. For a case with only two variables it is easy to find an
optimum solution. If there are three variables the situation is not so clear. (When hill climbing in fog, it is difficult to see
whether the hill one is standing on is the highest in the neighborhood.)
With more than three variables, the picture may get obscure. In complex studies, the true optimum may never be found
with the conventional parametric study. There are many numerical optimization algorithms known from the literature.
They can be divided basically into the following two major groups:
Methods that use gradient information;
Others.
In the program GasTurb, there is one method from each group implemented. The gradient search algorithm implemented
in GasTurb follows the procedure illustrated in Figure 3.
Start 1
POTENTIAL BENEFITS
With a conventional parametric study, it is often very difficult to find the optimum solution for a problem as soon as four
or more design variables and several constraints are involved. With the help of numerical optimization algorithms, one
can easily find the mathematically correct solution to the problem. Extensive parametric studies around the solution will
help to understand why this combination of design variables is the best choice and how sensitive the figure of merit is to
small deviations form the optimum. The parametric variation is best suited for presenting the sensitivity of the results in
the neighborhood of the optimum cycle design point. Sometimes this leads to a redefinition of the figure of merit or the
constraints imposed on the solution. In rare cases an outstanding solution, which was overlooked while doing a
preliminary parametric study, may be found.
CITED EXAMPLE
5.
Kurzke, J. Gas Turbine Cycle Design Methodology: A Comparison of Parameter Variation With Numerical
Optimization, Journal of Engineering for Gas Turbine and Power, Vol. 121, January 1999, pp. 6-11.
A very common design task is to adapt an existing engine for a new application. It is quite obvious that in this case there
are more constraints than during the design of a completely new engine. The case study is an unmixed-flow turbofan
2-12
engine for a business jet. This type of engine has a rather low overall pressure ratio and a moderate burner exit
temperature when compared to the big turbofan engines used on commercial airliners. Besides the pressure ratios of the
new booster and the fan, among the design variables of the growth engine there will be the bypass ratio and the burner
exit temperature. A new low-pressure turbine will be required while the gas generator remains unchanged. The core
compressor of the new engine will not necessarily operate at the same operating point as in the basic engine. In fact, that
might even be impossible because doing that would require an increase in the mechanical spool speed beyond the limits
of the original design.
There are several constraints to be observed for the new engine design. The common core of the basic engine requires
that both high-pressure turbines have practically the same airflow capability. The Mach number at the core exit should
also be nearly the same, with the consequence that the flow capacity of the low-pressure turbine of both engines must
also be very similar. A further constraint is that the low-pressure turbine inlet temperature must be below 1150K to
allow for an uncooled low-pressure turbine, which can then be manufactured from inexpensive materials. Another
constraint may come from the nacelle in which the engine has to be installed. This will limit the fan diameter of the
growth engine. The figure-of-merit is the specific fuel consumption (SFC) for Max Climb rating and is to be
minimized. This will automatically result in low fuel consumption for cruise.
The optimum growth engine chosen for this cited example was influenced by three of the design constraints. The growth
engine has a fan diameter of 0.75 m, which conformed to the largest fan allowed in this exercise. The second constraint
that had an impact on the design of the growth engine was the compressor exit temperature, which was limited to 750-K
for the hot-day Take-Off case. The third constraint was the minimum high-pressure turbine flow capacity. All design
variables remained within the predefined range during the optimization. The thrust increase for Max-Climb rating at
altitude is 25% and at Take-Off even 29% as illustrated in Table 1 - Cycle parameter summary
Note that both engines run during Take-Off with 7% more mechanical high-pressure spool speed than at Max-Climb in
this example. The specific fuel consumption at altitude is nearly 5% better for the growth engine.
Basic Engine
Growth Engine
Max Climb
Hot Day Take Off
Max Climb
Hot Day Take Off
3.61
13.10
4.50
16.94
Thrust (kN)
19.65
14.23
18.93
13.33
SFC (g/(kNs)
4.5
4.65
5.06
5.23
Bypass Ratio
1.775
1.62
1.73
1.6
Fan (P13/P2)
0.781
0.886
0.726
0.839
Ideal jet Vel. Ratio
1.5
1.33
1.80
1.61
Booster (P24/P2)
12
11.36
12.3
11.67
HPC (P3/P25)
1350
1479
1393
1530
T4 (K)
1.35
1.35
1.31
1.31
W41 (Rstd)
4.98
4.96
5.01
5.00
W45 (Wstd)
610
708
649
750
T3 (K)
973
1076
1000
1108
T45 (K)
Table 1 - Cycle parameter summary
3.5.
REFERENCES
6.
Stricker, J. M., The Gas Turbine Engine Conceptual Design Process An Integrated Approach, Design Principles
and Methods for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999
7. Schaffler A., and W. Lauer, Design of a New Fighter Engine The Dream in an Engine Mans Life, Design
Principles and Methods for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999
8. Khalid, S. J., Role of Dynamic Simulation in Fighter Engine Design and Development, Journal of Propulsion and
Power, Vol. 8, No. 1, January-February 1992, pp. 219-226
9. Horobin, M., Cycle-match models used in functional engine design - an overview, Design Principles and Methods
for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999
10. Kurzke, J. Gas Turbine Cycle Design Methodology: A Comparison of Parameter Variation With Numerical
Optimization, Journal of Engineering for Gas Turbine and Power, Vol. 121, January 1999, pp. 6-11.
2-13
assessments are considered and included as necessary to reduce risk, so that technology, manufacturing and support risks
are well understood before the next decision point. Cost drivers, life cycle cost estimates, cost-performance trades,
interoperability, and acquisition strategy alternatives are considered.
The phases represent a logical continuum of progress, and all programs tend to go through each of the phases in order,
provided they survive to see each phase. The exact location of program decision milestones may be dependent on the
program contract structure. For example, an engine EMD contract may start after the completion of a Concept
Exploration contract. This means the aspects of a Design and Verification phase have been integrated into either one of
the contracts, or into both. The contract structure may vary depending on whether the customer has contracted directly
with the airframe manufacturer for everything, or whether there are separate airframe and engine development contracts.
Moving from the Preliminary Design phase to the Design and Verification phase represents the transition from the
consideration of many concepts to the commitment to refine and select from a few of the best. At this point models are
driven primarily by customer requirements. The tradeoffs being considered can be broad; and while they may focus on
hardware characteristics, they do not yet focus on a detailed hardware configuration. Issues such as overall cycle
performance, cost, manufacturability, supportability, and tolerable technical risk (e.g. blisk or Integrally Bladed Rotors
vs. bladed rotors) are important.
Throughout this phase, specific hardware configurations emerge. Tradeoff studies and risk assessments need to be
conducted. This is a very important time for the use of models, because the outcomes of the studies will refine the
hardware configuration pursued. The need to make major configuration changes in later phases may kill the program
because of the large cost and time required. This phase contains the majority of component testing that will be
conducted. Rig tests will typically be used to gauge a realization factor on various technologies scaled to the engine
size. The rig tests are used in conjunction with the models and to refine the models. Engine testing may also be
conducted during this phase.
The goal of this phase is to select, from the possibilities that have already been extensively studied, a single
configuration to pursue in detail in the Development and Validation phase. In order to select wisely, one must
understand the strengths and weaknesses of the proposed system, especially in terms of what the customer expects.
Customer involvement is essential during this phase because the tradeoffs are often complex and intertwined.
4.1.
When defining the cycle parameters and calculating the performance of a real engine, there are numerous practical
constraints to be taken into account. These fall into two main categories: the limitations of available component
technologies, and the operational considerations that are dependent on the aircraft application. In applying
thermodynamic principles to specific applications, the gas turbine designer must take a multitude of practical factors into
account in order to select the most appropriate cycle parameters. Most obviously, the available component technologies
impose aerodynamic, thermal and mechanical limits that set upper bounds on cycle pressure ratios and temperatures.
Gas turbine research and development (R&D) is being pursued vigorously in government laboratories, manufacturing
companies, universities and other research institutes. This research is pushing back technical barriers with little sign that
a plateau of technology is being reached. While the acceptable limits may move with time, there remain firm limits
which the designer must observe, his only freedom being the judgement of precisely where to set them for the envisaged
application at the time of the design freeze. The application itself is equally important. The type of aircraft military
combat, civil or military transport, helicopter, etc and the planned service life and mission operating requirements will
all have a major influence on the choice of cycle parameters.
POTENTIAL BENEFITS
There is a multiplicity of factors, both technical and economical, that the engine designer must take into account and
which will affect his mechanical constraints, such as shaft torque loading and overspeed limits, vibration avoidance and
damage containment. The economic issues include perceived market size and the possibility of alternative applications,
availability and cost of raw materials, processing methods and fabrication techniques, etc. When all such aspects are
taken into consideration, the engine cycle may deviate significantly from the optimum indicated by simple design
practices.
CITED EXAMPLE
11. Philpot, M. G., Practical Considerations in Designing the Engine Cycle, AGARD Lecture Series, Steady and
2-14
2-15
efficiencies along the engine operating line. Both show a characteristic catenary shape, with minimum SFC occurring at
around 70% thrust. This is convenient because the engine will normally be throttled back at least to around 80% thrust at
the start of steady, level cruise and will be gradually throttled further back as fuel is burned off and the aircraft becomes
lighter. The shape of the curve stems from the opposing behavior of propulsive efficiency and thermal efficiency. As the
engine is throttled back and the turbine exit temperature (TET) drops, propulsive efficiency increases and dominates, but
below about 70% thrust, the thermal efficiency effect becomes increasingly dominant.
Thermal efficiency can be defined as the energy delivered by the core divided by the energy supplied by fuel. The
thermal efficiency turns out to be a function of total pressure ratio and total temperature ratio as presented in Figure 6.
At the cycle conditions appropriate to a modern high bypass engine at altitude, thermal efficiency will increase with both
cycle pressure ratio and temperature, although a law of diminishing returns operates for both parameters.
Figure 4 - Typical thrust and SFC performance for high bypass transport engine
MILITARY COMBAT ENGINES
The basic principles for transport engines thermal, propulsive and transfer efficiency and the effects of cycle
parameters apply with equal force to combat engines. However, both operational needs and cycle selection criteria are
quite different. Combat aircraft are required to operate effectively and efficiently over a wide range of flight conditions
as illustrated in Figure 5. While the transport engine has only about three critical flight conditions to satisfy, the military
engine may have to meet 20 or more cardinal points. In addition, the predominant aim is almost always to achieve high
aircraft thrust-to-weight ratio, in the interests of speed, agility and weapons carrying capability. This means engine
thrust-to-weight ratio needs to be high, but more importantly, engine specific thrust is emphasized. High specific thrust
means small engine cross-section and hence reduced aircraft fuselage cross-section. Any growth in engine size has a
considerable effect on airframe size and weight.
2-16
2-17
maneuvering.
Engine thrust-to-weight ratio is one of the commonly quoted measures of technical progress for combat engines. The
engines in current operational fighters, which are mostly based on designs dating from the mid-1970s, generally have
thrust-to-weight ratios around 7. This is based on the nominal maximum SLS thrust on full augmentor.
The current classes of engines being developed for the new generation of fighters have thrust-to-weight ratios of around
10. Foreseen technology developments will bring a thrust-to-weight of 15 within reach in the foreseeable future and
advanced research and technology programs are set 20 as a longer-term goal. Achieving high thrust-to-weight has a
significant direct benefit in terms of aircraft size, but it also symbolizes gains in other ways. One of these the
progressive increase in attainable cycle temperatures and hence in specific thrust means that more thrust can be
generated from a given size of engine. At the same time, improvements in internal aerodynamics have enabled the job to
be done with less turbo-machinery and this also contributes to an improved thrust-to-weight ratio. However, with stage
numbers now becoming quite low, there is less to be gained by this means in the future. Further direct weight reductions
will be gained through the use of radically new materials like metal composites and non-metallics. A generalized plot,
encapsulating the published trends and assuming constant airframe technology, is shown in Figure 7. Raising the engine
thrust-to-weight from 10 to 20 will save around 15% of the aircraft weight for the same mission. This represents
appreciable savings and helps to explain why the thrust-to-weight ratio attracts so much attention.
4.2.
HARDWARE IN-THE-LOOP
Real-time models (RTM) are models where the outputs of a transient performance computer program are generated at a
rate commensurate with the response of the physical system it represents. RTMs are required for a range of applications.
This synopsis attempts to summarize the real-time scene as described in detail in the reference document (AIR4548).
The following applications are summarized, and their requirements are examined in terms of 4 model attributes:
2-18
generic nature or of a pre-production standard. Either way, early real-time HITL testing can help reduce the risk of new
control technologies without placing an engine at risk. The various bits of hardware are connected to various slave load
generators (mechanical loads, pressures and temperatures) such that the operating environment of each component is
reproduced. The ability to rapidly re-configure a rig (e.g. to replace a simulated component with a real one) is desirable.
There is always a trade-off between model accuracy (consistency) and execution rate. The latter is also linked to
bandwidth, which for this application must encompass all the transient events that the controller can sense and control,
and is driven by its own operating cycle rate.
2-19
emergency procedures, in which case the engine performance representation may be secondary. The important
functionality is in the aircraft system, and so the engine model must provide the appropriate inputs into these system
models. In some cases these subsystems include real control boxes which will reject any unrealistic inputs. This may
place extra consistency requirements on the engine model. The need to model the simultaneous operation of up to four
engines may present execution time challenges!
CITED EXAMPLE
12. Aerospace Information Report : AIR 4548: Real-Time Modeling Methods for Gas Turbine Engine Performance,
Prepared by SAE S-15 committee 1995
This document states in its foreword: Current practices vary greatly in terminology and methods depending upon
application. The document is intended to provide a vehicle for presentation of model types and definitions to be used as
a basis for communication between customer and supplier. It is also intended to complement Aerospace Standard (AS)
681 - Gas Turbine Engine Steady-State and Transient Performance Presentation for Digital Computer Programs, and
Aerospace Recommended Practice (ARP) 4148 - Gas Turbine Engine Real Time Performance Model Presentation for
Digital Computers.
2-20
There are some certification and safety issues arising from the real-time use of engine models in engine controllers. For
example, reliance may be placed on the model to cash operating margins up to a critical point, or to indicate that
deterioration has reached a critical level. Either way - any inaccuracy in the model can impact on the overall safety
assessment. This is not so much a comment on the modeling technology, but on the application of engine models in
general.
4.3.
AIRCRAFT SIMULATION
A representative engine model was required to run in real-time as part of an aircraft simulation. The aircraft, a modified
F-18, featured thrust vectoring and flew as a research aircraft under the NASA HARV (High Alpha Research Vehicle)
program. This program was aimed at investigating the high angle-of-attack flight regime. The thrust vectoring enabled
sustained flight at angles of attack (AOA) of up to 70. 3 vanes per engine located in place of the divergent nozzle petals
provided the multi-axis (axisymmetric) thrust vectoring.
The model was used for the development of control-laws associated with the extended flight regime and for real-time
simulation (man and hardware in-the-loop testing) for evaluation of the modified control-system. The engine modeling
is very simple but was considered adequate for its purpose. The model is a top-level representation of the thrust effect
only. Components are not modeled. A few internal engine parameters are derived for use in the aircraft drag
calculations.
An attempt had been made to make use of the full, iterative, component-level model supplied by the engine
manufacturer (GE) but this more detailed model was found not to execute within the required timing criteria on the
target processor. It required 4ms per 20ms, which was deemed unacceptable. The engine model was required to execute
well within the flight control-system loop-execution rate. This became the primary drive towards a simpler model. The
accuracy constraints on an alternative model were that it should perform within 5% of the steady state, and 25% of the
transient response of the complete non-linear component-level dynamic model. The engine model had also to represent
the effects of thrust vectoring in terms of the increased drag and loss of axial thrust.
The simple mode also included tables of the following, which were looked up with the parameters generated by the
engine model section:
Inlet spillage drag (Dinl);
Nozzle aft-end drag increment (Dnoz).
The dynamic response of the engine was represented in the simple model by applying a variable response to the applied
PLA shaping model. PLA input was converted to PLA' using a transfer function and rate-limiter, and tuned using the full
model operating at the four corners of the flight envelope. PLA' was then applied to the look-up tables to generate
dynamic behavior. Note that PLA is used as an input to the model. This means that the engine control-system is also
being characterized. The time-constants associated with a control-system are likely to be insignificant in terms of gross
(long-range) engine thrust response.
2-21
Figure 8 - Incorporation of the simple dynamic engine mode into the thrust vectoring simulation
CITED EXAMPLE
13. Johnson, S. A. A Simple Dynamic Engine Model for Use in a Real-Time Aircraft Simulation with Thrust
Vectoring, NASA Technical Memorandum 4240, AIAA Paper # 90-2166.
The overall model structure is shown in Figure 8. On examination of the response of the full model to changes in PLA,
the judgement was made that the engines dynamic response was approximately 1storder on the parameters of interest.
Additionally the rate of response of the engine in afterburner mode was faster than in dry mode. Modulation in each case
was by a different mechanism - spool speed change in dry mode, and nozzle area modulation in afterburner mode. Also,
deceleration was faster than acceleration. (An engine might naturally exhibit this, but the response of the engine model
to power-lever represents a controlled intent.)
The approach described above was validated against the full model by applying similar PLA time histories, which
cycled through the power range at various flight conditions typical for an aircraft with thrust vectoring. There is
reasonable agreement between FG traces but the A8 and NPR terms do not line up so well. This is due to the complex
control laws used to control the nozzle (especially on reheat light up and shutdown).
Figure 9 and Figure 10 below show a time-based comparison between the full and simple models for the following PLA
input profile: Flight Idle to Military Power to Maximum Afterburner back to Minimum Afterburner back to Flight Idle
2-22
Figure 9 - Comparison of net propulsive force at 35,000 ft, Mach 0.2, no vectoring
The simple model took 0.3ms per 20ms to execute and was deemed acceptable for inclusion in the whole vehicle
simulation. The reduction was achieved through no iteration, a reduction in code (FORTRAN in both cases) to 25% of
the full model, and a similar reduction in storage memory requirement. The accuracy was inside the specification, with
the steady-state being within 3%, and the transient agreeing within 20-25% of the full model.
Figure 10 - Comparison of nozzle pressure ratio at 35,000 ft., Mach 0.2, no vectoring
In situations where detailed dynamics are required, this method may not be most appropriate, and reversion to
component-level representation may be essential. In these cases, measures must be taken to reduce execution time either by reducing program size or extent of iteration, or by specifying a faster target processor.
2-23
4.4.
Gas turbine engine components such as compressors, combustors, and turbines are usually tested in rigs prior to
installation into an engine. In the engine, the component behavior is different due to a variety of reasons. The installation
effects are caused by small geometric differences due to non-representative rig operating temperatures and pressures, by
different gas properties and Reynolds numbers, and by radial as well as circumferential temperature and pressure
profiles at the inlet to the component. For highly accurate performance predictions, these rig-to-engine effects have to be
taken into account.
Traditionally, the term, installation has been also used for describing all the differences in engine operation and
behavior between testbed and aircraft. Intake and afterbody drag, power offtake and bleed, intake pressure losses and
inlet flow distortion all have significant impact on airflow, thrust, specific fuel consumption and compressor stability.
Using modern performance synthesis programs all these effects can be simulated reasonably well.
POTENTIAL BENEFITS
Engine manufacturers are mainly interested in the operating conditions of the engine components. During engine
development there needs to be a mechanism to determine which component is responsible for performance shortfalls and
how the components are matched. To predict the effects of a variety of installation effects, computer modeling will be
used throughout an engine development program. Aircraft manufacturers are interested in the aircraft performance in
terms of achievable turn rates, specific excess power, and mission fuel consumption. For that purpose they need
computer programs which calculate installed engine performance.
CITED EXAMPLE
14. Kurzke, J., Calculation of Installation Effects Within Performance Computer Programs, AGARD Lecture Series,
Steady and Transient Performance Prediction of Gas Turbine Engines, AGARD-LS-183, May 1992.
For accurate performance synthesis one needs to make corrections to the component characteristics for any difference
between the in-engine operation conditions and the conditions for which the characteristic originally was set up. The
cited reference gives a complete summary of all rig-to-engine effects. The author shows how introducing those effects
into the performance model should minimize the need for adjustments to component performance and thus provide the
best performance synthesis model of an engine. The author discusses installation effects from a rig-to-engine component
perspective then the integration of the components into the simulation where effects that cannot be attributed to a single
component are described.
Under rig-to-engine differences there are differences due to geometric component differences. The author describes
models for tip clearances, blade untwist, thermal expansion, bleed offtake, and cooling air injection. The effect of tip
clearance on compressor characteristics as an example of a rig-to-engine difference is presented in Figure 11.
In the second part of the cited reference, the author takes a look at those effects that cannot be attributed to a single
component. Areas of concern are thrust-drag bookkeeping, power offtake, bleed effects, intake losses, flow distribution,
and inlet flow distortion on stability and performance.
As an example of these types of effects, lets look at the effect of power offtake and bleed air. It is no problem at all to
simulate the effect of power offtake within a performance synthesis program. One should be aware that taking off a
fixed power can have quite different effects on the cycle. At high altitude flight conditions such as 50000 ft and Mach
Number of 0.7, the power offtake is only 14% of the power available at sea level static conditions. Such a power offtake
has a severe effect on the cycle: the operating point in the high-pressure compressor is moving towards a lower spool
speed and in direction of the surge line, see Figure 12.While at sea level static conditions, the power offtake is limited by
mechanical restrictions within the gearbox, at high altitude it is limited by the compressor stability limit. For calculations
at high altitude the power consumption of the engine gearbox and the attached accessories need to be modeled in a better
fashion than just with a constant mechanical efficiency of the high pressure spool.
2-24
Figure 12 - Power offtake and bleed effects on HPC operating points at sea level and at altitude
On a modern civil engine it can be necessary to have two customer bleed air offtakes which are switchable. At idle
compressor exit bleed is used, and at max climb rating an inter-stage bleed port is used. Bleed air conditions at aircraft
and engine interfaces have to be calculated with sufficient care. The local pressure where the air is taken off is
dependent on the specific design; it is seldom the total pressure. The pressure losses within the bleed manifold and pipe
system are very much mass flow dependent.
2-25
4.5.
STATISTICAL ANALYSIS
Gas turbine performance models are normally used in a straightforward manner. Unambiguous direct questions about
the operating conditions, intended engine modifications, modified control schedules etc. yield equally unambiguous
answers. However, the data from a single engine cycle are sometimes not sufficient because the statistical distribution of
a parameter is also needed.
When some input data for a cycle performance model are statistically distributed, the calculated parameters will also be
statistically distributed. The correlation between the standard deviation of the input parameters with those of the results
can be calculated with the widely used root-sum-squared approach, which uses influence factors for the correlation
between the input and the result. This procedure implies that the influence factors are all statistically independent from
each other. In practice this is not always the case, especially when control system interactions must be taken into
account.
An alternative to the root-sum-squared approach is the Monte Carlo method. This method works such that random
numbers with prescribed statistical distributions are generated and fed into a mathematical model of arbitrary
complexity. The statistical distribution of the calculated parameters is then analyzed, and the statistical outputs form the
result of the Monte Carlo simulation.
POTENTIAL BENEFITS
The Monte Carlo method can handle simulation models of arbitrary complexity. It will automatically take into account
all interactions between the input parameters and the calculated results correctly. Contrarily, the root-sum-squared
approach assumes that all influence factors are independent of each other, and this is not always true. For example, the
control system interacts in a very complex manner with the individual engines of a series.
CITED EXAMPLE
15.Kurzke, J Some Applications of the Monte Carlo Method to Gas Turbine Performance Simulations, ASME Paper
97-GT-48, Presented at the ASME International Gas Turbine Institutes Turbo Expo, Orlando, FL, June 1997
Engine tests are performed to evaluate the overall characteristics in terms of thrust and specific fuel consumption.
However, during the development phase the main purpose of performance testing is to find the efficiency of the engine
components. The analysis result has a tolerance, which is affected by both random and systematic measurement errors.
This tolerance can be found with the help of the Monte Carlo method as will be shown for the example of an unmixed
flow turbofan.
RANDOM ERRORS
When a measurement is repeated several times, the instrument readings will not agree exactly but will show some
scatter. In gas turbine tests this scatter is caused not only by random effects in the measurement chain, but also by small
changes in engine geometry and operating conditions. A running engine is never absolutely stable because of small
changes in inlet flow conditions, variable geometry settings, thermal expansion of casings and disks etc. The unintended
(and undetectable) changes to the engine during re-assembly for a back-to-back test can also be regarded as random
errors of an experiment.
Obviously the magnitudes of these errors have an effect on the confidence interval for the test analysis result.
Compressor efficiency, for example, is calculated from the four measurements of inlet and exit total temperature and
pressure. How should the individual measurement precision errors be propagated to an efficiency confidence interval?
Abernethy describes some alternative solutions to the problem, and reports that the recommended method has been
2-26
checked with Monte Carlo simulations. This recommendation implies that a Monte Carlo simulation is superior to other
methods.
SYSTEMATIC ERRORS
In a carefully controlled engine performance test the random errors mentioned above are not negligible, but are smaller
than the systematic errors caused, for example, by incorrect positioning of the probes. There is seldom space in an
engine to put enough pressure and temperature pickups at the component interface plane. Although every effort is made
to correct the measurements for all known effects, an uncertainty remains.
The difference between the measurement (after applying all known corrections) and the true mean value is called a bias.
There is no data available to calculate the magnitude of the bias and therefore it is usually estimated from experience
with component rigs. For example, Abernethy discusses the problem in some detail and concludes that bias limits should
be root-sum-squared when the confidence interval must be estimated for a quantity that is calculated from several
individual measurements.
Using the root-sum-squared approach for estimating confidence intervals implies that the bias of every element to be
combined in the sum has a normal distribution. Both normal and non-normal bias errors can be simulated with the
Monte Carlo method.
ENGINE SIMULATION
During an engine performance test all temperatures and pressures on the cold side of the engine (stations 2, 13, 25 and 3,
see Figure 13) are measured. Furthermore, fan mass flow and fuel flow, and inlet and exit total pressures of the lowpressure turbine are normally available for test analysis. Temperatures that are eventually -measured around the lowpressure turbine are not used in the analysis because the severe gradients, both circumferentially and radially, make the
mean value rather inaccurate. For the cycle program one needs as additional input data, the internal air system and
mechanical losses, which must be estimated or analyzed separately from detailed measurements.
21
25 13
16 18
44 45
41
3 31 4
5 6
HP leak to LPT
b
c
handling
a
LPT
leakage from
a HP leakage to
b NGV
c HPT
overboard
Figure 13 - Nomenclature
From the pressures and temperatures on the cold side of the engine all compressor efficiencies can be derived. For the
analysis of the turbine efficiencies one needs to know the turbine shaft power. The power balance with the compressors
yields that information when the bypass ratio is known.
CORE FLOW ANALYSIS
In the test analysis process one can find the bypass ratio in several ways by iteration. For example, one can calculate the
bypass ratio in such a way that the continuity at the bypass nozzle exit is fulfilled. Such an analysis method needs the
exact dimension of the bypass nozzle area A18 and the nozzle discharge coefficient.
Often the high-pressure turbine nozzle-guide-vane throat area is known more precisely than the effective bypass nozzle
area and is therefore used as a basis for the core flow analysis. However, this method requires a good knowledge of the
secondary air system, because the amount of air that bypasses the turbine throat influences the result found for the
bypass ratio.
2-27
As input data for the Monte Carlo simulation the standard deviations of the individual measurements is needed. Those
data complemented by assumptions for the engine internal measurements are used here for the simulation of tests on a
sea-level testbed (Table 1).
Parameter
Inlet Pressure
Inlet Temperature
Fan Exit Pressure (Core)
Fan Exit Pressure (Core)
Fan Exit Temperature (Core)
Fan Exit Pressure (Bypass)
Fan Exit Temperature (Bypass)
HP Compressor Exit Pressure
HP Compressor Exit Temperature
Fan Mass Flow
Fuel Flow
HP Turbine Exit Pressure
LP Turbine Exit Pressure
Symbol
P2
T2
P21
P21
T21
P13
T13
P3
T3
W2
WF
P45
P5
2-28
Component
Bypass Ratio 4
Bypass Ratio 7
Fan
0.80%
1.40%
Booster
0.82%
0.82%
0.50%
0.50%
0.52%
0.52%
0.82%
1.19%
4.6.
Turbomachinery aerodynamicists have long realized that the flow within multistage turbomachinery is complex. In
addition, to the non-deterministic, small-scale chaotic unsteadiness due to turbulence, at a large scale the flow is also
unsteady and aperiodic from blade passage to blade passage. The flow features associated with these large scales are
deterministic. It is because of the unsteady deterministic flow that turbomachinery is able to either impart or extract
energy from a flow. The challenge is to develop an analytical model of sufficient fidelity to address key design issues.
At the same time, the cost and time of executing a simulation based on the model, and the cost of acquiring and
maintaining the empirical database that underpins the model must be compatible with the design environment.
2-29
the average-passage flow field. The effect of the unsteady deterministic flow field on aerodynamic matching of stages is
accounted for by velocity correlation within the momentum equations associated with the average-passage flow field and
by a velocity total enthalpy correlation within the energy equation associated with the average-passage flow field. The
term unsteady deterministic refers to all time-dependent behavior that is linked to shaft rotational speed. All unsteady
behavior not linked to shaft rotational speed is referred to as non-deterministic.
POTENTIAL BENEFITS
Up until recently (the late 90s), the aerodynamic design of most axial flow multistage turbomachinery has been
executed using various axisymmetric flow models. The use of these models is iterative. A design has been executed
based on the existing database. Data obtained from tests of the fabricated hardware are used to update empirical
correlations embedded within these models. The updated axisymmetric flow model is then used to support the next
design of the configuration. This bootstrap approach to the aerodynamic design of axial flow multistage turbomachinery
results in some truly impressive machines, as evidenced by the aerodynamic performance they achieve. However,
because of strong economic forces, the turbomachinery industry has been forced to re-examine this iterative approach to
aerodynamic design. These economic forces mandated that industry reduce the time and cost of developing a
turbomachinery component and, at the same time, required the design of machinery whose performance goals lay
outside the then-existing experience base. As a result, a strong need to develop a new methodology, for executing the
aerodynamic design of axial flow multistage turbomachinery, has arisen. The foundation of this new methodology
requires aerodynamic models whose resolution is greater than that of the axisymmetric flow models. These models have
to allow for overnight turnaround using computer resources compatible with the design environment. Finally, the model
has to be capable of addressing the aerodynamic issues associated with the design of advanced configurations.
CITED EXAMPLE
16. Adamczyk, J. J., Aerodynamic Analysis of Multistage Turbomachinery Flows in Support of Aerodynamic
Design, 1999 International Gas Turbine Institute Scholar Lecture, Transactions of the ASME, Journal of
Turbomachinery, Vol. 122, April 2000 pp. 189-217.
Figure 15 shows the measured pressure rise characteristic along with simulation results at a flow coefficient of 0.395 and
0.375 for the Lewis low-speed axial compressor (LSAC). LSAC is a four-stage machine with an inlet IGV, which is
representative of the rear stages of a high-pressure (HP) compressor. The compressor is of a modern design employing
hub-shrouded stators with end-bends. The four stages are geometrically identical. The simulation accounted for the rotor
tip clearance. The simulation did not include the stator hub cavities nor did the simulations account for stator hub
leakage.
2-30
good. For the flow coefficient of 0.395, Figure 17 shows plots of the total and static pressure coefficient, the axial and
absolute tangential velocity, and the absolute and relative flow angle as a function of span for the simulation and the
experiment. The plots are for an axial location behind the second stator. Once again, the agreement between the
simulation results and the data is good. The slight difference between the static pressure coefficient derived from the
simulation and that measured inboard at 40 percent span is unknown.
2-31
4.7.
REFERENCES
17. Philpot, M. G., Practical Considerations in Designing the Engine Cycle, AGARD Lecture Series, Steady and
Transient Performance Prediction of Gas Turbine Engines, AGARD-LS-183, May 1992.
18. Aerospace Information Report: AIR 4548: Real-Time Modeling Methods for Gas Turbine Engine Performance,
Prepared by SAE S-15 committee 1995
19. Johnson, S. A. A Simple Dynamic Engine Model for Use in a Real-Time Aircraft Simulation with Thrust
Vectoring, NASA Technical Memorandum 4240, AIAA Paper # 90-2166.
20. Kurzke, J., Calculation of Installation Effects Within Performance computer Programs, AGARD Lecture Series,
Steady and Transient Performance Prediction of Gas Turbine Engines, AGARD-LS-183, May 1992.
21. Kurzke, J Some Applications of the Monte Carlo Method to Gas Turbine Performance Simulations, ASME Paper
97-GT-48, Presented at the ASME International Gas Turbine Institutes Turbo Expo, Orlando, FL, June 1997
22. Adamczyk, J. J., Aerodynamic Analysis of Multistage Turbomachinery Flows in Support of Aerodynamic
Design, 1999 International Gas Turbine Institute Scholar Lecture, Transactions of the ASME, Journal of
Turbomachinery, Vol. 122, April 2000 pp. 189-217.
2-32
Through the use of modeling and simulation technology, coupled with the baseline information provided by current
wind tunnel and test cell test procedures, a fusion of computational and experimental data can be accomplished. This
will make more information available to the design engineer for system development and risk reduction. Such an
approach is key to the successful understanding of high performance engine phenomena such as High Cycle Fatigue
(HCF), and inlet and integration issues associated with tomorrows highly integrated flight systems.
The use of modeling and simulation in conjunction with test information results provides the following capabilities to
the T&E community:
Enhanced aircraft inlet and turbine engine capability for integrated test and evaluation on military weapon system
tests.
Improved ground test operational efficiency by providing accurate pre-test predictions and post-test analysis to
support pre-and post-test decisions and optimize test matrix.
Enables correlation of isolated inlet and engine simulated ground test results to provide a preflight release assessment
of integrated airframe and engine performance in a dynamic flight test environment.
Provides affordable or reduced test costs through improved planning, optimizing test matrix, and smarter test
analyses and decisions.
Improves knowledge-based IT&E applications to provide integrated aircraft in-flight performance assessments.
5.1.
PERFORMANCE
One of the first requirements for an engine program is to predict the performance of the engine over the full range of
conditions from take-off to a variety of flight regimes such as low altitude penetration, high altitude cruise and combat
maneuvering with maximum afterburning. This requires sophisticated modeling which can be done with a high level of
confidence before the engine has run; both design point and off-design-point performance must be accurately predicted.
This must be done before any component testing has been carried out and the component performance changes between
test rigs and actual engine operation will be discussed.
2-33
POTENTIAL BENEFITS
The two programs described above provide an ease of use to the user that makes any analysis task easier. These
programs are ideally suited for basic studies and for getting a fundamental understanding of the principles underlying the
design and operation of gas turbines. They are suited to both professional applications in the gas turbine industry and to
the student in an educational setting. The flexibility built within the codes will provide ease of adaptation to future
applications such as performance analysis of complex recuperated intercooled cycles, multi-stage combustion, detailed
simulation of STOVL propulsion systems and tilt-rotor propulsion system simulation.
CITED EXAMPLE
23. Kurzke, J., Advanced User-Friendly Gas Turbine Performance Calculations on a Personal Computer, ASME
Paper # 95-GT-147, June 1995.
24. Visser, W. P. J., and M. J. Broomhead, GSP, A Generic Object-Oriented Gas Turbine Simulation Environment,
ASME Paper #2000-GT-0002, May 2000.
Using the simulation by Kurzke, steady state performance can be obtained for a variety of configurations. One such
configuration cited in the reference was a turbojet and the matching of components. The calculation of each off-design
point requires iteration. Several input variables for the thermodynamic cycle must be estimated. The number of variables
is calculated as follows. The result of each pass through the cycle calculation is a set of errors. Inconsistencies are
introduced through the use of imperfect estimates for the variables. The number of errors equals the number of variables.
A Newton Raphson Iteration scheme is used to drive the errors to zero. One way that off-design information can be
presented is by plotting lines of constant mechanical speed on an altitude-Mach number flight envelope as illustrated in
Figure 18.
Figure 18 - Lines of constant spool speed overlaid on the altitude Mach number envelope
A relatively simple example of using GSP is the analysis of off-design performance of a typical high-bypass turbofan
engine. The GSP demo model, BIGFAN is depicted in Figure 19. Generic component maps were used and then scaled to
the BIGFAN design point. The code can calculate sea level take-off performance at varying ambient temperatures and
compressor bleed flows. This was accomplished by the calculation of steady-state points at a series of different ambient
temperatures, with the engine running at either maximum total turbine inlet temperature, Tt4 (or TIT) or maximum
burner pressure Ps3 (i.e. a flat rated engine).
A flat rated temperature (FRT) was assumed, in which the engine is at both maximum Tt4 and maximum Ps3 for an inlet
temperature of 288K. The procedure calls for the calculation of a design point, with a Tt4 of 155K. An ambient
temperature parameter sweep from 280K up to 320-K was then performed, while maintaining the Tt4 at 1554K. The
parameter sweep was performed both for no-bleed and for 5% compressor bleed. The results in Figure 20 show the
typical turbofan trends for fan speed N1, compressor speed N2, engine pressure ratio EPR, and net thrust FN, used for
specifying take-off performance at temperatures above FRT.
2-34
2-35
COMPRESSOR
BURNER
BURNER
EFFICIENCY
PRESSURE
RATIO
CORRECTED
SPEED
BURNER PRESSURE
CORRECTED AIRFLOW
TEMPERATURE RISE
CONTROL
NOZZLE
AUGMENTOR
TURBINE
FLOW
FUNCTION
CORRECTED
SPEED
FLOW
COEFF.
AUGMENTOR
EFFICIENCY
AREA
RATIO
AUGMENTOR
PRESSURE
PRESSURE RATIO
POTENTIAL BENEFITS
ATEST provides the means to quantitatively understand the following, during the engine start process:
ATEST can be used for arbitrary engine configurations and can simulate turbine engine operation continuously from
startup to shutdown. The approach expands widely accepted component-matching principles to simulate sub-idle,
windmill, and engine-starting operations and preserves the existing one-dimensional steady state and transient
capabilities and simulation accuracy for above-idle operations.
ATEST was validated using test data for:
The processes required to characterize gas turbine engine starting, and to predict the boundary between starter-assisted
and windmill starts, prescribed by the engine manufacturer's performance specification, were successfully simulated.
CITED EXAMPLE
25. Chappell, M. A. and P. W. McLaughlin, Approach to Modeling Continuous Turbine Engine Operation from
Startup to Shutdown, Journal of Propulsion and Power, Vol. 9, Number 3, May-June 1994, Pages 466-471.
A generalized approach, to modeling gas turbine engine operation continuously from startup to shutdown, for arbitrary
engine configurations, is described in this article. The approach preserves existing steady state and transient capabilities,
and simulation accuracy for above-idle operations. This article will focus on the approach used in applying conventional
component-matching simulation principles to the starting process. The approach is applied to a military flight-type twospool after-burning turbofan engine, and simulation results are compared to engine test data.
The ATEST model was applied to a military-type two-spool after-burning turbofan engine and included a simulation of
the engine control. Model results were obtained by prescribing throttle position as a function of time, and prescribing
flight conditions in terms of altitude and Mach number. Model results were compared to engine test data for steady state
windmill operation and windmill, starter-assisted, and spool-down starts.
The ATEST model results were compared to engine test data to evaluate the ability of the model to simulate enginestarting phenomena as illustrated in Figure 22. The more important aspect of the evaluation was the ability of the model
2-36
to reproduce the general shape of time-varying performance, including the existence of inflections, breakpoints, and
overshoots. The comparison of absolute levels of performance was evaluated, but was less important in demonstrating
the capability of simulating fundamental engine start processes.
Model
Test
Max Power
100
200
Idle
Engine Cutoff
Compressor
Speed
Percent
Burner
Pressure
Psia
Compressor
Speed
Idle
Throttle
Advance
Ignition
Burner
Pressure
0
0
10
20
30
40
50
60
70
80
90
0
100
Time, Sec
2-37
POTENTIAL BENEFITS
Hundreds of individual sensors produce an enormous amount of data during developmental turbine engine testing. The
challenge is to ensure the validity of the data and to identify data and engine anomalies in a timely manner. An
automated data validation, engine condition monitoring, and fault identification process that emulates typical
engineering techniques has been developed for developmental engine testing. The result is an ability to detect data and
engine anomalies in real-time during developmental engine testing. The approach is shown to be successful in detecting
and identifying sensor anomalies as they occur and distinguishing these anomalies from variations in component and
overall engine aerothermodynamic performance.
CITED EXAMPLE
26. Malloy, D. J., Chappell, M. A., and C. Biegl, Real-Time Fault Identification for Development Turbine Engine
Testing, ASME Paper # 97-GT-141, Presented at ASME International Gas Turbine Institutes Turbo Expo,
Orlando, FL, June 1997.
A new capability for real-time fault identification during developmental turbine engine testing has been developed and
demonstrated. A schematic of the overall process is presented in Figure 23. The capability automates a model calibration
process and emulates traditional manual approaches. A parallel computing approach permits real-time operation of the
processes. The model-based fault identification approach was demonstrated on steady state and transient test data,
successfully detecting and identifying anomalous measurements and distinguishing these from unusual variations in
component and overall engine performance. A component-level model provided the basis for fault identification and an
automated model calibration process ensured adequate model fidelity. In addition, the component-level model enables
automation of fault detection and diagnostic techniques that generally rely on engine cycle-matching principles.
ENGINE
ENGINEMODEL
MODEL
EVENT
EVENT
DETECTION
DETECTION
RULE-BASED
RULE-BASED
EXPERT
EXPERTSYSTEM
SYSTEM
FAULT
FAULT
DETECTION
DETECTION
and
and
DIAGNOSIS
DIAGNOSIS
High Performance Computer Platform
CALIBRATE
CALIBRATE
MODEL
MODEL
COMPARE
GOOD
BAD
TIME
2-38
1%
5%
STEADY-STATE
OPERATION
FAULT
IDENTIFIED
10
TIME, SEC
Figure 24 - Abrupt change in compressor exit pressure during steady engine operation
POTENTIAL BENEFITS
The effects of environmental variances can be determined for gross and net thrust, time-to-thrust, airflow, fuel flow,
turbine temperature, combustor pressure, rotor speeds, and fan and compressor operating lines, using a calibrated
component-level simulation of a gas turbine engine. A transient data adjustment method based on simplified turbine
engine cycle interrelationships was employed to compensate engine performance parameters for the combined effects of
variances in inlet pressure and temperature and ambient pressure. The adjustment method was applicable to both steady
state and transient operation.
2-39
CITED EXAMPLE
27. Chappell, M. A., and R. McKamey, Adjusting Turbine Engine Transient Performance for the Effects of
Environmental Variances, AIAA Paper # 90-2501, July 1990.
In general, at altitude test facilities, either referred parameters are used to adjust transient data or no adjustments are
made. Referred parameters address the effects of variances in inlet conditions (P2 and T2) but fail to address the effects
of variances in exit conditions (P0). The adjustment method that was applied for the cited example was based on
fundamental gas turbine cycle interrelationships. These relationships were developed specifically for determining
performance changes with respect to changes in the environment (or in the control) for steady state and transient engine
operation. As a result, the relationships were easily adapted as a transient data adjustment method for transient engine
testing in a direct-connect altitude testing facility. The adapted method addresses variances in inlet pressure and
temperature and exit pressure. Typical environmental variances for sea-level-static conditions are illustrated in Figure
25.
2-40
1%, and 32%, respectively. The adjustment method may be applicable to larger pressure variances. However larger
variances were not investigated. Applicability of the adjustment method was limited to small (< 1.0-%) temperature
variances because larger variances produced unacceptable results. The demonstration was restricted to choked nozzle
operation and to environmental variances that have a minimal effect (< 1.0-%) on fan rotation speed. The adjustment
method was found to be applicable to both steady state and transient engine operation.
5.1.5.
During development testing, it is necessary to verify the design quantified by the various measures of merit and to refine
the design. The test engines are usually instrumented with both steady state and transient instrumentation to evaluate
engine performance and operability. Determining figures of merit from engine measurements in not always easy and
practical. For example, turbine inlet gas temperature used in engine performance analysis is based on a flow-path area
weighted average of a spatially varying environment. To determine the flow-weighted average from measurements is a
very difficult task. It would require a large number of temperature probes covering the gas path station and each probe
must be able to survive in a high temperature environment. By using flight test data as the input to a modeling technique,
these difficult engine merits of performance can be calculated.
POTENTIAL BENEFITS
Engine transient simulations, which properly account for non-equilibrium conditions, can provide system analysis and
tradeoff study tools for judicious configuration selection and control-mode design, thereby effectively preventing
operation problems and reducing costs. System design and optimization can be particularly challenging for an
afterburning turbofan fighter engine with fast rotor response and augmentor transient requirements. Data driven engine
models aid design verification by calculating hard-to-measure parameters. The calculated average flow-path parameters,
such as in-flight thrust and turbine inlet temperature provide a means to quantify engine performance from rudimentary
flight test instrumentation.
CITED EXAMPLE
28. Khalid, S. J., Role of Dynamic Simulation in Fighter Engine Design and Development, Journal of Propulsion and
Power, Vol. 8, No. 1, January-February 1992, pp. 219-226.
2-41
The modeling technique described in the cited reference was successfully used to calculate the in-flight transient thrust
of a modern fighter engine installed in the F-15 flight-test aircraft. The analysis was of snap engine acceleration from
flight idle to military power, performed at an altitude of 40,000 ft. and Mach number of 0.52. The installation effects of
compressor bleed and horsepower extraction were included in the analysis. The measured flight-test data for flight
conditions, engine fuel flow, variable geometry, and exhaust nozzle area were used as input to the transient engine
model. Model calculations for transient thrust, exhaust nozzle inlet pressure, (PT6M), and fan rotor speed (N1) are
presented in Figure 27. The model calculations of PT6M and N1 are compared to flight-test data and show excellent
agreement. Since PT6M is reflective of gross thrust and N1 is reflective of total engine airflow, it may be inferred that
the calculated net thrust, FN, is representative. It may also be inferred from Figure 27, based upon the accuracy of the
compressor exit total pressure, PT3, and the use of actual fuel flow rates, that the calculated combustor exit temperature
TT4 is also representative.
Figure 27 - In-flight transients, Idle-to-Max Snap, Altitude 40,000 ft., Mach 0.52
2-42
POTENTIAL BENEFITS
The success criteria for starting include: good ignitability, combustion stability, adequate stall margin, ability to spool-up
from a low compressor speed at acceptably low airspeed, and short acceleration time to idle to prevent excessive altitude
loss in the engine-out condition. Meeting all these conflicting requirements in the far off-design sub-idle region is the
systems engineers challenge. A model that is suited for this type of off-design condition is a tremendous aid to the
analysis process.
CITED EXAMPLE
29. Khalid, S. J. and R. T. Legore, Enhancing Fighter Engine Airstarting Capability, International Journal of Turbo
and Jet Engines, Vol. 10, 1993, pp. 225-233.
The modeling technique described in the cited reference was successfully used to calculate an in-flight spool-down and
re-light. The fidelity of the empirical model is evident from Figure 29 which shows a comparison of model results with
flight-test data for the 25% spool-down airstart at 30,000 ft., 250 KIAS. The results showed good agreement with flight
test data and thus the empirical model was found adequate to obtain pre-test prediction of start transients with start logic
revisions.
2-43
Fouling, due to minute dust particles, pollen, salt spray and insects, which gets deposited on blade surfaces;
Erosion of blade sources caused by particulate ingestion;
Tip clearance increase of blade tips caused by particulate ingestion;
Water ingestion during rain;
Foreign Object Damage (FOD) caused by hailstones, runway gravel and bird ingestion.
2-44
of the compressor at all speeds, with the largest influence at 100% speed. Increased tip clearance has a more pronounced
effect on the compressor adiabatic efficiency and a lesser effect on the pressure ratio.
The paper by Singh, et al, uses a simple mean line method to model the effects of increased blade roughness and tip
clearance due to erosion, on compressor performance. The model can predict the compressor stage performance, given
the blade inlet and exit metal angles, blade stagger, camber, chord, solidity, thickness to chord ratio and hub to tip
diameters. The model developed by Singh was validated using experimentally measured performance data obtained
before and after erosion caused by the ingestion of 25 Kg. of sand. The model was then used to predict the effect of
increased blade roughness and tip clearance due to erosion on two other single stage compressors with higher blade
loading.
The paper by Lakshminarasimha, et al, uses a stage stacking method where the stage characteristics are synthesized from
a generic stage characteristic which is then modified by the design point information.
POTENTIAL BENEFITS
From operational, economic and safety considerations, gas turbine performance deterioration has emerged as a very
important topic of research. From an economic point of view, a one percent increase in specific fuel consumption (SFC)
for a 45 aircraft fleet of single engine aircraft, could increase the expenditure due to additional fuel by a million dollars a
year. Using models to aid in the determination and prediction of deterioration effects is very economical way of
analyzing this problem.
CITED EXAMPLES
30. Lakshminarasimha, A. N., Modeling and Analysis of Gas Turbine Performance Deterioration, ASME #92-GT395, June 1992.
31. Singh, D., et al, Simulation of Performance Deterioration in Eroded Compressors, ASME Paper #96-GT-422,
June 1996.
In general, blade erosion is a complex function of the physical properties of the particles, blade material and
aerodynamic parameters such as particle mass flow, particle size, shape, velocity and direction of impingement, the
geometry, and the material of the blade row. In view of this complexity, the prediction of erosion and the subsequent
change in compressor performance is a complicated task.
The method as described by Lakshminarasimha, was initially validated against a single stage compressor as illustrated in
Figure 30. Considering the simplicity of the method, a good comparison can be seen between measured and simulated
results.
2-45
2-46
operator, any simplified map generation technique would be extremely useful to the operator to aid in meaningful
maintenance scheduling and trending of the performance variation of the component.
The stage stacking method takes into account interrelationships among stages through compatibility of speed, mass flow
and energy. Thus, it provides a logical basis for examining the behavior of a multistage compressor subjected to
deterioration in one or more stages. Additionally, it is possible to study the effect of different types of deterioration using
this method. Usually, a gas turbine user has neither the compressor performance maps nor the stage characteristics for
performance simulations. In such a situation, one can use the on-site performance measurement values of baseline
compressor flow, efficiency, and the annulus area of the compressor gas path together with the generalized stage
compressor characteristics for compressor performance simulation. However, this means of obtaining performance can
only be used for trending since the accuracy is in doubt.
5.2.
OPERABILITY
An important component of the gas turbine engine is the compression system. In today's military turbine engines, the
compression system consists of one or more axial compressors. These axial compressors must operate in a stable manner
even with severe inlet pressure or temperature distortion. Many experimental and analytical investigations have been
conducted in the past three or four decades to separately quantify the effects of pressure or temperature distortion on the
compression system. While little experimental work has been performed on combined time-variant total pressure and
temperature distortions, some investigations have been carried out to examine the effects of combined steady-state
pressure and temperature distortion on compression system stability. With the advent of highly agile maneuvering
aircraft with weapons release near the engine inlets, there will exist a requirement to quantify the combined effects of
severe pressure and temperature distortion, both transiently and in the steady state.
2-47
Pressure Coefficient
1.8
Normal
Operation
Rotating
Stall
1.6
Reversed
Flow
1.4
1.2
1.0
-1.0
Zero Flow
-0.4
-0.2
0.2
0.6
1.0
Flow Coefficient
Temperature Coefficient
1.0
2.0
0.6
0.2
Rotating
Stall
Reversed
Flow
Normal
Operation
-0.2
-0.6
Zero Flow
-1.0
-1.0
-0.4
-0.2
0.2
0.6
1.0
Flow Coefficient
d ( FX )
+ FX = FX ss
dt
The time constant, , is used to calibrate the model to provide the correct post-stall behavior. The inflow boundary
during normal forward flow is the specification of total pressure and temperature. The exit boundary condition is the
specification of exit Mach number or static pressure. During reverse flow the inlet is converted to an exit boundary with
the specification of the ambient static pressure. Therefore, both the inlet and exit boundaries function as exit boundaries
during a surge cycle.
POTENTIAL BENEFITS
Mathematical compressor models have become a major tool for understanding compression system behavior during
dynamic events, such as inlet distortions. A validated compression system model can be used to extend the range of the
2-48
experimental test results to the untested regime. Using a validated stage-by-stage compression system model, a
parametric investigation can be conducted to determine the qualitative effects of various recovery actions or design
changes on system operability.
CITED EXAMPLES
32. Hale, A. A., and M. W. Davis, Jr., Dynamic Turbine Engine Compressor Code, DYNTECC Theory and
Capabilities, AIAA Paper -92-3190, June 1992.
33. Davis, M. W., Jr., et al, Euler Modeling Techniques for the Investigation of Unsteady Dynamic Compression
System Behavior, Loss Mechanisms and Unsteady Flows in Turbomachines AGARD-CP-571, January 1996.
DYNTECC was configured to aid in the analysis of a ten-stage compressor rig test conducted at the Compressor
Research Facility to investigate the boundary between surge and rotating stall for high-pressure ratio compressors. The
model was executed at the experimentally determined stall and surge boundary where the recovery hysteresis was most
severe. DYNTECC was calibrated at the stall and surge boundary by adjusting calibration constants to produce rotating
stall. The constants were held fixed for subsequent analysis that investigated possible hardware modifications to reduce
the sensitivity of the compressor to rotating stall.
Calibration of DYNTECC with experimental data matched the overall performance and individual stage performance of
the 10-stage compressor rig. As illustrated in Figure 35 the compressor initially experiences a partial surge cycle, then
transitions to rotating stall. DYNTECC individual stage performance is also compared to that obtained experimentally.
For both overall and individual stage behaviors, DYNTECC reproduced the experimental results fairly accurately during
the dynamic event.
2-49
information. If blade shapes are known, a mean-line or streamline curvature technique can be used to generate
characteristics. However, these types of codes rely on empirical inputs based upon cascade test results and generic
correlations. In addition the DYNTECC code, even though the single spool version has a parallel compressor capability,
does not handle radial distortion well and the dual-spool version of DYNTECC is currently not configured for parallel
compressors.
COMPRESSOR MODEL
DISTORTION
PATTERN
LOW PRESSURE
HIGH PRESSURE
POTENTIAL BENEFITS
Mathematical compressor models have become a major tool for understanding compression system behavior during
dynamic events such as inlet distortions. A validated compression system model can be used to extend the range of the
experimental test results to the untested regime. Using a validated parallel compressor model, a parametric investigation
can be conducted to determine the qualitative effects of certain combinations of transient and steady state pressure and
temperature distortions on system operability.
CITED EXAMPLE
34. Davis, M. W., Jr., et al, Euler Modeling Techniques for the Investigation of Unsteady Dynamic Compression
System Behavior, Loss Mechanisms and Unsteady Flows in Turbomachines, AGARD-CP-571, January 1996.
DYNTECC can be configured to analyze inlet total pressure distortion using a modified parallel compressor theory.
2-50
DYNTECC was used to analyze the effects of inlet distortion on a two-stage, low aspect ratio fan. The model was
initially calibrated against the experimental clean inlet performance of the compressor. The modified model was then
validated using the compressor performance with a pure circumferential inlet distortion pattern. The inlet distortion
pattern was generated from a 1/rev circumferential distortion screen shown in Figure 37.
Experimental Data
Clean
Pressure Ratio
4.4
Distorted
4.2
4.2
Model Results
Clean
3.8
3.8
Distorted
3.6
3.4
3.4
Distortion
Transient
3.2
94
94
95
95
96
96
97
97
98
98
99
99
100
100
101
101
102
102
% Corrected Airflow
Percent Corrected
Airflow
2-51
HTO300
3D CFD Algorithm
TEACC
POTENTIAL BENEFITS
Distortion imposed on a circumferentially swirling flow has been shown to be three-dimensional (3-D) in nature. Design
or analysis engineers are interested in understanding the details of the flow field to determine the effects of inlet total
pressure distortion on the compressor. One way to quantify the effects of distortion is to test for that effect in a ground
test facility. Currently, the inlet and engine are tested separately. Typically, the aircraft fuselage is too big to fit in a wind
tunnel. A forebody simulator is used in conjunction with the inlet to characterize its flow field. The forebody simulator is
designed to produce a flow field at the inlet reference plane (IRP) similar to the flow field produced by the aircraft.
Screens are constructed to capture the most severe dynamic patterns produced by the inlet and are then placed in front of
the engine to measure the loss of stall margin produced by the steady-state inlet distortion. However, it is expensive to
instrument a compressor and perform the necessary number of tests to adequately understand the compressor flow field.
Numerical simulations have been developed to support the testing community in this area.
CITED EXAMPLES
35. Hale, A. A., and W. F. OBrien, A Three-Dimensional Turbine Engine Analysis Compressor Code (TEACC) for
Steady-State Inlet Distortion, Journal of Turbomachinery, Vol. 120, July 1998, pp. 422-430.
36. Hale, A. A., Chalk, J., Klepper, J., and K. Kneile, Turbine Engine Analysis Compressor Code: TEACC Part II:
Multi-Stage Compressors and Inlet Distortion, AIAA Paper # 99-3214, Presented at the 17th AIAA Applied
Aerodynamics Conference, June 1999.
A three-stage fan (Ref. 36), representing a modern, high-performance military fan was modeled. It consists of a
structural strut, a variable inlet guide vane (IGV) attached to the back of the strut, and three rotor-stator pairs. Only
overall experimental performance data were available for comparison. No blade-row by blade-row or radial distributions
of flow quantities were available. Consequently, only overall performance was compared with data. An axial-radial
representation of the grid used for the three-stage fan is shown in Figure 40. The three-dimensional grid on which the
flow field solution was obtained was uniformly spaced in the circumferential direction at 15-degree intervals.
The fan was executed with clean, standard-day inlet conditions. The results of the TEACC fan simulation are at 68% of
the design corrected mass flow on the 80% speed line. Both the normalized pressure ratio and the normalized
temperature ratio achieved at this point were within 0.5% of the experimental data. These results were achieved with
2-52
uncalibrated loss and deviation correlations obtained from the open literature. As expected, there is an increase in both
total pressure and total temperature through the machine. Figure 40 provides an insight into the general character of how
the total pressure and temperature vary radially and axially throughout the machine.
Strut IGV
R1
S1
R2 S
2 R3 S
3
0.98
Normalized Total
Temperature
1.27
1.00
Figure 40 - Grid and performance for three-stage fan at 80 percent corrected speed
Illustrated in Figure 41 is the calibrated clean inlet performance of the three-stage fan at various speeds and a
corresponding flow-field for one particular point on the map. Illustrated in the flow-field view are streamlines, colored
with Mach number, through the fan. This figure clearly illustrates the rotors turning the flow and the stators
straightening the flow. The total turning of the flow through the entire machine is about 45 degrees. The hub has the
highest Mach number through the entire fan. The streamlines illustrate that the flow is almost completely axial at the exit
of the fan.
Total Pressure & Streamlines
TEACC (Calibrated)
4.00
Data
3.50
e
Lin
3.00
101 %
Pt
ll
Sta
2.50
3.6
F
90 %
l
2.00
o
w
80 % Corrected Speed
1.0
1.50
150
170
190
210
230
250
270
290
Figure 41 Point inlet performance and corresponding flow-field for a typical three-stage military fan
This same three-stage fan was exercised with a total pressure distortion flow-field imposed upon its inlet as illustrated in
Figure 42. The distortion was simulated with as a steady state pattern produced by an inlet distortion screen, a standard
ground test method for simulating distortion at the engine inlet plane. Although, the model results are not compared to
experimental results there is enough information to make an engineering judgement as to the validity of the results.
2-53
4.00
2.00
Distorted
2.50
i n e in e
lL
t a l ta l l L
S
e t let S
I nl
an ted In
e
l
C
to r
Dis
90 %
3.50
3.00
1.50
150
101 %
80 % Corrected Speed
170
190
210
230
250
Corrected
Mass
Flow
270
290
5.2.4. ROTATING STALL AND SURGE INVESTIGATION USING THREE-DIMENSIONAL EULER MODELING
Compressor instability is a major limiting factor on gas turbine engine operating range, performance, and reliability. The
instability, either in a form known as rotating stall or surge, occurs at an operating point with low mass-flow and a large
pressure rise. To avoid such instabilities, the compressor has to run at an operating point corresponding to a lower
pressure ratio so that an adequate stall margin is maintained. The stall margin can be badly reduced in operating
environments for which the inlet conditions are non-uniform. Predicting the condition at which instability will occur in a
compressor requires an understanding of the flow process leading to the onset of the instability. The transition from
initial disturbance to final stall or surge can usefully be divided into three stages, (1) inception, (2) development, and (3)
final flow pattern. The inception stage is the period when disturbances start to grow (flow becomes unstable). It defines
the operating point and conditions at which instability occurs. In practice, the disturbances will take a finite amount of
time, ranging from a few to several hundred rotor-revolutions, to grow into final stall or surge. So, the inception stage
can be viewed as the early development of the unstable flow. The development stage, which includes all the processes
after the inception stage through to the final flow pattern, is usually of less importance. It is often the case that one final
form of instability in one compressor could be the pre-stage of the final form in another compressor.
2-54
POTENTIAL BENEFITS
Having a model to emulate the inception of rotating stall or surge provides a means for improving the operational
characteristics of any compression system. Understanding the mechanisms of rotating stall inception will provide
compression system designers a means to prevent or delay the onset of this type of instability. A strength of this type of
modeling technique is that the model can be used to investigate the interactions between the compressor and other
components. Some types of investigations that could be conducted are:
Interaction between the inlet and the compressor with distortions and its impact on the performance and stability
margin;
Hot gas ingestion into the engine and its impact on stability;
The behavior of an inlet vortex in an intake and its impact on the performance and stall margin;
The model could become a component in an engine system to model the dynamic behavior of a whole engine under
various dynamic situations.
CITED EXAMPLE
37. Gong, Y., A Computational Model For Rotating Stall and Inlet Distortions in Multistage Compressors, GTL
Report #230, March 1999, Gas Turbine Laboratory, Massachusetts Institute of Technology, Cambridge, MA.
2-55
38. Longley, J. P., et al, Effects of Rotating Inlet Distortion on Multistage Compressor Stability, ASME Paper #94GT-220, June 1994.
In the work by Gong, the model is used to simulate stall inception. Two major inception types have been experimentally
identified: modal waves and spikes. Modal waves are exponentially growing long wavelength (length scale comparable
to the annulus) small amplitude disturbances. Modal waves penetrate the whole compressor in the axial direction, so
they can be detected by sensors at any locations at the inlet, exit, or with the compressor. The other inception mechanism
is the growth of localized non-linear short wavelength (with length scale of several blade pitches) disturbances, often
referred to as spikes. The inception starts as one or several spike-shaped finite amplitude disturbances within the tip
region of a particular stage. Usually, the disturbance develops into a large full span stall cell within three-to-five rotor
revolutions. The simulation was executed such that a short wavelength stall inception was initiated by a spike-shaped
disturbance.
The flow coefficient traces shown in Figure 45, as taken from the tip region of the first rotor inlet of a four-stage GE
low-speed compressor, show that the disturbance is sustained and the disturbance leads to compressor stall subsequently.
The overall inception is similar to the measurements as illustrated in Figure 42. A quantitative comparison shows that
the stall inception initiated by the spike-shaped disturbance has an initial disturbance rotating speed of 83% of rotor
speed, and a transition time of about three-rotor revolutions and compares reasonably well with experimental results.
Figure 45 - Model prediction of stall inception process for short wavelength (spike) stalls
Figure 46 - Experimental data of stall inception process for short wavelength (spike) stalls
In multi-spool engines, rotating stall in an upstream compressor will impose a rotating distortion on the downstream
compressor, thereby affecting its stability margin. Inlet distortion can modify the flow rate at which rotating stall occurs,
degrading the range over which stable operation is possible. The investigation reported by Longley (Ref 38) addresses
the relationship between speed and direction at which the inlet distortion pattern rotates around the compressor annulus
and the severity of the degradation in stability margin. Imposition of a rotating distortion can also be thought of as a type
of forced response experiment to probe the compressor dynamic behavior. The measured mean flow coefficients at stall
inception as a function of screen rotation speed, the means by which rotating distortion was imposed, for the GE fourstage compressor is presented in Figure 47. The calculations from the modeling technique, described in the paper by
Gong are shown in the figure as solid lines. Both the data and the model indicated that there is less stability margin
2-56
POTENTIAL BENEFITS
The ATEC model and simulation can simulate on and off-design steady-state operation, as well as transient and dynamic
engine responses to perturbations in a wide range of operational and control conditions. By example, it has been shown
that the ATEC simulation can handle a wide variety of conditions that occur during normal and abnormal gas turbine
2-57
engine operation. The benefits of the variable time-step routine, which uses a combination of the explicit and implicit
numerical solvers, were also demonstrated. Test cases were presented that demonstrated that the ATEC simulation
could: be calibrated to a steady-state data set; extended the steady-state calibration to a transient fuel variation; presented
results from an engine operation that resulted in compressor surge; and addressed a turboshaft engine going through the
start process. The ATEC results were shown to agree closely with test data where available.
CITED EXAMPLE
39. Garrard, G. D ATEC: The Aerodynamic Turbine Engine Code For the Analysis of Transient and Dynamic Gas
Turbine Engine System OperationsPart 1: Model Development, ASME Paper #96-GT-193, June 1996.
40. Garrard, G. D ATEC: The Aerodynamic Turbine Engine Code For the Analysis of Transient and Dynamic Gas
Turbine Engine System OperationsPart 2: Numerical Simulations, ASME Paper #96-GT-194, June 1996.
This example demonstrates the real benefit of a dynamic simulation, by computing post-stall operation of a compressor
and engine system, which cannot be modeled with a cycle-type simulation. The test case simulated a transient throttle
movement using the T55-L-712 that resulted in the gas generator portion of the engine decelerating from 100 percent
speed to approximately 90 percent speed. After a brief pause at the 90 percent speed, the engine was accelerated back to
the 100 percent speed condition. The change in fuel flow rate during the acceleration was fast enough to force the
compressor into surge cycles. The relative compressor pressure ratio as a function of time is illustrated in the figure
below. The multiple surge cycles cause a significant drop in the total pressure throughout the engine and a corresponding
flow reversal as indicated in Figure 48.
1.0
0.8
Surge
Cycles
0.6
0.4
0.2
0.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Time (Seconds)
5.2.6.
COMBUSTOR DYNAMICS
Due to the inherent dynamics of the fluid system, transient performance of a gas turbine engine can differ significantly
from that predicted from quasi-steady operating assumptions. The consequences of these can be quite dramatic,
including unexpected crossing of the compressor surge line while changing operating points. Beyond the surge line,
compressor rotating stall and surge serve as the forcing function for a complex dynamic interaction between engine
components. Because these unsteady operating cycles produce substantially reduced performance and durability the
recovery from the instability is an important issue facing the gas turbine designer. Because of this, significant efforts
have been made to accurately simulate the performance of a compressor undergoing a surge transient. As these
techniques have matured, the focus increasingly has shifted to an extension of these methods to encompass the entire
engine, and thus capture the important compressor-combustor interactions that occur during engine surge.
2-58
POTENTIAL BENEFITS
Dynamic modeling of combustor and compressor interaction can lead to a better understanding of the relationship
between combustor blowout and re-light during engine surge. When a multi-zoned model is used, combustor flow within
the primary zone can be modeled with finite-rate chemistry to allow predictions of blowout and the effects of
perturbation in boundary and operating conditions.
CITED EXAMPLES
41. Rodriguez, C. G., and W. F. OBrien, Unsteady, Finite-Rate Model for Application in the Design of Complete
Gas-Turbine Combustor Configurations, Design Principles and Methods for Aircraft Gas Turbine Engines, RTOMP-8, February 1999.
42. Costura, D. M., et al, A Model for Combustor Dynamics for Inclusion in a Dynamic Gas Turbine Simulation
Code, AIAA Paper # 97-3336, Presented at the 33rd AIAA/ASME/SAE/ASEE Joint Propulsion Conference, July 69, 1997, Seattle, WA.
Indicated in Figure 49, is a discretization of a generic reverse flow combustor used in an analysis as presented in Ref. 41.
The main issues associated with the layout of the grid from the point of view of a one-dimensional theory are division of
the main flow into primary and annular paths, and interaction between flow paths through the presence of dilution holes.
Once the grid has been selected, analysis can be conducted to determine the effects of perturbations in boundary and
operating conditions. Indicated in Figure 50, are the effects of an imposed fuel flow oscillation on total pressure.
2-59
b. Combustor Pressure
2-60
2-61
inlet with a gas turbine engine. The supersonic inlet was modeled using the Large Perturbation Inlet (LAPIN) computer
code, and the gas turbine engine was modeled using the Aerodynamic Turbine Engine Code (ATEC) as illustrated in
Figure 54.
APPL
LAPIN
ATEC
Translated
Centerbody
Figure 55 - Cross-section view of the 4060 inlet and J85-13 turbojet installation
The cited paper (Ref 43) describes the general modeling techniques used in the simulations, and the approach taken to
implement the simulations under the APPL environment. It presents results from selected test cases.
POTENTIAL BENEFITS
Traditionally, aircraft inlet performance and propulsion performance have been designed separately and latter mated
together via flight-testing. In todays atmosphere of declining resources, it is imperative that more productive ways of
designing and verifying aircraft and propulsion performance be made available to the aerospace industry. One method of
obtaining a more productive design and evaluation capability is with numerical simulations. Numerical simulations can
provide insight into physical phenomena that may not be understood by test data alone. Simulations can fill information
gaps and extend the range of test results to areas not tested. In addition, once a simulation has been validated, it can
become a numerical experiment and the analysis engineer can conduct what-if studies to determine possible solutions
to performance or operability problems.
CITED EXAMPLE
43. Garrard, G. D., Davis, M. W., Jr., Wehofer, S., and G. Cole, A One-Dimensional, Time Dependent Inlet/Engine
Numerical Simulation for Aircraft Propulsion Systems, ASME Paper # 97-GT-333, June 1997.
The simulation system was exercised using a supersonic inlet with sixty percent of the supersonic area contraction
occurring internally, and a GE J85-13 turbojet engine. The inlet-engine simulation combination of LAPIN and ATEC
was compared to experimental results. A transient event was initiated at a flight Mach number of 2.5 by pulsing the
bypass doors in the closed direction. The result was an inlet unstart followed by an engine compression system stall.
During the given transient, the majority of the system instabilities can be traced to the fact that the normal shock, located
initially downstream of the inlet throat, was expelled outside of the inlet. The location of the shock is plotted as a
function of time in Figure 56. The shock location is normalized by the inlet cowl lip radius, and referenced to the
centerbody tip. The cowl lip is axially located two cowl-lip-radii downstream of the centerbody tip. The act of closing
the bypass valve forces the shock structure to be expelled from the inlet. Moving the centerbody forward in conjunction
with proper modulation of the bypass doors allows the shock to be reingested.
2-62
6
5.5
5
4.5
4
3.5
3
2.5
2
1.5
1
0
0.5
1.5
2.5
Time (Seconds)
4.50
ycle
Surge C
4.00
Pre-stall
Operating
Point
3.50
3.00
2.50
2.00
Final
Operating
Point
1.50
1.00
0.50
-20
-10
10
20
30
40
50
60
70
80
Figure 57 - Compressor pressure ratio as a function of compressor inlet corrected mass flow rate
5.3.
Life cycle costs have emerged as a primary factor in the design of gas turbine powerplants. The recognition of repairs
and rebuilds and lost availability as major recurring costs suggests that some means of getting more use from the engine
would be attractive. The concept of on-condition repair and maintenance has emerged as a common ownership
philosophy for both military and industrial equipment. Such a philosophy can only be put into practice if the tools to
accurately assess engine condition can be assured. Furthermore, these tools must be tailored to the needs and capabilities
of the personnel who must use them.
5.3.1. INTEGRATED ANALYSIS TOOL FOR GAS TURBINE COMPONENT LIFE ASSESSMENT
A major part of aircraft engine operating costs is related to maintenance. Maintenance cost would be significantly
2-63
reduced if inspection intervals could be extended and component service life increased. Inspection intervals and service
life are commonly based on statistical analysis, requiring a limited probability of failure (a certain level of safety) during
operation. However, in many cases this approach leads to conservative inspection intervals and life limits for the
majority of parts or components. Having an appropriate analysis tool offers a way to attempt to reduce maintenance
costs and improved safety by applying usage monitoring to predict operation component condition and thereby
facilitating on-condition maintenance.
CFD model
Flight
data
h(t,x)
FACE
MARC
thermal
GSP
Wf
Aj
Mach
Alt
Temp
N2(t)
gas:
T(t)
p(t)
q(t)
MARC
stress/strain
N2(t)
p(t)
q(t)
comp:
T(t,x)
LIFING MODEL
stress
hold time
frequency
Life
fraction
n/Nf
FACE
GSP
CFD Model
MARC
Lifing Model
The FACE system consists of both on-board and ground-based hardware. In the aircraft two electronic boxes are
installed: the Flight Monitoring Unit, (FMU) and the Data Recording Unit (DRU). The FMU is a programmable unit that
determines which signals are stored and how they are stored. The relevant signals stored by the DRU are engine
parameters from the engines Digital Electronic Engine Control (DEEC) and avionics data. The DEEC signals can be
sampled at a maximum frequency of 4 HZ. The following signals, which together fully describe engine usage are stored.
These parameters, as a function of time, are used as input to the GSP model, which is the next tool in the sequence.
The Gas Turbine Simulation Program (GSP) is a tool for gas turbine engine performance analysis, which has been
developed by the NLR. This program enables both steady state and transient simulations for any kind of gas turbine
configuration. The simulation is based on one-dimensional modeling of the processes in the different gas turbine
components with thermodynamic relations and steady-state characteristics (component maps). GSP can be used to
calculate gas temperatures, pressures, velocities, and composition at relevant engine stations from measured engine data.
This particularly applies to stations for which no measured data is available, such as the critical high-pressure turbine
entry temperature. Also, GSP is able to accurately calculate the dynamic responses of these parameters (critical to engine
life) where measured data is not available or has unacceptably high time lags or low update frequencies. The GSP output
is used for further processing by the CFD and MARC finite element models.
The Computational Fluid Dynamic (CFD) model is used to accurately calculate the heat transfer from the hot gas stream
2-64
to the component. For this calculation it is important to have detailed information on the geometry of both the flow
channel and the different components that disturb the flow (blade vanes). From CFD analysis of the gas flow through the
gas turbine, values for the heat transfer coefficient are obtained at specific locations in the component. The heat transfer
coefficient value varies significantly along the flow path, due to variations in the flow conditions (gas velocity, type of
flow laminar or turbulent, etc). An engineering approach, which results in a number of functions that describe the
approximate distribution of the heat transfer coefficient across the blade surface, is followed. This heat transfer model is
based on, and validated with, heat transfer results obtained by the CFD model.
The Finite Element (FE) model consists of two interrelated models. The thermal model calculates the temperature
distribution in the component, based on the heat input from the hot gas stream. The mechanical model calculates the
stresses and strains in the component, caused by the varying temperature distribution and the externally applied loads.
The finite element code used is MARC, which is a commercially available, multipurpose finite element package.
The thermal model calculates the temperature distribution in the component. For each finite element on the surface of
the component, the heat-transfer coefficient follows from the CFD model. Given the thermal conductivity of the
material, the temperature distribution in the component can be calculated. A transient thermal analysis can be performed
for the complete flight under consideration with the time-variant ambient gas temperature as input.
The mechanical model calculates the stress and strain distribution in a component. Both rotational frequency and the
temperature distribution as functions of time are input for the model with stress and strain distribution as output. The
temperature distribution is obtained from the results of the thermal analysis and the values of the rotational frequency are
obtained from GSP.
A lifing model generally calculates either total time to failure or number of cycles to failure for a certain component
subjected to a specific load sequence. A large number of specific life prediction models have been developed over the
last twenty years, where each model is appropriate for a specific application. The major division in lifing models is
between total life models and crack growth models. Total life models only calculate the time to failure and do not
consider the way failure is reached. On the other hand, crack growth models represent the Damage Tolerance
philosophy, which accepts the presence of material defects and aims to monitor crack growth and suggests removing the
component before the crack become unstable. In the end, the choice of the lifing model depends on the expected failure
mechanism of the component under consideration.
POTENTIAL BENEFITS
The potential of the analysis tool presented within the cited example is twofold. Firstly, the tool can be used to examine
and support on-condition maintenance. The load history of every individual component can be tracked and can be used
to determine the inspection interval or actual life limit of that specific component. The general and mostly very
conservative life limits supplied by the manufacturer are based on an assumed usage, to which a safety factor has been
applied to account for heavier usage. This safety factor can now be quantified and possibly decrease, leading to a huge
saving in spare parts and inspection costs. Secondly, the tool can be used to compare different mission types or
maneuvers with respect to life consumption. The results can be used to optimize operational use of the aircraft.
CITED EXAMPLE
44. Tinga, T., et al, Integrated Lifing Analysis Tool for Gas Turbine Components, ASME Paper # 2000-GT-646, May
2000.
The analysis process described in the cited example has been demonstrated by applying it to the F100-PW-220 engine of
the RNLAF F-16 fighter aircraft. The component selected for analysis is the 3rd stage turbine blade, which is the first
stage rotor of the high-pressure turbine (HPT) module. To demo this capability, a random mission was selected and
illustrated in Figure 59.
4.00
9000
3.50
7500
3.00
6000
altitude
2.50
4500
2.00
3000
1.50
1500
1.00
fuel flow
0.50
-1500
0.00
-3000
4000
500
1000
1500
2000
Time (s)
2500
Altitude (m)
2-65
3000
3500
GSP 8.1
1400
deteriorated
FTIT [K]
1200
1000
800
600
400
0
500
1000
1500
2000
Time [s]
2500
3000
3500
4000
2-66
5.0E-06
3100
accumulated creep strain
4.0E-06
2800
deteriorated
3.0E-06
2.0E-06
normal
1.0E-06
0.0E+00
500
-1.0E-06
2500
2200
1900
1000
1500
2000
2500
3000
1600
3500
1300
blade temperature
-2.0E-06
1000
-3.0E-06
700
-4.0E-06
400
Temperature (degC)
From the GSP output (FTIT), the finite element models determine the temperature distribution, its variation in time and
the stress and strain distribution together with its variation in time. In the mechanical model the creep phenomenon is
incorporated, because creep is assumed to be the life-limiting factor for the 3rd stage turbine blade. For a new component
the amount of creep strain is nil and after sustaining a certain amount of creep failure occurs. Therefore, the creep strain
is a damage parameter and the evolution of creep strain represents damage accumulation. The results of the integrated
analysis is shown in Figure 62, which shows the accumulation of creep strain in the blade during the mission for both a
new engine and a deteriorated engine. The creep strain accumulation appears to be faster in the deteriorated engine,
which implies that the rate of life consumption is higher by a factor of 1.9.
Time (s)
2-67
POTENTIAL BENEFITS
With this coupled approach, a numerical tool is available to the aeropropulsion community that will allow fully
interactive analysis between the traditionally uncoupled aerodynamics and structural disciplines. A system level
approach, which would allow internal engine component and sub-components (compressor stage or blade row) to
interact with the full turbine engine system and thus with external aircraft structures, is envisioned. If analysis of an
internal engine component were desired, a more traditional CFD approach would be available through zooming.
CITED EXAMPLE
45. Nazir, J., Couch, R., and M. Davis, An Approach for the Development of an Aerodynamic-Structural Interaction
Numerical Simulation for Aeropropulsion Systems ASME Paper # 96-GT-480, June 1996.
An engine-inlet configuration associated with the HSCT program was analyzed in terms of inlet unstart and the effect of
the regurgitated shock wave. Inlet start is a complex three-dimensional phenomenon where a supersonic flow, which
comes through the inlet, is stabilized to some acceptable subsonic condition before entering the fan. This produces a
shock wave that sits strategically somewhere in the HSCT inlet, thus creating a transition zone from supersonic flow to
subsonic flow within the inlet itself. If, for whatever reason, the engine undergoes a transient such as an engine surge,
the stable shock wave will be disrupted, and may become unstable, possibly spilling out around the engine. This bubble
or plume will spread and produce loading on the surrounding structure, such as the wing, and can affect the aircraft
attitude-control surfaces. These will try to compensate for such pulse loading, as illustrated in Figure 63. If large enough,
the bubble or plume could be sucked in by the adjacent engine, causing it also to unstart. This will obviously intensify
the dynamics for the control system to compensate, thus requiring a thorough understanding of this phenomenon.
2-68
Aircraft
Wing
Area of Max Stress
Figure 63 - Effect of inlet unstart on aircraft wing assembly
The initial approach has been to analyze this coupled aerodynamic-structural interaction with a de-coupled numerical
technique. The approach that was taken was to model the surge cycle frequencies and intensity using a one-dimensional
compression system model, DYNTECC, for a typical high-pressure compressor. This scaled pressure loading at the fan
face was then introduced as a boundary condition to ALE3D, which characterized the inlet steady state shock location in
a three dimensional inlet. The appropriate boundary conditions were applied and the equilibrium flow was obtained as
described in the previous section. The surge conditions were then applied as a time-variant one-dimensional pressure
boundary condition. The initial spike and rapid drop-off occurs within the first 10 to 15 milliseconds after the event is
initiated. The highly cyclic nature of the blowdown part of the cycle was not represented by the boundary condition.
However, the cyclic oscillations cease to play a role in the inlet unstart once it has begun.
5.4.
ENVIRONMENTAL ISSUES
A gas turbine engine must often operate in non-ideal conditions. This section describes two such non-ideal
environments: Adverse Weather and Hot Gas Ingestion.
2-69
POTENTIAL BENEFITS
Through modeling, the effects of water ingestion on engine performance can be calculated and potential control actions
may be built into the engine control. A model can be used to determine the changes in performance due to water
ingestion. Further, it can show the effects of the design characteristics of the compressor, the water mass fraction, the
residence time of the mixture in the individual blade passages, and the operating conditions.
CITED EXAMPLES
46. Haykin, T., and S. N. B. Murthy, Transient Engine Performance with Water Ingestion, Journal of Propulsion, Vol.
4, No. 1, Jan-Feb 1988, pp. 81-88.
47. Ludorf, R. K., et al, Stage Rematching as a Result of Droplet Evaporation in a Compressor, ASME Paper #95GT-194, Presented at the IGTI Turbo Expo in Houston Texas, June 1995.
An illustration of the effects of water ingestion on key performance parameters is presented in Figure 64.
The following effects were also observed in the cited papers.
The extent of changes in compression subsystem performance due to water ingestion is significant and varies
nonlinearly with the amount of water ingested.
The effects of changes in compression subsystem performance on engine performance can become unmanageable
with the ingestion of large amounts of water, especially during engine deceleration.
Evaporation of water in the burner, during water ingestion into an engine, can have a serious effect on performance.
This includes surging of the compressor and a drastic fall in power output combined with large increases in specific
fuel consumption, especially when the phase change occurs at the exit of the burner.
Errors in input to the control system can arise due to the effect of water on sensors. Thus, the thermocouple (located
at the casing of the core compressor), which provides an input to the control for determining the setting of the stator
vane angles may register a temperature close to that of the water instead of that for the gas phase. The result is a
drastic change in the performance of the control and, hence, the engine.
Figure 64 - Effect of 4% humidity on major engine parameters, thrust, fuel flow, and SFC
2-70
2-71
Figure 67 - Ops line migration due to indicating stalling stag temperature ramps
2-72
POTENTIAL BENEFITS
Mathematical compressor models have become a major tool for understanding compression system behavior during
dynamic events such as inlet distortions. A validated compression system model can be used to extend the range of the
experimental test results to the untested regime. Using a validated stage-by-stage compression system model, a
parametric investigation can determine the qualitative effects of gas ingestion effects on system operability.
CITED EXAMPLE
48. Abdel-Fattah, A. M., Response of a Turbofan Engine Compression System to Disturbed Inlet Conditions Journal
of Turbomachinery, Vol. 119, No. 4, October 1997.
Using a stage-by-stage compression system simulation, the following observations were made with regard to the TF30
gas ingestion investigation:
The stability limit of the system in response to inlet temperature ramps improved with increasing low rotor speed;
The stability limit, in terms of temperature rise required to surge the compressor, was found to be independent of the
rise rate as illustrated in Figure 66.
At the time of instability because of inlet temperature ramps, the DYNTECC model predicted the possibility of the third
stage of the fan as the critical stage responsible for compression system surge initiation as illustrated in Figure 67.
5.5.
CONTROLS
With advanced engine types which feature many control inputs viz. fuel flow, compressor variable geometry, final
nozzle area etc. it is particularly important to ensure the correct setting of these inputs to attain optimum performance
through the life of the engine. When the engine and airframe control inputs are optimized as a single mathematical
problem (rather than being controlled independently as two distinct systems), there are highly significant gains. In order
to optimize these control inputs, the particular engine and airframe characteristics must be understood.
The PSC system was implemented on an F-15 supersonic research airplane that was powered by two PW1128
afterburning turbofan engines. The engines are controlled by Full Authority Digital Electronic Control systems
(FADEC). The FADEC provides control by:
Open-loop scheduling of compressor (LP and HP) variable guide vanes (VGV);
Closed-loop control of engine pressure ratio (EPR) via nozzle area (A8);
Closed-loop control of corrected fan speed (NLRT) via fuel flow (WFE).
The PSC system operates by applying trims as follows:
For subsonic conditions, trims are applied to LPVGV, HPVGV, NLRT and A8;
For supersonic conditions, trims are applied to LPVGV, HPVGV, EPR, calculated airflow (W2calc), afterburner fuel
flow (WFAB), inlet geometry: cowl (1st ramp) and 3rd ramp, and also stabilator (horizontal tailplane) position.
2-73
[x]T = [HP speed, LP speed, turbine metal temp, 5 component deviation parameters];
[u]T = [control inputs] (as measured);
[y]T = [pressures, temperatures, speeds] (as measured).
Modeling derives the engine outputs required for the optimal solution. This requires formulation of the Compact
Propulsion System Model (CPSM) which combines the Compact Engine Model (CEM) and the Compact Inlet Model
(CIM).
The CEM consists of a linear steady-state variable model (SSVM). This is essentially a matrix of exchange rates linking
the input vector comprising: measured fuel flows, turbine exit pressure, compressor variable position, and component
deviations obtained from the estimator, to the response vector comprising: speeds, nozzle area, pressures, temperatures
and airflows. Non-linear parameters e.g. reheat parameters, thrusts, effective nozzle area, ram and nozzle drag, are
calculated via analytical expressions and from data tables using outputs from the SSVM and engine measurements.
These are linearized in real-time with respect to the input vector. These partials are used in the optimization process.
The CIM consists of equations, which relate inlet geometry and calculated flow to observed inlet conditions.
The airframe model is only invoked at supersonic conditions (the geometry is near optimum for subsonic), and is used to
derive compensation in pitch control (via the stabilator) arising from changes in inlet geometry. The model comprises
tabulated relationships between pitching moments and drag effects for the cowl and stabilator positions.
Optimization is performed using linear programming to determine the local optimum in terms of control inputs and
output variables, within the accuracy of the models and defined constraints. Some iteration between the optimization and
modeling functions is necessary. The primary constraints are set by the particular mode of operation, such as minimum
fuel burn, and engine hardware, such as limits on actuator travel. The basis of the optimization is the Propulsion System
Matrix (PSM) which is formed by combining the linear (steady-state) models from the CEM and CIM.
CITED EXAMPLE
49. Gilyard and Orme, Performance-Seeking Control: Program Overview and Future Directions NASA Dryden
Research Facility ; 1993
50. Orme and Schkolnik Flight Assessment of the Onboard Propulsion System Model for the Performance Seeking
Control Algorithm on the F-15 Aircraft NASA Dryden Research Facility; 1995
The subsonic test program was conducted in 1990-91 and was followed by the supersonic program during 1992. Only
one engine was subject to the PSC algorithm. The ability of PSC to compensate for deterioration was demonstrated by
using a degraded engine in the subsonic test phase. Most of the testing was carried out under cruise conditions. Trim
update rate was slower in supersonic phase owing to the increased computing load for the inlet model.
The claims for each of the three modes can be summarized as:
2-74
5.5.2.
Reheat or afterburning is used in military gas-turbine engines as a lightweight means to boost thrust by increasing the
exhaust gas temperature and consequently its velocity. The reheat system (fuel flow and nozzle area) has direct and
indirect influences on the fan operating point. The amount of fuel burned in the reheat system is critical to safe operation
of the engine. If too much is burned, there is a risk of fan surge. If there is too little, there is a risk of combustion
instability and wasted performance through low fan pressure ratio.
Two reheat control approaches have been taken on engines: closed-loop where the fuel is controlled to achieve a
prescribed fan operating point, or open-loop where fuel is metered in accordance with the measured final nozzle area
(which can also be controlled using open-loops). Both methods have advantages and disadvantages, this paper deals with
the challenges in the open-loop method, and is linked to work carried out in the development of the Eurojet EJ200
turbofan for Eurofighter 2000 (Typhoon). Derivation and refinement of control laws is conveniently carried out using an
engine simulation.
In order to conserve fan safety (freedom from surge), it is necessary to build-in safety margins in the reheat fuel control
laws, commensurate with the accuracy of the simulation. Clearly, a greater confidence in the model will allow
minimization of these margins, which can lead to better performance levels. Although all elements of the engine model
contribute to the definition of the fan operating point, for reasons of clarity and brevity, only the building and the
calibration of the nozzle and reheat components of the whole-engine model are covered here.
2-75
The simulation model is used to analyze dry and reheated test data from a variety of flight conditions. The process used
is known as ANSYN (Analysis by Synthesis), which is a form of model calibration. The models thermodynamic
assumptions are automatically varied to match selected measured parameters. There are several ways of approaching
this, each of which may have different inherent assumptions. Examples include placing more emphasis on certain
measurements, or placing greater dependence on the quality of modeling for a certain component. For analyzing reheat
performance, P7 and P161 are used as anchors in the ANSYN process. In reconciling test data with the model, care
must be taken to reasonably account for uncertainties. In this example, bypass stream pressure loss coefficient is used as
a dump.
POTENTIAL BENEFITS
An accurate model is essential to derive open-loop reheat schedules. Schedules must be set to reflect the uncertainty in
the model. Consequently the better the model, the smaller are the margins that are required to guarantee thrust for the
worst engine and to guarantee a safe fan operating point. Using an accurate model to derive schedules can decrease
expensive altitude testing time, and help reduce program time-scales.
CITED EXAMPLE
51. Kurzke, J. & C. Riegler, A Mixed Flow Turbofan Afterburner for the Definition of Reheat Fuel Control Laws,
May 1998, -- Design Principles and Methods for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999.
The main idea of Analysis by Synthesis (ANSYN) is to match an engine simulation model to test data automatically.
Scaling factors are applied to the component models, to close the gap between the calculated and the measured
component performance. For example, corrected spool speed and specific work give the operating point of a compressor
on its map. Both can be derived from measured data. The corrected mass flow and the efficiency in the model are
calculated using the compressor map and corrections for Reynolds Number effects. The scaling factors then result as the
particular ratio of measure to calculated value. Illustrated in Figure 69 is an example of the scaling factors for the HPC
efficiency versus corrected HPC spool speed resulting from the analysis of many scans. The small scatter of the scaling
factor shows that the HPC efficiency model accounts for the most important physical phenomena and may be used for
high quality performance predictions over a wide range of engine ratings and flight conditions.
Figure 69 - Component efficiency scaling factor, found from engine test analysis
The control schedules for a mixed flow turbofan with an open-loop afterburner fuel control system, can only be derived
from a performance model of the engine. Any inaccuracy of the simulation must be covered by safety margins in the
afterburner fuel flow schedules. Poorly designed fuel schedules can cause a loss of thrust and may require additional
turbine inlet temperature clearance for compensation. The key to a good simulation is a precise model of the convergentdivergent nozzle over the full range of pressure ratios encountered during flight. The nozzle flow characteristic has a big
impact on the fan surge margin, while understanding the thrust characteristic is a prerequisite for optimizing the
performance of the engine.
2-76
5.6.
REFERENCES
52. Kurzke, J., Advanced User-Friendly Gas Turbine Performance Calculations on a Personal Computer, ASME
Paper # 95-GT-147, June 1995.
53. Visser, W. P. J., and M. J. Broomhead, GSP, A Generic Object-Oriented Gas Turbine Simulation Environment,
ASME Paper #2000-GT-0002, May 2000.
54. Chappell, M. A. and P. W. McLaughlin, Approach to Modeling Continuous Turbine Engine Operation from
Startup to Shutdown, Journal of Propulsion and Power, Vol. 9, Number 3, May-June 1994, Pages 466-471.
55. Malloy, D. J., Chappell, M. A., and C. Biegl, Real-Time Fault Identification for Development Turbine Engine
Testing, ASME Paper # 97-GT-141, Presented at ASME International Gas Turbine Institutes Turbo Expo,
Orlando, FL, June 1997.
56. Chappell, M. A., and R. McKamey, Adjusting Turbine Engine Transient Performance for the Effects of
Environmental Variances, AIAA Paper # 90-2501, July 1990.
57. Khalid, S. J., Role of Dynamic Simulation in Fighter Engine Design and Development, Journal of Propulsion and
Power, Vol. 8, No. 1, January-February 1992, pp. 219-226.
58. Khalid, S. J. and R. T. Legore, Enhancing Fighter Engine Airstarting Capability, International Journal of Turbo
and Jet Engines, Vol. 10, 1993, pp. 225-233.
59. Lakshminarasimha, A. N., Modeling and Analysis of Gas Turbine Performance Deterioration, ASME #92-GT395, June 1992.
60. Singh, D., et al, Simulation of Performance Deterioration in Eroded Compressors, ASME Paper #96-GT-422,
June 1996.
61. Kurzke, J. Some Applications of the Monte Carlo Method to Gas Turbine Performance Simulations, ASME Paper
# 97-GT-48.
62. Hale, A. A., and M. W. Davis, Jr., Dynamic Turbine Engine Compressor Code, DYNTECC Theory and
Capabilities, AIAA Paper -92-3190, June 1992.
63. Davis, M. W., Jr., et al, Euler Modeling Techniques for the Investigation of Unsteady Dynamic Compression
System Behavior, Loss Mechanisms and Unsteady Flows in Turbomachines, AGARD-CP-571, January 1996.
64. Hale, A. A., and W. F. OBrien, A Three-Dimensional Turbine Engine Analysis Compressor Code (TEACC) for
Steady-State Inlet Distortion, Journal of Turbomachinery, Vol. 120, July 1998, pp. 422-430.
65. Hale, A. A., Chalk, J., Klepper, J., and K. Kneile, Turbine Engine Analysis Compressor Code: TEACC Part II:
Multi-Stage Compressors and Inlet Distortion, AIAA Paper # 99-3214, Presented at the 17th AIAA Applied
Aerodynamics Conference, June 1999.
66. Gong, Y., A Computational Model For Rotating Stall and Inlet Distortions in Multistage Compressors, GTL
Report #230, March 1999. Gas Turbine Laboratory, Massachusetts Institute of Technology, Cambridge, MA.
67. Longley, J. P., et al, Effects of Rotating Inlet Distortion on Multistage Compressor Stability, ASME Paper #94GT-220, June 1994.
68. Garrard, G. D ATEC: The Aerodynamic Turbine Engine Code For the Analysis of Transient and Dynamic Gas
Turbine Engine System OperationsPart 1: Model Development, ASME Paper #96-GT-193, June 1996.
69. Garrard, G. D ATEC: The Aerodynamic Turbine Engine Code For the Analysis of Transient and Dynamic Gas
Turbine Engine System OperationsPart 2: Numerical Simulations, ASME Paper #96-GT-194, June 1996.
70. Rodriguez, C. G., and W. F. OBrien, Unsteady, Finite-Rate Model for Application in the Design of Complete
Gas-Turbine Combustor Configurations, Design Principles and Methods for Aircraft Gas Turbine Engines, RTOMP-8, February 1999.
71. Costura, D. M., et al, A Model for Combustor Dynamics for Inclusion in a Dynamic Gas Turbine Simulation
Code, AIAA Paper # 97-3336, Presented at the 33rd AIAA/ASME/SAE/ASEE Joint Propulsion Conference, July 69, 1997, Seattle, WA.
72. Garrard, G. D., Davis, M. W., Jr., Wehofer, S., and G. Cole, A One-Dimensional Time Dependent Inlet/Engine
Numerical Simulation for Aircraft Propulsion Systems, ASME Paper # 97-GT-333, June 1997.
73. Tinga, T., et al, Integrated Lifing Analysis Tool for Gas Turbine Components, ASME Paper # 2000-GT-646, May
2000.
74. Nazir, J., Couch, R., and M. Davis, An Approach for the Development of an Aerodynamic-Structural Interaction
Numerical Simulation for Aeropropulsion Systems ASME Paper # 96-GT-480, June 1996.
75. Haykin, T., and S. N. B. Murthy, Transient Engine Performance with Water Ingestion, Journal of Propulsion, Vol.
4, No. 1, Jan-Feb 1988, pp. 81-88.
76. Ludorf, R. K., et al, Stage Rematching as a Result of Droplet Evaporation in a Compressor, ASME Paper #95GT-194, Presented at the IGTI Turbo Expo in Houston Texas, June 1995.
77. Abdel-Fattah, A. M., Response of a Turbofan Engine Compression System to Disturbed Inlet Conditions, Journal
of Turbomachinery, Vol. 119, No. 4, October 1997.
78. Gilyard and Orme, Performance-Seeking Control: Program Overview and Future Directions NASA Dryden
Research Facility; 1993
79. Orme and Schkolnik, Flight Assessment of the Onboard Propulsion System Model for the Performance Seeking
Control Algorithm on the F-15 Aircraft NASA Dryden Research Facility; 1995
80. Kurzke, J. & C. Riegler, A Mixed Flow Turbofan Afterburner for the Definition of Reheat Fuel Control Laws,
May 1998, - Design Principles and Methods for Aircraft Gas Turbine Engines, RTO-MP-8, February 1999.
2-77
6.1.
The user environment consists of various activities. Each of them provides an opportunity for the use of EHM (Engine
Health Monitoring) methods but each, at the same time, has different requirements. These activities can be categorized
as flight line, engine repair shop and engine test cell, and are briefly discussed in the following.
The flight line is primarily the domain of the aircraft and engine technicians. It is organized around the concept of a
quick fix. Engine designers make every effort to make line replaceable units (LRUs) accessible and easy to change. In
this environment, the technician is compelled to work with some combination of pilot reported problems and whatever
recorded limit-exceedance or alarm information is available. The latter implies that the aircraft has been fitted with a
flight recorder, which is usually the case for modern aircraft.
Time is of the essence at the flight line. The technician typically has only one or two hours in which to effect a repair
and put the airplane back into service. Based on the available information, backed up with on-wing inspections, the
technician makes a preliminary diagnosis and conducts whatever tests he deems necessary, to substantiate his
hypothesis. It is also common practice to begin simply by changing a suspect LRU. On-wing testing follows this and, if
successful, the aircraft is released to fly. If unsuccessful, the technician may elect to replace another LRU and try again,
or he may elect to replace the entire engine and send it to the engine repair shop for more involved repair. In either case,
the flight line actions trigger demands on spares and consume engine life during on-wing tests.
From an EHM requirement viewpoint, a number of things are clear:
Unless the problem is quite obvious, the flight line repair is a process with very poor diagnosis success rates.
Diagnosis must be quick and very convenient for the technician. The technician is not necessarily an engineer.
Diagnosis must focus on those problems for which exchanging an LRU will effect a repair or else clearly indicate the
need for engine removal.
The engine repair-shop is, in effect, the nearest available repair area that provides service to the flight line. This service
consists of stripping the engine and replacing life expired parts (scheduled repairs) and dealing with all engines removed
from the aircraft because a quick fix could not be effected at the flight line (unscheduled repairs).
The repair-shop management is judged on its ability to complete repairs in a timely manner. This, in turn, is quite
dependent on a timely flow of the right spare parts at the right time. Bearing in mind the high cost of spare parts, this
translates to just in time spares management. Clearly, the key to success in this endeavor is the accurate prediction of
the workload and the type of work in the engine shop.
The unscheduled engine repairs sometimes represent 50% or more of the engine repair-shop workload. Without
adequate means of assessment and prediction, this workload will appear quite suddenly and will require large spares
inventories and many personnel. It is, therefore, evident that some means of providing early warning of the arrival of
unscheduled engines would have a very substantial impact on the entire range of concerns of the engine repair shop.
From an EHM requirement viewpoint, the engine repair-shop is quite different from the flight line:
The time frames of concern are substantially longer, making parameter trending very useful.
A more sophisticated level of diagnosis is possible with the support of engineering personnel.
Proper data management offers a feed-forward information loop to the flight line.
The essence of EHM at the engine repair shop is proper planning of resources.
The engine test cell is an expensive facility provided for the primary purpose of ensuring safety. The engine undergoes a
go-no go test, which checks the engine in accordance with the repair level. Since the express purpose of the test is to
ensure flight safety, two major aspects of engine operation are examined. First, the mechanical integrity of the engine is
2-78
established. This may include possible hydraulic and oil leaks, loose bolts and vibration levels. Secondary, the static and
dynamic performance of the engine is examined. These tests consist of establishing throttle settings and maneuvers and
then recording speeds, pressures, temperatures etc for purposes of comparison with acceptable standards of performance.
During the course of the tests, allowable control adjustments are made and the engine is re-tested. If the engine passes all
tests, it is declared ready for installation. If, however, it fails and subsequent adjustments do not cause it to fall within
acceptable limits, it must be partially or completely stripped and reworked. The engine test itself is quite straightforward.
Analysis of the data follows the normal practice recommended by the manufacturer. However, the diagnosis used in
current test cells is little better than that available on the flight line. The requirement for diagnostic techniques in the test
cell is self-evident:
Test cells provide a larger complement of engine measurements than most flight recorders.
Diagnosis requirements are similar to those for the flight line, progressing from simple adjustments to the
replacement of LRUs while the engine is on test.
Test cell diagnosis can and should progress to fault identification deep within the engine. It can help to eliminate
unnecessary testing and direct attention to repairs.
Data obtained during engine testing will provide quantitative assessment of the available engine margins. These data
are important in establishing a first estimate of when the engine will next need work, and for what reason.
6.2.
2-79
Diagnosis of a mechanical condition is the grasp of knowledge of the condition of parts of a machine, from information
coming to the engine exterior, without dismantling the machine or getting direct access to the parts. The field of
engineering science covering the techniques for achieving a technical diagnosis is called diagnostics.
The aim of diagnostics is to detect the presence and identify the kind of faults appearing in a machine. But what do we
mean by fault? A fault is a condition of a machine linked to a change in the form of its parts, or in its way of operation,
from what the machine was originally designed for and was achieved during its initial operation. A fault manifests itself
in the following ways:
Change of the geometrical characteristics of parts of a machine. Such a change is inevitably linked to all commonly
experienced faults, as for example when a part is broken or deformed.
Change of the integrity of the material of engine parts. Typically, such a problem is the occurrence of cracks inside
the material, which are not associated with any geometrical change but can nevertheless result in catastrophic
consequences.
Diagnose is a Greek word, literally translating to know through, and actually meaning gaining knowledge about
something that is not obvious but lies behind some barrier preventing direct access. Diagnosis is the outcome of the mental
process of diagnosing. These terms have been used extensively in medicine, because humans are a typical case of systems
which do not allow direct access to their interior. If one has to conclude about the condition of parts inside the human body,
one has to do so from observations from the outside. The doctor has therefore to diagnose the causes of an illness from the
external symptoms, or at least by observations which are made without getting direct access to the interior of the body.
2-80
Change of operating condition and entry to regions of unsafe operation. Although this is a situation in which there is
no geometry or material problem, such problems may quickly appear as a result of bad operation. Stalled operation
of a gas turbine is a situation, which can be characterized as faulty operation.
An engine free of faults is characterized as a healthy engine. An engine is usually healthy when it is initially
manufactured.
A distinction must be made between Machine Diagnostics and Inspection:
Machine diagnostics is a procedure applied to a machine in operation and does not require that the machine is either
stopped or disassembled.
Inspection, in contrast, refers to a procedure that involves direct access to the item of interest. Usually inspection
requires either stopping machine operation or even dismantling it. Obviously, external parts of a machine can be
inspected while it is in operation. While in order to inspect its internal parts, access to its interior must be gained,
requiring engine stoppage in almost all cases.
The fact that diagnostic techniques provide information from a running engine is important for two particular reasons:
Information is gathered while the engine is in operation. This is vital for engines in the process industries or energy
production, as they must run without interruption for long periods.
Incipient failures may be detected while running. This will lead to taking action necessary to prevent a catastrophic
failure, which might follow.
2.
3.
4.
5.
Get the basic mechanical and operational data for an engine. The full engine layout and details, and the values of
various parameters of operation, for example parameters describing vibration, performance, and mechanical speed,
for a healthy state, must be known. This information describes the reference condition of an engine.
Measure the values of the variables that are necessary for describing the operating condition of an engine. The
measured variables must be sufficient for producing all the information needed for diagnosis.
Reduce the measured values to others having diagnostic value. For this task, an appropriate set of programs, which
includes data reduction but modeling ones as well, must exist.
Derive diagnostic information by combining the reduced values. This information can be for example, an array of
values of differences from baseline, a parameter of a best fitting technique, a point in a feature space or other,
according to the method which is followed.
Compare the diagnostic information to existing knowledge on failures and their symptoms, and conclude about the
existence of a failure. Application of this step requires previous experience, in the form of a database, and decision
rules, of the symptoms of the failures that can be detected.
These stages can be followed when some kind of measurement data can be obtained from an engine. Depending on the
kind of data, the techniques for the individual stages differ, as do the diagnostic techniques. An example is vibration
diagnostics; namely diagnostics based on vibration measurement data.
Techniques based on data from measurements of aerothermodynamic quantities and engine performance parameters are
known as performance diagnostics techniques. Engine performance models support such techniques in different ways,
as will be discussed in detail in the following sections.
2-81
and fault signatures, while they provide information supporting the set-up of a diagnostic procedure. A brief description
of these tasks follows, while the role of performance models will be illustrated further when various diagnostic
techniques are presented.
6.2.1.4.
DERIVATION OF BASELINE INFORMATION
First, important pieces of information that can be derived by a performance model are the values of engine performance
variables and parameters for the entire operating range of a healthy engine (baseline values). This information serves
the following purposes:
It provides the baseline values for measured quantities or diagnostic parameters. It establishes therefore the reference
for the diagnostics.
Variations of diagnostic parameters for reasons other than faults can be studied. A typical example is the
establishment of the expected variation of such parameters for different operating conditions. A change in the
operating condition causes variations, which can lead to false conclusions if attributed to faults.
6.2.2.4.
DERIVATION OF KNOWLEDGE BASES
Fault signature is a general term, referring to the differentiation caused by the presence of a fault. A particular fault
results in changes of parameters in a certain way. The set of these changes is the signature of the fault. Fault signatures
must be available in order to identify the faults. The existence of a database of fault signatures is an essential part of a
diagnostic system. Building up such a database is a difficult task. The main alternative approaches that can be used are:
a)
b)
c)
Approach a) may prove too costly, since occurrence of some of the failures can be catastrophic and engine users would
have preferred it not to happen. On the other hand, if failures occur in an uncontrolled manner it is not certain that all the
necessary information can be collected for constituting the signatures. Finally, it is possible that long periods may be
needed before a reliable knowledge base, with a reasonable coverage of fault cases, is assembled. These drawbacks can
be overcome by setting up experiments, as in approach b). Faults representative of realistic cases expected in field
operation are implanted in engines and their influence on engine operation is studied. While they provide information
that is useful for particular engines and faults with specific characteristics and severity, such experiments are still very
costly.
An efficient alternative way of producing fault signatures is via the application of approach c); namely by computation.
For this approach, the quantities employed for diagnosis are obtained by modeling both healthy and faulty operation,
which are combined for the derivation of the signatures. The computed quantities and corresponding signatures should
be obtained in direct correspondence to the actual measured quantities on an engine and the subsequent processing
applied to the experimental data.
6.2.3.4.
DERIVATION OF BACKGROUND INFORMATION
Performance models can be further used to derive information useful for setting up a diagnostic system or data to
support diagnostic techniques. An example of the former type of application is the assistance they can provide, for the
selection of quantities to be measured and the measuring locations, when setting up a diagnostic system. The quantities have
to be selected in such a way that they are sensitive to the presence of faults. The same holds for the location at which the
measurement is effected. Certain locations are more suitable than others are. Calculation of influence coefficients for linear
gas-path analysis methods is an application that supports diagnostic techniques (see the following section). Models are also
used for measurement evaluation. Some typical applications are the checking of the consistency of test data and the
generation of virtual measurements for sensor validation.
2-82
formulation of the above correspondence is expressed by the well-known linear equation, which is valid at each
particular operating point.
y = C x x
(1)
(2)
The vector u determines the conditions that define the operating point (ambient conditions, load, control settings etc).
Linearization of equation (2) with respect to x leads to the equation:
y = C(u) x
(3)
2-83
The conventional formulation of the GPA (equation 1) is simply the application of the equation (3) at a particular
operating point. If we apply equation (1) at k discrete operating points we can write with the assumption of unchanged
x:
y1 C1
M = M x
y n C n
(4)
where
Ci = C(ui), i = 1, k
We see that we now have k x n measurements while the unknown parameters remain the same. If the rank of the
augmented coefficient matrix (now containing k x n rows) is greater or equal to m, Equation (4) can then be used to
determine x. In general, this condition is fulfilled for the case of a jet engine, because of the non-linear dependence of
the influence coefficients on the engine operating point. This will be shown later. By choosing therefore a suitable
number of discrete operating conditions, we can increase the number of equations in (4), allowing thus the determination
of the desired m elements of x even though n < m (parameters to be determined are more than the measured variables).
The final equation is:
x=Mk-1 CTiRiyI
(5)
i=1
where M is the so-called information Matrix k
M k = CiTRi-1Ci
(6)
i=1
and Ri is the typical covariance matrix of the measured variables.
x1
x2
Performance
xi
Model
y1
y2
yj
yM
xN
x1
x2
xI+ xi
Performance
y1
y2
yj+ yj
Model
yM
xN
y j
xi
Using this procedure, we can calculate the elements of the influence coefficient matrix at different operating points. We
2-84
also have the possibility of producing and adding the information needed by equation (5) for fault diagnosis. At this
point, we must emphasize the requirement for high accuracy in the calculation procedure. In fact, the corresponding
performance model must be able to predict accurately the engine behavior over the whole operating range, in order to
compute with acceptable accuracy the influence coefficient elements. It should be mentioned that generalized engine models,
especially when built by the engine user, do not always fulfill this requirement. In order to overcome these disadvantages the
models should be customized to the particular engine, as discussed in the following section.
(1)
This relation is usually derived through the conservation laws for mass, energy, and momentum, and from existing
experience in component operation. It can be an analytic relation, possibly including empirical constants (e.g. duct
pressure loss), or a set of curves (e.g. compressor map). Compatibility of component functioning imposes matching
conditions, as for example power balance and speed between turbines and compressors. A set of simultaneous equations,
which have to be satisfied by the fluid parameters, is thus formed. Solution of this system for a single operating point
gives the full cycle details. The solution to the system of equations is obtained numerically, since they are highly nonlinear.
The approach usually followed is first guess the values of some suitably chosen variables vi, and then explicitly solve the
equations, giving the full set of parameters. Error terms are then formed from the differences of quantities calculated
from different equations:
ei=|Pi1-Pi2|
(2)
where Pi1,Pi2 are the values of a parameter Pi calculated by two different equations.
The data that are given as input to a model can be divided into two types:
I - Data related to the particular operating condition (for example ambient conditions, fuel calorific value, speed and
load), which define the engine operating point.
II - Data related to the performance of the engine components, corresponding to equation (1), as for example, the
compressor and turbine performance maps.
For a given set of type-II data, all cycle details and performance parameters are uniquely defined for a choice of
operating condition, through a set of type-I data. This means that, once the component data are specified, for each
operating point we have a unique set of calculated parameters. If these data do not represent exactly component
operation of a particular engine, then the predictions will differ from actual measured values. (It is noted that type-II data
are usually not available to the user, while for an engine manufacturer, the available data usually represent an average
engine.) It is useful to understand why an available set of maps may not lead to accurate predictions for a particular
engine. The maps may differ from the ones of a specific engine depending on their origin:
Maps measured on isolated components - such maps may differ from actual on engine maps, due to interactions
with other components or different operating environments (e.g. heat transfer effects), inlet non-uniformity of
pressure and temperature, etc.
Maps predicted by computer programs - such maps may differ from real maps due to insufficient modeling
capabilities or lack of the necessary physical data.
2-85
Maps measured on a different engine - they may be different because of engine-to-engine dissimilarities. Such maps
are sometimes difficult to measure because of practical difficulties, such as high temperatures of the hot components.
6.2.2.7.
MODEL ADAPTATION
The performance maps of each engine component are derived from the functional relations between its characteristic
performance parameters, of the form of equation (1). These relations can be in an analytic form or in the form of a chart.
If a particular parameter has a value Xref on the reference map and a value Xact on the actual on engine map, then the
correspondence between the two can be expressed by means of a modification factor MF defined as follows:
MF =
X act
X ref
(3)
It must be noted here that modification factors can be introduced either as scalars, multiplying the reference value as
above, or as adders, namely values that are added to the reference value. Knowledge of the reference performance map
and the values of MF offers the possibility of reproducing the actual maps. Care must be taken however in the way that
these factors are introduced. They must be consistent with existing representation of the maps and the set of equations
used. For example, in the case of the compressor, we can employ the following definitions:
MF1 =
Q
Q ref
MF 2 =
ref
(4)
The value given to any component parameter used by the model is thus introduced as a product:
Xact = MFXref
(5)
We see that a value of a component parameter is now defined by means of two numbers: its reference value and the
value of the corresponding modification factor. When the reference values are available, lets see how the values of the
modification factors can be determined. Having chosen a set of modification factors values MFi, it is possible to define
an optimization problem, which allows the determination of their values, needed for adapting the maps.
In a straight engine model the solution of the equations for a particular operating point, proceeds by guessing values for
some variables vi and calculating error terms ei, equation (2). Besides the parameters necessary to form the error-term,
values of all cycle details are calculated. They include the values of the quantities measured along the gas path during
experiments. For any measured quantity Ym there is a corresponding calculated Yc. We then form a cost function FC as
follows:
FC =
M
i =1
a ie i +
2
N
i =1
b i (Y ci - Y mi ) 2
(6)
where M is the number of error terms of the model, N is the total number of measurements and ai and bi are weight
coefficients depending on both measurement and desired model accuracy. If modified maps are introduced into the
model, by means of a particular set of values for MFi, the value of FC obtained, will be a function of the original guess
for vi and Mfi:
FC = FC(v1,v2,,vM;MF1,MF2,,MFN) (7)
The set of independent variables that minimizes the value of this function to zero satisfies the matching conditions for
the engine, while it ensures that the measured and predicted quantities are the same. A set of values for the modification
factors is produced for the particular operating point. We get thus a set leading to an optimal reproduction of measured
quantities, through the simulation model. The flow chart of the procedure is depicted in Figure 71. Covering the entire
operating range of the engine will give the full set of MFs needed.
An engine model which only solves for cycle, namely it zeroes the first of the sums of equation (6), will be termed a
straight model. A model that allows adaptation of component parameters to match the performance of an engine, namely
it minimizes the function FC of equation (6), will be termed an adaptive model.
The method described above can incorporate a variable number of measured quantities. The number of modification
factors that can be determined changes accordingly. This method can be characterized as internal to the model: the
adaptation procedure is embedded in the performance model itself, and the adaptation is performed simultaneously with
the solution of the engine-matching problem. Although the procedure in this form offers an effective method from a
computational point of view, it requires that the model be built for the particular engine studied.
2-86
Actual
Engine
Cost
Function FC
Is FC
minimum?
Model
Equations
End
Choose
best new
guess
Modify
maps
Reference
maps
2-87
performed from the clusters formed in a feature space (For example, a method of recognizing such faults by using
Neural Networks has been presented by Kanelopoulos et al, 1997).
ADAPTIVE
MODEL
2-88
POTENTIAL BENEFITS
The benefits of implementation of applications of this kind include all those expected to result from diagnostic
techniques in general, while some additional advantages can be provided. The use of on-line diagnostic capabilities and
through it, the on-line adaptation of the engine model, allows their incorporation into the engine controller. Performance
seeking control schemes can then be implemented and through them, engine usage can be optimized.
CITED EXAMPLE
81. Bushman M.A. , Gallops G.W., 1992, I-flight Performance Diagnostic Capability of an Adaptive Engine Model.
Paper AIAA 92-3746, 28th joint Propulsion Conference and Exhibit, AIAA/SAE/SME/ASEE. July 6-8, 1992,
Nashville, TN.
2-89
Figure 74 - Modification of map initially used, in order to adapt the model to engine data
Simulation of faults is achieved by introducing deviations in component parameters (as caused by the presence of a
fault) and producing the resulting deviations in measured quantities (fault signatures). Of course, such a possibility exists
also with a non-adapted model, but the improved reliability of the adaptive model gives more accurate results, as
demonstrated in Figure 75. This figure shows that signatures predicted by the adaptive model are much closer to those of
the manufacturers than those predicted by the core model.
6.3.
REFERENCES
82. Doel D., 1992, TEMPER-A Gas-Path Analysis Tool for Commercial Jet Engines, ASME paper 92-GT-315
83. Doel D., 1993, An Assessment of Weighted-Least-Squares based gas path Analysis, ASME paper 93-GT-119
84. Kanelopoulos K., A. Stamatis, K. Mathioudakis: Incorporating Neural Networks into Gas Turbine Performance
Diagnostics, paper ASME 97-GT-035, 42nd ASME International Gas Turbine and Aeroengine Congress and
Exposition, June 2-5 1997, Orlando, Florida USA.
2-90
85. Lambiris B., K.Mathioudakis, A.Stamatis, K.Papailiou, 1994 Adaptive Modeling of Jet Engine Performance With
Application to Condition Monitoring. Journal of Propulsion and Power, AIAA, Vol 10, No 6, Nov-Dec 1994, pp
890-896
86. Lunderstadt,R., Fiedler,K., 1988, Gas Path Modelling, Diagnosis and Sensor Fault Detection. Paper 34, AGARDCP-448, Oct 1988.
87. Smetana, F.O.: Turbojet Engine Gas Path Analysis - A REVIEW. AGARD CP 165, Diagnostic and Engine
Condition Monitoring, 1974.
88. Smith, C., De Hoff, G.L., Developments in Performance Monitoring and Diagnostics in Aircraft Turbine Engines.
SAE 821400, 1982.
89. Stamatis A., Mathioudakis K., Berios G.,Papailiou K., 1991, Jet Engine Fault Detection with Differential Gas Path
Analysis at Discrete Operating Points. Journal of Propulsion and Power, Vol. 7, No 6, Nov./Dec. 1991, pp. 10431048.
90. Stamatis A., Mathioudakis K., Papailiou K.D., 1992: Optimal Measurements and Health Indices Selection for Gas
Turbine Performance Status and Fault Diagnosis. Journal of Engineering for Gas Turbine and Power, ASME,
vol.114, No 2, April 1992, pp.209-216.
91. Stamatis A., Papailiou K.D.,1988, Discrete Operating Conditions Gas Path Analysisv. AGARD CP 448, paper 33,
1988.
92. Stamatis,A., Mathioudakis,K., Papailiou,K.D., 1990a, Adaptive Simulation of Gas Turbine Performance. Journal
of Gas Turbine and Power, ASME, vol 112,No2, April 1990, pp. 168-175.
93. Stamatis,A., Mathioudakis,K., Smith,M., Papailiou,K.D., 1990b, Gas Turbine Component Fault Identification by
Means of Adaptive Performance Modelling. paper ASME 90-GT-376, 35 Gas Turbine and Aeroengine Congress,
June 11-14,1990.
94. Urban L.A, Volponi A.J., 1992, Mathematical Models of Relative Engine Performance Diagnostics, Aerotech 92,
Anaheim California, October 5-8, 1992
95. Urban, L.A.: Gas Turbine Engine Parameter Interrelationships. HS VAL, Windsor Locks GK, 1969.
96. Urban,L.A., 1972, Gas Path Analysis Applied Turbine Engine Condition Monitoring, AIAA/SAE 8th joint
Propulsion Specialist Conference, New Orleans, 1972, AIAA-Paper 72-1082.
97. Volponi A., 1994, Sensor Error Compensation in Engine Performance Diagnostics, ASME paper 94-GT-58.
98. Wood,G.R.,1981, The Ruston Tornado. A 6Mw Gas Turbine for Industrial Application. Paper ASME No 81-GT171.
7.1.
ACADEMIC ROLES
Many of the models discussed in the previous examples can be and are used in educational settings. Many textbooks
such as those written by Jack Mattingly supply software (e.g. ONX, OFFX) for use in propulsion classes. Both
educational institutions and industry use GASTURB, the cycle model developed by Kurkze, for in-house training. An
extensive listing and description of the available models are provided at the end of the engine simulation description in
chapter 4.
Models in Education - Example Synoptic
7.2.
PURPOSE
To introduce the student to the following concepts:
General form of a dynamic model;
2-91
AUDIENCE
This education is aimed at those who have some experience of steady-state performance synthesis (and may have blackbox dynamic modeling experience), who want to understand some of the related issues. This particular module is part
of a series of tutorials covering the techniques used in interfacing control-systems and engine models.
FORMAT
The module uses a very simple engine model - hardly credible as a model itself but featuring clearly identifiable
elements. input, output, state and state-derivative i/o (handled as vectors by the modeling system)
A dynamic equation for shaft speed;
A basic fuelling characteristic (fuel flow for a speed);
Quasi steady-state characteristics (output parameters as a function of speed and overfuelling).
The model can be set up easily in any dynamic modeling system, e.g. MATRIXx. The student is encouraged to do this
rather than pick up the one already supplied.
Figure 76 shows the model used in the exercise.
INPUT [u]
OUTPUT [y]
steady-state calcs
P3
WFE
WFE
T6
T6, P3
dynamic equation
SS
fuelling
characteristic
XP
XP/(J.NH)
NH
NHdot
A SIMPLE DYNAMIC ENGINE MODEL (1 shaft)
STATES [x]
2-92
x( t + t ) = xt +
dx
t
dt t
This integration method may be viewed as over-simplistic for many gas-turbine systems. However, it is the principle
being explored here rather than the accuracy or stability of a particular method.
The student is encouraged to run the model and observe the effects of varying the simulation timestep. The results can be
compared with the model run using a more advanced explicit integration methods such as Runge-Kutta 4th order.
For control-system design, the model is used to generate the linear representation of the engine at a particular set
points. From examination of the model, it can be appreciated that a set of partial derivatives can be obtained by
parametric perturbation of inputs and states. Again, the student is encouraged to use the automatic linearization function
within the modeling environment to generate the linear engine model, which can then be simulated and compared with
the non-linear version.
The implicit form of Eulers method is often used in full dynamic cycle-match engine models.
xt = x( t t ) +
dx
t
dt t
The principles of implicit integration can be explored here at the simplest level. For this, the student must construct an
outer iteration loop as shown in Figure 77. Here the value of NH at t is varied until the two different derivations of Nhdot
are numerically equal.
OUTPUT [y]
INPUT [u]
steady-state calcs
P3
WFE
WFE
T6, P3
T6
dynamic equation
STATES [x]
SS
fuelling
characteristic
XP
XP/(J.NH)
NH
NHdot
A SIMPLE DYNAMIC ENGINE MODEL
STATE DERIVS [xd]
store matched NH
calc derivative
(NH_iterate - NH_last)/t
vary NH
ITERATION
ITERATIONSOLVER
SOLVER
Match NHdot
VALUE
Once these principles are understood at the simplest level, it is much easier to see how they are implemented in a more
complex modeling system. In such cases, the implementation is often obscure and tied closely to the advanced
thermodynamic calculations employed. Also, iteration is often used in the steady-state elements of the model, and its
specific role for implicit integration can be difficult to identify.
It should become clear that iteration and integration are purely means to an end and are not a real part of the engine
model. Ideally, these processes are kept transparent to the user. However, it is often found that iteration in particular is
an impenetrable subject, and a simple treatment such as this can be of value in explaining the operation of the more
complex multivariable solvers needed for proper engine models.
The exercise also serves as a hands-on introduction to the modeling tool - taking in formulation of the model, population
with data, use of associated tools (linearization & trimming) and data visualization. The student works from an intranet-
2-93
7.3.
PURPOSE
To introduce the thermodynamic interactions and dependencies within a gas turbine engine.
AUDIENCE
This education is aimed at the new-start in the performance discipline, although it can be used as refresher material for
those with more experience. It can also be used at a more basic level, for students aged around 17-18 who are
considering aeronautical engineering courses at university. Students such as these usually pass through industry on
specially arranged work-experience periods. Experience has shown that the same basic learning material can be used for
each group of learners - the difference is in the depth of discussion which is a fundamental part of the education.
The following issues are covered:
FORMAT
There are three training modules
2-94
operating throughout its flight envelope and power range in a safe, stable and efficient manner.
The exercise uses a proprietary PC-based cycle synthesis program to examine the off-design performance issues of a 2spool mixed turbofan.
The following is an extract from the electronic module notes.
Objectives:
To investigate the effects of two engine rating methods, fixed Stator Outlet Temperature (SOT) and fixed HP
compressor aerodynamic speed (NH/T) on performance parameters such as net thrust (FN), over a range of
ambient temperatures at sea level static;
Investigate the effects of fixing the SOT and HPC aerodynamic speed on performance parameters over a range of
Mach numbers at sea level, ISA;
Produce HPC and fan-operating lines (i.e. flow function vs. pressure ratio). Assess the effect of HP turbine and HP
compressor deterioration on the engine;
Investigate transient performance effects.
As before, questions are set throughout the exercise and are followed up in the discussion phase. With all of these
modules, use of the computer could have been avoided. However, the hands-on experience (the generation of the data) is
considered valuable. The figures could be merely supplied and subsequently discussed. However, the student feels no
ownership of the data in this case.
VALUE
Objective 1 from module #2 leads the student to generate several figures, depicting the relationships between various
cycle parameters. The following figure is such an example and is generated by synthesizing cycles at various values of
Stator Outlet Temperature, bypass ratio, fan and HPC pressure ratio.
1.05
S OT = 2000K
1
0.95
S OT = 1850K
S OT = 1700K
SFC
0.9
increas ing OPR
0.85
0.8
0.75
0.7
23
24
25
26
27
28
29
30
31
32
S p e cific Th rust
2-95
Reduce operating costs of supersonic transport aircraft by modifications to existing engine only.
Further reduce operating costs by selecting new powerplant.
Add an extra constraint: minimize noise at take-off (development and retrofitting costs may be ignored)
Consider the implications (fuel vs. passengers) for transpacific SST capability
The students are given various data pertaining to the aircraft and the London to New-York flight. The Breguet range
equation and other simplified relationships are also supplied. The problem is fairly unconstrained which can encourage
the student to think out-of-the-box (e.g. consider variable cycle solutions - albeit at a very simple, conceptual level)
The students, usually operating in a small group of three or four, inevitably require some guidance on approaching the
problem. However, with help they emerge from the experience with an appreciation of the preliminary design process,
and of the difficulties inherent with providing large-scale supersonic transport. Perhaps most importantly, it brings the
foundation modules to life, and provides a focus for the lessons learnt previously.
3-1
Chapter 3
Whole Engine Systems
1
2
3
Introduction................................................................................................................................................................. 3-2
1.1
General Nomenclature ........................................................................................................................................ 3-2
1.2
Thermodynamic and Gas Properties................................................................................................................... 3-2
1.3
Installation........................................................................................................................................................... 3-3
1.4
Deterioration and Manufacturing Tolerance....................................................................................................... 3-4
1.5
Minimum, Average, New and Old Engine Models ............................................................................................ 3-5
Model Types ............................................................................................................................................................... 3-5
2.1
Introduction......................................................................................................................................................... 3-5
System Modeling Issues............................................................................................................................................ 3-25
3.1
Customer Bleeds and External Loads ............................................................................................................... 3-25
3.2
Lubrication and Fuel Systems........................................................................................................................... 3-25
3.3
Thermal Management Systems......................................................................................................................... 3-25
3.4
Control System Models..................................................................................................................................... 3-26
3.5
Noise ................................................................................................................................................................. 3-31
3.6
Emissions .......................................................................................................................................................... 3-31
3.7
Iteration and Numerical Methods ..................................................................................................................... 3-32
3.8
Other Numerical Methods................................................................................................................................. 3-38
Design and Off Design Performance Computation .................................................................................................. 3-39
4.1
Different Types of Off-Design Models............................................................................................................. 3-39
4.2
0-D Models ....................................................................................................................................................... 3-40
4.3
Steady State 0-D Performance Models ............................................................................................................. 3-40
4.4
0-D modeling technique.................................................................................................................................... 3-40
4.5
Detailed 1-Dimensional (1-D) Models ............................................................................................................. 3-47
4.6
Dynamic Engine Simulations ........................................................................................................................... 3-47
4.7
Benefits of 1-D vs. 0-D Models........................................................................................................................ 3-54
4.8
High Fidelity 2-D/3-D Models.......................................................................................................................... 3-54
4.9
3-D Models ....................................................................................................................................................... 3-55
4.10 Benefits of 2-D/3-D Models v 1-D models ...................................................................................................... 3-55
4.11 Non-Component Based Parametric Models ..................................................................................................... 3-55
Validation and Calibration ........................................................................................................................................ 3-56
5.1
Performance Instrumentation............................................................................................................................ 3-56
5.2
Conventional Test Analysis Method................................................................................................................. 3-57
5.3
Accuracy ........................................................................................................................................................... 3-58
5.4
Analysis by Synthesis ....................................................................................................................................... 3-60
5.5
Summary ........................................................................................................................................................... 3-64
References ................................................................................................................................................................. 3-64
3-2
INTRODUCTION
1.1
GENERAL NOMENCLATURE
The general nomenclature used in engine modeling is covered in Aerospace Recommended Practices (ARP) and
Aerospace Standards (AS) published by the Society of Automotive Engineers (SAE). They have been created in a
cooperative effort among the developers and users of engines to simplify the exchange of models and data. ARP 755B
defines the station definition within an engine, the application to various engine types and the nomenclature for
properties and fundamental parameters. AS681 provides definitions, requirements and assumptions for a class of models
most often by engine manufacturers, between engine manufacturers and customers, often-called customer decks. ARP
1210 provides additional guidance on interface requirements for models dealing with test data, ARP1211 for status
models that have been matched to specific test data, ARP 1257 for transient models and ARP 4148 for real-time
applications. ARP 4868 and ARP 4191 are being developed to cover the needs of newer computer systems and to define
application program interface (API) standards to facilitate use with object oriented software or in an event driven
environment.
1.2
Most physics based engine models make some assumptions for calculating the state properties and energy balance of the
various fluid streams. These include converting typical flight conditions into the boundary conditions required by the
model, and the capability required to support engine component models.
1.2.1.
ATMOSPHERE DEFINITIONS
Most component-based propulsion system models use a standard atmosphere reference for defining operation
conditions. Atmospheric properties for general use are specified in SAE AS681F. The current standard atmosphere
definition is ISO 2533. This is generally consistent with the U.S. Standard Atmosphere, 1976, which was an extension
of the US Standard Atmosphere, 1966 to higher altitudes. Earlier models were generally based on Keenan and Kayes,
1945 atmosphere definitions. For non-standard atmospheric conditions the most common reference is MIL-STD-210C
which defines standards for extreme conditions such as arctic, desert and tropical days. Differences in air composition
between these various sources is small but can be noticeable when comparing absolute properties and emissions due to
the assumed percentage of CO2, as shown below:
ISO 2533
US Standard Atmosphere, 1976
NASA TP-1906
Keenan and Kayes, 1945
0.030%;
0.03140 to 0.0322%;
0.0319%;
0.0000%.
1.2.2.
THERMO PROPERTY PACKAGES
Multiple thermo representations may be used in a single simulation or be an integral part of component models.
However, it is common to use a single separable set of routines, often called a thermodynamic property package, for the
entire simulation. This is for simplicity and to ensure consistency. The package selection is typically based on the
requirements of the simulation, convenience and historical reasons. AS681 requires that the properties be consistent
with those provided in NASA TP-1906 and the associated computer code. The level of agreement required will depend
on the application. Some considerations when selecting a package are:
Speed;
Accuracy over the range of operation;
Accuracy needs of the engineering application;
Kinetics;
Heat transfer requiring transport properties;
Water vapor, multi-phase water or other constituents such as solid carbon or soot;
Non-air streams;
Alternate or non-hydrocarbon fuels.
The thermo property package should have no impact on the overall simulation accuracy. However, because property
calculations are so pervasive in engine simulations, seemingly small speed differences can be significant. When the
extra capability or range of accuracy is not required, the speed gain will often justify use of a simpler thermo package.
In a typical engine simulation, the thermo property package may account for up to 30% of the execution time.
Special application combinations are possible. Kinetics calculation can be based on curve fit properties, but these are
normally not seen in general use property packages. Figure 1 contains a comparison of the various approaches and basis
for selection based on the simulation requirements.
3-3
Constant
Property
Limited
Curve Fit
Full Range
Curve Fit
Speed
Fastest (1x)
Fast (10x)
Fast (20x)
Accuracy /
Range
Limited
Temperature
Range (+/- 500
R)
Limited
Temperatur
e Range (<
2500R)
Wide,
Limited by
Complexity,
Range
Flexibility
Low
Medium
High
Typical
Use
Simple 0D/1-D
Models, 2D/3-D
Models
Most 0-D/1-D
Models,
Some 2-D/3D Models.
Limited
Constituent
Equilibrium
Medium
(200x)
Full
Equilibrium
Full
Kinetics
Slow (500x)
Wide,
Limited by
Extreme
Disassociatio
n (< 6000 R)
High
Limited by
Physical
Model
Assumptions
Slowest
(500x
10000x)
Limited by
Physical
Model
Assumptions
Combustion
Models,
Special
Application
Cycle Models
Highest
Varies, may
be limited by
Kinetics
Options.
Combustion
Models
Figure 1 - Thermodynamic property representation effect on model capability and execution speed
1.2.3.
IMPACT OF MODELING ASSUMPTIONS
Besides the basic property representation, the assumptions made in applying these properties can affect the accuracy and
consistency of model results. Some of the key assumptions that can change the results of simulation models are:
Whether high temperature mixtures are assumed to be in equilibrium or frozen composition during expansion
processes;
How cooling flow mixing with the main flow is modeled;
Behavior with incomplete combustion products or in fuel-rich conditions.
In special circumstances, information on assumptions relative to compressibility effects, combustion product estimates,
alternate fuels or operation with other fluids (water, ice, vitiated air, nitrogen diluted test stand air, etc) may be
important. When using the model for data analysis or comparison with on-line performance data generated in test
facilities, the model assumptions must match those used in creating the data.
1.2.4.
AERO-THERMO PROCESS CALCULATIONS
Independent of the thermodynamic process package, standard procedures for calculating aero-thermo processes are
often used throughout a simulation and can affect the applicability of results. Reverse flow, supersonic conditions, mixer
and ejector processes, swirl calculations, expansion and contraction processes are examples that may affect the range of
use or accuracy of models. Models that use a thermodynamic process package may make alternate assumptions during
some of these local calculations, for speed or simplicity.
1.3
INSTALLATION
The modeling of the installation boundary conditions can be handled either directly by the engine model or by an
external application using an engine model. When included as part of the engine model, standard requirements are
defined in AS681.
1.3.1.
INLET RECOVERY
Inlet recovery is an indication of the pressure drop in the air before it enters the engine. For supersonic aircraft this
includes the losses associated with the shock and airflow capture process that may be internal or external to the physical
inlet. Basic models for both sub-sonic and supersonic operation are generally correlations with a flight condition for a
particular aircraft inlet system. MIL-STD-5007/8 contains default super-sonic recovery curves for a standard inlet and is
used for studies and comparison purposes. More detailed inlet models will calculate the pressure loss at a more detailed
level and may include the dynamic response of the inlet to changes in the engine. An example of this at a 1-D level is
the NASA LAPIN code, which allows modeling of the high frequency inlet response to both external environment and
engine inlet changes. This type of detailed model is particularly useful for examining the start/unstart process in a mixed
compression inlet or in evaluating the stall and surge initiation and recovery process in conjunction with a dynamic
engine model. More detailed 2-D and 3-D models allow calculation of the pressure and temperature variations that are
3-4
1.4
Manufacturing tolerance effects on component performance are rarely measured for production engines. However, the
uncertainty in blade and seal clearances, and coating and surface finish contribute to significant variation, even in a
brand new engine. During operation, the severity and duration of use affect these characteristics. Some performance
changes can be related directly to operating condition (over-speed, over-temperature, maneuver, water wash, and sanddust-saltwater environment) while others simply follow a general long-term trend. Deterioration is typically based on
some combination of continuous and cyclic operation measurement. Cyclic use measurements include throttle
movement (TACs in US military engines), take-off and landing, speed excursions, augmentor light-ups). Continuous
measurement can include the number of operating hours or hours of operation in a particular condition (hot time, IRP
time).
1.4.1.
UNCERTAINTY OF COMPONENT PERFORMANCE
Uncertainty in measured component performance generally depends on the level of instrumentation, instrumentation
accuracy, repeatability of the test conditions and the level of correction required to go from the measured conditions to
the conditions at which the component performance will be compared and quoted.
Uncertainty in component performance prediction prior to test is generally based on accuracy of design and analysis
tools and historical information. For derivative turbo-machinery component designs with some previous calibrated
agreement, it may be possible to quantify uncertainty. Predicting turbo-machinery aero performance and operability is
one of the most difficult problems in CFD, a subject of on-going research. For new concept aero designs, the accuracy
improvement over the historical spread has not been established.
1.4.2.
CHANGE OF COMPONENT PERFORMANCE
As the engine passes through its usage life, component performance changes in typical, if not predictable, ways. Engine
level deterioration is easily observed. Determining the underlying component performance change is difficult, and
usually impossible, without special instrumentation or analysis. In general, opening of clearances, increased leakage,
surface roughness, etc. combine to reduce component performance. Although temporary improvement measures (such
as a water wash to remove residue from compressor airfoils and recover performance) are possible, most component
deterioration occurs gradually over the engine life. It is modeled via hours of use, or other factors such as time at high
3-5
1.5
Due to production scatter and component aging, two engines of the same type, having two different usage histories, will
have different performances. However, the engine manufacturer guarantees to his customer a minimum level of
performance for a given Time-Between-Overhauls (TBO), or at least defines overhaul criteria that are periodically
checked (as part of health monitoring).
To do this, the engine designer has to build minimum, average, new and old engine performance models. This
terminology may lead to severe misunderstandings between component designers, performance engineers and
customers. The key point for a good understanding of these different models, is that the only representative and accurate
model that can be established by the engine manufacturer is the model for a new average engine. This is when all engine
components have their average production characteristics, it is sure that the resultant engine has average performances.
Conversely, the minimum performances are not necessarily obtained with an engine that has all its components at
minimum level. The word minimum is in fact relative to an engine-level pass-off criterion (thrust, fuel consumption,
TET) for global parameters and not engine components.
There are numerous combinations of component performance reductions that will cause failure of one or more of the
pass-off criteria (one with highest TET, one with smallest thrust, one with maximum SFC). A true minimum engine,
at the threshold of failure for each of the criteria, will rarely if ever exist. It is better to talk of minimum performance
model. Such models give global performance levels that any new engine will achieve. These models are obtained either
by applying deltas directly to the performances of the average engine model or applying deltas to each component of the
average model. The same issue arises for aged or old engine models. There is an infinity of deterioration types that
depend on the environment, the mission profile, and other variables that can be envisaged. Thus a general deteriorated
engine model may not be possible. It may be better to talk of an aged performance model because such models set limits
to global parameters throughout the TBO in normal operating conditions.
Although not general, deterioration models can be issued when a database of in-service engines is available. It is then
possible to derive statistical deltas on either main engine parameters, or component characteristics, as a function of the
number of running hours or cycles. These models are becoming more and more important because they are a basis for
any diagnostic models that allow for example:
Damage detection;
Engine fleet management.
A typical question is, Is it better to overhaul this engine now or later? If the answer is now, the engine recovers its
initial performance (even when not necessary for its mission, as in cold day conditions) with an economic penalty
(overhaul cost). If the answer is later, the engine keeps a higher SFC and lower general performance, which can lead to
a reduction in payload, range, or operating conditions, constituting an equally severe economic penalty.
MODEL TYPES
The purpose of this section is to describe the key issues in engine performance simulations. The focus is to make an
appropriate selection of model type and component models to meet simulation needs, and to understand the limitations
and potential of various types of simulations for potential applications.
2.1
INTRODUCTION
The different types of models used in the prediction and simulation of gas turbine engines operations can be classified
by application and capability of the model. The range of potential applications and the types of models are shown in
Figure 1. Figure 2 shows the relative requirements for some of these models in terms of accuracy, fidelity and detail.
3-6
Detail involves how much of the engine is simulated. Fidelity refers to the depth and sophistication of the analytic
representations within the model. Accuracy is the ability of the model to match tested or target values for engine and
component performance or internal conditions.
Conceptual
Detail Design
Test & Validation
Fleet Support
Accuracy
Fidelity
Detail
Low
Medium
High
High
Low
High
Medium
Low
Low
High
Low
Medium
Figure 3 - Model fidelity, accuracy and detail needs through the engine life cycle
2.1.1.
SELECTING THE APPROPRIATE MODEL TYPE
The selection of an appropriate engine model depends on the phase of both the engine development cycle and the
application. Most users of engine simulations are interested in overall engine performance levels, internal conditions that
have a direct influence on the aircraft and other internal engine conditions necessary to properly operate and maintain
the engine. The most sophisticated models focus on the details of internal components for design purposes or special
purpose analysis that is primarily of interest to engine manufacturers or government or academic researchers.
Models used in the development phase are generally on the low end of the fidelity or depth of analysis spectrum,
because of the uncertainty in the engine being modeled. However, for believability, these models may be required to
hold very precise performance agreement with existing detailed models from which the component models are derived
or will be compared. Similarly, once in production use, the primary modeling is again at the lower fidelity level but at
high levels of accuracy due to large amount of data available and the need to maintain close agreement at least the
3-7
overall engine performance level. The primary use of detailed models is by those involved in design improvements,
failure analysis or technology development.
Although model detail tends to follow the model fidelity, the level of detail required may change with the model type
shown in Figure 3. A high fidelity 3-D model is often limited to just the primary flow-path while 0-D models may
include each cooling flow circuit and the incremental changes in temperature and pressure in the internal cavities of the
engine.
2.1.2.
COMPONENT CHARACTERISTICS REQUIRED BY ENGINE MODELS
When a component-based engine model is used, the model capability and fidelity is generally limited by the component
model characteristics available for the engine. The simulation will often be a trade-off between what is desired for
system model capability and fidelity and what can be achieved with the available component models.
Component model characteristics important for consideration are:
Accuracy - Does it match reality for overall component-performance?
Detail - Does it provide all parts of the model at the required level? (Cooling Circuits, Seal Leakage, Purge Flows,
Localized Transient Effects,)
Fidelity - Does it model all of the pieces of model at the required revel of detail? (Average, Radial Profile, 3-D,
Boundary Layer)
Functionality - Does it have the required capability? (In-Stall Performance, Variable Stators and Bleed, Low Water
Vapor, Vitiated Air, Rain or Ice, Clearance and Heat Transfer Effects)
Complexity - Execution speed, data required, model expertise required, computer limitations;
Range of Operation - Off-design, low and high speed;
Compatibility with System Model and Other Component Models - Installation and use in system simulation,
development overhead, consistency with accuracy and functionality of other models.
Figure 4 and Figure 5 summarize the requirements for component models, for effective use in various types of
simulations. Figure 6 identifies component model requirements for special or secondary effects within these component
models. This section contains a brief summary for each of the primary component models used in turbine engine
simulations.
Functionality
Complexity
Range of Operation
Compatibility with
External Models
High Fidelity
Medium Fidelity
Low Fidelity
Limited
High
Limited
Low
Limited to High
Medium to High
Limited to Full Range
Low to High
Limited to High
Low
Limited to Full Range
High
Conceptual
Detail
Design
Test and
Validation
Fleet
Support
Functionality
Limited
High
Limited
Complexity
Low to
Medium
Limited to Full
Range
High
Medium to
High
Full Range
Limited to
Medium
Low
Limited
Limited
Low
Low
Range of
Operation
Consistency with
External Models
High
Low
Figure 5 - Table of component model requirements and characteristics by engine life cycle application
3-8
Effects
Fan
Booster
Compressor Turbine
X
X
Distortion Effects
Deterioration Effects
COMPRESSORS
2.1.3.1
MODELING
Knowing the downstream conditions (T, P) and rotational speed, and eventually variable guide vane positions in case of
a variable geometry compressor, compressor modeling predicts the outlet conditions (W, T, P) and surge margin.
2.1.3.2
FIXED GEOMETRY COMPRESSOR
Compressor operation is characterized by the velocity triangles, on which the enthalpy drop (thus pressure ratio and
efficiency) and the mass flow depend.
Dimensional analysis shows that a velocity triangle similitude must be based on a Mach numbers similitude.
For this reason, the classical approach to describe the compressor operation uses the two following reduced parameters:
W2R2 =
W2 T2 101.325
P2
288.15
Inlet corrected mass flow, which depends only on axial Mach number at compressor inlet (in non-viscous flow).
XNR2 =
XN 288.15
T
Corrected speed is proportional to the tangential velocity and approximates the blade tip Mach number.
When XNR2, W2R2 are fixed, the compressor state is determined and the pressure ratio P3Q2 and the efficiency E23
are known.
This way, two maps may characterize a fixed geometry compressor:
P3Q2
F(XNR2, W2R2).
E23
F(XNR2, P3Q2).
Very often, instead of using E23 as reduced parameter, the corrected enthalpy rise H3D2/T3 is employed.
An important feature of compressor modeling is the surge line, which sets an upper limit to the pressure ratio at a given
corrected speed because of aerodynamic instability. This surge line can be described within the preceding maps as one
of their limits or separately. Showing the surge line separately allows secondary effects such as Reynolds number,
distortion or production-scatter to be taken into account, without modifying the maps.
2.1.3.3
IMPORTANCE OF COMPRESSOR MAPS IN 0-D MODELS
The compressor map is key data in 0-D models, and one must be aware of the consequences of using insufficiently
representative maps. The compressor characteristics are generally well known between 75% and 110% of the design
speed. Below 75%, theoretical compressor models may lack precision. For over-speeds, compressor modeling is often
limited by the inability of the partial test bench to reach these over-speeds in standard conditions, either because of
insufficient driving power or mechanical risks.
Cold weather and high altitude computations often lead to operation at 125% or more of the design-reduced speed.
Therefore compressor maps must be accurate in this region where efficiency is decreasing quickly. Accurate maps are
also necessary at low speeds to determine the null power speed or the idle ratings; these speeds need to be accurately
3-9
estimated because their influence on transient operations (acceleration times for example) and the controller design.
In a more general manner, the compressor map, like turbine maps, greatly determines the shape of the relationship
between specific fuel consumption and thrust or power, and thus the range of the aircraft.
For all these reasons, the major improvements to be made to 0-D compressor models are linked to a better account of
secondary effects described in Figure 6.
2.1.3.4
FAN REPRESENTATIONS
Because of the downstream splitter and booster sometimes present in high bypass turbofan engines, different model
representations are often used for fans rather than trying to make a conventional compressor representation work. A
separate map for the fan hub, booster and tip may be used. A common approach is to create a single model of the fan
hub and booster, and a separate component model of the tip. This corresponds to the way test data is typically taken and
simplifies matching test data.
2.1.3.5
SECONDARY AND ENVIRONMENTAL EFFECTS
Secondary and environmental effects are those which are not addressed in the component model but which have a
significant impact on the engine simulation, when the engine is not operating at the nominal condition assumed in the
component model. The best way to address these effects is to include them directly in a component model. This is often
impractical so that adjustments to the component model or model results are made to account for these operating
condition differences. These models are generally closely tied to the basic component model in both methodologies.
Usually they must follow a consistent approach, and provide the required accuracy
2.1.3.6
VARIABLE GEOMETRY COMPRESSOR
In a variable geometry compressor, 1, 2 or more stator grids have variable settings. Very often, these settings depend on
compressor reduced speed, and actuators commanded by the engine controller apply the setting laws.
When the variable grid number is low (1 or 2), it is possible to add 1 or 2 dimensions, corresponding to these
supplementary degrees of freedom, to the maps described in Figure 6.
This way, we have:
P3Q2
E23
The setting laws CAL = F(XNR2) are then defined separately. This description is the most complete one because it
allows the stator setting laws to be easily changed and optimized, without requiring a new map each time.
This first solution can be hard to apply if experimental data is lacking, or when the number of variable grids is
important. In such cases, fixed geometry maps are used. Including the setting law for each stator is easier, because the
data format is simpler and the data number is lower but the description is valid only for one setting law.
The more common arrangement is for all variable stators to be ganged, with changes in map performance based on a
single reference stator position. There may be a series of maps for different reference stator positions relative to the
nominal schedule. Alternatively, there may be a single base map for the nominal stator position and numerical
corrections for other stator positions.
2.1.3.7
BASIC ALGORITHM TO DETERMINE OUTLET CONDITIONS
At a given corrected speed, the reduced flow is limited between surge flow and blockage flow. To avoid computation
problems during iterations, it is often preferred to use another parameter like W2RQPR= W2R2/P3Q2 (close to surge
margin) as the input parameter, rather than W2R2.
The outlet conditions can be calculated via the following algorithm, using T2, P2, XN1, as input data:
T2, XN1
XNR2.
W2RQPR
P3Q2, W2R2.
P3Q2, XNR2
E23.
P2, P3Q2
P3.
T3.
W2 (= W3 if no bleed).
3-10
2.1.3.8
PRECAUTIONS - MAP CONSTRUCTION ASSUMPTIONS
Compressor maps are generally issued by either of two means:
Theoretical aerodynamic models,
Rig testing.
In any case, the maps and surge line are often calculated as follows:
For reference constant inlet conditions: T2 = 288.15 K, P2 = 101.325 kPa, that is to say Reynolds index equal to 1;
With clean air inlet, that is to say, with low space-time pressure and temperature distortions;
For given clearances: either constant for theoretical model or with a given (and identified) schedule for a partial test
bench;
In steady thermal operation;
For a rated and identified engine ventilation.
Many of these assumptions can be false in real engine operation and require specific corrections.
2.1.3.9
PRECAUTIONS - SECONDARY AND ENVIRONMENTAL EFFECTS
Reynolds effect - It was said before that: reduced parameters for compressors find their origin on a Mach number
similitude. But this similitude does not account for viscous effects. It is clear that boundary layer thickness and wakes
have an effect on both flow and efficiency. To take into account this phenomenon, the Reynolds index is widely used:
Re (T2, P2)
IR = Re (288.15 K, 101.325 kPa)
The Reynolds index is 1 in standard conditions and decreases quickly when altitude increases. Therefore, the following
correction is of major importance for altitude operation and is generally calibrated after either flight tests or altitude test
bench trials.
The maps calculated for standard inlet conditions are modified as followed (EPOL denotes the compressor polytropic
efficiency):
EPOL23 (IR) = 1 - (1 - EPOL(IR=1)) . IR-x 0.05 < x < 0.15
EPOL23 (IR)
W2R2 (IR) = W2R2 (IR=1) . EPOL23 (IR=1)
P3Q2 (IR) is then calculated at a specific work, that is to say:
H3D2
H3D2
T2 (IR) = T2 (IR=1)
It is often necessary to correct the surge line also, especially in cases where the altitude domain is extended, as for
turbofans.
Real gas effects - The real gas effects are partially taken into account because the enthalpy, specific, and entropy
functions are estimated with experimental data. However, the compressor map often uses XN/T2 and W2T2/P2 as
reduced parameters and these expressions only approximate Mach similitude. A better choice to account for the real
characteristics of the gas would be XN/(RT2) and (W2T2)/P2 R/, for example. Even with adjustment to the
corrected parameters, changes in gas properties may require additional correction as described in the AGARD report
Recommended Practices for the Assessment of the Effects of Atmospheric Water Ingestion on the Performance and
Operability of Gas Turbine Engines, AGARD AR 332 1995.
Distortion effects - Distortion effects are difficult to account for because their analysis requires extensive testing and
aerodynamic computation. Furthermore, very complete air inlet instrumentation is needed to get all distortion
characteristics. Such trials allow the establishment of tabulated rules, describing the effect of distortion on maps and
surge line. Such corrections are critical in cases where high angle maneuvers may happen, as with combat aircraft and
certain missiles, because the aircraft body may mask the air inlet. There may be also high distortion when the engine
installation is not optimal, as for APUs.
Clearances and ventilation effects - The compressor performance is highly dependent on blade tip clearances, which
depend on operating conditions. These effects are not accounted for with the use of reduced parameters (aerodynamics
origin), because clearances depend on the mechanical and thermal state of solid pieces. Furthermore, clearance changes
correspond to a geometry modification. The main parameters influencing the clearances are compressor speed, the
temperatures of the fluid in the compressor (blades and disk temperatures), out of the compressor (ambient conditions
for casings temperature) and of course, the compressor ventilation. Mainly derived from thermo-mechanical
computations, tabulated laws may be usefully added to correct maps and surge line.
3-11
2.1.4.
TURBINES
2.1.4.1
MODELING
Turbine modeling aims at determining outlet conditions, from knowledge of inlet conditions. This is the same as
compressor modeling.
2.1.4.2
FIXED GEOMETRY TURBINES
From an aerodynamic point of view the turbine operation depends only on the velocity triangle. Therefore, for nonviscous flow, a good similitude is based on Mach numbers as for compressors. In fact, compressor and turbines
modeling are very similar. The classical turbine description relies on the use of the following reduced parameters:
XNR4 = XNH/T4
reduced speed;
P4Q45 :
pressure ratio;
WR4
E445
efficiency.
In most operating conditions, the turbine nozzle guide vane is choked. Therefore, WR4 is constant. To avoid
computational problems, XNR4 and P4Q45 are preferred as input parameters for modeling. As for compressors, two
maps are used:
WR4 = F(XNR4, P4Q45);
E445 = F(XNR4, P4Q45).
The reduced enthalpy rise H4D45/T4 is sometimes used instead of efficiency.
These two maps are sufficient to calculate the turbine outlet conditions, as will be shown later in section 2.1.4.5.
Nevertheless, generally, a third map is often used in combinations with these two maps to more accurately calculate the
pressure losses in the ducts behind the turbine.
Contrarily, when compared to compressors, which mostly have axial outlet flows (both axial and centrifugal
compressors), the flow at the turbine outlet presents an important swirl depending on XNR4 and P4Q45 (swirl itself is a
reduced parameter because it characterizes the velocity triangle). Thus, the third map: SW45 = F(XNR4, P4Q45) is
added.
2.1.4.3 VARIABLE GEOMETRY TURBINES
This component is relatively rare because of the difficulties due to combining variable geometry and high temperatures.
Nevertheless, it can be found in free turbine turboshaft engines that include a recuperator. In such a case, a free turbine
with variable nozzle guide vanes permits optimization of the recuperator, by maintenance of a constant high temperature
at its inlet. This applies even at part loads, and results in a very flat SFC = F(RWSD) curve.
As indicated by the former example, the NGV position is not necessarily governed by the turbine speed. Thus it is not
possible to reduce the modeling of 2-D maps. By chance, the number of variable NGV stages is generally lower than
two. This allows the method used in certain cases for compressors, by adding dimensions (corresponding to the variable
settings) to be applied.
2.1.4.4
TOTAL-TO-STATIC TURBINE MAPS
Generally component maps are given for total parameters, which avoids having to know the sections at inlet and outlet.
Let us consider the example of a free turbine turboshaft engine.
A diverging nozzle, for slowing down the exhaust gases, follows the power turbine. The power created by the free
turbine results from both total-to-total efficiency and pressure losses in the nozzle. Very often, it is difficult to evaluate
the losses in the nozzle due to swirl, and the presence of struts, thus the total to total pressure ratio of the turbine is not
accurately known.
To avoid this conflict between power turbine and downstream losses, a map modeling both turbine and nozzle, which
uses a total to static pressure ratio P45Q59 = P45/PS9, is often used. This ratio is known more easily because PS9 is
equal to the ambient static pressure. In such cases, the swirl map is not given because the nozzle losses are already
included in the efficiency E4559.
This total to static map is thus given in the following form at:
WR45 = F(XNR45, P45Q59);
E4559 = F(XNR45, P45Q59).
The associated drawback is that any change in nozzle geometry requires new total to static maps.
3-12
2.1.4.5
BASIC ALGORITHM TO DETERMINE OUTLET CONDITIONS
The turbine outlet conditions can be calculated by the following algorithm using T4, P4, XN, P4Q45 as input data:
T4, P4, WR4
W4 = W45.
P45.
The power created by the turbine can be calculated by: PW = W4H4 - W45H45.
2.1.4.6
PRECAUTIONS - LIMITATIONS
Most remarks made for compressor maps, can be applied to turbine maps. Like compressor maps, turbine maps are
issued with reference inlet conditions and a reference environment expressed as clearances, ventilation The
translation of these performances to the actual engine environment is a critical problem for 0-D models, which minimize
the geometry description. This translation is even more difficult for turbines than for compressors because:
The design point for a turbine is at high temperature and pressure while turbine maps are often measured at partial test
bench using cold and low-pressure air. Consequently, the corrections due to inlet conditions are much bigger than for
compressors.
The temperature inlet condition for a turbine varies through a much wider range than for compressors because there is a
combination of both ambient temperature range, due to the flight domain, and the power or thrust level range. Thus, the
corrections are not only bigger than for compressors, but they have higher amplitudes within the operating envelope.
Nevertheless, there is one feature of turbines that in certain cases can ease their modeling. When a turbine stage is
located between two choked fixed sections (either nozzle guide vane or exhaust nozzle), it is easy to prove, thanks to the
critical mass flow formula, that the pressure ratio of the stage is fixed and does not depend on the operating conditions.
Therefore the specific work H4D45/T4 is also constant.
Generally, the turbine drives a compressor for which H3-H2/XN2 is quite constant (velocity triangle), so XN/T4 is
nearly constant. That means that such a turbine has a single aerodynamic operating point because both XNR4 and
P4Q45 are constant. In reality, this operating point moves a bit inside the maps because of the real gas effects and the
wide temperature range. Thus, HP turbine modeling is much more of an environmental modeling problem rather than an
aerodynamic one. Of course, turbines such as free turbines in turboshaft engines cumulate all difficulties because their
exhaust nozzle is not choked.
Most corrections mentioned for compressors in Figure 6, such as Reynolds effect, clearances and ventilation corrections,
can be applied to turbines. The last of these has a major impact on efficiency, because cooling flows generate local
distortion and aerodynamic disturbances.
Concerning the effect of distortion mentioned for compressors, the problem is slightly different for turbines because it is
not due to installation or flight conditions. Here it is due to engine design, at the combustor outlet, where there is a nonuniform temperature profile. This is because of:
It is very difficult to account for that distortion due to lack of experiments and the severity of the environment. As with
compressors, more accurate reduced parameters may be used for Mach number similitude by accounting for and R
variations:
XN/RT4 instead of XN/T4;
W4T4/P4 R/ instead of W4T4/P4.
2.1.4.7
IMPORTANCE OF TURBINE MAPS IN 0-D MODELS
As said before, what we may call a HP turbine has a fixed operating point. The operating point of a power turbine in a
turboshaft engine, for example, is variable because the pressure ratio is not fixed. For that reason, turbine maps wont
have the same criticality.
Fixed operating point turbines - From the point of view of the customer, who considers the engine with a given
turbines matching, the HP turbine map is not so important as the environment and secondary effects. This is
because the operating point is nearly constant. However, for the engine designer, it is important to have a good
prediction of reduced speed and pressure ratio effects, in order to optimize the initial matching, or to study eventual
engine re-matching by NGV section changes, or simply to evaluate the production scatter effect.
3-13
Variable operating point turbines - This case may be encountered in two configurations:
There is one critical section behind the turbine but this section varies according to operating conditions. This is
the case for an LP turbine in a military turbofan with a variable exhaust nozzle.
There is no critical section behind the turbine. This is the case for a power turbine in a turboshaft engine.
In both cases of variable operating point turbines, the operating point may vary across a wide range within the maps and
such maps will become, like compressor maps, key features of the 0-D models.
2.1.5.
BURNERS AND AUGMENTORS
The heat addition component models typically include the primary combustor between the compressor and turbine
components, and the augmentor or afterburner downstream of the turbomachinery. The component models typically
differ according to the design intent and relevant operating range.
Main burners are designed for high efficiency, long life, low-pressure drop and flat temperature profiles to
accommodate the downstream turbine. They must operate reliably over the aircraft envelop and all power settings.
Synonymously with Afterburner, the terms Augmentor and Reheat are used to describe an auxiliary burner downstream
of the turbines in which the gas is reheated to provide extra thrust. These are used for limited periods, only at high
power and for limited duration. Although there are many common issues, individual designs of main burner and
augmentor have significant differences. A key difference between main burners and augmentors is the large variation in
momentum pressure drop in the augmentor with operating condition, compared to the relatively fixed pressure drop in
the main burner.
2.1.5.1
BURNERS
At the 0-D level, combustor representations generally model the energy rise and pressure drop across the burner, based
on operating conditions and any special operating modes (number of fuel nozzles fired, number of flame-holder rings,
etc.). Models may also include kinetics models for emission calculations. More detailed models may include detailed
information on the temperature and pressure fields, the dilution air mixing and the fuel injection and dispersion
processes.
2.1.5.2
MODELING
Modeling of combustors is relatively easier than modeling of rotating components, at least for 0-D models of classical
burners, because the burning efficiency is close to 1 in many conditions. This is because they do not have any pollutantemission constraints or non-afterburning chambers.
The combustor is considered as a black box that receives:
Hot compressed air from an HP compressor characterized by W3, T3, P3 and eventually a non null water-air ratio
WAR3;
Fuel characterized by its mass flow WF and its lower heating value FHV. This mass flow may be split in the case of
a staged combustor.
The combustion is usually modeled as a heat addition at quasi-constant pressure (there are pressure losses). The real
added heat depends on the burning efficiency of the combustor defined as:
EFB =
W4H4 - W3H3
WF . FHV
This is the ratio of real heat over theoretical heat due to a perfect combustion. For classical combustors, EFB is close to
0.995 because the combustor is designed to have a quasi-stoichiometric primary zone, which guarantees both high
efficiencies and stability.
In the case of low-emissions chambers required by new legislation for land based turbines, new concepts are being
experimented with. Examples are LPP (Low Premixed Pre-vaporized) and RQL (Rich Quench Lean) where the primary
zone is either poor or rich. This leads to lower burning efficiencies although still generally higher than 0.97.
To model the changes of burning efficiencies according to the chamber inlet conditions, the aerodynamic load of the
chamber is often used:
W3
= P31.8 eT3/300 VOL ,
where VOL denotes the volume of the chamber.
This parameter represents EFB as a decreasing function of :
EFB = EFB 0 - , with 10-3 and 1.4.
3-14
For a high fidelity transient afterburner simulation the ignition process, and the time needed to fill the fuel injectors that
are empty during dry operation to avoid fuel coking, must also be modeled.
3-15
The following paragraphs deal mainly with the simulation of steady state afterburner operation in mixed flow turbofans.
Most of the methods described are also applicable to the afterburners of straight turbojets. Practical afterburner
simulation models are all of the semi-empirical type.
2.1.5.5
DRY (NON-BURNING)OPERATION
When no fuel is injected, the afterburner behaves as a mixer, and the calculation procedures described in chapter 2.1.8.2
apply. However, the pressure losses are significantly higher than in an engine without reheat because the geometry of an
afterburner must be optimized for best burning stability and efficiency, and not for minimum pressure losses.
2.1.5.5.1 TURBINE EXHAUST GUIDE VANES
The pressure losses of the exit guide vanes depend on the swirl downstream of the low-pressure turbine rotor and the
Mach number. This can be modeled with a loss characteristic that employs these parameters.
Alternatively, the exhaust guide vanes can be regarded as a part of the low-pressure turbine. The efficiency, as read
from the turbine map, would then include the losses of the exhaust guide vanes.
2.1.5.5.2 DIFFUSER, SPRAY BARS AND FLAME-HOLDERS
Flame holding and propagation requires a flame-holder system that creates a low velocity re-circulating air region. Such
a system produces, as a byproduct, significant pressure losses. Additional losses are created by the fuel injectors (spray
rings or spray bars) placed upstream of the flame-holders, and the diffuser. All these losses are often lumped together
and modeled as one. Since in this part of an engine the flow Mach number is subsonic under all conditions, the pressure
losses of the core stream will vary proportionally to the turbine exit corrected flow squared.
Modeling the pressure losses of the bypass flow can be more difficult, especially in engines with a low bypass ratio. In
this case, a significant part of the flow will pass behind the liner and only the remainder enters the afterburner at the
bypass exit. Eventually the pressure downstream of the bypass flame-holders can be derived from the bypass exit
pressure only with the help of an empirical correlation.
While all turbofans in series production employ afterburners with flame-holders, it should also be mentioned that there
are alternatives. Mixing and burning in two-stream systems can be enhanced by swirl, and no flame-holders are then
required. Turbofan swirl augmentors are described in some detail in Egan, 1978.
3-16
liner cooling air with the main stream causes another pressure loss in the jet pipe. Mostly these two losses are not
accounted for separately, and they are combined with the bypass flame-holder loss characteristic.
2.1.5.5.5 FLOW DISTRIBUTION AND TURBOMACHINERY MATCHING
In a high bypass engine without an afterburner the full bypass flow joins the core flow at the edge of the mixer. When
the mixer is of the confluent type, then both streams are flowing essentially in parallel, and the static pressures of both
streams are equal.
The shape of a forced mixer is quite complex, and so is the flow field in the region where the two streams join.
However, for the purpose of simulating the turbomachinery-matching of such an engine, it is sufficient to calculate the
mean static pressures from the invariable effective flow areas, and request that these mean pressures are equal.
In the cycle model of a turbofan the operating points of the compressors and turbines are found with an iterative
algorithm. In the first pass of an iteration through the mathematical model the operating points are only estimated
values, and consequently some conditions (flow continuity and energy balance, for example) are not fulfilled. In
particular, the mixer inlet conditions will be such that the static pressures of the core and the bypass stream at the mixer
edge are not equal, and this results in the so-called mixing error.
In turbofans with afterburner, Figure 7, a significant amount of the bypass air passes behind the liner and joins the main
stream successively through the screech damper and the liner cooling air holes. The nozzle cooling air does not enter the
afterburner at all. Thus, only a part of the bypass air enters the mixer. Moreover, at the entry to the mixer the geometry
is often very complex. The flame-holders create re-circulation zones of significant size just within the region where the
simple mixer model assumes static pressure balance. The question arises, which are the effective mixer areas, and are
they invariable for all operating conditions?
For modeling real engines some empirical corrections to the simple static pressure balance assumption are unavoidable.
These are dependent on the details of the afterburner design and no generally applicable advice can be given.
When iterating for an off-design operating point, it may happen that the mixer inlet conditions become quite unrealistic,
because the operating points in the turbomachinery maps were estimated badly. The mathematical model of the
afterburner must be able to cope with these inlet conditions and calculate the mixing error in such a way, that the
iteration can converge. In other words, the mixing error must change continuously when for any fixed core inlet
condition the bypass inlet pressure, temperature, and mass flow change from very low to very high. It must also change,
when for fixed bypass inlet conditions the core inlet pressure, temperature and mass flow vary from very low to very
high values.
2.1.5.6
Fuel-air-ratio;
Pressure;
Inlet temperature;
Residence time.
There are complex interactions between these parameters. For example, the fuel droplet diameter and the time needed to
evaporate the droplets depends on all four of them. The fuel distribution within the afterburner has also a major effect on
3-17
3-18
chamber.
When the burning efficiency is defined as mass ratio of injected fuel over ideally burning fuel then the true temperature
rise is the result of this calculation. Otherwise, the true temperature rise is found as a fraction of the ideal temperature
rise.
3-19
EFFICIENCY CORRELATIONS
3-20
In the flight regime where the fan stream air flow is cold (i.e. 350-400K) and the pressure is low (50 - 100kPa), that is,
in the upper left hand of the flight envelope, fuel vaporization is poor. Hence, uniform gaseous fuel-air mixtures are
nonexistent, and two-phase mixtures create ignition and combustion problems that may result in rumble blowouts.
2.1.5.14
SCREECH
Screech is a high frequency pressure oscillation with 1500-3000 Hz in radial direction, caused by variations in the heat
release process. It can be very destructive to the hardware and must be avoided under all circumstances. The oscillations
can be damped, by incorporating a perforated or corrugated shield in the liner just downstream of the flame-holders.
2.1.5.15
BUZZ (RUMBLE)
Rumble is a low frequency (50-100 Hz) pressure oscillation in a longitudinal direction caused by intermittent rich
extinction. The explanation given in [3] is that locally the flame goes out and a volume of unburned fuel-air mixture
travels down the afterburner, explodes at some point and sends a pressure wave upstream. The pressure wave hits the
flame-holder and restores light up, due to the higher pressure restoring local combustibility.
Buzz can be a problem when its frequency is in resonance with a low-pressure spool torque vibration mode.
2.1.8.1
SPLITTERS
Splitter models are often simple bookkeeping models and require no more component-modeling effort than for simple
ducts. The flow split will be iterated to match continuity conditions elsewhere in the simulation. An exception is when
the splitter model is related to upstream or downstream systems. A common example is the behavior of the splitter
downstream of a fan component. Here the flow split or bypass ratio may affect the fan performance via the relative axial
location and other geometry details. For models addressing hail or rain ingestion, modeling of the splitter effect on
capture of the liquid/solid particles is the primary driver in determining engine operability limits.
2.1.8.2
MIXERS
Mixers are employed with high bypass turbofans for thrust increase and for noise reduction. In low bypass engines there
is a mixing of two streams within the afterburner. In both cases the flow conditions within the mixer will have a
dominating effect on the matching of the two spool speeds of a turbofan.
The thrust improvement due to mixing of two streams ideally i.e., without friction pressure losses depends on the
3-21
difference in total temperature. When both streams have the same total temperature then there will be no thrust gain.
Figure 12 shows the ideal thrust gain by mixing two streams with equal total pressure.
When the mixer is applied to a turbofan engine then at cruise the ideal gain in net thrust is typically twice as big as
shown in the figure because 1% change in gross thrust is equivalent to 2% in net thrust.
Mixers, like any other component, perform less than ideally in practice. To fully mix two confluent streams needs a long
pipe. When forced mixers (chutes) are applied, the required length becomes shorter. In practice, confluent mixers
achieve about 20-30% mixing and forced mixers approximately 60-80%. Moreover, the mixing process and the
additional wall friction cause total pressure losses that decrease the benefit of mixing.
Figure 12 - Thrust gain from ideal mixing of two streams with equal total pressure
2.1.8.3
DEFINITIONS
We have spoken of the degree of mixing without defining what we mean with that term. For performance simulations
the degree of mixing is defined through thrust:
mix =
Fg Fg ,unmixed
Fg , fullymixed Fg ,unmixed
The evaluation of the thrust for unmixed and ideally mixed streams must be based on a clear definition. Figure 13 shows
the nomenclature that we will use here.
3-22
mixed and expand as a single stream without any losses through the nozzle, which might be convergent or convergentdivergent.
The total pressure and temperature of the cold stream in station 163 are the same as at station 16 and the equivalent is
true for the hot stream. The ideal mixing is taking place in a frictionless duct with constant area. The following laws of
physics are applied:
Conservation of mass: W63 + W163 = W64 ;
Conservation of energy: W63 * H 63 + W163 * H163 = W64 * H 64 ;
Conservation of momentum: W63 * V63 + Ps ,63 * A63 + W163 * V163 + Ps ,163 * A163 = W64 * V64 + Ps ,64 * A64 ;
The mixing shall take place in a duct with constant area, which gives us two more correlations:
Figure 14 - Total pressure after mixing two streams with equal pressure (200kPa)
When the total pressures in both streams are different, then in most cases the pressure at station 64 will be between P16
and P6 and there is no simple way to quantify the pressure losses due to ideal mixing. However, the change in thrust gain
due to ideal mixing is a good measure for the influence of the total pressure imbalance at the entrance to the mixer. In
Figure 14, we see from an example with T16 = 400K and T6 = 1200K, that any deviation from equal pressures at the inlet
will reduce the ideal thrust gain due to mixing.
Unequal total pressures at the inlet to the mixer can cause significant reductions in the thrust gain potential of mixers.
Note that in a bypass engine the pressure ratio P16/P6 will increase when the engine power is reduced. Therefore, the
aerodynamic design point of a mixer should normally be at values of P16/P6 between 0.9 and 0.95.
During all previous discussions we have assumed that there are no friction losses. In reality, these obviously must be
taken into account. In Figure 15 (With the assumptions: equal total pressures at the mixer inlet, T16 = 400K and T6 =
1200K, bypass ratio = 5) the effect of cold and hot friction pressure losses on the thrust gain are shown. It becomes
3-23
obvious from that figure, that friction pressure losses above approximately 1% in the bypass stream will decrease the
benefit of a mixer in such a way, that the weight of the mixer is not justified from a performance point of view. Losses
in the hot stream have less effect in this example with bypass ratio 5, because they affect only the smaller mass flow.
Figure 15 - Effect of friction pressure losses on the thrust gain due to mixing
2.1.8.5
THRUST GAIN OF REAL MIXERS
In a practical mixer we will not achieve a fully mixed flow. As mentioned above, in confluent mixers we will get
typically 25% and in mixers with chutes roughly 70% of the ideal thrust gain.
We have not yet mentioned the effect of nozzle pressure ratio on the performance potential of a mixer. During cruise
with a subsonic high bypass engine, the nozzle flow will be sonic (P8/Pamb>1.85), while during take off the nozzle
pressure ratio will be significantly lower. As can be seen from Figure 16 this will again reduce the potential thrust gain
due to mixing.
3-24
mixer areas A161 and A61, and the nozzle area A8. The upstream conditions for the cold and hot mixer inlet areas are
clearly defined and the pressure ratios P161/Ps,163 and P61/Ps,63 can be easily evaluated.
The situation for the nozzle is different. The total pressure P8 at the inlet to the nozzle is only clearly defined when the
two streams are fully mixed. For partially mixed flow there is no generally accepted procedure available for the
calculation of the mean total pressure at the inlet of the nozzle.
Any formulae that employ the degree of mixing, for the calculation of the nozzle inlet pressure, are not fully satisfactory
because the degree of mixing is defined as a thrust ratio. Thrust, however, is dominated (at a given pressure ratio) by the
velocity term and thus by the temperature of the fluid, not by the total pressure.
The dilemma with the total pressure downstream of a practical mixer can be bypassed, if the nozzle discharge
coefficient is defined in such a way that the ideal flow through the nozzle is calculated from the total pressure of a fully
mixed flow.
2.1.8.7
THRUST COEFFICIENT
The thrust gain (or loss) due to partially mixing two streams can be described with the help of the degree of mixing as
defined above. That requires three nozzle thrust calculations: two for the unmixed streams and one for the fully mixed
stream. In each of these nozzle calculations, one applies a coefficient that takes care of nozzle thrust losses, to the onedimensional calculation result.
In principle, with this approach we use two additive thrust corrections within the calculation.
Alternatively one can use the fully mixed flow as a reference, do the nozzle calculation only once and afterwards apply
an empirical correction factor. This is equivalent to the standard procedure for any turbo-machine, where we do an
isentropic calculation first, and then apply an efficiency to correct the result. Such an approach eliminates any debates
about the nozzle inlet total pressure for partially mixed flow.
2.1.8.8
STATIC PRESSURE BALANCE
When two parallel streams are mixed then there will be equal static pressures in both streams. This fact is used within
any custom cycle calculation for mixed flow turbofan engines, as a condition for finding the match of the low and highpressure spool speeds.
In real engines the flow in both streams will not be exactly parallel. However, this is mostly neglected and it is
postulated in Figure 13, for example, that the static pressure Ps61 equals Ps161.
For high bypass engines with corrugated or confluent mixers the simple static pressure balance is fully appropriate. In
low bypass engines with afterburner, however, the complex flow in the region of the flame-holders can cause a local
static pressure imbalance. The cycle match of such an engine must employ some empirical corrections to the
conventional static pressure balance assumption.
2.1.8.9
MIXER TEST ANALYSIS
The analysis of mixer-component tests is rather difficult because the thrust differences between alternative
configurations are rather small. For more details see Rowe, 1982. Comparing various mixer designs with the help of
engine tests is extremely difficult. The reason is, that there are not only differences in the degree of mixing, but also (at
least potentially) in the effective areas at the mixer inlet and the nozzle. These differences will cause the engine to
rematch and this in turn will cause a change in fan efficiency, for example. In the end, the engine with the superior
mixer design might have a poorer SFC because its effective areas are not correctly matched to the cycle.
2.1.9.
DUCTS
Most component-based engine models include a large number of identified stations within the engines. Many of these
differ only by a pressure loss or a simple extraction of flow that is not explicitly modeled. These pressure loss models
are usually modeling a friction loss (fanno line) process or a loss due to sudden expansion, separation or even an
unmodeled low level mixing process. Often these are modeled as an empirical function of velocity head or flow Mach
number. Most models are explicit, based only on entrance quantities and thus do not require a compatibility condition
closure like nozzles, compressors and similar component models.
If flow in the duct can be choked in any operating regime of interest, this is no longer true. The duct model must address
this choking and the simulation must address the implied compatibility condition or the simulation must be configured
to address or avoid the issue. For higher fidelity models, the duct model must address the transfer and modification of
the more detailed entrance conditions as well as the higher fidelity modeling of the processes inside the duct.
2.1.10.
INTERNAL AIR SYSTEMS
Modeling of air streams other than that of the primary flow-path varies widely between engine models. Many simply
model these as fixed percentages of the main flow-path air with minimal physical representation. Higher fidelity models
or special purpose 0-D models may model the detailed internal flow circuits, and the processes associated with mixing
3-25
and extraction of purging, cooling, thrust balancing and other secondary air-streams.
Performance models generally focus on primary flow-path components since they predominate in determining the
engine system behavior. However, many non-flow-path components or external modules are significant to the
simulations. The sections below describe how these non-flow-path components are treated and how they influence
engine system models.
A key need for internal system modeling is to provide the flexibility required without a significant overhead burden to
the primary focus of the modeling of primary engine flow-path and overall system performance.
3.1
Most component-based engine models include the direct effect of bleed between components as a standard feature. For
empirical models, the greater the flexibility in bleed level and location, the more tables or correlations are required to
achieve the desired effects. Internal bleed must consider the impact on component performance and the method of
determining the conditions of the bleed air. In trying to match bleed supply conditions to user requests, a single bleed
location may not be adequate. Bleeds from multiple locations may be selected or mixed to create the required bleed with
minimal impact on the overall engine performance.
External load models are generally fixed or simple relationships, and are treated by simply including them as one of the
power-outputs from the rotating shaft. For some applications, these models become more complicated and are included
as part of the basic engine simulation. This is particularly true where control systems are part of the model and customer
load affects the ability to maintain the required conditions.
3.2
In models where the fuel and lube system fluid conditions are of interest, these components are modeled similarly to
other components except for the use of a non-air working fluid. Most performance simulation applications are not
interested in these internal details. In these, fuel and lubrication systems are often empirically modeled as parasitic loads,
even in fairly detailed models. In models where conditions in the engine are dramatically affected by these loads, or
where the secondary systems interaction are of interest, more detailed models of the pumps and processes in these
systems may be required.
3.3
Military aircraft often have fairly extensive arrangements to reject excess heat from the aircraft and aircraft sub-systems.
The engine can be part of this process, and engine models are used in optimizing these systems.
Apart from the turbine, which requires significant attention in terms of managing the cooling flows necessary to
maintain component temperature levels, many other parts and systems in a gas turbine engine require management of
thermal properties. There are several parts which, without special measures, would get overheated and fail. The major
function of the oil system is cooling of bearings; many other parts are cooled by secondary airflows. Fuel often has to be
heated to maintain favorable viscous and lubricious properties and to prevent ice formation in the fuel system, or used as
a heat sink for excess heat loads. On some classes of aircraft, such as supersonic transports, rising fuel temperature may
limit flight duration.
3-26
3.4
Early engines featured extremely simple fuel systems, often with limiting by pilot observance of cockpit gauges.
Accordingly, the engine internal margins (compressor working lines) had to be large, thereby wasting performance. As
engine complexity increased, the control-systems became more refined to ensure accuracy and safety as margins were
cashed for greater performance. In recent years there has been a progressive shift away from hydromechanical systems
to electronic systems, which have greater flexibility.
Modern engines have many control inputs, most of which can be categorized into two main groups:
Fuel flow to main and reheat combustors. (Fuel may be distributed into different zones within each combustor to
satisfy local combustion constraints.)
Geometry final nozzle area, both convergent and divergent, compressor inlet guide vanes, bleed valves, blocker
doors, variable mixer, bypass injector, and turbine throat areas.
Other control inputs may include water or methanol injection.
The fundamental requirement for any control-system is to deliver the required (rated) level of thrust (or shaft power) at a
particular flight point (e.g. Mach no, altitude combination). This is achieved by controlling certain engine parameters to
prescribed levels that are related to pilot input (pilot-lever angle PLA). Direct prescription (open-loop control) of some
control inputs is clearly inappropriate, for example, relating fuel flow to pilot demand. So, closed-loop control is
employed for most control inputs.
Open-loop control of some variable geometry such as convergent nozzle area, or more correctly nozzle control ring
position, is common. However, it relies on the actuator position accuracy to achieve the required engine condition.
Because the optimum geometry setting is unlikely to remain constant over life, open-loop schedules must be
periodically trimmed, if optimum operation is to be maintained. In addition, engine deterioration and scatter, airframe
off-takes (bleed and shaft power) and intake distortion, all of which are random (within acknowledged limits) conspire
to make open-loop control undesirable. This said, a mix of closed and open-loops is often found; judgement on the mix
for a specific application is made on the basis of requirement, complexity (and therefore cost) and feasibility. Multiple
closed-loops can interact, and although there are standard multi-variable control techniques for compensating for this
3-27
3-28
leading to some compromised operation at some conditions. Independent (2-parameter nozzle) control of divergent area
may be justifiable. In these cases, A9 could be scheduled against nozzle pressure ratio and A8, or controlled in such a
way to maximize whole vehicle performance in conjunction with the aircraft autothrottle (again, the operation within
open-loop safe limits issue arises).
3.4.2.
AFTERBURNER CONTROL
The primary purpose of reheat is to provide a thrust boost with a low mass penalty. The main premise is to increase the
jet velocity by burning the remaining oxygen in the jetpipe, while ensuring stable engine operation, particularly with
respect to the fan working line. In order to maintain the dry engine operating point: as jet velocity is progressively
increased by modulation of reheat fuel flow, so must the nozzle area increase to compensate. Thus two extra,
interrelated control loops are introduced for reheated operation. As with the main engine control-system, it is desirable
to eliminate reliance on open-loop fuel metering. Closed-loop reheat control-systems could be based on EPR control of
the nozzle and some means of controlling fuel flow. Direct scheduling off the actual nozzle area is common, or a closedloop control might be employed, based on some indication of jetpipe gas velocity e.g. total-to-static pressure, or
though difficult to measure jet pipe gas temperature. Reheat staging to achieve optimum combustion and emissions
requires control of fuel distribution. As with main combustion, the distribution of the fuel may affect the overall engine
operating point.
The speed at which reheat is allowed to modulate depends on how well the fuel and nozzle area modulation can be
coordinated. Out-of phase modulation can upset the fan!
Reheat light-up is also potentially fan-unfriendly. Open-loop limits on fuelling may be employed to safeguard the fan
operating point. Automatic sensing and compensation for dangerous reheat instabilities (buzz and screech) may also be
required.
3.4.3.
VARIABLE CYCLE ENGINES
So-called variable cycle engines (VCEs) could more correctly be called variable bypass-ratio engines. They employ
variable geometry devices to progressively change the engine from a low bypass ratio engine (high specific thrust) to a
higher bypass ratio engine (low specific thrust). This is to meet performance criteria at widely differing flight
conditions. Bypass blocker doors and variable bypass exit mixers are primary features of such engines. The doors may
be of the 2-position type in which case the control task is: when to switch and ensure safety in the transition (when
doors close others might be opening and vice versa). Alternatively, doors may be of the continuously variable type, in
which case the task becomes more involved. The variable mixer is positioned to ensure correct HP to LP matching for
the two types of engine at either end of the operating range.
3.4.4.
PERFORMANCE-SEEKING CONTROL (PERFORMANCE OPTIMIZATION)
Judicious selection of control-loops and derivation of suitable schedules is seen as the basic optimizer of engine
performance, however smarter approaches have been demonstrated which employ searching techniques to find the
minima or maxima of a particular cost function. The cost functions are typically: lowest temperature at a thrust,
minimum fuel burn, best thrust etc. Such techniques are reliant on the use of embedded engine models in the control
system. Models included in this way are therefore also available for on-board diagnostics and fault detection. The use of
smart control techniques may reap most benefit when combined with the flight control-system (which often includes any
variable intake geometry) to optimize the whole vehicle.
3.4.5.
MODELING CONTROL SYSTEM COMPONENTS
Although it is often classed as an engine accessory the control-system has a fundamental role in defining what the
engine does, that is, it dictates functionality. The individual components of the control-system have differing levels of
influence on the functionality. However, the engine is subordinate to the control-system.
Whole-engine models (engine + control-system) are used for various purposes:
The level of model detail required for different activities may vary. For instance, actuator and sensor dynamics can often
be simplified when the effects of long-range engine transients are being studied. Greater dynamic fidelity may be
required in other cases such as selection of control gains to meet specific stability margins. Similarly, the requirements
placed on the engine model vary for each application. Engine model requirements are covered in detail in Chapter 2.
3-29
The applications listed above are focused on different parts of the whole system. In some cases the reaction of the
engine in response to the controller is of interest, whereas for other studies the converse applies. Accordingly, the
computing environment and analysis facilities may vary. Control-system models and engine models invariably emerge
from specialist departments, sometimes using different tools and programming languages. Interfacing models and
systems can be an issue, although the use of CORBA-compliant tools will facilitate the co-execution of differing
systems. This is discussed further in Chapter 4.
Any dynamic system model requires initialization. Initialization can be achieved by iteration or trimming, although this
can be problematic. As initialization is concerned with the initial values in numerical integrators, the initialization task
can be simplified by reducing the number of integrators present, by reducing the level of dynamic modeling. Such a
trade-off may be acceptable for some applications - usually those that require a whole-engine model. Detailed dynamic
models, with fewer shortcuts, may then be needed for detailed design and optimization of the fuel system. Such models
are usually run in isolation. Fuel-systems have fast dynamic terms, which are usually insignificant in terms of overall
engine response. However, where the combustion process is modeled in detail - such as for an advanced combustor type
- the fuel-system dynamics may significantly interact with combustor dynamics and hence affect engine response.
3-30
3.4.6.
DESCRIPTION OF A TYPICAL CONTROL SYSTEM
Figure 18 below shows typical control-system architecture for a combat engine.
ELECTRONIC
CONTROLLER
DEMAND
ACTUATION
ACTUATION
SENSORS
SENSORS
ACTUATION
ACTUATION
ignition
Metering
& distribution
(main)
FUEL TANK
Boost
Pump
Backing
Pump
Main
Pump
Metering
& distribution
(afterburner)
Afterburn
Pump
-Shaft speed;
-Gas temperature;
-Gas pressure (total and static);
-Metal temperatures (static and rotating);
-Acoustic resonance;
-Flame detection;
-Fuel flow-rate;
-Position.
Accuracy, reliability, cost and response constraints will dictate the sensing method used for each parameter.
3.4.6.2
VARIABLE GEOMETRY ACTUATION
Actuators can move variable guide vanes, final nozzle area, bleed valves and other variable geometry features on an
engine. The common types are:
-Hydraulic;
-Pneumatic;
-Electrical.
3-31
3.4.6.3
PUMPING
Fuel is injected into the engine under high pressure. Pumping systems are required to overcome the back-pressure on,
and the losses within the system, and to generate the required spray pattern. High-pressure fuel can also be used for
servo power within the fuel system and for actuation (fueldraulics).
There are normally 3 pumping stages:
Boost pump - usually electrically driven and part of the aircraft system, it ensures a constant supply of fuel to the
engine fuel system at all aircraft attitudes;
Backing pump - to provide suitable inlet conditions for the HP pump;
Main (high-pressure) pump - provides the main pumping effort.
Centrifugal and positive-displacement pumps are generally used for aero-engine applications, although other types such
as air-jet extractor types are used for low pressure duties e.g. accessory cooling by fuel circulation. Pumps may be
driven mechanically (shaft power via gearbox), by air-turbine (using bleed or ram air) or electrically. Electrically driven
pumps offer greater control of fuel flow by varying the motor speed.
3.4.6.4
METERING
The rate of fuel flow to the engine is the prime component of the definition of engine power output. Stable and fast
responding metering is required - especially for afterburners and staged combustion systems which involve the fast
selection and deselection of burners. The most common metering system is based on a variable orifice with a regulated
pressure drop - the orifice being varied by hydromechanical or hydroelectrical means. Valve position feedback is
required for the determination of fuel flow or for closed-loop control of the valve position. In order to reduce the overall
system mass, there is an effort to remove the provision for fuel metering and to provide a system based on a relatively
simple tap that operates in a relative sense (more fuel or less fuel) rather than metering fuel in an absolute sense. This is
discussed further in Chapter 5.
3.5
NOISE
As noise has become a more important limitation on aircraft use, engine models incorporating noise estimates have
become more common. The purpose is usually to examine the effect of various power level and flight path combinations
on specific take-off restrictions in the potential areas of operation. The primary outputs are dB levels at specific
operating points and total dB over a period of time and distance. Although the noise contributors are cumulative and
vary over the frequency domain, most engine system models limit noise estimates to correlations that include:
1) Jet Velocity;
2) High Frequency Blade Passing;
3) Low Frequency Combustor Noise.
Detailed acoustic models usually require geometry and air velocity information to create estimates of emitter source
strength. These more detailed models may also require information on the nacelle and aircraft installation and the
attenuation and propagation characteristics for the installation.
3.6
EMISSIONS
This is a key area of detailed combustor component models. The modeling is usually done at engine level using
correlations based on overall performance.
With the increasing attention to gas turbine exhaust gas pollution, exhaust gas emission levels must be predicted at
varying operating conditions. On the manufacturers side, the processes in the combustor are modeled in detail, with
CFD, in order to develop new technologies to reduce emissions, such as LPP and RQL combustion . On the operational
side, there is interest in how to minimize emissions by optimizing operating conditions such as engine condition, aircraft
flight procedures, and fuel type and water and steam injection. The latter two variables mainly relate to ground based
gas turbines, using LNG, LH2 or fuel obtained from gasification of coal or bio-mass. However, it must be noted that
LNG and LH2 fuels for aircraft are already being considered.
ICAO tests are done for all commercial engines prior to certification. They include data for 4 points roughly
corresponding to take-off, ground-idle, descent idle and max climb. The test conditions are defined in terms of a fixed
percentage of certified engine thrust and may not correspond to the actual engine operation at these power settings on
the aircraft.
These are done at sea level static but are used to generate the total emissions for a typical landing and take-off cycle.
The ICAO points have been generalized to be applicable across all engines and may not correspond to the actual engine
power settings in use. For example, the ICAO setting at 7% of rated power may not correspond to the actual descent or
ground idle power setting. Furthermore, the ICAO points do not represent high altitude cruise conditions at which
emissions exist for most of the time. To determine cruise condition emission levels, research is directed at both in-flight
3-32
3.7
3.7.1.
LOCAL ITERATIONS
3.7.1.1
0-D MODELING OF A FREE TURBINE TURBOSHAFT ENGINE
The example hereafter shows the loop nesting required by local iterations when that method is applied to the resolution
of the compatibility equation set described in for a free turbine engine turboshaft. The algorithm is given in Figure 19. It
describes four nested loops, one for each compatibility equations (the speed compatibility is directly obtained, because
XNH is data) but the equations are solved one after each other.
In such a model, the calculation at given gas generator speed is privileged. In order to calculate the engine cycle at given
output power, or at given HP turbine inlet temperature, etc., it is necessary to add one more loop in which XNH varies
until the data parameter reaches its given value.
3.7.2.
0-D MODELING OF A TWIN-SPOOL MIXED FLOW TURBOFAN ENGINE
This example illustrates the complexity of the local iteration method: numerous loops are nested in order to satisfy the
compatibility equations one after each other. Furthermore, in this case, a strong coupling between the fan and the mixer
will slow down convergence, which would not be the case for a separated flow turbofan.
3.7.2.1
ADVANTAGES
The local iteration method was, historically, the first one used because, it is very close to the manual procedure to
3-33
3-34
data
TS0, PS0, XM0
T1, P1
data
XNH
T2
Initial
23
T2, P2
XNR2
Compressor
model
23
1st
iteration?
23
modification
W2R2/WR4
compatibility?
WR2
PW23
1st
iteration
WF
Combustor
model
WF
modification
Gas generator
power balance?
data
XNH
T4, P4
T4
Initial
P4Q45
XNR4
HP turbine
model
P4Q45
P4Q45
modification
1st
iteration?
WR4/WR45
compatibility?
WR4
data
XNPT
T45
Initial
P45Q5
T45, P45
PW445
XNR45
Power turbine
model
P45Q5
modification
1st
iteration?
PS9/PS9
compatibility?
WR4
T5, P5
Nozzle model
PS9
3-35
modeled components, or by using an approximate tabulated working line. The successful convergence and the
calculation time are dependent on the quality of this step. Let us denote Xo as the initial unknown vector.
The second step is to apply the Newton-Raphson method. This consists of linearizing the function F around the point
Xo, such that:
F(X) = F(Xo) + A(X - Xo),
where A denotes the Jacobian matrix of F (dimension n x n). ach element Aij of A, represents a correlation or influence
coefficient of the j-th unknown on the error committed for the i-th equation. These n2 partial derivatives are obtained by
finite difference calculation:
This matrix contains a lot of zeros, which represent all the local iterations that were exploited in the method local
iteration approaches described in section 3.8.2.
To achieve the goal, which is to have F(X) = 0, we must have:
X = X - Xo = -A-1 F(Xo)
By inverting the Jacobian matrix A, it is possible to calculate a variation of the unknowns X and thus to determine a
new point X1. The linearization process can then be repeated at point X1. This procedure defines a series of values Xn
that converges to the solution point X such that F(X) = 0.
The following remarks apply:
If the Jacobian matrix is singular or ill conditioned, it means that the choice of the unknown parameters and
compatibility equations is not appropriate. One of the unknowns may be not representative or an equation may be
absurd, or the unknowns are not independent.
It is often necessary to limit the variation of the unknowns at each iteration because the problem is non-linear and the
solution may diverge for big variations of the unknown parameters.
The following algorithm may summarize this method. Other methods are available to solve a non-linear set of equations
but Newton-Raphson is very often the most robust.
Unknowns
initialisation
Xn+1 = Xn + X
Xn
X limited
Error computation at
point Xn
Unknowns variations
limiter
Jacobean matrix
inversion
F(Xn)
A-1
X
Unknowns variations
computation
X = - A-1 F(Xn)
No
Convergence?
Yes
END
3-36
3.7.4.
0-D MODELING OF A FREE TURBINE TURBOSHAFT ENGINE:
The example hereafter shows an application of the global iterations method when it is applied to the resolution of the
compatibility equation set described in for a free turbine engine turboshaft. Let us take the same example as in Section 2.
In order to compare both methods, we will analyze the same calculation case, that is to say, taking the gas generator and
power turbine real speeds as power setting parameters.
The unknown vector X is:
XNR 2
b23
WF
X = XNR 4
P 4Q5
XNR 45
P 45Q5
The set of equations to be solved is:
Mass flow: HP compressor and HP turbine
WR 2( XNR 2, b 23) P 2
T2
+ WF
WR 4( XNR 4, P 4Q 45) P 4
T4
=0
WR 4( XNR 4, P 4Q 45) P 4
T4
=0
1)
2)
[
[
3)
XNR2
XNR2
dXNR2
XNR2
4)
XNR2
23
d23
23
5)
XNR2
23
WF
dWF
WF
6)
XNR2
23
WF
XNR4
dXNR4
XNR4
7)
XNR2
23
WF
XNR4
P4Q45
dP4Q45
P4Q45
8)
XNR2
23
WF
XNR4
P4Q45
WF
WF
XNR4
XNR4
P4Q45
P4Q45
XNR45
XNR45
P45Q5
P45Q5
]
]
+ WF
XNR4
P4Q45
XNR45
P45Q5
+ XNR4
P4Q45
XNR45
P45Q5
+ P4Q45
XNR45
P45Q5
+ XNR45
P45Q5
+ P45Q5
XNR45
dXNR45
XNR45
P45Q5 + ]
dP45Q5
3-37
data
TS0, PS0, XM0
PS0
XNR2
23
Compressor
model
PW23
XNPTdata
E
R
R
O
WF
T3, P3
R
S
XNR4
P4Q45
Combustor
model
XNR45
P45Q5
T4, P4
F(X)
WR4
HP turbine
model
PW445
F(X)
C
A
L
T45, P45
C
WR45
Power turbine
model
U
L
PW445
A
T
I
T5, P5
O
N
S
PS9
3.7.5.1
ADVANTAGES
The advantages of the global method are clear when comparing the algorithms of the two method types used for the two
former examples. In the global method, the cycle and components calculation is clearly separated from the mathematical
resolution instead of being mixed because of numerous nested loops in the local iteration method.
3-38
Thanks to this separation, the resolution algorithm does not depend on the engine architecture under consideration.
There are only small differences between the algorithms of the two former examples:
Size and unknown types of the vector X;
Size and equation types of the vector F;
Component calculation.
Therefore, the global method allows multi-architecture 0-D performance programs to be built easily, which was
impossible with the local iterations method.
By separating component calculation and resolution, it is easier to modify the engine modeling and thus to maintain the
program because the resolution part remains unchanged. This includes accounting for secondary effects.
The global method allows a simultaneous convergence of all the unknowns instead of accumulating nested loops in
which only one unknown converges. This generally results in timesaving as soon as the modeled engine becomes
complex or presents cross couplings.
3.7.5.2
DRAWBACKS
The global method does not have really drawbacks but its usage requires more attention:
Being based on non linear equations resolution, the global method uses more sophisticated mathematics. Thus it requires
more development work than the local iterations method.
In particular, the development of the initialization step and variation limitation step requires a very careful analysis in
order to ensure convergence in the whole range of computation cases (high max variation of the unknowns reduce
iteration number but also stability of the method).
3.7.6.
RELAXATION
Relaxation techniques use the same derivatives that are created for either local or global iteration. However, instead of
solving the system at each step, the information is used to project a solution without explicitly resolving the errors. A
damping factor, to reduce the sensitivity of the change in independent variables to the change in error term, drives the
solution to the same result as direct iteration solution but with much less mathematical calculation. A relaxation
approach is often more effective in highly non-linear regions of the simulation.
3.7.7.
CONSTRAINT HANDLING
The iteration schemes described above are used to satisfy continuity and the implicit nature of the simulation. They also
assume the dependent and independent variables in the iteration are within the limits of the component and engine
simulation. Options for constraining the iteration for these limits or to add additional iteration variables to match data or
maintain other limitations are:
3.8
A range of numerical utilities is commonly used in conjunction with the component models and iteration scheme for in
engine performance models. Some of these are listed below.
a)
b)
Integration
Transient simulations and mission simulations require integration of values over time. Simple Euler integration
may be used in high frequency simulation while more accurate predictor-corrector or Runge-Kutta techniques will
be used when larger time steps are used.
c)
3-39
agreement with the most likely estimates for component performance. Often the goal of a model is to predict not
just the most likely engine operation but also the level of performance that can be guaranteed at a confidence level
consistent with the acceptable risk for the application. Here both Monte Carlo and more analytic uncertainty
evaluations may be used. Optimization can also be used as an alternative approach to the iteration techniques
discussed previously.
The gas turbine performance simulation can be sharply divided into two categories: design point analysis and off design
modeling.
The first category mainly involves the engine designer because it consists in selecting the best thermodynamic cycle in
order to achieve a performance goal: delivered shaft power for helicopter engines, net thrust for airplane engines, bleed
air flow at given pressure, temperature conditions for Auxiliary Power Unit (APU) starts.
This analysis is led at a single working point, the so called design point, which is supposed to be representative of
typical customer use. The design point analysis allows optimization of the cycle and preliminary design of the engine
components by:
Selection of compressor pressure ratios and turbine inlet temperatures;
Comparison of different engine configurations such as single spool or twin spool;
Varying the number and type of stages for compressors and turbines.
It is only after this analysis that a first engine geometry is defined. At this step, the engine performances are known only
at design point. In order to estimate the performances under various ambient air conditions and power or thrust settings,
it is necessary to create an off design model which has the ability to describe the behavior of the engine components in
conditions other than those at design point. Such modeling involves both designers and engine users (aircraft
manufacturers and aircraft operators) because it is common to predict the performances during development and to
define them in the whole working domain once the product definition is frozen.
It must be underlined that the importance of such off design models has increased since the beginning of engine
performance simulation. In the late sixties it was found sufficient to optimize the engine at a single working point and
extensive testing permitted qualified the technologies and established the performance. But nowadays, mainly for
economic reasons, an off design model is required at a very early stage of engine design:
To match the customer specifications: it is necessary to optimize the engine within the whole working envelope and not
just at one design point. For example, a turboshaft engine for a helicopter has to be optimized at both take off rating and
economical cruise power ( 50% take off) to satisfy both the maximum take-off weight, and the range requirements.
The meaning of the word performance is becoming wider and wider. It initially dealt with power or thrust and fuel
consumption in steady state conditions, but the following topics are now fully part of engine performance.
Steady state:
Exhaust gas emissions
Noise
Transient:
Starting time
Acceleration time
All these types of performance have their own dimensioning points in the working domain, thus there is an increasing
need for global optimization of the engine performances that implies extensive use of off-design modeling.
To reduce testing costs, many tests have been replaced by theoretical analysis in fields such as mechanics, and engine
control. These analyses need off design results as input data.
To reduce development cycles, it is necessary to design all aspects of the engine in parallel. This means that all
specialists in aerodynamics, mechanics, control, and external equipment have to be provided with accurate data before
the engine has any material existence, and even before the passage of components to partial test bench. The only
theoretical source for these data is an off-design model.
For all these reasons, an accurate off design model is at the heart of the engine design process and becomes increasingly
critical.
4.1
All off design models aim at computing the fluid state in different locations of the mainstream in the engine, from these
results, mainly W, T, P, it is possible to derive powers, thrusts, fuel consumptions and all characteristic parameters of
the components.
3-40
The off design models can be split into different categories according to the level of discretization of space. Historically,
the first models belong to the 0-D category because the averaged fluid characteristics are computed at discrete positions
inside the engine, generally at the inlet and the outlet of each component such as compressor, combustion chamber,
turbine, and exhaust nozzle.
The next generation of model is the 1-D type and introduces continuity in the computation: the fluid characteristics are
still averaged in each plane, where plane means a fluid section perpendicular to the engine axis. They are computed
quasi-continuously, limited by the space discretization step, along a mean line representing the average trajectory of the
fluid inside the engine.
2-D and 3-D models extend this description by discretizing the whole flow path inside the engine, and not just the mean
line of 1-D models. 2-D models consider there is symmetry of revolution for the stream, while 3-D models make no
simplification and use the complete equations of conservation.
The goal of this chapter is to describe the 0-D models, which are the simplest and most widely used in the industry, by
analyzing the assumptions, the phenomenon modeling, the computation methods and the limitations of such models.
4.2
0-D MODELS
These models are the most widely spread in the world of turbo-machinery for many reasons:
Historically, they were the first kind of performance model that was used.
They do not require a detailed description of the engine geometry. Thus they can be used very early in the engine
development.
The engine description is simple and close to reality by considering it as a set of black boxes, one for each major
component of the engine.
The calculation methods are simple because the number of unknowns needed by the modeling is reduced, the goal
being to compute the fluid characteristics only at the interfaces of the black boxes. The number of unknowns has the
same order of magnitude as the number of individual components considered as black boxes.
The calculation methods are natural because the fluid characteristics are computed plane by plane, or station by
station, in the same order as the one used by the fluid to pass through the engine. The calculation begins at the air
inlet, continues with the compressors, the combustion chamber, the turbines, and ends with the exhaust.
Thanks to the simplicity of 0-D models, they can be run on all computer types, from the small portable PC to the
workstation or the main frame. This is a very useful feature of 0-D models, because of the wide range of people who
can have to run engine performance models. Such models constitute a common tool for the designer, the integrator,
the customer; and the engine designer. As the main model provider, the engine designer finds issuing customer
decks, based on the same approach as the one he uses for development, very convenient. This results in time and cost
savings and quality gains.
Firstly, we will analyze the steady state models which were the first developed, and which allow simulation of the
engine in steady state operation, that is to say, in non time-dependent working conditions.
We will then show how to derive a transient model from the steady state one. Steady state and transient modeling are
fully linked but transient modeling raises specific problems, in taking into account the physical transient phenomena.
Nevertheless, the performance programs used today by designers generally allow both types of computation. Those
issued for external use, engine computer decks for example, are following the same tendency albeit with some delay, as
proved in AS 681 revision F aerospace standard which unifies the presentation of computer programs for both steady
state and transient operation.
4.3
4.4
The general technique for obtaining a 0D solution is illustrated below by way of a simple single-spool turbojet example.
The calculations employ iteration; the global iteration approach is described. The calculations can be broken into the
modules described earlier in the chapter. Consider an engine represented as follows:
Compressor
Pressure Ratio, Efficiency, Non-dimensional Flow = f(NT, )
Combustion
T4 = f(T3, FAR, W4)
Turbine
WT/P4 & Effy = f(H/T, NT4)
Nozzle
3-41
3-42
More complex representation of the same engine architecture would typically include matching pairs involved with the
modeling of re-circulating bleeds, intakes, afterburning etc.
4.4.2.
TRANSIENT SOLUTION
This 3x3 matching scheme is easily extended for a simulation which models the dynamics associated with the shaft. In
this case, required power-level must be specified as a function of time. For example WFE might be a ramp function
against time.
We now have an extra descriptor of engine behavior, this time a dynamic equation that describes the behavior of a statevariable in the time domain:
dN PW
=
dt
J .N
where:
This is the rotational form of F=ma and concerns the conservation of angular momentum.
The engine calculation process is largely unchanged. The difference lies in the treatment of shaft speed. Numerical
integration is required to obtain the speed at each timestep. There are several methods of numerical integration, in the
interests of clarity, Eulers explicit method is use in this example.
4.4.2.1
SOLUTION BY EXPLICIT EULER INTEGRATION
Eulers explicit integration method is:
x(t + t ) = xt +
dx
t
dt t
dN
dt
t
now
dx
dt
t
t + t
3-43
xt = x (t t ) +
dx
t
dt t
dN
dt
t
now
At any particular timestep, the speed at time = t-t is known and Euler's equation can be solved if the current speed and
acceleration (N_dot) can be determined. If the speed is an iteration variable, then state-derivative can be obtained by
looking backwards (rearranging implicit Euler equation):
dx
xt x (t t )
=
t
dt t
The iterative/implicit approach requires the matching condition to be:
State-derivative from dynamic equation = state-derivative from Eulers implicit equation:
Nnow Nlast
PW
=
t
J .Nnow
In the interests of iterative convergence, this is best rearranged with the excess power expressed in its constituent terms:
J .Nnow PWc J .Nlast PWt
+
=
+
Nnow
Nnow
t
t
... Where PWc is the total power requirement (i.e. includes compressor power, losses and offtakes) and PWt is the
turbine power.
The bandwidth of a dynamic model, which only contains the shaft dynamic equation, has a bandwidth of around 5Hz.
For wider bandwidth, other dynamic events must be modeled. Better representation in the higher frequency ranges
requires the modeling of gas dynamics (the behavior of the air in each area of the engine). For better representation in
the lower frequency range, heat soakage (the transfer of heat to and from the blades and carcass) must be modeled. Gas
dynamics are fast events and numerical stability of the integration method is a significant issue. The implicit approach
described above is perhaps most suitable given that the cycle match process is iterative by nature. Non-iterative 0D
modeling techniques are more constrained to use explicit approaches and in these cases a more sophisticated integration
method is required (e.g. Runge-Kutta 4th order).
4.4.2.3
EXTENSION OF SIMPLE EXAMPLE TO INCLUDE GAS DYNAMICS
In the example so far, it is assumed that the gas is incompressible, that the flow into any control-volume at any instant is
equal to the outlet flow. In reality, there is some storage of gas in a volume. There is therefore an imbalance of gas
properties (inlet to outlet) relative to the quasi-steady values encountered at each timestep in the example above. This
imbalance implies a rate of change of properties within each volume.
The dynamic equations that describe these phenomena are listed below. The terms refer to inlet - outlet.
Conservation of mass
d
W
=
d t Volume
Momentum
d W ( pA + Wv)
=
dt
Length
Energy: enthalpy form
d
(W .H )
( H - p) =
Volume
dt
Energy: entropy form
d
(W .s )
( s ) =
dt
Volume
Some models may use simplified forms of these equations that assume low Mach number flow. A typical equation set is:
3-44
Mass
dP
W
= R.T
dt
Volume
Momentum
dW
A P
=
dt
Length
Energy
d T (W .T ). T .W
=
dt
M
M
where M =
P.Volume
R.T
The state variables in the equations above refer to the average component properties although in some model
implementations, the same equations are used in an actuator volume sense. Rather than being applied to the average
component properties, the gas dynamics are applied across a dummy volume at exit to the quasi-steady-state (qss)
process.
Returning to the example, the engine can be broken down into several control-volumes, corresponding to each separate
component or several components lumped together. For each control-volume there are now 3 more unknown quantities
(variables) these are:
P: 'stored' pressure in volume (cons. of mass)
W: 'stored' flow in volume (cons. of momentum)
T: 'stored' temperature in volume (cons. of energy)
For implicit integration, the matching quantities are set up in the same manner as for the shaft dynamic above.
Essentially, the matching condition is the equality of the two expressions of state-derivative. As before, some algebraic
re-arrangement helps iterative convergence.
Conservation of mass
Volume.
Pqssnow
Pqsslast
+ Woutnow = Volume.
+ Wqssnow
t.R.Tqssnow
t.R.Tqssnow
Conservation of momentum
Length.
Woutnow
Woutlast
+ A.Poutnow = Length.
+ A.Pqssnow
t
t
Conservation of energy
M .Toutnow
M .Toutlast
+ outnow.Woutnow.Toutnow + Toutnow.Wqssnow =
+ qssnow.Wqssnow.Tqssnow +Toutnow.Woutnow
t
t
Pqssnow.Volume
where M =
R.Tqssnow
It is not necessary to solve all 3 conservation equations for each volume. For around 10Hz bandwidth, only the
modeling of mass conservation may be necessary.
Note that if the momentum equation is solved, then a length term is introduced into the model, hence the model can be
regarded as a 1D representation.
4.4.2.4
FURTHER EXTENSION OF THE EXAMPLE TO INCLUDE HEAT SOAKAGE
The heat transfer between metal and gas can be highly significant. The long-term stabilization of the engine onto a
temperature limit at a power level is important from an operational point of view, as is the behavior of the engine on hot
reslams (i.e. soaked high power idle high power). In the latter case, the metal is relatively hot compared to the
shaft speeds. The aerodynamic speeds can therefore differ greatly from normal idle conditions. Compressor stability can
be affected.
A cold slam (soaked idle max power) modeled without heat-soakage effects can be misleading especially for large
engines, as in addition to the aerodynamic speed shifts mentioned above, power lost to the heating up of engine parts
3-45
(k . Tblade _ now)
(k . Tblade _ last )
= Tblade _ now
t
t
Tg
(k . Tcarcass _ now)
(k . Tcarcass _ last )
= Tcarcass _ now
t
t
3-46
3-47
accurate [consistent], and numerically stable results. Of course, bandwidth and timestep are linked. The advantage of
cycle-match models is that if a low bandwidth is required, then the model may be stable at the large integration
timesteps commensurate with the dynamics of interest. e.g. an aircraft simulator may only require outputs at 50ms.
The prime issue at stake here, is the use of iteration in a model on which is placed strict constraints on execution time.
Clearly, no specific guarantees can be made on the number of iterations to match the cycle and so a limitation has to be
imposed in order to give a model that has predictable execution times. It has been shown that a model limited to as few
as 2 passes is viable. Truncation of iteration in this way may lead to loss of consistency, however this discrepancy is
likely to be less than the difference between the iterative model (which is taken as the reference computer definition of
the engine cycle) and a separate, non-iterating real-time model.
This being so, there are also potential run-time advantages to be cashed for some model applications. A model requiring
4 passes every 50ms uses less computing power than a non-iterative model running every 1ms (where 1ms may be
required for stable running). The computing effort advantage amounts to a run time factor of around 10. The advantage
is not so marked for applications requiring high bandwidth RTEMs. An iterative model would run 2-4 passes at 1ms; a
non-iterative model would typically run (1 pass) at 0.5ms or less. The comparison in this case may not be so
spectacular!
Correct handling of gas dynamics in a non-iterative model may need extremely small timesteps, perhaps 0.1ms or less.
Such a small timestep is not desirable for real-time work, a larger timestep is required. The usual way of ensuring
numerical stability with a larger timestep is apply factors to selected volumes. Consequently, a larger timestep can be
tolerated but at the expense of dynamic fidelity. The run time comparison for correct dynamics representation is still
therefore in the iterative model's favor. It must be said, however, that with the advance in computing power, run-time
advantages become less of an issue; it is the benefits of commonality and versatility that come to the fore.
Why stop at real time? There is perhaps the potential to further exploit Monte-Carlo type methodology to engine and
controller design where ultra-fast non-linear models could be used effectively.
4.5
A 1-D model typically tries to extend a 0-D cycle model representation by physically modeling an additional dimension.
The most commonly used 1-D models are for dynamic simulations where 0-D models require the addition of a length
dimension to adequately model high frequency behavior. The next most common use is where components or local
areas of the engine are modeled at a higher level of fidelity in either the radial, circumferential or axial direction to
address specific concerns. Often these models remain at 0-D fidelity in other parts of the engine. 0-D and 1-D models
may be used predict 2-D and 3-D variations in engine conditions and may use a physical model to do it. But the physical
models stay at the lesser level of detail. An example is the use of a more detailed compressor model, which predicts the
variation in driving pressures at the hub and tip, and then uses these to model internal leakage and cooling circuits.
4.5.1.
STEADY STATE 1-D MODELS
Steady state 1-D models are typically used to provide detailed information inside one or more components such as:
Blade row, cooling passage or cavity level information. In some cases this may not be a true 1-D model but merely a
0-D model taken to a greater than normal level of detail.
Tracing radial or circumferential information through the engine. This could be inlet pressure or temperature
distortion, water or fuel cloud, or a non-uniform condition created inside the engine. This might be in the burner due
to a fuel nozzle variation, or in a compressor due to a locally off-angle sector of stators.
When providing more detail inside a component, more detailed boundary conditions may not be required. For models
addressing radial or circumferential behavior through the engine, creation and transfer of the additional boundary
conditions is the key part of transient 1-D models.
4.5.2.
TRANSIENT 1-D MODELS
Often these models are identical to a steady state 1-D model and only include the transient extensions found in 0-D
models. This includes models of low frequency phenomena such as rotor accelerations, heat transfer and active or
passive geometry changes. Active geometry changes modeling can include dynamics associated with variable stators,
variable bleed-valves, or modulated cooling flows such as anti-ice air or variable exhaust nozzles. Passive geometry
changes can include turbine or exhaust nozzle areas, which change with temperature or tip clearance. Heat transfer
effects must typically cover a range of time constants (from a few seconds to a few minutes) to accurately model engine
behavior. These models may include volume dynamics to address leakage or cavity flows or to meet other accuracy
needs. However, these models are generally not intended to address acoustic level phenomena such as stall, surge or
combustion instability.
4.6
The purpose of this section is to describe recent advances in modeling gas turbine engine dynamic behavior. A dynamic
3-48
turbine-engine combustor model and simulation, VPICOMB, which was developed at Virginia Polytechnic Institute and
State University, will be discussed. The integration of the VPICOMB combustion model equations with the DYNamic
Turbine Engine Compressor Code (DYNTECC), conducted at the Arnold Engineering Development Center (AEDC)
will then be discussed (Hale and Davis, 1992). Finally, a full gas turbine engine model and simulation, the Aerodynamic
Turbine Engine Code (ATEC), also developed at AEDC, will be described. ATEC currently has the ability to simulate a
turbojet engine with the compressor system operating post-stall, and a turbofan engine operating up to the point of
compressor stall.
The governing equations are derived by the application of mass, momentum, and energy conservation to the elemental
control volume where:
u 2 + P
F=
Q
u
S=
1
P
2
1 2
+
+
u
e
u
e + 2 u
2
;
;
w B x
Fx
Q + S H
x
Bx
x
P = RT
is used. The assumption of constant specific heats (Cp) and ratio of specific heats ( ) in the property calculations, while
computationally efficient, does result in predicted temperature levels in the combustor that are too high during rich
combustion. This effect will be shown in the following sections. The distributed turbomachinery source terms
wBx , Fx , Qx + S x H Bx
are supplied to the overall model by the user. Models for the source terms will be discussed in each of the respective
sections.
The time dependent flow field within the system of interest is obtained by solving the time dependent system of
equations using one of two numerical approaches. The VPICOMB model uses an explicit Roe's flux-differencing
scheme adapted to the Euler equations with source terms to evaluate the face fluxes, and then uses a second-order four
step Runge-Kutta algorithm to solve for the time dependent equations. DYNTECC and ATEC use a flux-differencesplitting scheme based upon characteristic theory (Varner, et. al., 1984) to solve for the face fluxes. A first order Euler
method is used to solve for the time dependent equations.
4.6.1.
COMBUSTOR COMPONENT VPICOMB
The VPICOMB model and simulation provides a one-dimensional tool running on the personal computer for analyzing
dynamic gas turbine engine combustor operation. With VPICOMB, the user can analyze the influence of varying inlet
conditions, fuel pulses, exit flow restrictions, and many other possible dynamic events. The user provides the simulation
with appropriate initial and boundary conditions, plus any time dependent variations in the boundary conditions, and
then exercises the program to determine the time dependent flow field.
For the inlet boundary conditions, the user specifies inlet total pressure, total temperature, and flow rate. The user can
also specify the friction factor along the wall of the combustor. No pressure loss due to combustion is assumed. The exit
boundary condition assumes a constant value for the mass flow parameter:
Wex Tex
Pex Aex
= Constant
The value of the mass flow function is obtained during the initial condition calculations. It is assumed that the fuel mass
flow addition occurs in the first control volume. The heat release due to combustion occurs in the control volumes
specified by the user as the zone of heat release. The heat release is equally distributed across the zone of heat release
control volumes. The amount of energy released in the zone of heat release control volumes is a function of fuel flow
rate, the combustion efficiency, the lower heating value of the fuel, and the combustor flammability limits.
The combustor flammability limits are determined by using steady state engineering correlations developed by Herbert,
1957. In order for stable combustion to occur, the primary zone equivalence ratio (PZ) must fall within a rich and lean
limit:
L PZ R
Based on experimental data, Herbert defined a Combined Air Loading Factor to calibrate the light off and blow off data.
A polynomial curve fit of Herberts flammability data for a generic can type combustor is used in the VPICOMB model.
Combustion efficiency is determined by using steady state engineering correlations developed by Lefebvre, 1985.
3-49
Lefebvre assumed that the efficiency of fuel evaporation and the reaction efficiency limit the overall combustion
efficiency. Further modification to the Lefebvre work was done by Derr and Mellor, 1990.
Because of the dynamic operation of the combustor, it is possible for heat release to occur for a short period of time
even though the combustor equivalence ratio may lie outside the steady state flammability limits. Likewise, the heat
release process may not resume immediately after the combustor equivalence ratio reenters the flammability bounds. To
account for these effects, a first order lag on the heat release rate has been incorporated in the model:
o
o
dQ o
+ Q = Qss (t )
dt
3-50
1.0
1.8
Normal
Operation
Rotating
Stall
1.6
Reversed
Flow
1.4
1.2
1.0
-1.0
Zero Flow
-0.4
-0.2
0.2
0.6
1.0
Temperature Coefficient
Pressure Coefficient
2.0
0.6
0.2
Rotating
Stall
Reversed
Flow
Normal
Operation
-0.2
-0.6
Zero Flow
-1.0
-1.0
-0.4
Flow Coefficient
-0.2
0.2
0.6
1.0
Flow Coefficient
d ( FX )
+ FX = FX ss
dt
(2)
The time constant, , is used to calibrate the model to provide the correct post-stall behavior. The inflow boundary
during normal forward flow is the specification of total pressure and temperature. The exit boundary condition is the
specification of exit Mach number or static pressure. During reverse flow the inlet is converted to an exit boundary with
the specification of the ambient static pressure. Therefore, both the inlet and the exit boundary function as exit
boundaries during a surge cycle.
An explicit split-flux finite-difference algorithm is used to numerically solve the area weighted quasi-one-dimensional
Euler equations. The quasi-one-dimensional Euler equations with source terms (equation 1) are written in conservation
Cartesian form and applied to a fixed grid. A finite difference representation of equation 1 can be applied over an
interval between grid points j and j+1 with the fluxes evaluated at the nodes and the sources evaluated at the center of
the volume given by
U
+ U
U
1 + I 1
= I 1
t j
j + t j + 1
j t j
2
2
2
2
(3)
where
( F j F j 1 )
U
;
G
=
t j 1 j 1 ( x j x j 1 )
2
( F j +1 F j )
U
= G 1
t j + 1 j + 2 ( x j +1 x j )
2
+
Characteristic theory is used to develop weighting terms ( I , I ) for splitting the time derivatives to the adjacent
nodes as illustrated in Figure 26.
3-51
n+1
t
U
t j 1
t j
t j + 1
j-1
j-1/2
j+1/2
j+1
C OM PRESSOR M ODEL
D ISTORTION
P ATTERN
L OW P RESSURE
H IGH P RESSURE
3-52
Flow
Compressor
Gas
Combustor Generator
Turbine
Power
Turbine
Flow
Compressor
Combustor
Turbine
Location
Mfg
ATEC
%Delta
Mfg
ATEC
%Delta
Inlet
0.29
0.29
0.00
0.26
0.26
0.00
Compresso
r Exit
Burner Exit
2.38
2.36
0.73
0.51
0.51
0.72
2.29
2.25
1.76
1.16
1.15
0.47
Gas
Generator
Turbine
Exit
Power
Generator
Turbine
Exit
0.81
0.87
-6.27
0.93
0.96
-3.63
0.30
0.30
-0.75
0.75
0.79
-5.22
3-53
4.6.3.
COMPRESSOR COMPONENT DYNAMIC SIMULATION
To demonstrate dynamic operation of ATEC, the same test case as was used to demonstrate the integration of the
VPICOMB model equations into DYNTECC has been exercised.
Variation of the relative total pressure in the engine is shown in Figure 29. During the initial steady state operation, the
total pressure increases through the compressor system. In the combustor, a small total pressure loss occurs. Work
extraction in the turbines reduces the pressure back to near atmospheric before the flow exits the engine. As with the
DYNTECC test case, the fuel flow pulse forces the compressor into surge. Rather than being driven by a constant Mach
number exit boundary condition, however, the pressure increase is tied to the turbine choking as explicitly defined by
the turbine steady state operating characteristics. Steady state operation is not reestablished until the fuel flow rate is
decreased back to the original flow rate. The frequency of the surge cycles is reduced due to the increased volume of the
calculation domain.
Relative total temperature in the engine as a function of time is also shown in Figure 29. As the compressor enters the
surge cycle, the reduction of air mass flow rate causes the combustor temperature to increase dramatically, until the
equivalence ratio rises to the rich flammability limit. After the surge cycle is completed, the combustion process is
reestablished until the next cycle forces the equivalence ratio to rise above the flammability limit.
Relative mass flow rate in the engine as a function of time is shown in Figure 30 In addition to the bleed extraction
occurring in the axial compressor and the fuel addition occurring in the combustor, turbine-cooling bleed is injected into
the flow in the last gas-generator turbine control-volume. As with the DYNTECC test case, during each of the surge
cycles, the mass flow rate in the front section of the engine reverses and becomes negative. In the back section of the
engine, the flow rate is greatly reduced, but it does not reverse.
To demonstrate the capability of the model and simulation to simulate rotating stall in the compressor system, the above
test case was repeated with a modification to the time-lag constant used with the compressor characteristics in the
rotating stall regime. By increasing the time constant of each stage of the compressor system, the operational
characteristics change. This results in the compressor system being unable to return to the normal operating mode. The
effect is shown by comparing the compressor pressure ratio curves as a function of mass flow rate through the system
for both cases. This comparison is shown in Figure 31. With a relatively small time constant, the dynamics of the system
allow the compressor to return back to the normal pre-stall operating speed line before reentering the surge cycle, as is
shown in Figure 30. As is shown in the various figures, once the perturbation in fuel flow rate is removed, the system
returns to normal, steady state operation. An increase in the compression system time constants forces the system into
rotating stall, as is shown in Figure 31. Even when the fuel flow rate is reduced back to the original level, the system is
unable to recover back to the original steady state operating point.
3-54
4.7
1-D models provide additional boundary condition details not covered by 0-D models. The reason for using a more
detailed model in place of a purely empirical correlation based on 0-D conditions is to obtain greater accuracy or better
fidelity in response to changing conditions. Correlations to extend 0-D models are often difficult to create and require
significant effort to recreate if engine operating conditions or component representations change.
4.7.1.
COMPUTER HARDWARE ISSUES
Detailed 1-D models are generally well adapted to desktop PC and unix workstations. They do not represent a
significant computational barrier unless near real-time transient performance is required or they are to be part of a larger
simulation or a study requiring a large number of runs.
4.8
2-D and 3-D models not only predict but also attempt to model the physics at this higher level of detail.
0-D or 1-D models may be used predict 2-D and 3-D variations in engine conditions, and may use a physical model to
do it, but the physical models stay at the lesser level of detail. Engine system level models at this level of detail are just
becoming practical from both computation and component modeling needs.
Detailed models provide potential for closer modeling of the physics, reducing the need for empiricism and to provide
insight into the component operation and variation. The primary near term benefit of using 2-D and 3-D models of an
engine system is to provide insight into the component operation details that 0-D and 1-D models cannot.
4.8.1.
ZOOMING
4.8.1.1
2-D AXI-SYMMETRIC MODELS
2-D Axi-Symmetric models have been the workhorses for compressor and turbine design for many years. 3-D tools are
gradually replacing them. The primary benefit of using 2-D system models is to facilitate the design and detail analysis
of a particular component across the range of possible boundary conditions. In the past, running lower fidelity models
did this. More detailed boundary conditions were created through an empirical process, and then the 2-D code was run
in isolation. Computing capacity made this process the only reasonable option.
It is now possible to run a 2-D component simulation, in conjunction with a simplified model of the rest of the engine, in
a few minutes and to even run the entire engine in less than an hour. The choice now falls between:
Modeling a component or the entire engine at the 2-D level, which must be justified in terms of the extra model
development complexity and effort required to add value to the results;
Obtaining estimates that can be made based on a simpler model.
Currently, it is only in the design and detailed insight into the turbo-machinery where the added information justifies the
more detailed modeling. The engine level performance impact of this greater detail is generally lost in the overall
empiricism and modeling uncertainty. For example, 2-D modeling of clearance effects can be of interest in the analysis
and design of the compressor. However, the predictive accuracy and stability of currently available models does not
justify their use in a whole engine model. Figure 32 indicates typical characteristics for different types of model.
3-55
Component Design
Component Analysis
Engine Analysis
Accuracy
Stability
Computation Speed
High
Medium
Low
Low
Medium
High
Low
Medium
High
4.9
3-D MODELS
Full physics 3-D CFD models for single blade rows and components are beginning to see use outside of research areas.
For system models, virtually all require some level of empiricism or simplification such as use of source terms or body
forces for the turbo-machinery or an average passage or mixing plane assumption. These techniques are discussed in
chapter 5.
4.10
4.10.1.
COMPUTER HARDWARE REQUIREMENTS
It is now possible to model engines at a 2-D level or even limited 3-D representations using a desktop PC or single user
engineering workstation. However, even modeling a single component at full 3-D fidelity generally requires a RISC
workstation computer definition. Full physics 3-D modeling of the primary flowpath typically requires multi-processor
parallel machines to achieve overnight execution. Even this may require extensive simplification or great care in
defining the computational structure.
4.11
These simple, fast models are used as surrogates of more complex thermodynamic models in a variety of applications
that span the full range of gas turbine design, development and operation. Parametric models are employed in
applications where speed is valued over accuracy or where uncertainty in the engine component characteristics or engine
environmental conditions does not warrant thermodynamic model complexity. Generally, they achieve computational
simplicity and speed at the expense of accuracy and resolution relative to the thermodynamic models from which they
are derived. However, they typically offer equal or better accuracy than a simplified thermodynamic model of
comparable speed. The formulation of parametric model always involves some form of data fitting such as polynomial
functions or piece-wise linearization. In addition to speed, these models also provide a means of conveying engine
performance characteristics while protecting proprietary component characteristics and modeling methodologies.
4.11.1.
APPLICATIONS
Some typical applications of parametric models are discussed below. Model formulations and synthesis procedures are
discussed in the following sections.
4.11.1.1
CONCEPTUAL DESIGN
The first parametric models were simple tabular listings or graphs of dependent versus independent performance
variables. These continue in use today and in the foreseeable future in the form of multivariate computer databases,
which are used in conceptual vehicle design. The objective of the conceptual design process is to effectively converge
on an optimum vehicle design including engine system requirements for engine size, performance and operability.
Detailed component requirements are defined later.
Parametric models (whatever the formulation) offer several advantages in the conceptual design process:
A parametric model may be employed as a living engine specification, conveying the anticipated steady and dynamic
characteristics of an engine.
4.11.1.2
CONTROL DESIGN & VALIDATION
Parametric dynamic models have long been used in the design of control laws where linear control theory required
representation of the engine dynamic behavior as a Laplace, state-space, or other linear formulation. Today, sets of these
single point dynamic models are assembled with the corresponding steady-state models to provide full range transient
models that are employed in the control validation process as well. These models are especially effective in the real-time
validation of integrated flight/propulsion controls systems where fast execution is a requirement.
3-56
4.11.1.3
MODULE PERFORMANCE ANALYSIS
Module performance analysis (also called gas path analysis) is a process employed in fleet (on-wing) monitoring and
trending to determine the health of gas turbine components and forecast operational impacts and logistics requirements.
Today, statistical parameter estimation algorithms are employed to determine component performance indices that can
not be computed directly from the small number of on-wing measurements. These estimation algorithms are founded in
linear control theory and typically are based on state-space parametric models as described in the control application
above. In deriving the estimation algorithm, the normal cause-effect relationships represented in the model are
transposed mathematically into effect-cause relationships.
Module performance analysis can be applied to either steady state or transient data and can even be embedded in the
real-time control.
4.11.2.
MODEL FORMULATION
A parametric model is a surrogate for a physics-based parent model and must represent the important characteristics of
the parent for the intended application. In general, the physics of the parent model are represented implicitly rather than
explicitly in the surrogate model. Parametric models map the complex relationships between the dependent and
independent variables in the parent model into simple relationships between outputs (fluid pressures, temperatures,
flows) and inputs (fuel flow, variable geometry, bleed) and internal dynamic states (rotor speeds, metal temperatures), if
the model is dynamic.
The choice of input, output and state parameters is important because it determines the model accuracy and the
extensibility to operating conditions beyond those where model was synthesized. The choice of parameters is commonly
based on intuitive insight and experience, but can also be based on numerical analysis of parent model characteristics as
described in the references. Conventional gas turbine correction factors are commonly used to account for inlet
conditions.
There are a variety of mathematical formulations for parametric models ranging from simple multivariate tables for
steady-state models to state-space formulations for dynamic models. In general, the state space formulation offers the
greatest accuracy and flexibility and is well supported by commercially available software toolkits (XMath, MatLab).
Multiple state-space models can be assembled to cover the full engine operating range. Simple transfer function
representations, are a popular alternative model formulation often employed in preliminary design when engine are not
well characterized. This formulation does not even require a parent model, and can be based on expected characteristics.
4.11.3.
MODEL SYNTHESIS PROCEDURES
A parametric model synthesis process can be used. In this case, the appropriate steady state and dynamic characteristics
are extracted from the parent model. These characteristics may be computed using a variety of techniques such as simple
mapping, factorial experiments, and system identification procedures. It is desirable to automate the synthesis
procedures to reduce development time and eliminate errors. This is important because model synthesis is normally a
recurring process that must be performed whenever there are substantial revisions to the parent model.
The overall representation of the parametric model must be validated with the parent model. Typically, validation is
performed by simply comparing the parent model with parametric model outputs for a set of input cases. Dynamic
validation can be by comparing the frequency response of the parent and parametric models.
During an engine test a force, spool speeds, many pressures and temperatures at various locations along the flow-path
and positions of the variable geometry (guide vanes, nozzle area) are measured. However, these values alone are of
limited use without special interpretation. The task of the engine test analysis is to find the operating points of the
compressors and turbines in their maps, from the measured data.
This section deals mainly with turbofans because tests of this type of engine are especially difficult to analyze. It is
difficult to get a precise value for the mass flow that enters the core. This mass flow, however, is required for the
calculation of the burner exit temperature, which cannot be measured directly for a variety of reasons.
After a thorough test analysis the performance model can be calibrated and - if necessary - improved with newly found
correlations.
5.1
PERFORMANCE INSTRUMENTATION
For a dedicated engine performance test to check the specific fuel consumption, it is not sufficient to analyze only the
thrust and fuel flow. The engine inlet conditions in terms of total pressure and temperature are also needed. A precise
value for the mass flow entering the engine is very important. The measured force must also be corrected for testbed
specific effects, like cradle drag, to finally obtain the thrust.
Between and downstream of the compressors there must be a sufficient number of total pressure and temperature
probes. In the hot part of the engine the total pressure measurements at the inlet and at the exit of the low-pressure
3-57
turbine are very important. The total temperature rakes at these locations are of limited use for engine performance
analysis. There are severe temperature gradients both radially and circumferentially making it extremely difficult to get a
representative mean value for the total temperature of the main gas stream.
5.2
A conventional engine test analysis is quite simple and straightforward. The power required to drive the fan is calculated
from the engine mass flow and the total temperature increase. The fan efficiency for the core and the bypass stream can
be found from total pressure and temperature measurements. Similarly the efficiency of the high-pressure compressor is
derived.
21
24 25
16 18
13
44 45
41
3 31 4
b
c
handling bleed
LPT cooling
overboard bleeds
a HP leakage to bypass
b NGV cooling
c HPT cooling
3-58
After taking into account gearbox drag, power off-take from the high-pressure spool and the cooling and leakage air
mass flow, the work done by the turbine can be calculated. We get the ideal work from P41 (as calculated above) and the
measured total pressure P45. Turbine efficiency can be evaluated with this information.
Besides the efficiency we get a calculated value for the high-pressure turbine exit temperature T45. The power balance
for the low-pressure spool allows calculation of the low-pressure turbine exit temperature T5. These two temperatures
will not be identical to the measured data T45m and T5m. Actually, the measured hot end temperatures are ignored when
doing a turbine capacity core flow analysis.
5.2.2.
HEAT BALANCE
The heat balance method is based on a measured hot end temperature and ignores the eventually known turbine
capacity. This method makes sense when T45 or T5 is measured with many rakes. On rare occasions a rotating rake is
used downstream of the low-pressure turbine which allows calibration of the standard instrumentation with only a few
rakes for T5.
Similarly to the turbine capacity method, the core-flow analysis starts with an estimated value for W25. The highpressure turbine inlet temperature T41 is found with the same assumptions as described above. The energy balances for
the high and the low-pressure spool will yield T45 and T5.
The estimated value for W25 is modified, in such a way that after convergence the calculated value for either T45 or T5
equals the measured value.
5.2.3.
ISA CORRECTIONS
During a normal engine test at a standard sea level testbed, for example, both the inlet pressure and the inlet temperature
will deviate from ISA sea level standard conditions (T = 288.15K, P = 101.325kPa). The test results must be corrected
to these engine inlet conditions to make the data comparable. This is often done on the basis of the Mach number
similarity.
The operating conditions of a turbomachine are similar when the Mach numbers are the same everywhere in the
flowfield. For a fluid with known properties of isentropic exponent and gas constant, this is the case when the corrected
flow WT/P and the pressure ratio are the same. Also the temperature ratios and the corrected spool speed N/T will be
identical for strictly similar flow fields. Note that the variable geometry settings (guide vanes, nozzle area, bleed valve
positions) must remain unchanged during any data correction on the basis of the Mach number similarity.
It can be easily shown that when the Mach numbers are the same, the terms FN/P0 (corrected thrust) and SFC/T0
(corrected specific fuel consumption) will also be the same.
With the help of these and other simple correction formulas for flow, pressures, temperatures, spool speeds, thrust,
power off-take and SFC one can easily derive all engine data for a standard day.
The correction procedure on the basis of the Mach number similarity is not very accurate because the formulas are
strictly valid only when: the gas properties do not change, thermal expansion of the engine has no effect on tip
clearance, and Reynolds number effects do not exist.
The quality of the correction procedure can be improved empirically by slight modifications of the original formulas:
FN
= FN * P01 FN * P0 x
P0
The exponent x in this formula is adjusted empirically to give the best fit to measured or calculated data. The same
approach can be used with other quantities like SFC. For example:
SFC
= SFC * T 0,5 SFC * T y .
T0
After having corrected all measured data from several engine tests one has a sound basis for the calibration of the engine
performance model. Engineering judgement, experience, patience and many trials are necessary to get a good match of
the model to the data.
5.3
ACCURACY
5.3.1.
GENERAL REMARKS
Engine tests are performed to evaluate the overall characteristics in terms of thrust and specific fuel consumption.
However, especially during the development phase, the main purpose of performance testing is to find the efficiency of
the engine components and to prove that the design assumptions were valid. For such an analysis one needs to know the
total pressure and temperature at all component interfaces as well as the mass flows.
3-59
Any measurement (as for example a temperature probe on a rake between two components) has an uncertainty that is
affected by both random and systematic measurement errors. When it is repeated several times, the instrument readings
will not agree exactly but will show some scatter.
In gas turbine engine tests random effects in the measurement chain are caused by this scatter, and also by small changes
in engine geometry and operating conditions. An engine is never running in absolute stability because of small changes
such as inlet flow conditions, variable geometry settings, and thermal expansion of casings and disks.
In a carefully controlled engine performance test the random errors mentioned above are not negligible, but smaller than
the systematic errors caused, for example, by non-ideal positioning of the probes. There is seldom space in an engine to
put enough pressure and temperature pickups at the component interface plane. Besides that, instrumentation intrusion
effects must be minimized.
Every effort is made to correct the measurements for all known effects. However, an uncertainty remains.
The difference between the measurement (after applying all known corrections) and the true mean value is called a bias.
There are no exact data available to calculate the magnitude of a bias and therefore it is usually estimated from
experience with component rigs, for example.
5.3.2.
EXAMPLE: LOW PRESSURE TURBINE EFFICIENCY ANALYSIS
The efficiency of a low-pressure turbine (LPT) is:
H actual
H ideal
T5
T45
P 1
1 5
P45
H LPT =
W
W2
H Fan + 21 H Booster .
W45
W45
Any error in the analysis result for the engine inlet mass flow W2 will have an impact on the LPT efficiency result.
Getting a precise value for the specific fan work HFan is very difficult for high bypass engines. It must be derived from
the total temperature measurements upstream and downstream of the fan. Figure 34 shows in the left part the typical
temperature rise in the fan as a function of bypass ratio. In the right part one can see that a measurement error of 1C
will cause an error in specific fan work of 1% for engines with bypass ratio of 4 and nearly 3% for very high bypass
engines.
3-60
110
100
2.5
90
80
70
1.5
60
50
40
30
4
10
12
Bypass Ratio
0.5
10
12
Bypass Ratio
5.4
ANALYSIS BY SYNTHESIS
The conventional test analysis as described in the chapters above makes no use of information, which is available from
component rig tests, for example. It will give no information about the reason why a component behaves badly. A low
efficiency for the fan may be the result of either operating the fan at an aerodynamic over-speed or a poor blade design.
To improve the analysis quality in this respect is the aim of Analysis by Synthesis (ANSYN).
5.4.1.
PRINCIPLE
When doing analysis by synthesis a model of the engine is automatically matched to the test data. This is done with
scaling factors to the component models, which close the gap between the measured efficiency and the model. For
example, efficiency scaling factors greater than one indicate that the component performs better than predicted.
3-61
measured
N/ T map
measured
ling
sca ine
l
N/ T measured
Pressure Ratio
3-62
pickups, for example - that the mean static pressure is evaluated to be higher than the total pressure. The model will not
allow for that and the consequence is that the iteration will fail to converge.
The same can happen at minimum afterburner rating when the pressure losses due to heat addition become insignificant.
The model will never produce negative pressure losses. The measurements, however, might require such a result.
Other reasons for convergence problems can be that the actual hardware is different to the model assumptions.
While on the one side, it is very annoying when iteration does not converge, on the other side a very important hint is
given that something is wrong either with the measurement or with the model. The conventional test analysis method
does not give this information.
5.4.3.
BEST MATCH
Analysis by synthesis can be formulated as a turbine capacity method or as heat balance method. However, these are not
the only methods to find the core flow for a turbofan engine test. In fact, there are quite a lot of options:
The laws of physics require that all methods for core flow analysis give the same answer. However, the unavoidable
problems with measurement biases, random scatter and misinterpretations will cause every core flow analysis method to
yield a different result.
Selecting only one core-flow analysis-method, means that some information is ignored. It is very advisable to run
several different methods and to compare the results. This will give many hints about the quality of the measurements
and the model. The measurements should be consistent with the pre-test uncertainty-analysis.
Another option is, to use several analysis methods simultaneously. We can combine, for example, the turbine capacity
method with a heat balance. Remember that the turbine capacity is based on the measured value of the turbine throat
area and the heat balance method on a measured value for T5. From a turbine capacity analysis we will get a difference
between the measured and the calculated T5 and from a heat balance analysis we get a difference between the measured
and the calculated turbine throat area A41.
We can set up the analysis in such a way, that we minimize the weighted sum of [T5-T5,m] and [A41 - A41,m]. The
weighting factors will be selected in such a way, that they take care of the confidence that we have in the temperature
and turbine throat area measurements.
Obviously we need not restrict ourselves to the turbine capacity and the heat balance methods for core flow analysis. We
can use all available measurements simultaneously, and thus we will get a compromise between all conflicting
indications of the true core flow. The resulting set of ANSYN scaling factors describes the best match of the model to
the test data.
5.4.4.
SIMULTANEOUS ANALYSIS OF SEVERAL DATA SETS
Up to now we have discussed only the analysis of a single scan, at a single steady state operation point. For each scan,
we will get a set of scaling factors for all component models. When we have to analyze a full performance curve we will
get many sets of scaling factors.
When we plot the scaling factors over corrected fan speed, for example, we will normally observe that they are not equal
to 1.0 (which would indicate perfect agreement between the model and the test results) and in addition to that there is a
trend in the data.
There are two ways to deal with these deviations between the model and the measurements. The first is, we attempt to
find the reason for the deviation. When we have found it then we can improve the model by introducing revised or even
new correlations. It might also happen, that we have to modify the way we interpret the measured data.
The second option for closing the gap between test data and model is to use representative curves for the scaling factors
and thus calibrate the model.
5.4.4.1
DEVELOPING MODEL IMPROVEMENTS
When we are looking for the reason why the model deviates from the test data then a straightforward core flow analysis
like the turbine capacity method should be used. A best match core flow analysis as described in the previous chapter
would make trends in the data less visible because it will, to some extent, distribute the deviations between
3-63
Gearbox drag;
Fuel, oil and hydraulic pump power;
Changes in gas properties;
Reynolds number effects;
Thermal expansion of rotors, blades and casings.
3-64
We have discussed how the results from a single scan can be corrected to ISA conditions. However, the rated
performance also has to be derived from engine performance tests. With the conventional test analysis this requires a set
of scans which include the power range of interest. Then a curve fit is applied to the ISA corrected data and the resulting
curve is read at the exact value of the rating parameter. This might be a rated temperature, a spool speed or an engine
pressure ratio.
With the ANSYN approach one can easily evaluate the rated performance by just running the calibrated model (which is
either based on single or on multiple scans) at rated power. During this evaluation one can even simulate an engine
without rakes by setting the rake pressure losses to zero in the model.
5.5
SUMMARY
The calibration of an engine performance model with test data is a time consuming task. Traditionally, special test
analysis programs are employed for deriving ISA corrected performance data that are then compared with the results of
a cycle program. In a second step the model is manually adjusted in such a way, that the simulation results match the test
data.
The Analysis by Synthesis approach integrates the simulation task with the test analysis. It gives a better insight into the
differences between rig and engine test results and allows automation of the process of matching the model to the test
data.
6
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
REFERENCES
Charles R. King, A Semi-Empirical Correlation of Afterburner Combustion Efficiency and Lean-Blowout FuelAir-Ratio Data with Several Afterburner Inlet Variables and Afterburner Lengths, NACA RM E57F26, 1957
D.J. Stromecki, An Assessment of Gas Turbine Engine Augmentor Technology and Needs for the 80s, AIAA80-1200
H.V. Hattingh et al, The Design and Development of an Afterburner, ISABE 93-7041, 1993
J.C. Mayer et al, Some Aspects on the Joint GR/Volvo Development of the F404/RM12 Augmentor, ICAS-905.3.4, 1990
William J. Egan, James H. Shadowen, Design and Verification of a Turbofan Swirl Augmentor, AIAA 78-1040,
1978
Joachim Kurzke, Claus Riegler, A Mixed Flow Turbofan Afterburner Simulation for the Definition of Reheat Fuel
Control Laws, RTO Symposium on Design Principles and Methods for Aircraft Gas Turbine Engines, Toulouse
1998
Derr, W. S., and Mellor, A. M., Recent Developments, in Design of Modern Turbine Combustors, edited by A.
M. Mellor, Academic Press, Harcourt Brace Jovanovich, New York, NY, 1990.
R. Kevin Rowe, A.P. Kuchar, Energy Efficient Engine, Scaled Mixer Performance Report General Electric
Company, NASA CR-167947, 1982.
Hale, A. A. and Davis, M. W., DYNamic Turbine Engine Compressor Code DYNTECC - Theory and
Capabilities, AIAA-92-3190, Presented at the AIAA/SAE/ASME/ASEE 28th Joint Propulsion Conference and
Exhibit, Nashville, TN, July 6-8, 1992.
Garrard, G. D., ATEC: The Aerodynamic Turbine Engine Code For The Analysis Of Transient And Dynamic Gas
Turbine Engine System Operations Part 1: Model Development, ASME Paper #96-GT-193, June 1996.
Herbert, M. V., A Theoretical Analysis of Reaction Rate Controlled Systems - Part 1, Chapter 6 in Combustion
Research and Reviews, 1957, Agardograph No. 15, Butterworths Scientific Publications, London, England,
February, 1957.
Kneile, K. R., and Hale, A. A., Appendix C: Numerical Solution to the Governing Equations, Appendix C in
Garrard, G. D., ATEC: The Aerodynamic Turbine Engine Code for the Analysis of Transient and Dynamic
Turbine Engine System Operations, Ph. D. Dissertation, The University of Tennessee, Knoxville, August, 1995.
Lefebvre, A. H., Fuel Effects on Gas Turbine Combustion - Ignition, Stability, and Combustion Efficiency,
Journal of Engineering for Gas Turbines and Power, Vol. 107, January 1985, pp. 24 - 37.
Shahrokhi, K. A. Application of Modified Dynamic Compression System Model to a Low-Aspect Ratio Fan:
Effects of Inlet Distortion. MS Thesis, Mechanical Engineering, Vanderbilt University, Nashville, TN, 1995.
Visser, W.P.J., and Kluiters, S.C.M., Modelling the Effects of Operating Conditions and Alternative Fuels on Gas
Turbine Performance and Emissions Research and Technology Organization, RTO-MP-14, 1999.
The SAE, Real-Time Modeling Methods for Gas Turbine Engine Performance, SAE AIR4548, 1995.
Volponi, Allan J., Gas Path Parameter Corrections, ASME 98-GT-347, 1998.
Doel, David L., An Assessment of Weighted-Least-Squares Based Gas Path Analysis, ASME 93-GT-119.
Kerr, Laura J., et al., Real-Time Estimation of Engine Damage Using a Control Based Kalman Filter Algorithm,
ASME 91-GT-216.
4-1
Chapter 4
Computer Platform and Software Implementation
1
4
5
6
7
Overview
1.1
Computer Platforms
1.2
Hardware
1.3
Operating Systems
1.4
Development Environments
Trends and New Technologies
2.1
General
2.2
Computing Power
2.3
Computing Costs
2.4
Parallel and Distributed Computing
2.5
Interfaces
2.6
Object Orientation
2.7
PC Technology
2.8
Zooming
2.9
Development Environments
2.10 Architectures
2.11 Configuration Management
2.12 Windows versus UNIX
Challenges
3.1
General
3.2
Reducing Development Effort
3.3
Generic Tools
3.4
Standardization
3.5
User Interfaces
3.6
Visualization
3.7
Maintainability
3.8
Grid Generation
3.9
Distributed Parallel Computing
3.10 Probabilistic Analysis
Future
Examples of Gas Turbine Simulation Systems
5.1
A FORTRAN-Based Modeling System
5.2
MOPS (Modular Performance Synthesis Program)
5.3
GasTurb
5.4
The GSP Object Oriented Modeling Environment
5.5
TERTS (Turbine Engine Real Time Simulator)
5.6
Simulation Models for Engine Diagnostics
5.7
Other 0-D modeling systems
5.8
Multi-Dimensional Modeling Environments
5.9
Numerical Propulsion System Simulation (NPSS)
References
Acronyms
4-2
4-2
4-2
4-3
4-3
4-3
4-3
4-3
4-5
4-5
4-6
4-7
4-8
4-8
4-8
4-8
4-9
4-9
4-9
4-9
4-9
4-10
4-10
4-10
4-10
4-10
4-10
4-11
4-11
4-11
4-12
4-12
4-15
4-16
4-19
4-24
4-24
4-27
4-28
4-28
4-31
4-32
4-2
OVERVIEW
1.1
COMPUTER PLATFORMS
The computer platform is the combination of the hardware and software needed for gas turbine performance
calculations. The hardware is the actual computer; the operating system represents the software required to use the
hardware. A specific gas turbine simulation application is implemented on the platform using application development
software. Together, they form the development environment.
Many varieties of development environments exist for gas turbine simulation. A gas turbine model type can be
characterized by three needs:
Application
Model fidelity
Computing performance requirements
The level of model fidelity directly depends on the type of simulation application (see Chapter 2 Applications).
Application types include:
Currently, 2-D and 3-D simulations often focus on component R&D applications while 0-D models usually simulate the
whole engine for a large variety of purposes such as a customer cycle deck. 0-D models also include parametric, nonthermodynamic or non-component based models, which may be considered the more simple 0-D models.
Model fidelity directly relates to required computer performance. Full 3-D Navier-Stokes simulations still require
special high performance hardware while 0-D models can now be run on PCs. Consequently, computing performance
requirements strongly relate to the computer platform.
1.2
HARDWARE
High performance computers are generally used for high-fidelity simulation for R&D purposes, such as 2-D and 3-D
CFD. Use of 0-D or 1-D engine models for integration within a larger system simulation or for probabilistic analysis
may also require high performance computers. High-fidelity CFD has become indispensable for gas turbine R&D and
can be regarded as heading the (fidelity) frontier of the modeling spectrum. The most important limitation for high
fidelity computing is available computing power, both in terms of memory and processing speed. Mainframe computers
are rapidly being replaced by other systems such as PC and workstation networks, but are still used for running older
applications such as 0-D (cycle decks) and 1-D models. UNIX workstations are widely applied and used for medium
fidelity simulation or visualization and data processing of high-performance computing results.
PCs are rapidly increasing their share of the entire computing market. Due to rapid increase in computer power, PCs are
now able to run medium fidelity 0-D and 1-D models and to a limited extent (coarse grid) even 3-D CFD simulations
(e.g. FLUENT). An important issue at this end of the spectrum is the efficient use and development of new gas turbine
simulation applications for the operational field (e.g. maintenance and diagnostics tools). Powerful PCs may well be
considered workstations now, since they can easily be configured to match the conventional UNIX workstations in
performance but then require disk space and memory beyond that of a typical office PC, even with the fastest CPU.
Networked UNIX workstations or PCs can be used in parallel for some problems with properly configured software. In
some circumstances performance can equal or exceed that of a supercomputer.
Figure 1 presents the relation between model fidelity, computer platform hardware and application.
4-3
NUMBER OF
DIMENSIONS
(time and space)
1
2
3
4
SUPER/HIGH
PERFORMANCE
COMPUTING
MAINFRAME
Cycle decks.
Probabilistic.
Multi-disciplinary.
Cycle decks.
Maintenance or
Diagnostics.
Probabilistic.
Cycle decks.
Probabilistic.
Multi-disciplinary.
Cycle match CFD.
Cycle match CFD.
Component CFD.
SINGLE UNIX
WORKSTATION
SINGLE
PC
Cycle decks.
Maintenance or
Diagnostics.
Probabilistic.
Cycle decks.
Real-time.
Cycle decks.
Maintenance or
Diagnostics.
Probabilistic.
Multi-disciplinary.
Cycle decks.
Real-time.
Component CFD.
none1
none1
Cycle decks.
Real-time.
1.3
OPERATING SYSTEMS
With the retirement of the old mainframe systems, the number of different operating systems is reduced. In general it
can be stated that there are two main streams: UNIX which is commonly used from workstations up to higher
performance systems and Windows (Windows 95/98, NT 4.0 and 2000) for the PC based systems. The need to perform
simulations on legacy platforms (such as old mainframe systems) can become a barrier and must be identified early on.
Windows NT is also available on a number of high-performance 64-bit systems like the DEC-Alpha, providing a
combination of high-computing power with the ability to use the customary PC office software suites.
1.4
DEVELOPMENT ENVIRONMENTS
A large number of development environments exist, both for the UNIX and the Windows systems. Traditionally these
environments consisted of 3rd Generation Languages (3GL), the most widely used being FORTRAN in the scientific
world. Newer 3GL languages include C and the object oriented languages C++ and ADA. More modern are the 4th
Generation Languages (4GL). Often these are wrapped around a 3GL language in order to reduce developer effort when
building (graphical) user interfaces. Many 4GL tools automatically generate most of the user interface parts of the
application. Examples are Visual Basic, Delphi, C++Builder, JBuilder and Visual C++. There are also a number
of development environments dedicated to simulation in general or sometimes even to gas turbine simulation (i.e.
generic gas turbine simulation tools). Examples are MATLAB-Simulink and MathCad. Examples of turbo-machinery
CFD tools are CFX-TASCflow with Turbogrid, NUMECA Fine/Turbo.
FORTRAN is still the standard programming language for gas turbine simulation. The ARP standards (see paragraph
3.4) mostly apply to FORTRAN. If FORTRAN is used without platform specific code (such as user interface shells), it
can be compiled and run on most platforms. This will be the dominant advantage of FORTRAN until alternative
standards become widely accepted.
Section 5 lists a number of simulation systems including descriptions of the development environments.
2.1
GENERAL
The Internet and the PC have dominated computer related technological development since the nineties. Both
technologies require low cost platforms which offer great potential for gas turbine simulation by offering distributed
computing, a good user interface, and high power, especially at the lower fidelity end (i.e. operational use) of the
spectrum.
High performance computing technology may not have as much public attention, but is also developing at a rapid pace
and offering ever more power for high fidelity computing.
With distributed parallel computing technology improving, PC and high performance computer technologies may well
merge into a single type of environment or platform. High-speed networks have made remote distribution common in
simulations and remote computing is becoming a reality.
2.2
COMPUTING POWER
Since the introduction of the digital computer, computing power has increased at a rapid rate, see Figure 2. Interesting to
note is that the low cost PC is increasing its share of the entire spectrum, while high performance computing is
maintaining the top high-fidelity part of it.
4-4
As a consequence, computer power has ceased to be the bottleneck for all but the high-fidelity CFD gas turbine
simulations. For all types of simulations, implementation effort, user interface including visualization and code
maintenance have become critical for successful and efficient use of the models.
10
FLOPS
10
10
10
15
tin
mpu
l co
le
l
para
ted
ribu
t
is
D
12
Hig
h pe
man
rfor
ute
omp
c
e
c
N
s, U
tion
a
t
s
rk
Wo
C
le P
Sing
Multi-stage CFD
NS CFD
rs
Euler CFD
3/4-D
IX
2-D
1-D
0-D
10
1980
Figure 2 - Trends in computing power
For high fidelity CFD simulations a bottleneck remains in the available computer power, especially when the time
domain is added as an extra dimension for dynamic simulations. Simultaneous simulation at high spatial resolution and
high time domain resolution for instance remains limited as indicated by Figure 3.
100
1-D
10
2-D
3-D
1
0
50
100
4-5
2.3
COMPUTING COSTS
In general it can be stated that the cost/benefit ratio of using simulations for various purposes is decreasing rapidly. This
becomes evident from Figure 1 and the fact that the prices of state-of-the-art PC systems have not risen over the years.
1-D whole-engine thermodynamic performance calculations and simulations can be run at costs many times lower than
a few decades ago. This is demonstrated by the emerging PC applications for gas turbine simulation, which allow 0-D
simulations at very low cost.
At the other end of the spectrum, the increasing power of high-performance computers offers new opportunities to
optimize aero-thermodynamic designs with high-fidelity CFD. Especially when using distributed parallel computing
using clustered low cost workstations or PCs this can be done at relatively low costs. Many large gas turbine R&D
programs focus on greater CFD detail at limited costs, which is critical to gas turbine technology progress.
As a consequence, established simulation technologies and tools move to lower cost platforms, yet retain their speed and
fidelity. High-performance computer technology benefits from low cost technology in the form of distributed parallel
computing (see section 2.4) to satisfy the ever-increasing hunger for CFD calculation power. While computing power
remains critical for the high fidelity CFD challenges, the established simulation technology basically needs
improvements in order to make their use more efficient (i.e. at lower costs). This means greater attention to user
interface, code portability and maintainability aspects is needed.
2.4
Symmetric MultiProcessor (SMP) technology is becoming common in servers. This technology uses several processors
in a single computer, and offers the ability to run several tasks in parallel, under the control of the operating system.
These tasks are usually distributed at task or thread level. A thread is a small piece of a program that is capable of
performing a task that is largely independent of other threads. To distribute processing at thread level requires design
effort from the programmers. Whether using an SMP computer will provide worthwhile gains in speed depends on the
amount of data that is used, and the amount of processing that is done to it. If the application is data bound and the CPU
is not highly loaded, any investment may be better spent on a faster disk. Similarly, a processor that could run at twice
normal speed would not carry the operating system overheads associated with SMP and 2 normal speed processors.
To share tasks between multiple processors:
The CPUs must have SMP enabling features.
The operating system must support SMP.
Most new CPUs now are SMP enabled. Different OS versions support different numbers of processors. Windows NT
has crept up from 2 to about 8; Windows2000 supports up to 64 processors! Other versions of Windows do not support
SMP. Linux 2.2 theoretically supports up to 16 processors on Pentium, UltraSparc, SparcServer, Alpha and PowerPC
machines.
Parallel computing is a new development, in which tasks are shared between several processors. Ideally little effort is
required from the programmer, with the effort being provided by the operating system. This has great potential for highperformance computing. Such techniques are used internally in many current processors. Optimally, parallel computing
could offer a way to increase computing performance in direct proportion to the number of processors used. In reality,
the performance gain is less due to the problem of how to distribute the computing tasks over the processors.
Early applications that exploited parallel computing had to include the computing task distribution themselves, requiring
large efforts in software development. Now, the trend is to have the operating system or development environment
handle that task with solutions like the Parallel Virtual Machine (PVM) [18]. Although becoming easier to use, this still
requires that special actions be taken in defining the problem to facilitate parallel operation. Other examples are:
W(indows)PVM & Bulk Synchronous Parallelism (BSP) A new programming model for parallel processing simplifies
writing programs and promises code portability [19].
Multi-threading
Multi-threading is basically time slicing by the OS. The OS must make these features available to compilers. The
language and compiler must in turn make the necessary commands available to the programmer. All 32-bit Windows
platforms support multi-threading. This was necessary, so that memory could be shared between related processes, and
to prevent Windows applications from being slowed down (frozen) by a single very intensive process. Multi-threading
enables the processor to start another process parallel to the slow process and also to be able to interrupt or control the
slow process from another thread.
Whether the OS makes SMP available between applications and also between threads within an application, is OS
dependent.
With web servers, Windows based systems are typically thread based, usually via reference counted DLLs (ISAPI,
NSAPI), while Unix systems spawn new processes (CGI) that typically return data via files. The target platform
4-6
therefore affects the architecture of a new application, and the way in which efficient code is ported between platforms.
This means that multi-threading gas turbine simulation applications will be difficult to port to other platforms.
All modern programming languages have features to use parallel computing and the modern development environments
like Delphi come with tools to facilitate multi-processor and parallel computing using multi-threading. It is interesting to
note that recent computing performance records have been set with parallel computers using large numbers of cheap
processors like the Intel 386.
Parallel computing is often applied in super-computers and high-performance workstations operating multiple
processors. A new trend is to apply parallel computing to multiple computers that are interconnected over a network.
This distributed parallel computing requires special software controlling the distribution of different computing tasks in
a simulation.
It is expected that eventually software will become available to control parallel distributed computing using a large
number of ordinary network environment PCs. This would allow simulations, which could traditionally only be run on
super-computers, to run at a fraction of the current cost. This is already being done with networked UNIX workstations.
A critical new technology for distributed computing is object orientation (see section 2.6). Object orientation offers
modularity and common interface mechanisms required for distributed computing. Each computer in a network is
executing the simulation of an object as part of the entire simulation session across the network.
Most distributed computing has been limited to the same type of processor and operating system. Computers networked
over the Internet (or an Intranet) can be used to perform a distributed computation task with Suns JAVA technology
and the CORBA (Common Object Request Broker Architecture) technology. The JAVA gas turbine simulator [10, 11]
is an example of this new trend. Microsoft Windows uses a similar technology called variously ActiveX or DCOM
(Distributed Component Object Model).
An example of a distributed-parallel computing project is the Visual Computing Environment (VCE) project at NASA
Glenn Research Center [9]. One of VCE's objectives is to develop a visual computing environment for controlling
the execution of individual simulation codes that are running in parallel and are distributed on heterogeneous host
machines in a networked environment.... VCE was designed to provide a distributed, object-oriented environment
including a parallel virtual machine (PVM) for distributed processing. Users can interactively select and couple any set
of codes that have been modified to run in a parallel-distributed fashion on a cluster of heterogeneous workstations.
2.5
INTERFACES
User Interfaces
Most modern computer applications have replaced the command line interface with the graphical user interface (GUI).
This offers significant benefits in terms of user friendliness. The older gas turbine modeling environments, especially
the 3GL based ones such as FORTRAN, still use the command line interface. Many of them have been updated and
wrapped inside 4GL GUI structures.
4-7
To specify input-data for complex models, sophisticated user interfaces are required to prevent unacceptable timeconsuming data-entry tasks. Across the spectrum of modeling platforms, attempts are made to accomplish this with
advanced GUIs. As an example, component maps for 0-D simulation are usually presented to the program in tabular
format. To use the tabular format for user data entry (for specification of new or modified maps) is very time consuming
and therefore graphical tools are used to have the user edit the data using the graphical map representation to actually
draw the map. SmoothC and SmoothT [7] are examples of stand-alone Windows applications able to do that task.
With the increase in computing power, the size and detail of the results increases drastically. Graphical visualization and
sometimes animation tools are required for their analysis, such as the VCE [9] for example.
As a result the user-interface issue tends to become separated from the modeling issue. The modularization of the
simulation environments reflects this trend also. In programs like NPSS, sub-programs are defined to address userinterface issues such as visualization of CFD results.
Needs of the expert user or the user with specific highly repetitive tasks can conflict with the needs of the low-end user
who needs easy access without being confused by the features and options which arent relevant to simpler applications.
Some GUIs (such as GasTurb, see section 5.3, and GSP see section 5.4) are designed so those more advanced options
are hidden or separated from the low-end user options.
External Interfaces
Interfaces with data acquisition systems and measurement databases are often platform specific. The advantages of
having these systems on the same platform as the simulation system are often the reason for maintaining legacy systems.
2.6
OBJECT ORIENTATION
Object orientation (OO) is an approach in software development that was defined during the seventies. Before this, the
program design was entirely up to the programmer, and the relationship between data and the procedures that operated
on it could be unnecessarily complicated and inconsistent. For instance several procedures could operate on the same
data, causing a problem if one procedure was changed and another not. The basic idea of object oriented design (OOD)
is encapsulation. This means that everything is described as an object, and that every object has methods, properties and
data. For illustration, an object called airplane might have methods called take off, fly and land, properties called all-upweight, number-of-engines and maximum-number-of-passengers, and data called elapsed-flight-time, number-ofpassengers and current-speed. The key idea is that only the methods contained within the object can change the
properties and data, thus ensuring integrity. Depending on the programming language, objects may be known as types or
classes. Two additional principles of object orientation are:
Inheritance, which means that specific types of airplane may be defined by changing the properties of the generic
airplane object, and by adding new or subtracting existing methods, properties and data. In some languages an
object may inherit from more than one parent.
Polymorphism, which means that different methods may have the same name, but operate differently depending on
the context. For example take off could apply to the start of flight or the removal of equipment.
Inheritance and polymorphism offer significant extra benefits in terms of software design but are not fully included in
some development environments. Although OOD promised many benefits in terms of code development effort and
maintainability the OOD approach was widely adopted only during the nineties. One of the reasons was that the
requirement for software developer skills was underestimated. Most popular object oriented programming languages are
traditional languages extended with object oriented features like C++ and OOPascal. ADA is an object-oriented
language widely used by the US military, but does not include all OOD features (such as inheritance and
polymorphism).
Especially in 4GL languages, OOD is commonly applied for GUI development and OO code is generated automatically.
It is up to the developer to decide the extent to which the actual functional (in this case simulation) code will be object
oriented and event driven. For gas turbine simulation, there is great potential in object orientation since in many cases
4-8
the simulated process can be divided into objects directly. For example, in a non-dimensional whole-engine simulation,
engine components such as compressors, turbines, control systems etc. can easily be defined (encapsulated) as objects.
With the OOD principles of inheritance and polymorphism, code development effort, reusability, maintainability and
flexibility can be significantly enhanced. An example is the Visual Computing Environment VCE [9]. See paragraph 5.4
for an example of an OO gas turbine modeling architecture.
2.7
PC TECHNOLOGY
Since the eighties, PC technology has drastically changed the IT world. With its continuing and rapid increase in
performance the PC is increasing its share in the overall computer market. With the increasing need for computing
power for gas turbine simulation, this development offers great potential. Also the PC is becoming the platform on
which most efforts to improve development and user environments are focused. This means that all gas-turbinesimulation applications except the high-fidelity simulations requiring high-performance computers will probably be
most efficiently developed on PC platforms. If distributed-parallel computing technology becomes mature for
networked PCs (see paragraph 2.4) the high-performance simulation jobs may also benefit from PC technology.
The major corporate operating system used on PCs is Windows (W95/98/NT4.0/2000). An interesting development is
the use of UNIX on PCs such as SCO-UNIX and Linux, the Free Software Foundation open UNIX clone.
2.8
ZOOMING
The Zooming concept allows high-fidelity simulation of local phenomena of interest in a gas turbine, together with
lower fidelity simulation of the rest of the engine system. This approach is necessary to reduce computing power,
development time and complexity where higher fidelity analysis is not required. High-fidelity CFD simulation of the
aero-thermodynamic processes throughout the entire engine would require computing power far beyond what is
feasible. With the zooming concept, detailed CFD simulation of flow around specific compressor blades, for example,
can be performed while the rest of the engine is simulated with lower detail.
An example of the application of the zooming concept is the NPSS program, described in paragraph 5.9
2.9
DEVELOPMENT ENVIRONMENTS
There are trends towards using new developments like 4GL languages and C++ in the global IT world. The scientific
world still generally considers FORTRAN the standard, basically due to the lack of a new clear standard for more
modern environments. Only the 3rd generation (3GL) languages C, C++ and ADA seem to be able to receive confidence
enough to be adopted by some organizations as a standard for gas turbine simulation. Even these can become a problem
since they expose low-end computer users to details beyond their level of interest.
On the lower fidelity end of the spectrum with PC simulations and applications, for engine operators and maintenance
(e.g. diagnostic tools), new 4GL tools are applied. In the Architectures section, two examples are given of 0-D modeling
environments using the Delphi 4GL tool based on Object Pascal. C++Builder is another Borland product that uses the
same back end compiler as Delphi. This also runs on AS400 and Windows platforms. Linux (Unix) compatibility was
launched in June 2001 with the Kylix product. Cross platform independence depends on programmers not using
platform specific Application Programming Interface calls in the code that they write. A graphical library called CLX
(pronounced clicks) has been made available for both Windows and Linux to allow this. Other 4GL tools that could be
used for 0-D modeling are Microsoft Visual Basic and Visual C++, although Visual Basic is not an object-oriented
language. At the time of writing Microsoft have started to introduce a new language called C# (C Sharp), which is
designed to compete with Java, but only to run on WinTel platforms.
With the large amounts of existing and proven FORTRAN code, 4GL environments based on modern languages are
often applied to encapsulate FORTRAN sub-routines with a modern front end.
Generic simulation tools like MATLAB Simulink and MatrixX have become popular for 0-D modeling for some types
of performance analysis (e.g. real-time modeling and control system design). An important advantage of these tools is
that they usually employ auto-solvers, hiding the details of numerical methods. The user only needs to specify the
required accuracy. However, under some circumstances auto-solvers may use inappropriate methods that produce
instability due to rounding and other computational errors and solvers specifically developed for gas turbine simulation
may give better results by applying plausibility checks.
In 5.5 an example is given of a Simulink real-time thermodynamic 0-D model. The figure only shows the whole engine
model level and hides the top level including the control system, and many sub-levels as well as.
2.10 ARCHITECTURES
The gas turbine model architecture represents the way the model is built up of sub-models, components, finite elements
etc. Several types exist; ranging from non-component based parametric models where no gas turbine components can be
identified, to high-fidelity CFD models of flows in particular sub-components like compressor blades. These may
include large numbers of finite elements.
4-9
The model architecture is important both to the model developer and the model user.
For the developer, for example, a modular approach may be adopted for 0-D and 1-D models with sub-modules
representing typical components like compressors and turbines. Generic sub-modules may be developed using object
orientation offering significant benefits in terms of software development and maintenance effort. This approach may
also be used when using generic simulation tools like MATLAB-Simulink.
In paragraph 5 the architectures of some simulation environments are described.
CHALLENGES
3.1
GENERAL
In general it can be stated that with the rapid increase in available computing power and high bandwidth networks at
low cost, the challenge is to efficiently use that power for gas turbine simulation. The growth rate of the PCs
performance/cost ratio indicates there are significant opportunities. Moores Law, reported in 1965 predicted a doubling
in the number of elements on a chip every 18 months. This was when the largest chip contained 64 elements. From 1970
to 1990 Intel chips doubled their complexity every 2 years. Since that time the doubling period for Intel has become 2.5
years (ref http://www.physics/udel/edu/wwwusers/watson/scen103/intel.html), and 28 million elements are contained in
a Pentium 3 cpu. To benefit from these opportunities, new technologies for (gas turbine) simulation software
development and user-interfaces software maintenance and new standards must be developed.
3.2
Closely related to maintainability is the aspect of development effort. With increasing complexity, advanced
development environments are needed for automating many tasks previously performed with line-by-line coding. This
implies distribution of software development tasks, e.g. 4GL tools to develop GUI and general software structure.
Ideally, the line-by-line coding should be limited to the implementation of the actual equations being used in the model,
4-10
but this will not easily be accomplished. With the hand written code being reduced to the actual equations, advanced
development environments should also enhance maintainability. However, this may be at the cost of code efficiency.
This is illustrated by the fact that some of the current 4GL systems, with automated code generation, carry significant
Safe practice coding overheads that are simply irrelevant to many applications.
Object oriented technology (see section 2.6) offers reduction in development efforts with the inheritance principle. This
has the potential to allow the engineer to quickly do tasks that are currently limited to Information Technology
professionals or methodology experts.
3.3
GENERIC TOOLS
An approach to reduce gas turbine model development effort is to use generic gas turbine or turbo-machinery specific
tools. For the 0-D models several tools already exist for modeling performance of any kind of gas turbine (see section
5). Also specific simulation tools like MATLAB-Simulink may be applied for certain simulation tasks (section 5.8).
For the entire spectrum, more attention can be expected to more intensive use of existing generic tools instead of
redeveloping code repeatedly. This may also be in the form of reusing generic objects in OO environments.
3.4
STANDARDIZATION
Standards for gas turbine modeling code and interfaces are critical for large, comprehensive multi-disciplinary models,
created by large numbers of developers. Portability is also enhanced when standards for development environments,
languages etc are observed. Current standards for gas turbine simulation are SAE Aerospace Recommended Practice
(ARP) 755 and 681 [7, 8]. These are based on shared FORTRAN common blocks. These lowest common denominator
standards severely limit simulation options for future simulation development. A new ARP recognizing the needs of
modern computer systems, development environment and providing the option for an Application Programming
Interface (API) type interface is being developed as part of ARP 4868 (Draft) limitations.
Standards are indispensable to interface gas turbine models with other models (e.g. aircraft system models) and to
benefit from modern development environments, such as CORBA and DCOM (see 2.4). These will reduce development
efforts, increase maintainability and improve user interfaces. Without them the R&D gas turbine world will remain
committed to FORTRAN, at least for implementation of the fundamental algorithms. As has happened before, the
market will probably define the new standard. So for now it seems the technical community will have to wait and see
what happens.
MATLAB-Simulink and similar environments are becoming common for lower fidelity models, non-component
models, real-time models and even simplified 0-D models. For the higher fidelity models, new standards may become
available to wrap interfaces and objects around existing FORTRAN codes in order to maintain backward compatibility.
New standards for modern OO languages may then be used for developing new codes.
So far, the lack of standards and legacy system compatibility needs have caused most model makers and engine
manufacturers to hesitate to move from FORTRAN to another modern development environment. The development of
new standards and the gradual disappearance of these legacy systems is causing this change to begin.
3.5
USER INTERFACES
In general, the different tasks involved in developing and using engine performance models become dispersed over a
large variety of applications. When using engine-modeling tools in the operational area, interface aspects are very
important. For example, an engine diagnostic tool based on an engine performance model must have a good userinterface, dedicated to a maintenance engineer instead of a research engineer. Often the engine model is isolated from
the user interface so that it can be tailored to specific tasks such as trending, testing and data visualization.
3.6
VISUALIZATION
Visualization, both static and animated graphical representations, will become more and more important for the
presentation of gas turbine simulation results, especially for high fidelity CFD. With the increasing fidelity of modeled
flows, new visualization technologies will be necessary to present results from the large amounts of data. For 0-D
models visualization needs are often driven by the needs of the data system and the application. Keeping the simulation
tool independent of the visualization package should be a goal for flexibility.
3.7
MAINTAINABILITY
Maintainability is a big issue for the entire IT world. General trends trying to enhance maintainability are new
programming languages (object oriented), CASE tools, documentation tools. Terms involved are: modularity, OOD,
portability, documentation, debugging tools.
Currently, much gas turbine code is still in the FORTRAN language. This allows maintenance by those with low-end
computer skills but often requires specialized knowledge of the simulation tool and application, which are impractical.
However, the need to exploit software technologies and move to new development environments and languages or to
complement FORTRAN models with new tools is growing.
4-11
Object oriented technology offers improvements in maintainability because the encapsulation principle offers highly
modular code. Used well this can capture the knowledge necessary for maintenance with the tool and the application.
Previously this only resided in the minds of those intimately familiar with the development of the application.
3.8
GRID GENERATION
For high fidelity simulations, grid generation software is used to specify hardware geometry and can therefore benefit
from direct coupling to a CAD system or geometry database. Developments in user interface technology are required to
improve the complex tasks of grid generation. However, it is usually handled outside of the engine simulation and is
therefore considered as a data input issue beyond the scope of this report.
3.9
A big challenge lies in exploiting the large amount of cheap computing power becoming available with PCs and PC
processors. The development of software, for efficient distribution of computing tasks over a large number of networked
PCs, will be critical. This is a need across all engineering computing tasks and is also being developed for more
powerful stress analysis.
An important issue is how data is managed among distributed parallel computing tasks. Remote computers have data
sharing limitations due to limited network transfer speed. This is, in essence, no different from disk accessing speed
limitations on a fast singleton PC. With distributed processing, a workflow management system is likely to be needed.
Such systems are already commonly used to coordinate complex commercial Customer Relationship Management,
Credit Checking, Accounting, Service Provisioning, Trouble Ticketing and Billing systems.
FUTURE
In general, developments in computer platform technologies will have a significant impact on the potential of gas
turbine simulations. The current rapid pace of developments like Internet technology, PC technology, and distributed
parallel computing indicate the importance of carefully monitoring these developments in order to continuously exploit
all possible benefits for gas turbine simulation. This however requires a significant effort, and few new trends can be
adopted with the certainty that they will become standards.
FORTRAN will probably remain the standard language for the higher fidelity R&D applications. C/C++ will continue
to make inroads. Moreover, C++ and other object oriented languages are used often in conjunction with 4GL
development environments, which offer significant benefits in terms of user interface development, code
maintainability, debugging and code documentation and readability. For the lower fidelity modeling applications,
especially for operational users, these newer environments will continue to be adopted.
Distributed parallel computing will provide large potential to satisfy the increasing hunger for computing power highfidelity simulations for R&D purposes.
With the rapid increase in available computing power, the traditional bottleneck in computer power becomes replaced
by implementation issues like development effort, visualization, and software for (distributed) parallel computing.
4-12
Modern computer platforms and development environments will be applied for new types of model such as those
applied in operational areas like maintenance, diagnostic tools or even customer cycle decks.
The past decades have shown it is hard to predict very far into the future of information technology. However, it is clear
that we can expect significant progress in the following areas:
Computing power: As is visualized in Figure 2, there is no reason to assume the current rate of progress in available
computer power will not be maintained. This means there is a great challenge in efficient use for even higher-fidelity
simulation for gas turbine R&D. Also, increasing power means the pressure to develop efficient code will decrease
and the focus can be moved to reducing development effort instead (see section 3.2).
Costs: The increasing computing power will become available at lower cost. PCs will be able to do higher fidelity
simulations, meaning current simulation tasks will become cheaper.
Distributed-parallel computing: High-performance computing will be done more and more, using clustered or
networked low-cost systems working in parallel. This will also lower the cost of high fidelity simulations.
Model development environment: Development environments for gas turbine simulation will move from traditional
3GL languages to environments specific to simulation (CFD) or environments including readily available user
interface and visualization tools. This will enable research engineers to concentrate on the actual modeling tasks
when developing gas turbine models.
Standards: Standards will evolve from those driven by legacy requirements to those consistent with commercial
software tools, databases and development environments.
0-D simulation:
Multi-dimensional simulation:
Multi-Dimensional Modeling Environments (section 5.8)
Numerical Propulsion System Simulation (NPSS, section 5.9 )
GSP, GasTurb and some others listed in sections 5.7 and 5.8 are generally/commercially available tools. The others are
proprietary or limited distribution programs used in the industry.
The descriptions will mainly address the software implementation related characteristics of these environments such as
architecture, platform, interfaces etc.
5.1
In this modeling system, the engine is viewed as a set of interconnected gas flow passages, while engine component
models are represented by subroutines. The major part of the definition of the engine performance is obtained from
knowledge of the conditions pertaining at a number of engine stations in the gas flow. At each station, an array of gas
conditions in terms of fuel-air ratio, mass flow, pressure, temperature etc. is defined and used to pass information from
one component to the next. The architecture is modular and provides a flexible tool to model a variety of gas turbine
configurations. However, the FORTRAN language has limited capabilities to apply modern software development
methodologies such as object orientation, modern data organization, databases and graphical user interface features.
The advantage of FORTRAN is that it is still the standard and, if no platform specific code such as user-interface shells
are included (which often is the case for a command line interface), may be compiled and run on most platforms.
Architecture
Figure 4 shows the elements of the modeling system. The subroutine containing the definition of the engine is a small
part of the total infrastructure. Each element is discussed below.
4-13
UTILITIES
BLOCK
DATA
D
A
T
A
B
A
S
E
basic data
default options
USER
INTERFACE
READ
DATA
PROGRAM
CONTROL
OUTPUT
CONTROL
COMPONENTS
ENGINE ROUTINE
other routines
& functions
Database
This is not strictly part of the modeling system but a fundamental part of the Information Technology (IT) electronic
infrastructure. It holds functional engine data, which is used for model definition, and also receives and stores data
generated by the model. It also holds engine test data, which is required as program input for model-based analysis.
Security features are essential on any database for model input or output data.
Utilities
Common utilities include data visualization (plotting, tabulation), data maintenance (deletion, addition, grouping,
security), data manipulation (creation, formatting of graphical functions etc.) and output definition (creation and storage
of instruction sets for standard output formats).
Block Data
Associated with an engine-modeling program is a set of data which underpins the basic program structure, and which is
independent of the standard of engine being modeled. Default options (controlled by data switches) can be set up in this
dataset which is compiled with the engine routine (see below).
User Interface
The user communicates with the engine program via standard AS681 interfaces and if working in-house by lowerlevel program input stores for greater flexibility. The user supplies the following information:
Some of these options may not be available to some users, especially if the model is issued externally (i.e. as a customer
4-14
deck)
Read data
The data is pulled in from the various sources: database, user, block data and presented to the program.
Program control
The call is made here to the engine subroutine.
Engine routine
This is the heart of the system. Here, the structure of the engine is defined in terms of its flow path, which is modeled at
whatever detail is appropriate for the programs application. The engineer-programmer is provided with a data structure
built around the station and component subroutine structure. The engine is viewed as a set of interconnected gas flow
passages, and a major part of the definition of the engine performance is obtained from knowledge of the conditions
pertaining at a number of stations in the gas flow. Each station is defined in terms of fuel-air ratio, mass flow, pressure,
temperature, velocity, area, flow function etc.
Some stations may be defined as total stations in which case velocity and area terms are zero. Other stations e.g.
associated with pressure losses or mixing, may be defined as static stations in which case the pressure and temperature
terms will be static values associated with the specified area or velocity. Calculations of other parameter values can be
added as required.
Thus stations are handled as vectors of information. AS755 is an internationally recognized standard for station
numbering. Whereas this nomenclature appears on the program standard output, the programmer is given flexibility
within his own program to use whatever definition is convenient.
Several FORTRAN arrays are available to the programmer. These can be used for internal working and program
interfaces. The system makes use of COMMON blocks for ease of communication between different subroutines. The
program structure is largely constructed to reflect the physical layout of the engine. Standard subroutines are used, with
customizing being required to handle data transfer and different user options.
Components
Component subroutines are grouped into classes such as intakes, compressors, combustors, turbines and nozzles.
Although not strictly components, the following are treated as such because they follow a similar control volume
construct:
Pressure changes;
Multi-stream mixing;
Bleed network (secondary air-system)
For each class of component, several options exist within each subroutine. For example, a pressure loss may be defined
in many ways. The component-subroutines model the steady-state performance of the feature. The dynamics associated
with heat-transfer (to and from the blades and casings), shafts (conservation of angular momentum) and gas dynamics
(conservation of gas steam mass, energy and momentum) are handled in separate subroutines.
4-15
Graph read;
Obtain gas or fluid properties.
The above are used within or without the component modules.
Output control
Figure 5 shows the standard output, which may generated per point. When all that may be needed is a plot of thrust vs.
SFC for a series of points up a running range, such a comprehensive output may be too cumbersome. In such cases,
selected parameters can be identified and extracted to the database for plotting later. Some customization of the full
output is possible. Each section of the output is mutable, or may be embellished with station descriptors in plain
(perhaps project-specific) nomenclature. To display a small subset of the data, an expert user may configure a summary
section. Special diagnostics may be required for problem tracing, and these can either be appended to the basic output or
diverted to a separate output channel.
confirmation of input data
confirmation of user input data
status flags
optional
user reference
e.g. limiting, invalid solution
FAR, W, P, T etc.
component summary
summary output
additional information
as defined by user
data destination, special diagnostics etc.
Future Developments
FORTRAN is an old computing language with limited capability especially in I/O and data organization areas. The
system described above is easily envisaged in a more modern language such as C++. This would allow a true objectoriented approach and compatibility with modern computing platforms featuring graphical user-interfaces (GUIs). This
said, C++ appears to have limited advantage over FORTRAN as far as the mathematical constructs required to model
gas-turbine engines are concerned. A hybrid approach is feasible, and is inevitable in the short-to-medium term.
5.2
Another example with a focus on flexibility is MTUs in house performance program MOPS (Modular Performance
Synthesis Program). Its development started in the early 80s, and the program is presently extended to be a
multidisciplinary pre-design tool. In the beginning FORTRAN IV was used, and later FORTRAN 77. Recently added
4-16
options make use of the new data structures offered by FORTRAN 90.
MOPS is used for a wide variety of tasks, including engine test analysis, cycle design studies, off-design and transient
simulations. Moreover, MOPS is the basis for all computer decks issued by MTU.
Before actually using the program, the engine configuration must be defined with the help of a special pre-processing
program. In this pre-processing the user composes his engine from modules that can be connected in any sequence. In
most cases, a module is directly representing an engine-module like a compressor or a turbine. Besides the turbomachine modules there are also other modules like ducts, shafts, control units etc.
The primary connection between the modules is the main gas stream, and secondary connections are the shaft power
transfer between modules, internal air system paths, heat transfer, control sensor signals and position commands.
The program modules are strictly isolated from each other. Normally, they can only communicate via their primary and
secondary connections. The program internal nomenclature follows the ARP 755 standard and all calculations are done
in SI units. There is a sophisticated error message system built into the program, and in most cases standardized
diagnostic methods allow the reason for any problem that may arise to be found rapidly.
The user has to set up an iteration scheme, which is specific for his engine configuration and the task to be performed.
There are variables to be selected and errors to be defined. For example, in a mixed flow turbofan design task, the
bypass ratio may be used as a variable and the difference in static pressure between core and bypass flow may be treated
as the corresponding error.
Typical turbofan simulations for cycle design tasks employ iteration schemes with only a few variables, while test
analysis by synthesis tasks can require over 50 variables.
Setting up the iteration scheme requires a thorough background of gas turbine theory. This is a certain disadvantage, but
on the other hand, with MOPS, gas turbines of arbitrary complexity can be simulated.
Diagrammtitel
MOPS
Define Iteration
Design
Off Design
Test Analysis
Transient
Inlet Distortion
Engine Model
Atmosphere
Inlet
Compressor
Burner
Spool
Turbine
Nozzle
Control System
Utilities
Gas Properties
Isentropic Flow
Table Reading
5.3
GASTURB
The big programs used within industry for performance simulations all have one common problem. They require an
experienced engineer to operate them. Mostly there is no user-friendly interface and the user has to deal with the
sometimes-complex component matching issues.
When predefined engine configurations are employed, it is possible to hide all the mathematics from the user of the
program. This makes the program applicable for a much wider audience than the traditional gas turbine performance
programs.
An example of a 0-D model with predefined gas turbine configurations is GasTurb [6], see Figure 7. GasTurb was
originally developed as a Turbo Pascal program and was later transferred to Delphi. It has a traditional Pascal program
structure (i.e. a main program with subroutines and data blocks) where each engine configuration is implemented as a
program unit.
GasTurb is run on common desktop PCs with the Microsoft Windows operating system (Windows95/98 or Windows
NT4.0/2000). The development environment is Borland Delphi, which is based on the Object Oriented Pascal
programming (OOPascal) language.
Apart from the user interface, GasTurb does not make use of the object-oriented features that are offered by Delphi.
4-17
4-18
Main
Select Configuration
Off Design
Design
Select a Task
Optimization
Single Point
Parameter Study
Design Effects
Monte Carlo
Steady State
Select a Task
Design
Iteration
Cycle
Design
Transient
Steady State
Schedule
Single Point
Parameter Study
Operating Line
Off-Design Effects
Flight Envelope
Mission
Inlet Distortion
Monte Carlo
Bleed
T5 = f(T2,Mach)
NH = f(T2,P2)
etc.
Select a Task
Transient
Schedule
Manual Control
Fuel Step
Use a Schedule
Transient
Iteration
Transient
Off-Design
Iteration
Inlet
Iteration
Off-Design
Gas Properties
Atmosphere
Compressor
Maps
Turbine
Maps
Table
Reading
Nozzle
NL=f(Time)
NH=f(Time)
Power=f(Time)
4-19
5.4
NLRs Gas Turbine Simulation Program (Ref 1) is a component based modeling environment for gas turbines and
related systems. Both steady-state and transient simulation of any kind of gas turbine configuration can be performed by
establishing a specific arrangement of component models in a model window as displayed in Figure 9.
Both the flexibility and user-friendly interface are owed to GSPs object oriented architecture, which has been designed
with primarily these two qualities in mind.
The flexibility is to a large extent reflected in the component modeling approach. With efficient ways to develop or
adapt component models, simulations of new gas turbine configurations and models with different levels of (local) detail
or fidelity can easily be realized. For this approach a solver is required that is able to handle any configuration of
components (and thus states) in a model. In effect, a generic solver is needed for a virtual set of abstract components
with an undefined number of states. This also implies a specific approach for the user-interface i.e. an interface focused
on the component level.
Object orientation offers an excellent mechanism for this problem. Inheritance is used to concentrate code common to
multiple component types (e.g. both compressor and turbines have some similarities) in abstract component object
types or classes. From these abstract classes, component classes are derived and instantiated as real gas turbine
components in an engine model. See Figure 10 for the GSP component class hierarchy.
Many publications on object oriented software designs (and also on object oriented gas turbine simulation tools [1, 10])
exist [16] and show the three basic principles of object orientation: encapsulation, inheritance and polymorphism. These
principles offer significant potential to efficient gas turbine simulation software development.
Encapsulation enhances code maintainability and readability by concentrating both the routine code (representing
behavior) and data block code (representing the properties and state of operation) of a particular component in a
single component object class. Contrary to conventional software design practice (i.e. FORTRAN etc.), all data
declarations and procedures (in OOD terminology methods, both for interface and simulation calculations) are
concentrated in a single code unit.
Inheritance facilitates concentration of code common to multiple component classes in one or more abstract
ancestor classes. This is to eliminate code duplication. The turbo-machinery component class in Figure 10 for
example represents all functionality common to compressors, fans and turbines. Code maintainability is also
enhanced because single code adaptations in ancestor classes are effected in all descendant classes.
4-20
Polymorphism is the ability of abstract parameters to represent different object classes. This principle is extensively
applied in GSP. Every component class for example has a Calc method for running the simulation code. The
system model code has an abstract (polymorphic) component object identifier able to represent any real component
object instance in the model. During simulation, the system model subsequently lets the abstract identifier point to
successive components, calling their Calc methods. The abstract component object has an abstract Calc method,
that is a token representation of the real simulation code. During runtime a mechanism called late binding replaces
this abstract Calc code with the actual Calc code of the component it is representing.
4-21
4-22
4-23
4-24
5.5
TERTS is an example of a real-time 0-D component stacking model. TERTS is built with MATLAB Simulink, and
offers a component-based predefined configuration. In order to comply with the requirement of limited computation time
per time step (for real-time simulation), the one iteration per time step method was applied, which offers good accuracy
with high update frequencies (time steps smaller than 0.02 s). Figure 16 displays the TERTS thermodynamic engine
model level.
1
S_NG
S_NG
errors
19
dS_N1/dt
States
12
S_N1
NGc_
23
S_PRgt
TQreact
24
S_m2c
Old States
S_ML
S_PRlp
Duct
S_BR
S_m2c
16
S_mfanc
17
S_PRlp
18
State
Extractor
14
S_PRgt
15
S_MLfan
State_Change
AngMtum
S_N1
13
S_MLcpr
S_BR
Pt0
2
Tt0
8
mach
Fan
Inlet
Gasgenerator
LP Turbine
Mixer
Afterburner
Nozzle
3
mf
4
Ps0
out
7
mfAB
OUTPUT
6
ENP
nozzle pos.
1 1 T41
10
*
FDrag
Vt [m/s]
Vtchk
To Workspace3
p41t
2 (Pb)
Output
20 m6
7 h45t
2 1 FAR6
6 Ts_throat
5 PRnozzle
9 NGc
m2
Mux
Mux3
Fdrag
To Workspace4
22
Fdrag
FG
f(u)
Pt6/Pt2
3
EPR
8 EGT/FTIT
5.6
Models offering the possibility of use for gas turbine engine fault diagnosis have been developed by the Diagnostics
Group of the Lab of Thermal Turbo-machines and the National Technical University of Athens [2, 3, 4]. The
TEACHES model has been built with VISUAL BASIC programming language for a building a shell and a Graphic
User's Interface (GUI), operating in a MS Windows 98 environment. An engine performance calculation module
performs the key aero-thermodynamic calculations. This is a dynamic link library (DLL) written in FORTRAN.
Information is passed between the Visual Basic shell and the performance calculation module whenever performance
calculations are requested, as shown schematically in Figure 17.
4-25
Measurement data
(Diagnostic Mode)
Outputs
Adaptive Engine
Model
(FORTRAN)
Figure 17 - The structure of a modeling environment offering the possibility of fault diagnosis
The FORTRAN code forms the core of the modeling system. It employs the Adaptive Modeling technique (described in
Chapter 2) to perform fault diagnosis. By appropriate selection of input data it can perform either direct simulation of
engine operation at any desired operating point (Simulation mode) or a diagnosis of the condition of the engine
components, once a set of measurement data is available (Diagnostic mode).
4-26
Engine
Performance
Operating point
data
Graphic output
MAIN
WINDOW
Measurement
data
Measurement
Input
Graphs
Import Data
Component
Condition
Settings
Compressor
Performance
Measurements
Used
Turbine
Performance
Health
Parameters
Results
Results
Table
Export
Results
Engine Performance
Examples of Results
Examining the Effects of Component Malfunctions.
Modifying the performance characteristics of the components simulates different faults. The engine performance for
these modified characteristics is then calculated. The deviations from nominal component performance are introduced as
percentages in the corresponding section of the main window, and using scalars, multiplying the component
performance-parameter effects map modifications. The modification factors are explained in Chapter 3. For example,
setting the value of modification factor f1 to a value of 0.98 represents a reduction in pumping capacity of the
compressor by 2%. The modified component characteristics can be visualized, in comparison to the initial intact ones, as
for example shown in Figure 5. Figure 21 shows a comparison between the design and modified compressor
performance maps.
4-27
5.7
There are a number of proprietary or limited distribution programs used in the industry:
4-28
ON-X/OFF-X
CASANDRA (Cranfield)
TurboMatch (Cranfield)
TERTS
ECAP
FAST (Honeywell Allied Signal)
TESS (JAVA, Univ. Toledo, [19])
ARTEMIS
ATEST (AEDC)
AIAA Series Programs, A series of programs primarily written by Jack Mattingly for student use in conjunction with
the AIAA textbooks
MOPS and TESS are similar to GSP in that they provide a configurable modeling system with a suite of components.
These systems typically include other features to allow productive use in the engine design and analysis process.
Examples are communication to test data systems, provisions for matching models to test data and configuration
management for the various component and engine models for multiple users of multiple engine models.
NPSS (section 5.9) is also a configurable modeling system for both 0-D and multi-dimensional modeling.
5.8
Two types of Multi-Dimensional Engine-Modeling systems may be encountered: Those that extend a component
analysis tool to multiple components and those that were originally developed to model the entire engine system. The
NASA Average Passage Code (APNASA) [24] is an example of a component code which has been extended to cover
all the blade rows in an engine. Simplified component models or boundary conditions inputs are used to address
portions of the engine that can not be addressed using the component code. In some instances two multi-dimension
components codes are joined to address the interaction between different component types. The NPARC and ADPAC
code have been combined to study Fan and Inlet interaction [20, 21]. Models based on combining or extending detailed
component codes require the highest level of computer resources. To run a single point, these codes may require in
dozens or even hundreds of workstations. Codes have been developed for the purpose of multi-dimensional simulation
throughout the engine. These include the dynamic TEACC code [25] and the steady state NASA ENG10 [22] code.
These codes generally use some modeling simplifications (source terms, streamline curvature methods) to allow solution
on a single workstation in minutes to hours. They can generally run multiple points or even transients. As computing
power grows, these limitations may be reduced. However, the need to accept simplifying assumptions to allow for the
desired level of full engine analysis will remain for some time.
Multi-Dimensional codes require a level of geometry detail not generally required for 0-D system performance models.
The detailed component models often require gridding and application to different disciplines may put additional
compatibility constraints on the solution. Often the bookkeeping of the geometry and gridding exceeds the complexity
of other parts of the simulation. This process is often facilitated by use of a CAD package to manage geometry and
geometry centered information. Linking this information to a CAD system designed to address these issues greatly
simplifies the already complicated job of creating and managing the simulation.
Two commercially available CFD tools for gas turbine component simulation are:
CFX-TASCflow (http://www.software.aeat.com/cfx/products/TASCflow.html)
5.9
Overview
The Numerical Propulsion System Simulation NPSS [17] is a concerted effort by NASA Glenn Research Center, the
aerospace industry and academia to develop an advanced engineering environment - or integrated collection of software
1
programs - for the analysis and design of aircraft engines and, eventually, space transportation components . Its purpose
is to dramatically reduce the time, effort and expense necessary to design and test jet engines. It accomplishes that by
generating sophisticated computer simulations of an aerospace object or system, thus permitting an engineer to "test"
various design options without having to conduct costly and time-consuming real-life tests. The ultimate goal of NPSS
is to create a "numerical test cell" that enables engineers to create complete engine simulations overnight on costeffective computing platforms. Using NPSS, engine designers will be able to:
1
Parts of this section were copied from (see for further information): http://hpcc.grc.nasa.gov/npssintro.shtml
4-29
NPSS
Work Breakdown Structure
Engineering
Applications
High-performance
Affordable Computing
Simulation
Environment
Modular Architecture
- National Cycle Program
Toolkits
- SSCD
- CM Manager
Library/Utilities
- NURBS
- PEV
- Data Standards
Coupled Aero-ThermalStructural (CATS)
CFD/Controls
Spectrum
VCE
MSAT
0-D Engine/
1-D Inlet
0-D Core/3-D LP
Subsystem
2-D Combustor/3-D
Engine
Clustered Workstations
High-Speed Networks
Parallel System
Software
Code Parallelization
4-30
Examples
Shown here is output from a software tool called ENG10, which was developed through the Numerical Propulsion
System Simulation (NPSS) Project at NASA Glenn Research Center. The ENG10 code is used to analyse the airflow
through modern jet engines. One of the strengths of this code is its ability to use the results of studies of individual
components of an engine to model how the overall engine system behaves and how various components influence each
other.
4-31
Some codes developed through NPSS analyze individual engine components. Shown here is a simulation of the Energy
Efficient Engine's combustor using ALLSPD-3D, a Glenn - developed combustor code. Component design teams use
simulation codes such as ALLSPD-3D to simulate and design engine components in detail.
6
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
REFERENCES
Visser W.P.J, Broomhead M.J.; GSP, A generic Object-Oriented Gas Turbine Simulation Environment, ASME
paper 2000-GT-2, May 2000.
Mathioudakis K., Stamatis A., Tsalavoutas A., Aretakis N., 1999, Performance analysis of industrial gas turbines
for engine condition monitoring, presented at: First International Conference On Engineering Thermophysics,
Beijing, China, August 18-21, 1999 (ICET '99).
Mathioudakis K. , Stamatis A. , Tsalavoutas A., Aretakis N., 2000, Instructing the Principles of Gas Turbine
Performance Monitoring and Diagnostics by Means of Interactive Computer Models, paper 2000-GT-0584, The
45th ASME International Gas Turbine & Aeroengine Technical Congress, Munich, Germany, 8-11 May 2000
Tsalavoutas A.,Aretakis N.,Stamatis A., Mathioudakis K., 2000, Combining Advanced Data Analysis Methods for
the Constitution of an Integrated Gas Turbine Condition Monitoring as Diagnostic System, 2000-GT-0034, The
45th ASME International Gas Turbine & Aeroengine Technical Congress, Munich, Germany, 8-11 May 2000
Visser, W.P.J., Gas Turbine Simulation at NLR, Making it REAL, CEAS Symposium on Simulation
Technology (paper MOD05), Delft, the Netherlands, 1995.
Kurzke, J., Advanced User-friendly Gas Turbine Performance Calculations on a Personal Computer, ASME
Paper 95-GT-147.
Kurzke, J., How to get Component Maps for Aircraft Gas Turbine Performance Calculations, ASME Paper 96GT-164.
Claus R.W., et al., Multi disciplinary Propulsion Simulation Using NPSS, AIAA-92-4709-CP.
Visual Computing Environment (VCE) project at NASA Lewis Research Center, is part of the NPSS program: see
[8].
Reed, J. A. and Afjeh, A.A., Computational Simulation of Gas Turbines: Part I - Foundations of Componentbased Models, ASME 99-GT-346.
Reed, J. A. and Afjeh, A.A., Computational Simulation of Gas Turbines: Part II - Extensible Domain
Frameworks, ASME 99-GT-347.
SAE Aerospace Recommended Practice (ARP) 681
SAE Aerospace Recommended Practice (ARP) 755
Sellers J.J., Daniels C.J., DYNGEN - A program for calculating steady-state and transient performance of turbojet
and turbofan engines, NASA TN D-7901, 1975.
Visser, W.P.J., Kluiters, S.C.M., Modeling the Effects of Operating Conditions and Alternative Fuels on Gas
Turbine Performance and Emissions, NLR Technical Publication NLR-TP-98629 or Research and Technology
Organization, RTO-MP-14, 1999.
Booch, G., Object-Oriented Analysis and Design With Applications, Addison-Wesley Object Technology Series,
2nd edition (February 1994) Addison-Wesley Pub Co; ISBN: 0805353402
Evans, A.L. et al, An Integrated Computed and Interdisciplinary Systems Approach to Aeropropulsion
Simulation,ASME paper 97-GT-303
Geist et al, PVM: Parallel Virtual Machine, A Users' Guide and Tutorial for Networked Parallel Computing,
MIT Press, ISBN 0-262-57108 (on the WWW: http://www.netlib.org/pvm3/book/pvm-book.html)
Reed J. A. and Afjeh A. A., Intelligent Visualization and Control System for Multidisciplinary Numerical
Propulsion System Simulation, AIAA paper 96-4034, 6th AIAA/USAF/NASA/ISSMO Multidisciplinary and
4-32
20.
21.
22.
23.
24.
25.
ACRONYMS
3GL
3rd Generation Language, conventional computer language for structured programming of line-by-line
code according to a specific syntax; requires a compiler for translation to machine code. Examples are
FORTRAN, C, ADA, PASCAL, ALGOL, BASIC.
4GL
4th Generation Language, employs code generation using a visual interface. Often generated 3rd GL code.
Code generation is usually focused on the user interface, database structures or other specific tasks.
Examples are Visual Basic, Visual C++, Jbuilder, and Delphi, which are focused on generating code for
the interface. They are also designed to make working with a wide variety of databases very easy.
MATLAB-Simulink may be regarded as a 4GL environment for simulation (it generates C code for offline and real-time simulation).
CAD
FLOPS
ICT
IT
PVM
5-1
Chapter 5
Recent Progress
I
INTRODUCTION
This chapter provides an overall view of the recent progress made on the prediction and simulation of aspects of gas
turbine engine component performance. These components are inlet, compressor, combustor, turbine, afterburner,
nozzle, splitters, mixers, and secondary flow systems. It also includes a section on control systems modeling. The
chapter is arranged as follows. We begin with the rotating components (compressor and turbines); this is then followed
by the non-rotating components.
TABLE OF CONTENTS
5-2
3
V
1
4
VI
1
2
3
4
6
7
VII
1
5-3
II
INTRODUCTION
The performance map for a specific compressor configuration can be obtained through both systematic experimental
measurements and computations. With the recent advances made in numerical techniques [1, 2], and the availability of
computational resources, it is now possible to evaluate the performance and the design changes in multistage
compressor, on a three-dimensional flow basis using a computational flow solver. It is intended to provide a concise
delineation of those factors that can potentially modify baseline compressor performance-characteristics, rather than to
discuss the details of the measurement and computation techniques for compressor performance maps. These factors
are commonly referred to as characteristic modifiers.
There are thus two items of interest in this chapter, shown in Figure 1:
The generation of a baseline compressor map for a specific configuration;
The change in the pressure rise characteristic, the efficiency, and the operability limit (the surge line and the flutter
boundary).
Figure 1 - Representative compressor-map, with surge-line, speed-line, efficiency contours, and flutter boundary
In particular the specific focus will be on compressor performance changes associated with:
Changes in compressor tip clearance;
Stationary and rotating distortion, either self-induced (due to asymmetric compressor tip clearance) or externally
imposed (due inlet distortion);
Changes in intra-blade row gap;
Blade surface roughness and Reynolds number effect;
Changes in bleeds;
Changes in stator schedule and position;
Blade untwist associated centrifugal effect (this is of particular significance in fan rotor);
Heat transfer to compressor.
In this, it is assumed that the baseline compressor-performance characteristics are known and that one would like to
determine the change in compressor characteristics and operability range, for instance, due to changes in compressor
tip clearance. As such, the material present here is not meant to be exhaustive and extensive; it is aimed at providing
the readers with an adequate knowledge base (in line with the objective of the monograph) to aid the readers in:
Interpreting data from gas turbine engine;
Implementing gas turbine engine system simulation.
The readers are strongly advised to refer to the wealth of information on compressor aerodynamics, in the excellent
book by Cumpsty [3] and ASME Journal of Turbomachinery.
This chapter is organized as follows, based on work reported in journal publications up to 1999, of which the authors
are aware.
Generation of the baseline compressor map;
Effects of changes in (axisymmetric) compressor tip clearance on compressor pressure rise characteristics,
efficiency and stall inception point;
Response of compressors to stationary and rotating distortion, induced by inlet distortion and asymmetric
5-4
A compressor map is the 0-D representation of its performance, showing the mass flow and efficiency of the
compressor for a range of operating conditions. It may also include stall line operating limits and various flutter
boundaries, as shown in Figure 1. The requirements for the map vary with their application. For engine and control
system design, maps may be required to run from near zero speed to extreme over speed conditions that are beyond
shaft break limits, and from stall down to windmill pressure ratios less that unity. For special studies, post-stall maps
may be required.
A map represents the performance for a nominal set of conditions that may include:
Stator position;
Bleed amount and location;
Entrance temperature, pressure and gas properties (including Reynolds number, gas composition changes such as
water vapor or combustion products);
Tip clearance.
Compressor tip clearance causes a leakage of gas across the blade tip from the pressure surface to the suction surface.
This tip leakage interacts with the primary gas stream and the wall boundary layer. Indeed, the tip leakage flow
dominates the aerothermodynamic behavior of the end-wall flow and the blade-to-blade flow in the tip regions. As
such it has a strong impact on compressor efficiency and stability. The influence of tip leakage flow manifests itself in
two ways:
Fluid dynamic blockage that effectively reduced the flow area;
A thermodynamic effect, in that the work done in the tip region is different from that in the free stream.
The measured impact on compressor or fan performance due to increased tip clearance is shown in Figure 2 [3] and
Figure 3 [3]. As shown in Figure 2 when the tip clearance was increased from the baseline value of 1.38 to 2.8 per cent
of span for a low speed multistage compressor [3]:
Peak efficiency reduced by 1.5 points;
Peak pressure rise dropped by 9.7 per cent
Operability range reduced from 17.5 to 9.7 per cent (a reduction in stall margin).
The parametric trend of loss in performance (in terms of efficiency, pressure rise or pressure ratio and stalling pressure
rise coefficient or stall margin) with increased tip clearance is further brought out in Figure 3. L Smith [4] has
developed a correlative approach for estimating the effect of end-wall flow on multistage compressor efficiency. This
works in terms of:
The analysis that derives this correlation is based on a repeating stage assumption, and uses data from a set of
multistage designs or builds.
While the approach does not provide information about the flow, it is apparently a useful approach as elucidated in the
results shown in Figure 4 [4]. More recently a slight more complex correlation has been developed by Khalid et al.[5]
for the pressure rise capability across a rotor v blockage associated with the tip leakage flow; as the results in Fig. 5
show it has a behavior analogous to the diffusion factor used for correlating cascade data. While these approaches
5-5
appear correlative, their utility could lie in enabling the post-processing of vast amounts of computed data (e.g. from
the use of CFD to simulate flow in multi-blade rows compressor). This data could then be used for simulating gas
turbine engine systems.
Figure 2 - Effect of increased tip clearance on overall compressor performance for a low-speed compressor [3]
Figure 3 - Effect of tip clearance on overall compressor and fan performance [3]
Useful measured information on tip clearance effects in high-speed compressors was given in a paper by Freeman [6]
and Schmucker [9]. Figure 6 (taken from a lecture at VKI by Freeman) and Figure 7 (taken from [9]) show the impact
of increased tip clearance on the efficiency and pressure ratio of high-pressure compressors. Not only do the efficiency
and pressure ratio deteriorate with increased tip clearance, but also the loss or gain in efficiency is not monotonic with
increasing or decreasing tip clearance. Indeed, for tip clearances below a threshold value, the efficiency deteriorates
and the surge line moves considerably to the right so that the compressor would surge at higher mass flows. All these
measured effects are highly detrimental to the operability of a compressor.
In a multistage compressor environment the effect of changes in tip clearance would generate additional blockage and
loss (say for instance to the front stages), altering the matching of the downstream stages. Likewise, the alteration in
5-6
the performance of a specific stage, due to changes in tip clearance, could influence the aerodynamic matching with
the upstream stage. These aspects of tip clearance together with the consequential impact on stage matching have yet to
be defined on a quantitative basis. This is required if they are to be incorporated into a simulation procedure for gas
turbine engine performance changes associated with changes in tip clearance.
Figure 4 - Effect of tip clearance on overall compressor and fan performance [4]
Measurement of detailed flow in the rotor tip region has always been difficult. The advances in computational fluid
dynamics and computer technology have enabled the simulation of unsteady three-dimensional flow for rotors with tip
clearance, both in isolation and in multistage environments. This opens up an entirely new way of probing the physics
of tip leakage flow, with the objective of establishing its role on compressor performance on a non-correlative basis.
Figure 5 - Correlation of rotor pressure rise capability against flow blockage associated with tip leakage flow [5]
To summarize, we have reviewed physical measurements to show compressor performance deterioration with
increased compressor tip clearance. This may be attributed to the response of the flow behavior in the tip region, which
is essentially dominated by the tip leakage vortices. While the overall performance of compressor is set by the pressure
rise characteristics, its change is determined by the fluid dynamic event local to the compressor blade passage and the
flow conditions set by the preceding and succeeding blade row.
5-7
Figure 6 - Effect of tip clearance on pressure ratio, surge line and efficiency of 6-stage, high-speed compressor
[6]
An operating compressor or fan may be subjected to various types of flow distortion. In general, the distortion is of the
combined radial-circumferential type. Purely radial distortion can be assessed or analyzed using standard streamlinecurvature methods for computing steady-state ax symmetric flows. Circumferential distortion introduces additional
fluid dynamic effects that are associated with flow unsteadiness. Thus, a different class of methods is needed to assess
the response of compressors to circumferential distortions.
In this section, we shall consider the impact of following types of distortion on compressor performance:
5-8
Stationary inlet distortion, such as that due to flow separation in an inlet (e.g. due to cross-winds or aircraft
maneuver) upstream of the fan or compressor;
Rotating distortion, which can occur in a multi-spool compressors when rotating stall occurs in an upstream
compressor, thus imposing a rotating distortion on the downstream compressor;
Self-induced stationary distortion, associated with tip clearance asymmetry due to an off-centered rotor or oval
casing, Figure 8;
Self-induced rotating distortion associated with rotating tip clearance asymmetry due to non-uniform blade height
or whirling rotor, Figure 8. Whenever a compressor is subjected to an externally imposed distortion or self-induced
distortion, its stall margin and performance are degraded and that there could be aeromechanical consequences as
well.
Figure 8 - Non-axisymmetric tip clearance: (a) stationary caused by off-centered rotor and oval casing; (b)
rotating caused by whirling shaft and non-uniform rotor heights
Because of the significant impact circumferential-flow distortion has on compressor performance, empirical and
correlative approaches (SAE ARP 1419 [7], 1420 [8]), based on laborious compressor testing, have been developed for
estimating the aerodynamic consequence. The parallel or multi-segments compressor model [10] provides useful
results in certain class of problems and gives an overall view of the associated fluid dynamics. For instance, for a
compressor subjected to circumferential distortion it yields a loss in stability margin and stalling pressure ratio. It also
gives the result that for effective attenuation of distortion; a given compressor should have a steeper slope in the
unstalled pressure rise characteristics. However, the method does not provide an adequate stall point prediction.
Advances in analytical and computational techniques for assessing compressor response to flow distortion will be
described in the section on methodologies.
4.1.
STATIONARY AND ROTATING INLET DISTORTION
The measured effects of inlet distortion on the performance of a 9-stage axial compressor are shown in Figure 9. The
measurements showed that the stall line with circumferential inlet distortion moved considerably to the right, a
degradation in compressor stall margin. The general trends of compressor performance with different inlet distortions
are elucidated through a series of experiments undertaken by Reid [11]. A representative set of these results is shown
in Figure 10, which shows the compressor delivery pressure at the surge line for different types of distortions. Two key
aspects may be deduced from these data:
As the angular extent of the spoiled sector (low inlet total pressure) is increased, there is an angle above which the
exit static pressure changes little (Figure 10). This angular extent is often referred to as the critical sector angle.
Fixing the total angular extent of the distortion, the effect of sub-dividing it into different number of equal sections
is shown in Figure 11. The greatest effect on the loss of peak pressure rise is observed when there is only one
region. This suggests that inlet distortion patterns, which have a longer length scale and a lower circumferential
harmonic content, are the most important.
While there is a considerable database on the response of compressor to stationary inlet distortion, only limited
physical measurements are available on the degradation in performance (particularly in stall margin) when
compressors are subjected to a rotating inlet distortion. Experiments by Ludwig et al. [13] for an isolated rotor show
clearly that the stall margin is strongly affected by the speed at which distortion rotates. Co-rotating distortions have a
larger impact than counter-rotating distortions, with the maximum loss in stability when speed of rotating distortion is
near that at which a rotating stall cell would propagate if the compressor were throttled to stall. Kozarev et al (1983)
[14] has obtained similar results for a two-stage compressor.
5-9
5-10
Figure 12 - Flow coefficient at stall versus distortion rate, for single-resonance-peak type of compressor
4.1.1 AERODYNAMIC COUPLING BETWEEN INLET-INDUCED DISTORTION AND COMPRESSOR
The above discussion has so far been confined to the response of compressors to a distortion pattern that was specified
far upstream. However such is not often the situation encountered in practice. The development of the flow within the
inlet can be substantially altered by the presence of the compressor. This aspect of the flow interaction has been clearly
elucidated in the set of measurements taken by Hodder [18] shown in Figure 14. In the absence of the compressor and
with the inlet at 30-degree incidence there was a large region of low total pressure flow due to flow separation. In the
presence of the compressor, and at an even larger incidence of 35-degrees, the region of low total pressure fluid was
much smaller because the compressor acted to equalize the velocities. This reduced the extent of the separated flow
region.
5-11
Figure 13 - Flow coefficient at stall versus distortion rate. Two resonance peak type of compressor
5-12
Compressor response to rotating inlet distortion can be linked to the disturbance structure at the onset of the stall. The
results provide a clear illustration of a situation in which the unsteady flow effects directly affect the time-averaged
performance of the compressor.
STATIONARY AND ROTATING ASYMMETRIC TIP CLEARANCE
4.2.
We have examined the change in compressor performance associated with changes in compressor axisymmetric tip
clearance. However, the circumferential distribution of tip clearance may become asymmetric due to engine usage and
overhauls. An operational question of engineering interest is thus Can one regard the asymmetry in tip clearance as
just an effect of time and circumferentially averaged clearance?
Experimental and analytical investigations have been carried out on a low-speed multistage compressor [19] where a
single and a two-lobed approximately sinusoidal variation in tip clearance had been separately introduced. The
baseline configuration had an axisymmetric clearance of 4% chord. The quoted clearances are relative to the baseline,
so that +2% and 2% chord clearance corresponds to tip clearance of 6% and 2% chord respectively. The asymmetric
clearance was introduced by varying the casing so that the circumferential average clearance was the same as the
baseline clearance.
Figure 15 - Effect of tip clearance distribution on pressure rise characteristics and stall margin
These results presented in Figures 15 to 19, show that clearance asymmetry reduces both the compressor stability
margin and efficiency. The single lobe variation has a larger impact than two and higher lobe patterns, which is a
frequency effect. The reduction in stability margin is closer to the maximum than would be given by the circumferential
average clearance. The sensitivity of performance to clearance asymmetry is a function of the steady state compressor
response and unsteady response. The computed results from the theoretical model are in general agreement with the
measurements. While no experimental data is available to depict the effect of rotating tip clearance asymmetry on
compressor performance, calculations, based on the model presented in the section on methodology, show that the
effect is analogous to that of rotating inlet distortion. In this case, the impact on compressor performance deterioration
is lgreatest when tip clearance is co-rotating at a speed close to that at which the rotating stall inception would
propagate.
The performance of axial compressors is known to depend on axial gapping between blade rows. Experimental data,
published from General Electric (Smith [4]), Rolls Royce (Hetherington & Moritz ([21]) and Pratt & Whitney
(Mikolajczak [22]) and shown in Figure 20, clearly show that the performance (efficiency) of axial flow compressors
can be increased by optimizing the axial gap between adjacent airfoil rows. The results of Figure 20, while interesting,
provide neither a quantitative guideline nor an understanding of how to improve the performance of the machine.
Smiths results showed a one to two per cent gain in efficiency, and a two to four per cent gain in the stage pressure
rise in a low-speed research compressor, by reducing the blade row gap from 0.37 to 0.07 chords. Mikolajczak
obtained similar results in a moderate speed compressor. Hetherington and Moritz, however, obtained a 2 per cent gain
in efficiency by increasing the gap in front of the rotor rows in a multistage compressor. The experiments of Smith and
Mikolajczak suggest that reduced axial gap improves performance, while Hetherington & Moritzs investigation
indicates increased axial gap can also be beneficial. A wake recovery model, developed by Smith [23], establishes a
connection between blade row interaction and performance and is further substantiated by the computed results of
5-13
Valkov and Tan [26] to include the tip leakage vortex flow.
Figure 16 - Changes in stalling pressure rise with axisymmetric and non-axisymmetric clearance
5-14
Figure 19 - Changes in peak efficiency with asymmetric clearance, with the flow condition corresponding to the
peak efficiency points in next figure
5-15
Figure 20 - Variation of compressor efficiency (top) and pressure rise (bottom) for closely-spaced (dashed lines)
and widely-spaced (solid lines) blade rows
The performance deterioration of a high speed axial compressor rotor due to changes in Reynolds number and blade
surface roughness and airfoil thickness variations has been investigated on an experimental and on an analytical and
computational basis by Schaffler [46], Koch [25] and K. Suder et al. [47]. Within the flight envelope for a typical
5-16
modern fighter aircraft indicated in Figure 22 [46], the operational Reynolds number can change by a factor of 10
(from 100,000 to 1,400,000).
Figure 22 - Range of operational Reynolds number of a high-pressure compressor within the flight envelope [46]
Three flow regimes corresponding to this range of Reynolds number can readily be identified. These are:
Low Reynolds number regime where laminar separation occurs at least in the front stages, resulting in reduced
flow and efficiency levels and stall margin. This corresponds to operation at high altitude.
Intermediate regime with a turbulent attached boundary layer flow and hydrodynamically smooth blade surfaces;
High Reynolds number regime where the middle and back stages of high pressure ratio compression systems
experience turbulent attached boundary layer flow with hydrodynamically rough blade surfaces. This corresponds
to operation at low altitude.
A multistage compressor can be expected to show the following behavior over a wide Reynolds number range:
Mass flow increases steadily with Reynolds number up to the choking condition in the blade passage;
The surge line is essentially unaffected by Reynolds number until there is severe flow separation corresponding to
operation at Reynolds number less than 100,000;
Polytropic efficiency varies with distinct slope changes that are dictated by specific boundary layer behavior.
The impact of boundary layer behavior on the functional dependence of efficiency on Reynolds number is shown in
Figure 23 [46] for an axial compressor at design point operation. As to be expected there are again three different
regimes demarcated by a lower and an upper critical Reynolds number that define the laminar separation boundary and
the surface roughness boundary. The consequence on compressor performance changes corresponding to each of the
regimes is indicated in Figure 24. If the Reynolds number is sufficiently high, the efficiency essentially becomes
independent of the Reynolds number. Under this condition the blade surface roughness elements protrude through the
laminar sub-layer of the turbulent boundary layer.
5-17
5-18
Figure 26 - Effect of variation in blade surface roughness on pressure ratio versus mass flow characteristics at
70 per cent span.
5-19
(a)
(b)
Figure 27 - Effect of Reynolds number on the performance of: (a) a three-stage intermediate pressure
compressor; and (b) a six stage high pressure compressor [46]
Variable stator vanes (VSV) are used to optimize performance and operating stability of compression components.
Compressors often operate with stator positions differing from the nominal schedule that is consistent with the map.
For single stage machines, the stator angle can be used directly although an intermediate stator position indicator is
often used rather than the actual airfoil angle. For multistage machines with multiple variable stators, the variable
stators are generally ganged together and one average stator angle is used to represent the overall stator movement.
Tabular 0-D compression component maps may address VSV effects by providing separate maps for different average
stator positions relative to a nominal setting and interpolating between these maps for intermediate stator settings.
Analytical 0-D component maps can do the same or use the implied change in the average stage characteristic and flow
coefficient to indirectly model the impact of stator movement based on a simplified physical model. Stator movements
5-20
change the overall performance of the component and the matching between stages, which can affect the predicted
inter-stage conditions for the same overall operating conditions. This can be important for bleed and cooling flow
extraction in full engine use. Figure 29 shows the impact of stator position on flow and efficiency as a function of
speed for a particular operating line.
Higher fidelity models that require specific geometry will generally model the effect of stator position directly as part
of the component representation. Stator position accuracy is critical to obtain accurate results from these high fidelity
models. Stators and their position sensors are a frequent source of difficulty in validating component models with real
data due to hysteresis and variation between the indicated and actual stator position.
Figure 29 - Impact of stator position on flow and efficiency as a function of speed for a particular operating line
8.1.
TEMPERATURE AND GAS PROPERTY EFFECTS
A compressor map generated for a specific set of entrance conditions may not be accurate at other entrance conditions
due to the effects noted above. Many of these effects are compounded. Temperature changes affect clearance as the
case grows. The relationship between the mechanical and corrected speed changes affects both blade untwist and blade
growth. Unless the entrance pressure changes to match, entrance temperature also changes the Reynolds number. Even
if all these effects are ignored, the use of temperature corrected parameters still ignores the change in the speed of
sound with gas properties due to temperature and gas composition. So-called gamma-R corrections are commonly
used with most 0-D models to adjust base map predictions to a constant average Mach number similarity parameter.
This is to allow for gas property changes. Due to lack of test data to separate these effects, it is common to select a few
key effects such as clearance, Reynolds number and VSV (variable stator vane) angle to cover empirically all the
effects not modeled hitherto.
8.2.
HEAT TRANSFER EFFECT
For 1-D, 2-D and 3-D models, the effect of heat transfer from the blades and static parts to the gas path flow can be
modeled directly. The change in temperature will modify the stage-by-stage performance and result in an overall rematching of the compressor. At a 0-D level, it is generally not possible to create an adjustment to the compressor
performance that can be distinguished from the empirical bulk heat-transfer effects in the overall engine model. The
heat transfer impact on compressor performance depends on the controlling stage in the machine and whether heat
transfer at that particular operating condition moves that stage to a less or more favorable condition.
Heat-transfer effects on 0-D compressor maps are difficult to model, because performance depends on the controlling
stage in the machine and whether heat transfer at that particular operating conditions moves that stage to a less or more
favorable conditions. However, simple approaches allow the main consequences of heat transfer to be caught.
5-21
8.3.
FAN AND BYPASS DUCT SYSTEM
As shown in 30, the fan-bypass duct system consists of a fan followed by a splitter that subdivides the fan flow into a
stream through the bypass duct and another through-the-core compressor. The core compressor may also include a row
of exit guide vane and supporting struts.
Fans have a hub-to-tip ratio ranging from 0.3 to 0.6, while core compressors have a much higher hub-to-tip ratio of 0.7
and above. Thus the outer portion of the fan generates the pressure ratio needed to maintain the required flow in the
bypass duct and the inner portion of the fan generates the pressure ratio for the flow into the core compressor. The
dividing streamline that demarcates the inner and outer stream of the fan is the stagnation streamline with the
stagnation point located on the leading edge of the splitter.
Fan OD map F(
Pt12.5
,
Pt12
t2
Wtot t 2
, fanOD ,surge line, see Figure 31
2
e
e Lin
Surg
5-22
P
W fan OD t12.5
N
, see Figure 32.
,
,
Fan exit guide vane (FEGV) map F
t12.5
P fegv
t12.5
Pt 2.5
W2 t 2.5
N
,
,
, LPC ,surge line, see Figure 33
2.5
Pt 2
t 2.5
e
e Lin
Surg
P
N
Super charging curve for primary stream F ( t 2.5 ),
, BPR, LPC , see Figure 34.
Pt 2
t2
5-23
BLADE UNTWIST
8.4.1 DESCRIPTION
Blade untwist is a mechanical phenomenon where the leading edge blade angle varies with operating condition.
Typically this is due to the loads from centripetal acceleration and changing pressure forces on the blade. It is primarily
of concern for low hub/tip radius ratio blades (high bypass fans and the first stages of multi-stage medium bypass
fans), but can be an issue for accurate detailed predictions of off-design behavior in highly loaded multi-stage
compressors. Figure 35 shows an example of untwist for a high bypass fan. The change with flight condition shows the
need to model more than just the centripetal accelerations. Figure 36 shows the impact of pressure and mechanical
loading on the blade. The fan is designed with some nominal untwist which changes with both engine power setting
and environmental conditions. Figure 37 shows the definition and conventions for untwist angle.
5-24
5-25
5-26
gm
c
The reduced damping, g/*, combines the mechanical damping, g, and the mass ratio, , into a single non-dimensional
parameter, useful for linear stability purposes. This parameter, is defined as
where m0 is the modal mass (per unit length), c is the chord, and 0 is the inlet density. The g/* label emphasizes the
dependence upon inlet density. For cases with frictional damping (e.g. inserted blades), the reduced damping is order 1
and has a significant impact on flutter stability.
5-27
The compressible reduced frequency, K0*, is defined as 0c / (RT)1/2, where 0 is the blade natural frequency at rest,
and (RT)1/2 is the inlet speed of sound (which varies with inlet temperature, T). The K0* parameter comes from a
decoupling of corrected performance effects from purely aeroelastic effects. Roughly speaking, for a constant structure
(i.e. constant o c) the parameter K0* varies with the temperature, a parameter for which the performance is
"corrected".
The first two parameters of the set, mc and Nc, account for the corrected performance, as normally measured for high
speed turbomachines. For constant structural parameters, the latter two parameters, K0* and g/*, span the inlet
temperature and density. These can be related to the flight conditions in terms of flight Mach number and altitude. An
implication of this four-parameter viewpoint is that the flutter clearance of a machine depends on the corrected
performance point, and the intended flight envelope of the aircraft.
These parameters were applied to the analysis of full-scale test data. Figure 40 shows the trends with K0* and g/*.
Both increasing K0* and increasing g/* (at constant mc and Nc) have stabilizing effects upon flutter stability. It is
proposed that these trends hold generally. The trend with g/* is analytically based in the equations of motion, and is
equivalent (by similarity) to the statement that increasing the mechanical damping is stabilizing. The trend with K0* is
ultimately empirical, but it is supported by several previous studies and is equivalent (by similarity) to the widely-held
design principle that increasing the natural frequency stabilizes flutter.
2.6
g/*
2.4
1.2
1
0.8
0.6
2.2
0.7
K*0
0.65
0.75
Increasing
g/
1.8
1.6
1.4
1.2
1
70
80
90
100
110
120
130
140
150
120
130
140
150
a)
2.6
g/*
2.4
1.2
1
0.8
0.6
2.2
0.7
K*0
0.65
0.75
Increasing
K*
0
1.8
1.6
1.4
1.2
1
70
80
90
100
110
b)
Figure 40 - Trends of performance map flutter boundary with K0* and g/
* from full-scale engine data
Figure (a) shows the effect of varying g/*, and (b) shows the effect of varying K0*. Increases in both K0* and g/*
are stabilizing.
With the relevant similarity parameters spanning a four-dimensional space, the range of comprehensive testing is vast.
The proposed trends with K0* and g/*, however, suggest a way to simplify the requirements for comprehensive
testing. Since increasing K0* and increasing g/* are both stabilizing, it follows that the worst case at a particular (mc,
Nc) are the minimum values of K0* and g/*. Thus, clearance for the engine at the minimum K0* and g/* implies
clearance throughout the flight regime.
5-28
METHODOLOGIES
We have so far been examining physical factors that impact the compressor performance and stability margin; these
have been referred as compressor characteristic modifiers. In this section we will discuss analytical or computational
models for assessing the effects of some of these characteristic modifiers on a quantitative basis. In principle one can
proceed to build a correlative database for defining the effects these have on compressor performance through
extensive testing. Such an approach, while of engineering utility, does not provide a rigorous scientific approach or
basis that would allow one to predict changes in performance due to changes in the modifiers. Computational
techniques have often been suggested as the ultimate means to predict component performance and its change with
subsequent alteration of parameters such as tip clearance and operating conditions from those it is designed for.
However as articulated by Adamczyk [1], the range of time scales and length scales encountered in flows through
multistage compressors are so broad that it is unlikely that direct computation is feasible in the foreseeable future.
Adamczyk suggested that one way to overcome this is to calculate fluid dynamic effects when the computational
timescales are acceptable, and to model them on other occasions. Such an approach has been adopted by Adamczyk [1]
and Rhie, et al.[2] to develop procedures to compute the performance of a multistage compressor. In the following,
methodologies to provide quantitative assessment of the impact of characteristic modifiers, via the simulation of flow
phenomena in multistage compressors, are described.
9.1.
APPROXIMATIONS AT GLOBAL LEVEL
In the development of techniques for simulating flow phenomena in multistage compressors with the aim of extracting
quantitative information on compressor characteristics and its modifiers, there are two levels of approximation:
Physical Approximations to problem under consideration; here we first ask the following question
- What is it that is important to include?
Numerical approximations to partial differential equations (PDE) for the problem
In general the various levels of physical approximation for flow phenomena can broadly be given as follows: Navier-Stokes equations (direct numerical simulation)
Large eddy simulation
Reynolds averaged Navier-Stokes equations - this would require the use of a turbulence model
Thin shear layer approximation
Inviscid Euler equations
Steady flow v unsteady flow
Three-dimensional v two-dimensional flow
Streamline curvature
Throughflow methods (hub-to-tip & blade-to-blade)
Potential flow equations
Incompressible flow
One-dimensional flow (control volume approach)
Because of the large range of length and time scales involved in the multistage environment, high-speed multistage
turbomachinery flows are not presently amenable to direct numerical simulation (Adamczyk). Thus it is still necessary
to develop multi-row geometry models that yield averaged descriptions. These provide useful information that can be
used to generate compressor performance maps and to quantify changes associated with characteristic modifiers.
Hence we have to address the basic question of What is the proper level to average at? Furthermore the flow in
multistage compressors is inherently unsteady. One is thus interested in assessing the time averaged impact of flow
unsteadiness on compressor performance. (The characteristic modifier associated with changes in axial gap is a good
example to illustrate this aspect of the problem.) In view of this it is useful to draw on the equation of hierarchy for
turbomachinery flow given by Adamczyk:
It is clear from the above that as one proceeds downward through this list, the degree of empiricism increases. Or more
positively, one needs to do an ever better job of flow modeling. Conversely, the requirement for computational
resources increases as one proceeds upward through the list.
5-29
5-30
Local to System
System to Local
5-31
distortion [48].
Local level CFD calculation at the component level: This is used to establish information for use in the
representation of the blade row as a body-force (i.e. local to system level). Likewise, this can also be used to design
the blade row to yield the desired component characteristic to result in a compressor with good operability
characteristics (i.e. system to local level).
To assess the impact of adjacent blade rows or adjacent stages, we can first implement unsteady 3-D flow in a discrete
rotor-stator environment. (The computed results in this situation can be post-processed to examine the influence of
adjacent blade rows on the resulting unsteady blade loads associated with rotor tip vortices and blade wakes.) The
computed results for a discrete rotor-stator can then are used to establish the equivalent body force representation in a
manner illustrated in Fig. 41
Now we can envision embedding a discrete rotor-stator pair in a multi-stage environment with the remaining stages
represented as body force distribution as indicated in Figure 42. Computed results from this situation can be assessed
against calculations from the discrete rotor-stator pair alone to quantify multi-stage effects on time-average
performance. One can take a step further to envision aeroacoustic and aeroelastic calculations within the framework
shown in Figs. 41 and 42. For instance, the aero forces and moments as dictated by the multistage environment on a
specific blade row can be viewed as an input for local aeroelastic calculation (as implied in Figure 41). The local
aeroelastic calculation would now redistribute the body force representation of the specific blade row under
consideration. This is then in turn used in the multistage environment to update the flow field and the associated blade
response. Thus, this provides a conceptual model for implementing multistage aeroelastic calculations on a consistent
basis. For application of this methodology for assessing the stability properties of multistage compressor, the reader is
referred to [48].
ROTOR
STATOR
....
Stage Represented
as Bodyforce
Distribution
Unsteady 3-D/2-D
Flowfield in a Discrete
Rotor-Stator Pair
5-32
10
OVERALL SUMMARY
In this section we have reviewed the impact of tip clearance, stationary and rotating flow distortion, flow unsteadiness
(axial spacing), Reynolds number, and blade surface roughness on compressor performance. We have also reviewed
effects of thermal origin, and effects associated with blade untwist in high-speed fans. Recent progress made in the
development of techniques for simulating and predicting flow effects that affect the operability of the compressor is
also discussed. These techniques take advantage of modern CFD but physically model those aspects of the flow
phenomena that are still beyond the feasibility of currently available computational resources.
11
1.
2.
3.
4.
5.
6.
REFERENCES
Adamczyk J A Model Equation for Simulating Flows in Multistage Compressor, ASME Paper No. 85-GT-226,
1985
Rhie, C.M., etal Development and Application of a Multistage Navier-Stokes Solver: Part I, ASME Journal of
Turbomachinery Vol. 120 No. 2, April 1998.
Wisler, D.C. 1988 Advanced Compressor and Fan Systems. GE Aircraft Engines, Cincinnati, Ohio, USA.
Copyright @ 1988 by General Electric Co. USA. All Rights Reserved. (Also 1986 Lecture to ASME
Turbomachinery Institute, Ames Iowa)
Smith L.H. 1969 Casing Boundary Layers in Multistage Compressors. Proceeding of the Symposium on Flow
Research on Blading. Brown Boveri & Co. Ltd, Baden, Switzerland 1969. In Dzung L S (ed) Flow Research on
Blading. Elsevier Publishing Company 1970.
Khalid A. etal 1998 Effect of Tip Clearance on Compressor Endwall Blockage and Pressure Rise Capability
Presented at 43rd IGTI Conference, Stockholm
Freeman C 1985 Effect of Tip Clearance Flow on Compressor Stability and Engine Performance. VKI Lecture
Series 1985-05.
5-33
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
SAE Aerospace Information Report, AIR-1419, "Inlet Total-Pressure Distortion Considerations for Gas Turbine
Engines", May 1983.
SAE Aerospace Recommended Practice, ARP-1420, "Gas Turbine Engine Inlet Flow Distortion Guidelines",
March 1978.
Schmucker, J. etal 1994 Errosion, Corrosion and Foreign Object Damage Effects in Gas Turbines. AGARD-CP558.
Graf M. B., Sharma O. P. (1996) Effects of Downstream Stator Pressure Field on Upstream Rotor Performance
ASME Paper No. 96-GT-507
Mazzawy, R.S., 1977, Multiple Segment Parallel Compressor Model for Circumferential Flow Distortion, ASME
J. Engineering for Power, 99, pp. 288-296.
Reid, C., (1969) The Response of Axial Flow Compressors to Intake Flow Distortions ASME Paper 69-GT-29
Ludwig, G.R., Nenni, J.P., Arendt, R.H., 1973, Investigation of Rotating Stall in Axial Flow Compressors and
Development of a Prototype Stall Control Ssytem, Technical Report USAF-APL-TR-73-45.
Kozarev, L.A. and Federov, R.M., 1983, Aspects of the Appearance and Elimination of Breakdown in an AxialFlow Compressor in the presence of a Rotating Non-Uniformity at the Inlet, Izevstinya vuz Aviatsionnaya
Tekhnika, 26(1), pp. 33-37 (translated).
Longley, J.P., Shin, H-W, Plumley, R.E., Silkowski, P.D., Day, I.J., Greitzer, E.M., Tan, C.S., Wisler, D.C.,
Effects of Rotating Inlet Distortion on Multistage Compressor Stability, ASME Journal of Turbomachinery
Silkowski P. D. (1995)Measurements of Rotor Stalling in Matched and Mismatched Multistage Compressor,
MIT Gas Turbine Report No. 221, Massachusetts Institute of Technology.
Day J. J. (1993)Stall Inception in Axial Flow Compressors, ASME Journal of Turbomachinery, Vol. 115, pp. 19.
Hodder, B.K., 1981, An Investigation of Engine Influence on Inlet Performance, NASA CR-166136.
Graf, M., Wong, T., Greitzer, E.M., Tan, C.S., Marble, F.E., Shin, H-W., Wisler, D., Effects of Asymmetirc Tip
Clearance on Compressor Stability, accepted for publication in ASME Journal of Turbomachinery, 1997.
Gordon, K. 1998, Private communication.
Hetherington R., Morritz R. R. (1977) The Influence of Unsteady Flow Phenomena on the Design and Operation
of Aero Engines, in Unsteady Phenomena in Turbomachinery. AGARD CP-144.
Mikolajczak A. A., The Practical Importance of Unsteady Flow Phenomena on Design and Operation of Aero
Engines, in AGARD CP-144, Unsteady Phenomena in Turbomachinery, 1977.
Smith L. H. (1966) Wake Dispersion in Turbomachines, ASME Journal of Basic Engineering, September 1966.
Smith L. H. (1996) Discussion of ASME Paper #96-GT-029 Birmingham, United Kingdom, June 1996.
Koch, C. C., 1981, Stalling pressure rise capability of axial flow compressor stages, ASME Paper 81-GT-3.
Valkov T. V. The Effect of Upstream Rotor Vortical Disturbances on the Time-Average Performance of Axial
Compressor Stators, MIT, GTL, Report #227, August 1997.
Walker G. J. (1974), The Unsteady Nature of Boundary Layer Transition on an Axial-Flow Compressor Blade,
ASME Paper #74-GT-135.
Pfeil H., Herbst R., Schroder T. (1983), Investigation of the Laminar-Turbulent Transition of Boundary Layers
Disturbed by Wakes, ASME Journal of Engineering and Power, 105
Hodson H. P. (1990), Modeling Unsteady Transition and Its Effect on Profile Loss, ASME Journal of
Turbomachinery, 112, October 1990
Sharma O.P., Pickett G. F., Ni R., Assessment of Unsteady Flows in Turbines, ASME Paper, #90-GT-150,
1990.
Hodson H. P. (1990), Modeling Unsteady Transition and Its Effect on Profile Loss, ASME Journal of
Turbomachinery, 112, October 1990
Dong Y., Cumpsty N. A. (1990a), Compressor Blade Boundary Layers: Part 1 - Test Facility and Measurements
with Incident Wakes, ASME Journal of Turbomachinery, 112, April 1990
Dong Y., Cumpsty N. A. (1990b), Compressor Blade Boundary Layers: Part 2 - Measurements With Incident,
ASME Journal of Turbomachinery, 112, 1990.
Halstead D. E., Wisler D. C., Okiishi T. H., Walker G. J., Hodson H. P., Hsin H. W. (1995), Boundary Layer
Development in Axial Compressors and Turbines. Part 1 of 4: Composite Picture, ASME Paper #95-GT-461.
Halstead D. E., Wisler D. C., Okiishi T. H., Walker . J., Hodson H. P., Hsin H. W. (1995), Boundary Layer
Development in Axial Compressors and Turbines. Part 2 of 4: Compressors, ASME Paper #95-GT-462.
Schulz H. D., Gallus H. E., Lakshminarayana B. (1990a), Three-Dimensional Separated Flow Field in the
Endwall Region of An Annular Compressor Cascade in the Presence of Rotor-Stator Interaction. Part 1: QuasiSteady Flow Field and Comparison With Steady-State Data, ASME Journal of Turbomachinery, 112, October
1990.
Kerrebrock J. L., Mikolajczak A. A. (1970), Intra-Stator Transport of Rotor Wakes and Its Effect on Compressor
Performance, ASME Journal of Engineering and Power, October 1970, pp.359-368.
Dawes, W. N. (1994), A Numerical Study of the Interaction of a Transonic Compressor Rotor Overtip Leakage
Vortex with the following Stator Blade Row, ASME Paper #94-GT-156.
Valkov T., Tan C. S. Control of the Unsteady Flow in a Stator Blade Row Interacting with Upstream Moving
5-34
5-35
III
The turbine system of the gas turbine engine is a complex integration and balancing of the disciplines of aerodynamics,
internal and external heat transfer, stressing, material properties, and manufacturing capability. During the design phase
of the engine it is important to consider the expansion system in totality, i.e. high pressure turbine, interduct, low
pressure turbine, exhaust diffuser and jet pipe. By so doing, the optimal engine may be derived. This may or may not
mean that each turbine stage is at its peak aerodynamic performance point. Turbine interducts are potential sources of
high losses, with high mach number swirling flow, highly rotational flow, and often with large radius changes. The
turbine turbomachinery behaves in similar ways to that of the compressors. Reynolds dependency on profile losses,
shock-boundary layer interactions, tip clearance effects, and many of the references given under the compressor section
are relevant to both environments
This section of the report aims to give the reader an understanding of the way in which turbine performance is predicted,
measured and analysed, highlighting areas of particular sensitivity.
1.1.
1.2.
GENERAL PERFORMANCE
High pressure turbines operate for most of the time at a fixed expansion ratio with small changes in aerodynamic speed.
Depending on the performance of the compressors, the LP turbine, the combustor pressure loss and efficiency, and the
capacity and efficiency of the HP turbine, the turbine may, or may not, operate at its design expansion ratio and
aerodynamic speed. In order to develop the engine it is important to have a characteristic of the turbine for use in the
synthesis of the engine. In early design stages and simplified cycle calculations it is sufficient to hold a constant
efficiency due to the relatively small variation in operating point.
The turbine has to convert energy from the hot gas stream to drive the compressor on the same shaft. In its simplest
form, assuming constant flow through the gas turbine, the energy balance is:
Compressor enthalpy rise = Turbine enthalpy drop
In practice this is moderated in varying degrees by the following:
Compressor:
External bleed bypassing the turbine;
Internal bleed , partially or wholly bypassing the turbine;
Leaks (i.e. from variable geometry vanes) bypassing the turbine;
Radiated heat from casings.
Secondary air system:
Windage losses on shaft and discs.
Bearings and power offtake shaft:
Heat to oil and windage losses; power extraction.
Turbine:
Position of cooling air return relative to rotor throat(s)
Energy state of returned secondary air
Heat losses through casings.
These modifiers of the simplified arrangement all need quantification for a real assessment of the turbine performance in
the engine. Some of the above are amenable to direct measurement, whilst others are reliant on calculation and separate
experimentation with subsequent superposition.
Two of the above are closely linked and are the two that are most difficult to assess. They are 1) the secondary air
system, and 2) the energy state of returned secondary air. Both these systems are driven by main gas stream pressure
differentials, which are themselves subject to changes throughout the running range, and are strongly subject to engine
deterioration. A typified cooling and secondary air system is indicated in the following figure
5-36
Platform
leakages
Casing
cooling
Main gas
throughflow
Leading
edge
cooling
Surface and
trailing edge
cooling
Rotor:
Bleed flows introduced in the preceding blade row are treated as available for work in subsequent blade rows.
Platform and disc bleed flows introduced upstream of the throat affect throat sizing, but are not available for work.
Airfoil bleeds upstream of the throat contribute to the work in the turbine.
Bleed flows downstream of the throat contribute no work in that blade row.
The astute reader will recognize that such a convention is not academically correct. This highlights the importance of
clarity of the turbine efficiency definition being used.
1.3.
CHARACTERISTIC GENERATION
The potential efficiency of the turbine stage, or stages, is dependent on the basic velocity triangles i.e. work, blade
speed, and throughflow velocities. These have to be considered in the context of the turbine physical size and the hub/tip
ratio of the stage concerned. Such physical considerations determine the effects of trailing edge thickness on the
potential efficiency, the influence of tip clearance, and the Reynolds regime that the turbine will operate in.
5-37
Optimization of the airfoil profile, shock and secondary losses leads to the design point efficiency of the turbine, when
the effects of cooling and secondary air flows have been incorporated. In order to generate a turbine characteristic for
use in the synthesis of the gas turbine engine one of three actions can be taken.
Scaling of a previously measured turbine characteristic which has similar leading parameters i.e. work, hub/tip ratio,
throughflow axial mach numbers, and reaction.
Prediction based on blade row stacking. Based on regression analysis of previous turbines, the generic performance of
the turbine can be predicted over a wide range of operation. Such methodology works adequately well with consistent
and progressive small changes in the designs. It may however lead to substantive errors when more radical changes in
design are pursued [62, 63, 64].
Quasi 3D/3D simulation of the turbine stage(s). This methodology has proven itself capable of simulating widely
different types of design with good results. The speed range and expansion ratio range is more limited than methods 1
and 2, normally due to the range of applicability of the boundary layer methods and calculation stability at off-design
flow conditions. The flow conditions along the hub and casings allow better analysis of the secondary air system, and
from the knowledge of the changes in lift distribution on the airfoil, so improved blade cooling assessments can be made
[65]. This method does have the following added potential benefits: identifying the effects of upstream wakes, and
downstream perturbations from struts or instrumentation; assessing the exit profile from the turbine at the various
operating conditions. This is particularly important if inter-turbine ducts (especially with large radius change) are used,
and if the downstream turbine is sensitive to pressure and temperature profiles.
1.4.
RIG AND ENGINE TESTING
The HP turbine lives in a hostile environment and hence there are associated difficulties of measuring, as opposed to
deducing, its in-engine performance. Consequently, much turbine development has used cold flow rigs, which
replicate the turbine unit and are operated at substantially lower temperatures. Without the complexity of combustor or
following turbine, more accurate measurements may be taken over wide operating regimes, but which need accounting
systems to transpose from the rig to the engine environment.
Four major types of rig have been used in the development of turbines, viz.
2D or annular cascades these facilities are capable of high accuracy measurement of the aerodynamics of the airfoil
over a wide range of operation and Reynolds number. The necessity to control the end wall boundary layer, and its
effect on the mid height aerodynamics led to development of methods such as porous walls, and each tunnel has its
own particular characteristic. Due to the potential of high quality test results these tunnels have formed the
foundation of improvement to the analytical approach to airfoil design [66].
This can be further subdivided into two sections
Internal cooling Investigations of the effectiveness of internal cooling passage designs, including the residual
pressure that is key to the selection of the location of the exhausting of the coolant , and its impact on the external
aerodynamics. [67]
External cooling Used to determine the optimum pitching and direction of the coolant ejection holes for
effective filming of the surface. Such rigs also indicate the sensitivity of the film to the hole shape and the
pressure ratio between the main gas passage and the source. [68]
Both of these rig types need to be able to evaluate the effects of corriollis forces from within the gas flow, and
superimposed centrifugal forces, and their consequential effects on the localised heat transfer rates.
Cold flow turbine rigs The cold-flow turbine rig simplifies the instrumentation and measurement difficulties that
are encountered in the high temperature in engine situation. It allows for easier full spatial inlet, inter-blade row, and
exit traverse, and can exercise the turbine over a wide of operation. As such these facilities provide the first insight
to the measured performance of the turbine, as opposed to the predicted performance. There are corrections to be
made to the measured performance, such as disc windage and bearing losses, and gas conditions from rig to engine.
In uncooled, or low flow cooled turbines the cold flow rig has close similarity to the engine environment. Most of
the modern engines now rely on relatively high proportions of the main gas flow being used for cooling and
secondary air systems. Consequently, the cold flow rig with solid blading has to attempt to emulate these effects.
The experimenter can elect to maintain the blades as per the engine, test the turbine over the range of expansion
ratios, and then manipulate the turbine results for the correct engine reaction. Alternatively, it may be preferred to
redesign the nozzle, such that the correct reaction is achieved at the design expansion ratio. Such considerations are
compounded with increasing stage numbers. Both approaches ignore the effects that cooling films have on the basic
loss of the airfoil, and potential shock-interaction changes, and further modification to the measured characteristics
must be made.
Warm flow turbine rigs These facilities have the advantages of the cold flow rig plus the capability to introduce
cooling and secondary air into the turbine. The spoiling effects of these return bleeds, their effects on disc windage
and the blade row to blade row throat distribution and hence stage reaction, lead to a better simulation. Some
facilities use vitiated cooling air that has been combined with other gases in order to better approximate the relative
densities between the cooling and the main stream air. They therefore represent very closely the in-engine
5-38
performance. Due to their complexity, the cost of the unit and plant on which the unit is tested is substantially
greater than that of cold flow testing. Since the airfoils contain either engine or emulated engine cooling systems, the
availability of these complex parts may be late in the lifespan of the engine development. So although good
measurement of the turbine characteristic can be accomplished, the timing may be so late that the only gain may be
an understanding of how to make improvements for the next mark of engine!
The experimenter must decide how best to run the unit to generate the required characteristics. Options include: running
the cooling or secondary air at constant feed to re-entrant point pressure difference; or holding feed pressure constant
over a small expansion ratio variation; varying the leading edge feed pressure in conjunction with trailing edge; or only
varying trailing edge feed as expansion ratio changes. It must also be noted that as the feed pressures change, so may,
the windage levels on the turbine disc and the thrust bearing losses. This depends on the detailed design of the rig.
Because of the cost of these units, their relative inflexibility, and the experimental accuracy available, the use of warmflow turbine rigs is often reserved for exceptional occasions.
1.4.1 ENGINE TESTING
In the engine environment the turbine experiences the radial and circumferential patterns generated by the combustor,
and the upstream influence of the LP turbine. Both of these can lead to differences between cold and warm flow tests,
usually in minor ways. The difficulty in engine testing is to achieve a high experimental accuracy on all of the
parameters (see above), such that the resulting assessment of the turbine is of high integrity.
In order to measure correctly the energy state of the gas entering and exiting the turbine, full area traverses are required.
These are difficult to achieve in the strictures of the engine architecture. This often leads to the use of separate engine
core testing, without the LP spool, with greater access for measurement, and yet retaining the temperatures and
secondary air systems and clearances as experienced in the whole engine. The quality of the primary factors in the
calculation of turbine efficiency, mass flow, work in the compressor and average inlet and exit temperatures can all be
substantially better than those of the engine.
Unlike the engine, where it is extremely difficult to change the operating point of the turbine, the HP spool vehicle
enables the experimenter to explore a wide range of operation. Changes to the downstream control nozzle change the
operating point, giving a limited turbine characteristic. If available, then alterations to the compressor variable-geometry
allow for a range of turbine aerodynamic speed to be assessed. Extra turbine load can be added by absorbing power at
the power off-take shaft. This is normally done by use of hydraulic motors. Whilst these investigations are underway it
remains important for the secondary air and cooling air system to be monitored, along with changes in tip clearance in
order to gain a correct assessment of the turbine.
The table below summarizes the methods of turbine characteristic measurement and gives guidance to the validity of the
resulting performance map.
Method
Representative
conditions
Measurement
or sampling
difficulty
Measurement
accuracy
Poor
Low
Engine Parts
Cold Flow Rig
Good
HP Spool
Nett Assessment
HP
Turbines
LP
(uncooled)
Turbines
High
No. 5
No. 2
Low
High
No. 3
Best
Intermediate
Intermediate
No. 2
N/A
Direct on
Engine
High
Intermediate
No. 4
No.4
On Engine with
performance
simulation
High
Intermediate
Best
No.3
1.5.
TRANSIENT AND ABNORMAL OPERATIONS
During rapid throttle movement, the combustor may not complete the combustion process. Hence, the gas entering the
turbine may also contain a relatively high degree of unburned fuel. This causes insignificant blockage of the turbine
throat and does not affect engine matching. The major area of concern under transient operation is that of tip clearance,
and the effect on turbine efficiency. Time constants for the tip clearance changes are substantially greater than those of
the gas flow are. Consequently, turbine performance can be treated as pseudo steady state with the effects of the tip
clearance changes added by superposition.
5-39
Whilst the aerodynamicist will be able to justify the use of continuous curvature on the blade surfaces, and the use of
polished surfaces to gain that last point of efficiency, once developed, turbines are generally extremely resilient to
damage. It is not infrequent that operators will only detect damaged turbines either by boroscope visual inspection or by
indicated out of balance. Missing blades, burnt trailing edges of nozzles, impact damaged leading edges, erosion by
volcanic ash, all are found in service and have to be severe to reduce the engine performance. It is essential though to
maintain the mechanical integrity of the engine.
1.6.
High specific thrust engines. These turbines, often single-stage units, are usually cooled, and due to the effects of
the reheat system, variable final nozzle and the wide flight regime, operate over a wide range of aerodynamics.
Low specific thrust engines. These multi-stage units are usually uncooled, at least in the airfoil, and changes in
expansion ratio reflect the changes in aerodynamic speed. Generally, the aerodynamic excursions of the blades are
substantially less than the high specific thrust units.
A modern exception to the above is the Remote fan solution for one of the contenders in the Joint Strike Fighter
Vertical landing scenario. In this gas turbine, a clutched fan is driven by a low-pressure turbine that also drives the main
engine low-pressure compressor. For up and away flight the turbine operates as a lightly loaded unit and is subject to
normal high specific thrust engine changes of operation. In the vertical land mode the turbine operates as a highly
loaded, but low specific thrust unit. At similar aerodynamic speeds the turbine changes its expansion ratio by
approximately 2:1 as the load of the Remote fan is demanded from the turbine. Similar operations have been
considered for civil low noise supersonic engines. These ranges of operation demand that representative characteristics
are available early in the project design phase as critical thrust conditions have to be met in both methods of operation.
The exhaust diffuser is the last aerodynamic unit before the gas enters the jetpipe. It has an arduous task, taking swirling
flow from the exit of the turbine, turning it to the axial direction and diffusing from typically 0.6Mn to 0.2Mn. As the
turbine aerodynamic loading varies so does the swirl and mach number into the diffuser. Since the performance analyst
is primarily concerned with jetpipe gas conditions, it is often convenient to incorporate the losses of the exhaust diffuser
into the LP turbine characteristic. Otherwise it is necessary for the diffuser losses to be expressed as a function of the
turbine operating point i.e. exit swirl, Mach number,
1.7.
CHARACTERISTIC GENERATION
Methods for estimating or measuring the LP turbine performance are essentially the same as for the HP turbine. Due to
the variation in aerodynamic operation of the LP turbine, the cooling flow source to sink pressure ratios are not as
constant as for the HP turbine. Consequently, it is important to include the variations in the warm flow testing that may
be done. As an alternative, the influence, normally by calculation, of the varying cooling flows can be superpositioned
on the test results of a solid bladed cold flow rig result.
1.8.
RIG AND ENGINE TESTING
Similar requirements and concerns exist for LP turbine testing as those of the HP turbine. The cooled LP turbine may be
subjected to the same options for rig tests as those for the HP turbine. Uncooled LP turbines, usually associated with the
high power civil engines generate large powers, and need large capital investment in facilities if they are to be tested at
engine conditions. It had been common practice to use scaled models of these units to overcome these difficulties. When
model rigs are used it remains important to emulate the blading of the turbine as it will exist in the full scale unit. This
means that representative features, such as trailing edge thickness and tip clearances are replicated in the rig. Due to the
change of centrifugal and gas bending loads it is usually necessary to design such that the scaled turbine is
representative at design aerodynamic speed, and account the small variances that will be present at off design
aerodynamic speeds. With modern blading methods that are strongly Reynolds number, turbulence level and
unsteadiness level dependant, the use of model LP turbines is reducing. [69]
Engine testing of the LP turbine has a variety of technical areas that need consideration. The necessary quality
measurements of the gas parameters, both inlet and outlet are difficult to achieve, as per the HP turbine. The work done
by the turbine is attributed from the analysis of the low pressure compressor, and the working mass flow is deduced
from the analysis of the core engine. These factors accumulate such that the of the turbomachinery analysis the LP
turbine performance has the highest uncertainty. However given these reservations, with careful analysis over a series of
engine tests, then changes in efficiency and capacity of the turbine over a reasonably wide operation can be established
5-40
1.9.
THE EXCEPTIONS
REFERENCES
5-41
IV
COMBUSTOR SYSTEMS
One of the main issues for combustor design and performance is compliance with the high-pressure distributor &
turbine specifications for durability and thermodynamic efficiency. Life and durability specifications are directly linked
with the limiting thermal & mechanical point-stresses of the turbine nozzle and rotating parts and their associated
cooling requirements. The circumferential and radial temperature and velocity profile envelopes and maximum
temperature threshold compliance conditions at the exhaust of the combustor derive from these limits. These
requirements are generally established in accordance with extreme design points for the engine operation and various
representative fuelling mode effects, especially for staged or double combustors.
Additionally, profile distortions at the exit of the combustor can be generated during transient phases of the engine
operation. These distortions due to air supply and air-to-fuel ratio variations can directly impact the global
thermodynamic efficiency of the engine, through combustor and HP-turbine efficiency variations. Of course, these
specifications have also to take into account the machining and tooling variations from both the turbine and combustor
points of view. The cost-effectiveness and technological compromise between the two modules also affects the outcome.
For design points, 1D & 2D engineering methods and 3D-CFD reactive calculations for isolated combustion, or coupled
with the turbine nozzle module, are currently used to determine the most effective interface compliance conditions.
Transient aspects are generally considered through 1D & 2D transient global modeling of the engine fitted on
experimental data resulting from core engine tests and combustion rig tests on other engines. Inert & reactive 3D & 2D CFD calculations can also be considered for fuelling transient considerations.
1.1.1 MODELING
Classically, the design requirements have been obtained through numerous tests, leading to the practice where all of a
manufacturers experience is recorded and codified and some rules are given as advice. In the last ten years, CFD has
been used to reduce the number of tests needed, but one is always kept as a reference point for calibrating the various
constants involved in the different models. Today the main goal for the manufacturer is to design new concepts for
advanced combustion, which usually represent technological progress regarding in-house experience and ability.
Consequently, a better understanding of all the involved physical phenomena, combined with related improvements in
CFD tools is greatly needed. Although all phenomena are linked in turbulent two-phase reactive flow the prediction of a
good temperature field is of primary importance. Without a correct temperature field, a correct flow field cannot be
obtained. [62] (Because the density is false the solution of the conservation equation for momentum is false.)
Two phenomena are particularly crucial in the prediction of all the performances related to combustion. These are:
Two phase flow in turbulent combustion;
The interactions between turbulence and chemistry.
Both need to be represented simultaneously, with accuracy, by the model to get reliable CFD tools.
The most attractive approach for modeling turbulent combustion is the laminar flamelet concept. [63, 64] Within the
laminar flamelet concept, a turbulent flame is regarded as an ensemble of stationary laminar, locally one-dimensional
flame elements, stretched and distorted by the turbulent flow.
Two assumptions are commonly introduced: the reaction zone is thin in comparison to the typical length of the turbulent
flow, and the chemistry is fast in comparison to diffusion and convective transport. Then the reaction zone can be
modeled as a one-dimensional reaction diffusion layer. This method is based on the local mixture fraction, and includes
an extinction mechanism due to high local strain.
In applications of flamelets, stationary laminar flame solutions are stored in a flamelet library. During the CFD
calculation, the flamelet library is consulted to determine the local distribution of reactants.
However, laminar-flamelet concept drawbacks are:
This method works well only if the reaction zone is small compared to turbulent lengths;
The turbulence influence appears through a presumed PDF (Probability Density Function) that required an a priori
knowledge of the composition PDF.
The second approach concerns the calculated PDF method. [65, 66] The main characteristic is that this method
combines an exact treatment of chemical reactions with the influence of turbulence. It does so by solving a balance
equation for the one-point composition PDF wherein the chemical reaction terms are in closed form. All PDF codes use
notional particles that obey stochastic differential equations, solved by a Monte Carlo method.
The PDF contains random variables representing all chemical species at a particular spatial location. However, it
contains no information concerning local fluctuations in the scalar gradient (2-point information). Thus, a model is
5-42
5-43
efficiency. Lets say that on a test bed there is an engine running at a very high turbine-entry temperature. If the frozen
assumption is used to analyze the test data, the energy available for the turbines is less than the energy available if the
local equilibrium assumption is used. Therefore, the turbine efficiency will be higher with the frozen assumption than
with the local equilibrium assumption. Of course, the engine SFC is the same, what changes is the trade off between
chemical composition and turbine aerodynamics. This means that there is no clear cut-off (station 4) between the
combustion (i.e. chemical reactions) and the turbine designer.
As has been mentioned above, the energy of dissociation becomes very important at high temperatures, making it is
necessary to determine how much energy is recovered by the recombination reactions downstream of the combustion.
There are few publications dealing with aerodynamics and chemistry through the turbine expansion. The cited
publication [62] is based on one-dimensional flow in the first nozzle (without cooling flows). This study shows that,
except for the NO concentration, local equilibrium is attained at the exit of the nozzle.
Because NO has a much slower reaction rate, in this reference it is assumed that the concentration at equilibrium cannot
be achieved at the end of the combustion. However, as the fluid progresses through the nozzle more NO is produced. In
order to guarantee low emissions, turbine designers should consider NO formation.
From a performance viewpoint, the combustion efficiency is not enough, the chemical reactions downstream of the
combustion cannot be ignored. The problem is very complex, aerodynamics and chemistry are involved, and 3-D effects
may be important. The turbine blade cooling technique may have a strong influence on the recombination mechanism.
1.1.3 MODELING
NOx is defined by the regulatory authorities as the sum of nitric oxide (NO) and nitrogen dioxide (NO2). Many
mechanisms have been developed to represent NOx formation under a variety of thermo-chemistry conditions. [68] In
classical combustion, characterized by the presence of diffusion flames Thermal Nox is the principal route to NOx
formation. It is well described by the Zeldovich equations.
N2 + O <=> NO + N (R1)
N + O2 <=> NO + O (R2)
N + OH <=> H + NO (R3)
A combustion kinetics model for NOx, with a classical hypothesis of steadiness and partial equilibrium, is applied as a
post-processing step after a converged two-phase turbulent reactive computation has been obtained. [69, 70] The model
for turbulent combustion is based on a presumed PDF approach where chemistry is very fast compared to the turbulent
micro-mixing and where liquid fuel is represented by an ensemble of droplets in a Lagrangian framework. From the
expression of the chemical reaction rate we can see that two characteristics of the flow field have to be very well
predicted by the reactive computation. They are the temperature and equivalence ratio fields. Temperature acts directly
on the ks constants while equivalence ratio is used to estimate the equilibrium values for all the radicals. This method is
currently applied in the design process at Snecma and the other engine manufacturers use similar methods. All the users
have demonstrated that this approach works well only for high level pressures (>30b), and that more information about
chemistry must be included in the combustion model as a first step, and in the NOx mechanism as a second step. The
reasons for the method breaking down when finite-rate chemistry effects become important (when the pressure is
reduced) lie principally in the combustion model. If we try to apply a very fast chemistry turbulent combustion model to
situations where we get a false temperature field the consequence is a bad flow field associated with a bad equivalence
ratio field. Super equilibrium concentrations for the O radical have been frequently observed or computed with more
powerful tools. This is directly the consequence of slow three body recombination reactions. Only the integration of
more detailed chemistry to the turbulent combustion model can give us solution to this problem. This is currently the
case at Snecma where work has been oriented toward the use of the PDF techniques associated with complex chemistry.
Applied to a research tubular combustor the simplified method has given the following results, which are a good
illustration of the considerations above.
5-44
15.6
8.6
28,6%
15.5
9.8
37,5%
5,5
10
15.3
13.7
45,0%
20
20
32.9
11.2
0,0%
22
27
32.6
15.3
18,5%
27
35
33.4
18.4
22,9%
28
28
37.6
12.8
0,0%
31
31
37.6
13.7
0,0%
32
38
37.3
16.8
15,8%
Improvements can be obtained in the framework of PDF transported methods using the coupled kinetics of kerosene and
NOx, and taking into account more information about the NOx production mechanism. Although Nox modeling is
compatible with flamelet model capabilities, it seems CO prediction is limited by this assumption. The flamelet
approach for CO prediction can be improved by the inclusion of turbulent and chemistry models. However CO
formation is very complex, and flamelet hypothesis is not necessarily appropriate for this pollutant.
In combustors, CO oxidation can take place in lean post-flame zones, far from the flame front, where the flamelet
approach may be incorrect. CO emissions are greatly influenced by high mixing rates in combustors, so the statistics of
fluctuations need to be well predicted.
Therefore, the PDF transport method, which is able to model the non-equilibrium effects, works well for CO
predictions, and should improve results. Finally, two aspects are important for CO modeling: the kinetic mechanism
must be enough detailed, and the PDF method must be accurate with scalar fluctuations, and account for finite rate and
non-equilibrium effects.
1.1.4 HAIL & WATER INGESTION
Exposed to various climatic conditions, the jet engine ingests different forms of water contained in the atmosphere:
vapor, liquid, hail, snow; and ice crystals. The operating conditions where these forms of water can be met are below
20,000 ft. For higher altitudes, the atmospheric water content is insignificant. Depending upon flight speed (i.e. the ratio
between the free stream area and the inlet area) and engine rotor speed, the fluid flow at the compressor outlet might
include water vapor, liquid or solid water. The lower the corrected speed, the more important these phenomena are. The
impacted flight phases are then low altitude cruise, taxiing, and descent. Engine performance predictions and risk
assessments for flight safety in severe climatic conditions require an understanding of the different factors
characterizing the behavior of the combustion chamber. If the working fluid has a high water content, the specific fuel
consumption (SFC) rises and the operational range of the aircraft decreases. With liquid water at the combustion inlet,
several cases must be considered. With low liquid water content, vaporizing takes place in the combustion chamber, but
combustion efficiency is deteriorated. The fuel flow rate must then be increased. Beyond some threshold, extinction
might occur.
Two modeling approaches are required:
Modeling of reductions in flame temperature and efficiency as a function of water content
Calculation of extinction limits due to the presence of liquid and solid water in the combustor.
1.2.
TRANSIENT & DYNAMIC MODELING
1.2.1 FLAME BLOW-OUT & RELIGHT
The engine operational envelope at high altitude or during fast transient or descent operations can be severely restricted
by flame blow-out and combustor re-light capabilities. For example, flame blow-out effects can lead to very high
thermal stress on turbine elements, nozzles and other parts located at the rear of the combustor.
Due to flight safety considerations and civil regulations, combustor ignition capability and operability requirements are
generally defined through sub-domains and included in the flight envelope specifications. Complementary specifications
on fuel temperature and atmospheric conditions such as temperature and hygrometry, rain and hailstone precipitation are
also taken into account. Thermal limitations of the pumps and injection systems require minimal fuel flow rates for
cooling and these can also reduce the operability of the combustor.
From the control system view of the combustor, these constraints are translated into fuelling and actuation laws that are
dependent on engine operating conditions. All these conditions must be integrated in the global engine performance
5-45
model in order to specify module performance and control systems, and help to define in-flight emergency procedures.
Sensitivity to various control options, especially for staged or double combustion, have also to be considered:
commutation laws between pilot, main and mixed modes in combustor fueling have to be considered in order to prevent
such effects. Sensitivity to power off-take and air bleed, which can induce fuelair ratio variations, also must be taken
into account.
A re-light control system needs simulation models of extinction and light up capacities of the combustion. These
transient models have to provide extinction detection and light up information. They integrate time-delay modeling for
fuelling and spark operations, and combustion energy efficiency, which is correlated with the compression efficiency,
fuel flow and engine power setting of the engine. Alarm temperature on turbine parts and engine rotor speed information
generally is considered in this loop in order to detect a blow out risk or light up failure.
A particular aspect of the problem is the windmilling starting process, which requires:
A precise knowledge of the airflow conditions at the combustor entry (pressure and temperature).
After re-light, a good understanding of thermal and kinetic inertia of all other parts of the engine.
Of course, the control system and engine performance models have also to take into account the global variations in the
quality of components such as the compressor, fueling system, combustor, sensor systems and others elements of the
core engine.
Regarding the operational limits of the combustion, operation capacities and flame-out performances can be expressed
in terms of fuel-air ratio and air-loading envelopes These stability and re-light performance envelopes are determined
from 3D-CFD combustion calculations, experimental results and engine tests. As spark energy intensity and frequency
at the ignition plug system has a direct influence on the combustion re-light and light up performance, it is also taken
into account by 3D-CFD. The windmilling envelope re-light capacity is directly derived from this, incorporating the
compressor effects through a simulation model of the engine for these particular conditions.
Unsteady and instability effects of the combustion preceding the flame-out event are currently incorporated in these
models. Due to recent progress in computing, Large Eddy Simulation CFD calculations are progressively supplying an
improving representation of these phenomena and providing new methodologies for building such simulation models.
[71, 72, 73, 74]
On another point, blow out and relight performances are directly affected by the variation in quality of components and
fueling and air flow distortions at the combustor inlet. CFD simulations and specific rig tests of the hydraulic chain
including pump, manifold, valves and injectors currently provide models for the fuel system. However, a multi-module
coupled CFD analysis from compressor to the combustor exit is needed to provide a sensitivity analysis of blow out risk
and relight capacity to component variations and flow distortions.
1.2.2 MODELING RELIGHT
Today, empirical rules are used to determine the primary zone airflow rate required for an acceptable altitude relight
capability. The size of an aircraft combustor is, for the major part, fixed by this capacity. Ignition begins when power is
supplied to the igniter plug to make a spark inside the combustor. We can then distinguish four steps, which are
necessary in order to restart the engine.
The kernel formed by the spark must be able to burn. This is the result of the competition between energy losses
(turbulent conduction outside the kernel, conduction to the droplets inside the kernel, evaporation, heat release
chemical kinetics ) and the source of energy for the kernel (density of energy injected by the spark, energy coming
from initiated combustion).
Once the kernel is burning it must be captured by a stable eddy to maintain stable combustion. This process may be
tuned by positioning of the igniter combined with the general flow organization.
The flame, which is anchored to one sector of the combustor, must propagate to the whole combustor.
When the flame has been established throughout the entire combustor, the combustion efficiency must be sufficient
to ensure that the engine will accelerate. (Inertia and turbine/compressor power balance).
Failure of one single step will cause failure of the engine restart or re-light. For example doubling the energy to the
igniter has no influence on the final ignition if step two fails. On the other hand optimizing the location of the igniter is a
bad strategy if step one fails.
With only empirical rules that do not lead to any understanding of the underlying mechanism, significant improvements
in ignition capability cannot be obtained. Note that if we are able to reduce the size of the combustor while keeping the
ignition capacity constant, or improving it, we will have gains related to mass and weight because combustor chamber
may be made more compact. For civil aircraft engines a reduction of combustor size can lead to reduction in NOx
emissions.
Making predictive models for this phenomenon becomes very hard when we try to connect the spark process through
5-46
the related kernel growth to the turbulent flow inside the combustion. [75] Many length scales are involved with at least
four orders of magnitude of difference. It is then impossible in the same computation to represent all those scales except
with DNS (Direct Navier-Stokes) tools. Unfortunately due to CPU limitations DNS is not at present a viable strategy for
the calculation of turbulent reactive two-phase flows with complex geometry in industrial applications, where Reynolds
numbers are always high.
Once again, two-phase flow has a major role in the ignition process. If we consider step one in more detail we can see
that the distribution of droplet size inside the ignition kernel will affect its future growth. This distribution is a function
of the injector system and the location of the igniter inside the combustor. As already mentioned, the precise
representation of the evaporation process inside the kernel is a key to success in building a reliable predictive
computational tool. The method chosen at Snecma consists of: computing a two-phase flow with 3D code to represent
all the fields in the combustor related to engine windmilling before starting; applying a 1D code at every point to
determine the minimum ignition energy.
Although it is theoretically possible to make those computations by means of the PDF transport equation associated with
detailed chemistry, we must keep in mind that we have to compute ignition curves corresponding to many points in the
plane (fuel flow, air flow). Therefore, simplified methods are preferred.
Pressure = CTE
Stable
Unstable
Ignition stability
boundary
5-47
The fuel and stator schedules during acceleration and deceleration, with the computation of the operating and surge
lines.
The available surge margin.
Process viability, for example fuel cutoff and re-light transients, for surge recovery
Stability margins are evaluated on the engine to ensure that sufficient margin is available to meet the operational
requirements. Compression system surge lines may differ on the engine from the rig test because of differing
dimensional, thermodynamic and dynamic properties. To characterize the engine installed HP compressor surge line
position or shift the following method is commonly used. The fuel step technique allows HPC operating point to be
moved very quickly towards the HPC surge line. By using the continuity equation in the choked turbine nozzle the
compressor mass flow is then solved.
With the aforementioned applications in mind, the following investigations are of interest:
Modeling of outlet temperature, efficiency, stability domain, relaxation time, and heat soak characteristics as a
function of gradients in fuel, FAR, and inlet conditions.
Method to analyze compressor operating line excursions during fuel step transients
REHEAT SYSTEM
2.1.
STABILITY & BLOW-OUT
As for main combustors, fast transient from idle to full throttle operations can be severely restricted by flame blowout
and instability in reheat system. There are different risks linked with this. From the pilots viewpoint, and depending on
the flight point considered, this phenomenon can be felt as an important and instantaneous lost of operability. This may
lead to safety issues during heavily loaded take off or dog-fight phases of flight. Less critical issues, such as thrust loss
during supersonic operations, may also appear.
From the engine point of view and depending on the engine control system loop, the criticality of instability or a
blowout varies. A local extinction or a local blow out of the combustion between different gutters of the flame holder
system may have limited impact on the combustion efficiency and the thrust efficiency. However, due to consequential
exhaust nozzle operation, the engine rating, the internal pressure of the reheat pipe, and eventually the compressor stall
margin, the internal cooling distribution and the integrity of the liner, may all be affected. Due to the fuel release, it may
also induce uncontrolled re-light phenomena in the nozzle or in the plume with undesirable effects on the infrared
signature.
Reheat operability requirements expressed in terms of fuel-rate-range are generally defined through sub-domains and
included into the flight envelope specifications, together with aircraft incidence parameters. These specifications have to
integrate for instance the compressor stall margin transcribed in terms of maximum fuel rate and nozzle section
amplitude. Complementary specifications on fuel temperature, and atmosphere conditions such as temperature and
hygrometry also have to be taken into account.
All these conditions must be integrated in the global performance engine model in order to specify module
performances, and control systems operation, and help to define in-flight re-light procedures. Sensitivity to various
control options, especially for staged fueling and distribution laws of the fueling of the various manifolds and gutters of
the flame-holder systems have also to be considered. Transient models have to integrate time-delay modeling for
fuelling, spark and nozzle operations, reheat system energy efficiency variation and correlation with internal pressure,
nozzle section, and fuel flow and engine speed.
Of course, these models have also to take into account the global variations in quality of components such as the fueling
system, flame-holder geometry, by-pass ratio and nozzle section variations, regulation measurement systems and other
elements of the engine.
With regard to the isolated module view of the reheat system, stability and blow out performance can be expressed in
terms of fuel-air ratio and air-loading envelopes or limits. Specific sub-domains concerning fuel-staging modes of the
different gutters and flame-holder regions are generally associated with these limits.
These stability and blow-out envelopes can be determined both from 3D-CFD calculations and experimental results on
sub-components and engine tests.
Unsteady and instability effects of the combustion preceding the blow out or the unsteady process are currently
incorporated in these models. Due to recent progress in computing, Large Eddy Simulation CFD calculations are
progressively supplying a better representation of these phenomena and providing new methodologies for building such
compartmental models. These may be further developed to take into account the influence of the aircraft incidence and
the consequences on fan efficiency and combustion quality.
Blowout and stability performances are directly affected by the variations in the components of the fuelling system, bypass ratio variations, and the vitiation and airflow distortions in mixer regions. CFD simulations and specific rig tests of
the hydraulic chain including pump, manifold, valves and injectors currently provide compartmental models for the
5-48
fuelling system.
2.2.
SCREECH & RUMBLE
Screech and rumble phenomena in reheat systems may impose severe mechanical stress on the engine and may rapidly
create safety issues for the engine and the aircraft. Due to combustion and acoustically coupled effects, the combustion
amplified energy release may induce extensive physical to the exhaust duct structure. Because of the closeness of fuel
pipes and fuel tanks, these problems can directly affect aircraft safety.
These instabilities can suddenly appear during transient operations, like fuel staging or engine acceleration, and continue
until the pilot or the control system switches off the reheat mode. These phenomena are generally amplified in regimes
of high fuel-air-ratio. Hence, free-screech & free-rumble operating conditions have to be precisely defined in the engine
cycle models. These conditions can be transcribed from terms of fuel rate and air-fuel ratio interaction limits into fuel
operating and staging regulation and intake airflow conditions. These limits have to take into account fuel flow
restrictions or staging options directly connected to the fuel-system operating-regime, and include margins for distortion
effects and component variations.
Screech & rumble margin safety domains can be transcribed into fuel-air ratio or air-loading envelopes or limits.
Specific sub-domains concerning fuelling staging modes of the different gutters and flame holder regions are generally
associated with these limits.
Experiments on full-scale engines and sub-component tests are required to determine these limits and build
compartmental models for the whole flight envelope. Large Eddy Simulation CFD calculations are also useful to
confirm these limits.
Screech and rumble sensitivity are affected by variations in components which may influence the fuelling regions, the
by-pass ratio variations, and the vitiation and air flow distortions at the core and bypass mixer. CFD simulations and
specific rig tests of the hydraulic chain including pump, manifold, valves and injectors currently provide compartmental
models for the fuelling system.
Transient aspects are generally considered through 1D & 2D transient global modeling of the engine. This is typically
based on experimental data resulting from full engine and reheat component tests on other engines. Inert & reactive 3D
& 2D -CFD calculations can also be considered for fuelling transient considerations.
2.2.1 MODELING
The problem of combustion instabilities is classical, difficult and critical. It is classical and has been under study for at
least sixty years. The first large effort on the subject was in the area of rocket propulsion with solid fuel during the two
decades following WWII. It is difficult because the same term in fact applies to a wide variety of mechanisms, coupling
several physical phenomena such as combustion itself, convection, fuel atomization and vaporization, mixing,
hydrodynamic instabilities and acoustics. It should be noted that instabilities are generally named after the way they
sound to the ear, and not after the underlying mechanism, which is often complex and ill understood. Examples are
groaning, screech, organ noise, chugging, and growl. The problem is critical because combustion oscillations may lead
to very large disturbances that jeopardize combustor operability, especially for advanced, low-emission designs.
It is useful to consider a classification of instabilities in three groups:
Intrinsic combustion instabilities,
Chamber instabilities,
System instabilities [3].
Intrinsic instabilities deal with the combustion process itself. The LBO limit, and flashback or auto-ignition phenomena
may be treated within this class. By definition these instabilities do not involve complex coupling between various
physical phenomena, and the modeling challenge is not as difficult as with the two other types, which will be our main
concern from now on.
Chamber instabilities involve, in a strong way, the acoustic eigenmodes of the combustor. Given the scales typically
encountered in aeronautical combustors, these instabilities can also be identified as high frequency instabilities, which
are based on a coupling between acoustics and combustion. Indeed the frequency range is related to the acoustic
characteristic frequency c/L, where c is the speed of sound and L the size of the combustor. The speed of sound is the
highest velocity scalar available, especially at high temperature, and frequencies tend to be high.
System instabilities involve the whole combustion system, which not only includes the combustor itself, but also the fuel
line, injection system and exhaust. In the case of a full engine, the system can go as far as including elements of the
compressor and turbine. This type of instability is the most difficult to tackle since the first difficulty is to choose the
limit of the system under study. These instabilities can also be referred to as low frequency instabilities. The time scale
is defined either by the acoustics of the system, with longer length scales than the combustor alone, or by slower
physical processes, such as hydraulics, convection or vaporization.
5-49
( c 2p )
2 p
= H (1)
t 2
Generally speaking the source term, H, is complex, and some techniques aim at providing a model for H. Before
detailing this, it is useful to derive an acoustic energy transport equation, with a source term relating to [80]:
S p' q& ,
&
where p ' is the acoustic pressure and q the unsteady heat release rate. By integrating this over time, the classical
Rayleigh criterion is obtained. A combustion oscillation tends to amplify if the heat release is in phase with the pressure.
One approach consists in writing the heat release and pressure correlation in the (n,) form initially proposed by Crocco
[81]:
q& (t ) = n. p(t )
where n is a sensitivity factor and a time lag accounting for the time it takes for the flame to respond to a pressure
perturbation. This time lag typically includes convection or vaporization time, and could be measured. In principle, it is
then possible to build instability models. The only information that needs to be evaluated is the (n,) couple. However,
the theory has the drawback that it implies a simple causal relationship between pressure or velocity fluctuation and
flame response, while in a real instability the coupling is highly non-linear. Therefore, even with more complex
definitions, the (n,) nomenclature remains a formal description, which is more useful in forming a visual representation
of the phenomenon of high frequency instabilities rather than in actually predicting it.
As high frequency instabilities are strongly linked to the acoustic eigenmodes of the combustor, these modes are useful
information in themselves. Computing the modes will give no information as to which modes will actually be excited,
but will at least identify the higher risk frequencies. Moreover, the spatial structure of these modes can suggest locations
where passive-damping techniques may be more efficiently applied. The extraction of the acoustic modes of a cavity is a
classical problem with well-known solutions for an homogeneous medium in basic geometric configurations. For
application in combustors, more general solving techniques have been developed for arbitrary geometric configurations
with a non-homogeneous base temperature (hence speed of sound) field [82]. Let us recall that the computed
eigenmodes are the solution of equation (1) with no source term: it is supposed that the combustion oscillation simply
locks onto a cavity mode, without strongly affecting it, which is of course a strong assumption.
A more difficult problem is to define boundary conditions, while it is relatively simple to define boundaries themselves.
The only boundaries that are easily handled are rigid solid boundaries, or zero pressure conditions, which are both
energy conserving. The nozzle throat or turbine in the case of rocket engines or afterburners, and the compressor and
turbines in the case of the main combustor in a jet engine, lead to complex boundary conditions. Generally speaking
they involve frequency-dependent impedance conditions, which are not so easily modeled. Models are available for
various types of problem [83] but the boundary problem remains a critical point in these approaches.
Given that the basic acoustic modes were accurately captured, attempts have been made to account for a complex non
linear interaction with combustion, namely by actually keeping the source term H in equation (1) and decomposing the
perturbed modes on the free acoustic eigenbase. This is essentially Culicks approach [88]. Given the free cavity modes
n(x) associated to frequencies n:
(c 2n ) + n2 n = 0
The solutions perturbed (by combustion) are decomposed on the n base:
p ( x, t ) = an (t )n ( x)
n
5-50
Applying equation (1), the amplitudes an(t) are solution of the following equation:
d2
+ n2 ) an (t ) = H .n dV
2
dt
Of course, with zero source term, an is a simple sine wave with frequency n. Let us assume that an has the form
an (t ) = exp(it )
With no source term we recover of course =n. The source term represents the influence of combustion and will
therefore lead to a frequency shift (real part of the frequency), and a growth or decay (imaginary part of the frequency)
of the eigenmode. Starting from an unperturbed acoustic mode computation, this method is therefore able to predict the
alteration of the combustor acoustic modes by interaction with combustion, and provides the amplification factor of each
modified mode. These are attractive features of the approach. On the other hand, all the difficulty consists in evaluating
and modeling H, and more precisely response functions in the form H/an.
2.3.
LINER COOLING
2.3.1 MODELING
Because of the interactions resulting from the reheat chamber pressure and the air flow distribution between the liner
and the reheat system, it is necessary to consider a coupled approach to modeling the liner cooling and the mixing of the
core and fan flows.
These two functions directly influence the reheat combustion efficiency, the global pressure drop performances and the
cooling efficiency. They also may influence the core rating and performances.
1D models can be sufficient to give an average representation of the exhaust and liner flow rates. However, because
recent non axi-symmetric flame-holder geometries and cooling configurations may also induce local pressure drop and
flows distortions in the exhaust pipe, 2D & 3D CFD analyses may be necessary to build a derived compartmental model.
REFERENCES
70. K. N. C. Bray The Challenge of Turbulent Combustion Twenty-Six Symposium (International) on Combustion,
The Combustion Institute, Pittsburg, 1996 pp 1-26.
71. H. Pitsch, M. Chen and N. Peters Unsteady Flamelet Modeling of Turbulent Hydrogen-Air Diffusion Flames.
Twenty-Seven Symposium (International) on Combustion, The Combustion Institute, Pittsburg, 1998 pp 10571064.
72. H. Barths, N. Peters, N. Brehm, A. Mack, M. Pfitzner and V. Smiljanovski Simulation of Pollutant Formation in a
Gas-Turbine Combustor Using Unsteady Flamelets. Twenty-Seven Symposium (International) on Combustion,
The Combustion Institute, Pittsburg, 1998 pp 1841-1847.
73. C. Dopazo, Turbulent Reacting Flows (Libby, P. A. and Williams, F. A., Eds), Academic Press, London, 1994,
pp. 375-474
74. J.C Larroya, C. Franois, M. Cazalens, L. Vervisch. Testing a new Monte Carlo method for solving pdf equation in
turbulent combustion. In Fifth International Conference on Technologies and Combustion for a clean
environment, vol I, pp 177-183, 1999.
75. Burkhard Leide and Pascal Stouffs, Residual Reactivity of Burnt Gases in the Early Expansion Process of future
Gas turbines. Gas turbine and Power
76. Visser, W.P.J., and Kluiters, S.C.M., 1999, Modelling the Effects of Operating Conditions and Alternative Fuels
on Gas Turbine Performance and Emissions Research and Technology Organisation, RTO-MP-14
77. F. Ravet, L. Vervisch, Modelling non-premixed turbulent combustion in aeronautical engines using PDF
generator AIAA 98-1027, 36st AIAA Aerospace sciences, January 12-15, 1998, Reno
78. M. Cazalens, F. Beule, E. David Design of Advanced Low Emission Combustor CEAS European Propulsion
Forum programme 2001
79. T. Schnfeld, M.A. Rudgyard, Steady and unsteady flow simulations using the hybrid flow solver AVBP. AIAA
Journal, 37(11), November 1999, pp 1378-1385.
80. T. Schnfeld, M.A. Rudgyard, COUPL and its use within hybrid mesh CFD applications. Proc. of the 10th Intl
Conference on Parallel CFD 98, pp 433-440. Eds A. Ecer, D. Emerson, J. Periaux and N. Satofuka. Elsevier
Science Publishers, 1998.AIAA 98-1027, 36st AIAA Aerospace sciences, January 12-15, 1998, Reno
81. C. Angelberger, D. Veynante, F. Egolfopoulos, T. Poinsot. Large eddy simulations of combustion instabilities in
premixed flames. Proc. Of the Summer Program 1998, Center for Turbulence Research, Stanford.
82. O. Colin, F. Ducros, D. Veynante, T. Poinsot. A thickened flame model for large eddy simulations of turbulent
premixed combustion. Physics of Fluids 2000. In press.
83. M. Cazalens, R. Lecourt, V. Quintilla Predicting Ignition Performance for Altitude Relight
84. XV ISABE 2001, 02-07 September 2001 Bangalore Inde.
85. A. H. Lefebvre Gas Turbine Combustion. Second Edition 1999 Taylor & Francis.
5-51
86. J. Odgers and C. Carrier Modelling of Gas Turbine Combustors; Considerations of Combustion Efficiency and
Stability. Journal of Engineering for Power pp 105-113 April 1973.
87. W. Tipler and A.W. Wilson Combustion in Gas Turbines 1959 Paper B9 in Proceedings of the Congrs
International des Machines Combustion (CIMAC) Paris pp 897-927 .
88. F.E. Culick & V. Yang 1995 Overview of combustion instabilities in liquid-propellant rocket engines. Progress in
Astronautics and Aeronautics, 169.
89. F.A. Williams 1985 Combustion Theory. Menlo Park: Benjamin / Cummings
90. L. Crocco & S.-I. Cheng 1956 Theory of combustion instability liquid propellant rocket motor.
AGARDOGRAPH no 8.
91. A. Laverdant, T. Poinsot & S. Candel 1986 Influence of the mean temperature field on the acoustic mode structure
in a dump combustor. J. Propulsion and Power, 2, 311-316.
92. F. Vuillot Acoustic mode determination in solid rocket motor stability analysis. AIAA Journal, July-August
1987.
The nozzle exhaust system is a major component of an aircraft gas turbine powerplant. It is the device by which the
energy from the gas generator is converted into useful thrust. It has efficiency just like other components. Turbofan and
turbojet engines are especially sensitive to nozzle losses in because they operate on all of the flow from both an airflow
and thrust standpoint. Careful design to keep losses low is necessary to achieve goal performance.
The exhaust nozzle component system serves the following functions in the aircraft gas turbine powerplant:
The exhaust nozzle component contains the exit throttling area (throat) of the powerplant. It meters flow and sets the
turbine back-pressure, thereby controlling turbine work. It is one of the primary metering areas in the engine,
regulating the cycle. For the case of a dual stream bypass engine, the exit nozzle is the principle controller of the
back-pressure to the fan thereby setting fan pressure ratio.
The nozzle functions as an aerodynamic component that accelerates the gas exiting the turbine to produce thrust
from available energy, Figure 45.
A variable area nozzle plays an important role in providing optimum performance and assuring engine operability.
Variable throat area is essential for control of engine matching and fan back-pressure during augmentation.
A variable geometry nozzle can be designed to generate pitch and yaw forces to control the air vehicle.
The exhaust nozzle can be designed to produce reverse thrust.
Turbine Inlet
4
5
Compressor
Exit
H ~ Enthalpy
9
9
Hideal
Nozzle Exit
Hactual
Available for
Propulsion
S ~ Entropy
5-52
ideal thrust can be defined on a basis of an ideal convergent/divergent nozzle, an ideal convergent nozzle, or variations
in between. It is recommended that the ideal thrust be defined based on fully expanded convergent/divergent nozzle
theory. Expressions to do this are given in Table 1 for unchoked and choked pressure ratios. Some programs prefer to
use the ideal convergent definition and expressions for these are given in Table 1. If the ideal thrust is defined as that
developed by a nozzle with a given area ratio, the formulae become too complex to be presented in Table 1.
The true performance of the nozzle system remains of course unchanged by the choice of the ideal definition; often the
ideal definition is simply a choice of a reference datum for bookkeeping purposes. However, it is also a measure of
maximum potential of the nozzle component to produce thrust. Insight into potential nozzle system improvements may
be gained by analysis and understanding of the various real gas and mechanical design effects causing the actual thrust
and nozzle coefficient to be less than ideal.
1.1.2 SINGLE STREAM EXHAUST SYSTEM
Gas turbine rotating machinery produces hot gas at the turbine exit, station 5, shown schematically for a single-stream
turbojet engine in Figure 45.
W g 7 = W2 Wbl + W f
Eq (1)
The total energy available for thrust is equal to Wg7 * ht7 where ht7 is the total specific enthalpy of the turbine exit gas.
The gas expands through the nozzle to ambient pressure at Station 9 for a fully expanded ideal convergent/divergent
nozzle. This acceleration process transforms available heat energy in the gas to kinetic energy in the expanding jet,
which produces useful thrust to propel the aircraft.
The flow velocity at the nozzle exit will be:
Eq. (2)
Fg 9,ideal = W g 7 * V9,ideal
Eq. (3)
The ideal gross thrust can also be evaluated as a function of nozzle pressure ratio:
Fg 9,ideal = Wg 7 * M 9 * a9 = Wg 7 * Tt 7 * f1 ( , R, Pt 7 / Pamb )
Where M9 and a9 are the corresponding Mach number and speed of sound for the exhaust gas expanded to ambient
conditions at the nozzle pressure ratio, Pt7/Pamb.
Eq. (4)
[Fg9/(Wg7*Tt7 )]ideal developed in Eq. (4) is called the ideal specific thrust parameter and is an analytic function of
isentropic exponent, gamma, gas constant and nozzle pressure ratio.
An alternate equivalent expression for gross thrust, based on full expansion, is
5-53
Eq. (5)
[Fg/(A8*Pamb]ideal in Eq. (5) is called the ideal thrust function and is a second analytic function of gamma, gas constant
and nozzle expansion ratio.
Expressions for these functions are given in Table 1 for ideal nozzles operating at choked and unchoked pressure ratios.
These definitions are used for both converging and converging/diverging (C/D) nozzles, both choked and unchoked.
The definitions assume that the controlling area of the nozzle is at station 8. However, for an unchoked C/D nozzle, the
controlling area shifts to station 9. Because of this shift, it is not unusual to calculate nozzle throat flow (discharge)
coefficients greater than unity for unchoked C/D nozzles.
The expansion process in real nozzles is not isentropic. Losses occur relative to the ideal, because of real gas effects and
other reasons listed below.
Leakage
Cooling
Thermal expansion
Reverser links
Steps and gaps
Surface roughness and friction
Acoustic treatment
Gas mixing
Temperature profile
Pressure profile
Non-axial exit flow vector
Swirl
Over or under expansion
Shock losses
Separation
Because of these losses, the actual thrust is less than the ideal.
The nozzle gross thrust coefficient that quantifies this loss is defined as:
Cg is called the gross thrust coefficient and can be expressed as
C g = Fg 9 actual / Fg 9 ideal
Eq. (6)
Eq. (7)
Eq. (8)
The actual gas flow will also be less in a real nozzle than in an ideal nozzle. The flow coefficient, Cd, quantifies this
loss, equation 9.
= W g 7 * Tt 7 /( A8 * Pt 7 )
actual
/ W g 7 * Tt 7 /( A8 * Pt 7 ) ideal
Eq. (9)
Eq. (10)
An alternate equivalent method to calculate gross thrust uses the specific thrust coefficient, Cv.
Fg 9 = [ Fg 9 /(W g 7 * Tt 7 )] ideal * C v * W g 7 * Tt 7
Actual gross thrust expressions defined by equations 8 & 11 are equivalent.
Eq. (11)
5-54
Cv =
(
/ (W
Fg 9, actual / Wg 7 * Tt 7
Fg 9,ideal
g7
* Tt 7
)
)
actual
= Cg *
ideal
(W
(W
g7
* Tt 7
g7
* Tt 7
)
)
ideal
actual
Cg
Cd
C g = Cv * Cd
Eq. (12)
From the definition of Cg, equation 8, the following correlation may be derived:
= C d * [V 9 , actual / V9 , ideal ]
Comparing equation 12 with 13, shows that:
C v = [V 9 actual / V 9 ideal ]
Eq. (13a)
Therefore the specific thrust coefficient, Cv, is also the ratio of actual to ideal jet velocity and is commonly referred to as
the velocity coefficient, equation 13. This relationship is very useful to adjust to actual thrust in cycle-match computer
simulations where the ideal jet velocity is calculated from energy deltas, equation 2.
1.1.2.1 THRUST DEFINITIONS
A control volume for a single stream turbojet is shown in Figure 47.
max
FIPF = A0 Pamb W0 * V0 + Ps * dA +
Ps * dA +
P * dA + A * P
9
s9
+ W9 *V 9+
max
P * dA
s
(Eq. 14)
plug
max
( Ps Pamb ) * dA +
+ A9 * ( Ps 9 Pamb ) + W9 *V 9+ ( Ps Pamb ) * dA
plug
W0 *V0
1
Additive Drag
(P P
s
amb
) * dA
(P
max
Pamb ) * dA
(Eq. 14a)
5-55
max
(P P
Lip Suction
amb
) * dA
(P P
amb
) * dA
max
A9 * ( Ps 9 Pamb ) + W9 * V9 +
(P P
s
amb
) * dA
plug
Spillage drag FINL is the sum of additive drag and lip suction.
FIPF = Wa 0 *V 0 +
( Ps Pamb )dA -
max
( Ps Pamb )dA +
max
( Ps Pamb )dA
14231442443
ADDITIVE DRAG
LIP SUCTION
EXHAUST
RECOMPRESSION FORCE
= FG9
Equations 14a and b do not account for friction forces. Including friction uses the definition of which describes the
axial force component of the combined pressure and friction force on a surface in axisymmetric flow:
[( P P
s
amb
) + * cot ]* dA
surface
where, = angle between free stream flow direction and plane tangent to the surface element ds
= shear stress acting on surface element ds
dA = sin ds = area of a surface element ds projected on a plane normal to the free stream flow direction
Rewriting Eq. 14b including friction forces leads to Eq. 14c:
1
Eq. (14c)
The installed propulsive force FIPF acting on the pylon consists of engine net thrust reduced by nacelle drag and is equal
to the airframe system drag, Dafs, equation 15. SAE AIR 1703, Ref. 74
Eq. (15)
Fnet = Fg 9 W0 *V0
Eq. (16)
(P P
s
plug
amb
) * dA
5-56
[ (
= Fg 9 / Wg 9 * Tt
)]
ideal
* Cv * Wg 7 * Tt 7
= Wg 7 , actual * V9,ideal * Cv
= Wg 7 * Cv * 2 * (ht 7 hamb )
A representative plot of the nozzle coefficients is shown in Figure 48.
Figure 48 - Representative thrust and flow coefficients vs. nozzle pressure ratio
1.1.3 DUAL STREAM ENGINE CONFIGURATION
The foregoing relationships can be extrapolated to dual-stream by-pass engines. See Figure 48. Station numbers in the
bypass stream are similar to those in the primary stream but increase by ten; e.g. station 9 primary becomes station 19 in
the by-pass duct. Force accounting and thrust definitions are shown in equation 17a, the no friction case.
( Ps P
0
amb
)dA -
max
( Ps Pamb )dA +
max
( Ps Pamb )dA
1442443 14424443
ADDITIVE DRAG
LIP SUCTION
14423 14444442444443
RAM DRAG
= FR
1442443
INLET
SPILLAGE DRAG = FINL
EXHAUST
RECOMPRESSION FORCE = FIXH
5-57
Eq(17a)
1444444444442444444443
BYPASS STREAM GROSS THRUST = FG19
1
max
(P
Pamb ) * dA +
(P
Pamb ) * dA
max
+ A9 * ( Ps 9 Pamb ) + W9 *V 9+ ( Ps Pamb ) * dA
Eq. (17a)
plug
(P
Pamb ) * dA
core cowl
W0 *V0
Ram Drag
(P P
Additive Drag
amb
) * dA
max
(P P
Lip Suction
amb
) * dA
(P P
s
amb
) * dA
max
A9 * ( Ps 9 Pamb ) + W9 * V9 +
(P P
s
amb
) * dA
plug
(P P
s
amb
) * dA
core cowl
Spillage drag FINL is the sum of additive drag and lip suction.
If friction forces are included, then Eq. 17a becomes
1
+ A9 * ( Ps 9 Pamb ) + W9 *V 9+ Plug
Eq. (17b)
5-58
5-59
5-60
Figure 55 - Sea level engine test facility to measure thrust, airflow and nozzle coefficients
Figure 56 - Altitude engine test facility to measure thrust, airflow, and nozzle coefficients
1.1.6 VARIABLE NOZZLE BENEFITS AND EXAMPLES
The calculation for thrust and performance for all foregoing definitions are applicable for variable geometry nozzles.
The nozzle coefficients are a function of both nozzle expansion ratio and jet area. The coefficients must be measured
during full-scale engine or sub-scale model testing. The F-100 military-turbofan engine, shown in Figure 54, provides
an example of a variable geometry nozzle.
A8
CV
A8
CD
5-61
Operability - The variable geometry nozzle is used to regulate the backpressure to the fan thereby maintaining stall
free operation throughout the operating envelope. This is especially important for an engine with augmentation. The
jet area must be increased when the afterburner is operating to maintain stall free fan operation. This accomplished
with modern control systems by closed loop logic, where pressure ratio measured with high response transducers, is
maintained by control of a variable nozzle area.
Performance optimization - Since the nozzle throat meters the exit flow, it regulates the backpressure of the gas
generator. Variation of the backpressure affects the cycle match and therefore powerplant thermal efficiency. The
throat area also affects jet velocity, which impacts propulsive efficiency. A variable throat area may be optimized to
provide best performance at a given flight condition. Generally, non-augmented engines require minimal or no throat
area variation to optimize engine thrust. Augmented engines, however, require throat areas as much as 100% larger
than the corresponding non-augmented area.
Improved performance may also be obtained from a variable expansion ratio (exit area) convergent/divergent nozzle
when a large operating range of nozzle pressure ratio (NPR) is required (most transonic/supersonic vehicles), see
Figure 57. The maximum thrust that the nozzle can produce for a given expansion ratio is obtained when the nozzle
flow is fully expanded to ambient pressure at station 9, See Figure 58. The potential improvement for a convergentdivergent nozzle relative to a convergent nozzle is a function of nozzle pressure ratio. It may be calculated as the
ratio of the ideal thrust functions for the nozzle types from Table 1, and this ratio is shown in Figure 59. The actual
improvement for a convergent /divergent section will be reduced because of additional pressure loss. Inspection of
Figure 57 shows that the convergent nozzle is better than the C/D nozzle at low nozzle-pressure ratios. The
improvement for a C/D nozzle occurs when the C/D nozzle over-expansion losses are less than the convergent
nozzle under-expansion losses
Commercial and transport engines are normally high bypass - low nozzle pressure ratio configurations to maximize
propulsive efficiency and reduce fuel consumption. They do not have augmentors or operate over a large range of
nozzle pressure ratio so variable nozzle areas are not required. The optimum nozzle is usually a convergent or fixed
convergent/divergent with a small area ratio.
High thrust-to-weight ratio military engine configurations operate over a much wider range of nozzle pressure ratios
and flight Mach numbers, and benefit greatly from variable-area-ratio nozzles.
Air Vehicle Maneuverability - The nozzle may be designed to turn the exhaust thrust vector to provide forces in the
pitch and yaw directions as well as the normal thrust direction. This feature may be used by the pilot to direct additional
pitch and yaw control to the aircraft. Special configurations may also provide lift control. An example is provided by the
F-119-PW-100 shown in Figure 60.
P7
P9
P8
PAMBIENT
PTOTAL = PT7
Pressure
P6
PAMB > P9
PAMB = P9
to Overexpansion
Loss due to
PdAUnderexpansion
P9
1
2
PAMB<P9
5-62
1.3.1 FLOW
The effective area of the nozzle is the product of the geometric area and the discharge flow coefficient, A8Cd, as defined
by Eq. (9) referenced to station 8, (nozzle exit and throat). The magnitude of Cd8 depends on the nozzle pressure ratio,
P7/Pamb, and the nozzle design features, e.g. relative petal length (L/D), and petal angle, . Figure 61. For the case where
the pressure loss between Stations 7 and 8 is small then P7/Pamb is approximately equal to P8/Pamb. The development
below is based on this assumption.
Note that in the text below, the total pressures and temperatures are designated without the subscript, t.
The flow at the nozzle throat W8 may be calculated from P8/Pamb, T8 and A8 from the formulae for isentropic flow.
1.4.
ENGINE TEST ANALYSIS METHOD
Design to calculate nozzle exit area, A8, from other measured parameters and sub-scale model correlations:
5-63
1.
2.
3.
4.
5.
Calculate T8 from measured W2, Wf, overboard leakage, and main burner efficiency utilizing conservation of
energy considerations. This calculation utilizes an iteration which begins with an initial value for petal angle
().See Figure 17
P7 and P8 are determined from sub-scale model correlations of P7/Ps7 = f() and Ps7/Ps7wall = f()
A8*Cd8 is then calculated from W8, T8, and P8.
Actual A8 is calculated from Step 3 and sub-scale model correlation of Cd = f(P7/Pamb, )
Solve (iterate) till A8 and petal angle,(), are consistent
Hinge Leakage
Slave Petal
Master Petal
Ps7
Petal
Leakage
5-64
causing an additional nozzle-area measurement error. A large resulting error in nozzle-area will propagate to
unacceptably large uncertainties in calculated and measured airflow and thrust.
1.5.3 NOZZLE POSITION SENSOR CALIBRATION
The nozzle position sensor requires calibration, which can be done utilizing the full-scale static pressure at the nozzle
charging station (inlet) Ps7wall, and the Ps7wall/Pam and Cd from sub-scale model tests (or alternatively CFD calculations).
1.5.4 THRUST
The gross thrust of a convergent nozzle may be defined as
Fg8 actual =A8 (Ps8-P0)+W8*V8,ideal*Cv8
Eq. (18)
This equation is consistent with the thrust-accounting control volume for a single stream convergent exhaust system
shown in Figure 46 and equation 14 adjusted to a convergent throat at station 8. The nozzle coefficient is defined here as
the ratio of the actual to the ideal jet velocity of a convergent nozzle, and is consistent with the correlations in th erhight
hand column of table 1.
For the example of Figure 61, the velocity coefficient, Cv8, is a function of nozzle design, the petal angle, the nozzle
pressure ratio, and many other loss causing effects of the type listed in section 7.2.1.
1.5.5 MIXING
Incomplete mixing of the gas streams in a turbofan engine will result in a thrust loss, which is properly assessed against
the mixer, not the nozzle. This loss has been accounted as a mixer efficiency, equation.19, see detailed discussion in
section 6.
mix = (Fg8-Fg8,unmixed)/(Fg8 fully mixed Fg8 unmixed)
Eq. (19)
The application of mix requires the calculation of the fully mixed and unmixed thrust, i.e. separate expansion of the
cold and hot streams. This makes the nozzle calculation complex because one must now define two nozzle entry
stations: one for the hot and one for the cold expansion. In addition, the nozzle area must be apportioned for the hot and
cold streams. Inconsistencies in throat Mach number for the separate streams also arise due to gas property effects.
When different turbofan cycles or different mixers are to be compared, the concept of partial mixing is very useful, both
to describe and to help understand effects on thrust. On an existing engine, although it is much simpler to book-keep the
effects of non-uniform temperature and pressure to the nozzle thrust coefficient, it must be recognized that a significant
contributor to the level of the thrust coefficient derives from an effect generated outside the nozzle.
1.6.
CLEAR DEFINITION OF IDEAL THRUST REQUIRED
It is important in any particular program to avoid ambiguity by clear understanding and agreement at the outset among
all stakeholders regarding the definition of nozzle ideal thrust.
A common definition for the ideal expansion process is to ambient pressure in an ideal convergent/divergent nozzle.
Other definitions are possible depending on how the averaging calculations at the nozzle charging station, the flow
mixing calculations, bleed and leakage calculations, and multi-stream evaluation are handled. The definitions may
become complicated especially as the complexity (dimensionality) of nozzle model increases.
The actual thrust for an existing engine will not change regardless of how the ideal nozzle coefficient is defined.
Understanding of the components including the nozzle can be improved by more complex physics based nozzle
component representations.
The addition of a divergent section to the exhaust nozzle increases the usefulness and in some cases need for one plus
dimensional analysis and simulation of the system. For the example discussed here, Figure 62, a precise description of
the geometry for all positions is necessary. The primary petal angle is an important parameter; the metal temperature of
the primary petal is also important. Thermal expansion of the primary master petals causes an increase in A9. This
happens because the length of the strut which holds the divergent petals remains unchanged because its temperature
does not change, Figure 62b. The multidimensional geometry model is necessary to properly evaluate the values for the
throat area, A8 and the exit area, A9, and thus nozzle area ratio A9/A8 for all operating conditions.
5-65
Nozzle Geometry
Primary
Petal
Angle
c) Rear View
= Wg *
T t /( A * Pt )
actual
/Wg *
T t /( A * Pt )
Eq. (19)
ideal
In order for the definitions of equation 9 to be valid, flows, temperatures and pressures must be consistently calculated
at Station 8.
When the ideal flow, Wg8,ideal, is calculated from W8-Wleak, the total temperature T8 and pressure ratio P8/Pamb, a flow
coefficient calculated from Cd= Wact/Wideal will provide coefficients greater than 1.0 at low pressure ratios. Proper
physics in the computation of the coefficient will always result in a value less than 1.0. More realistic computation, of
the flow-conditions at the throat, is no additional burden for this simulation because it is needed for the calculation of
forces and nozzle leakage.
At sufficiently high nozzle pressure ratios (when the flow is under-expanded or only slightly over-expanded) the
discharge flow coefficient can be closely approximated as a function of petal angle () only, equation 20.
Cd = -1.25*10-5 * 2 0.001425 +0.995
Eq. (20)
When the pressure ratio is low, the flow in the divergent section of the nozzle detaches and Cd becomes a function of
both and P8/Pamb.
The nozzle exit discharge flow coefficient CD9 is very nearly constant and has a value of ~ 0.995, for under expanded
operation and becomes a function of pressure ratio and area for low P8/Pamb.
2.2.
LEAKAGE CONSIDERATIONS
The leakage through the hinges and between the master and slave petals depends on the difference between the
pressures within and around the nozzle. A description of how to evaluate the pressure distribution inside the convergent
part of the nozzle can be found in the foregoing description.
As long as the pressure outside the nozzle is lower than inside, the effect of leakage can be directly evaluated. However,
there are many operating conditions, sea level static for example, for which static pressure in the throat and divergent
sections is lower than pressure outside, and air is consequently sucked in.
5-66
During engine testing of the EJ200 (a mixed flow turbofan engine), for example, one could easily identify when the
pressure inside the divergent nozzle section is lower than outside. Inspection of Figure 63 shows that the slave petals are
sucked in and a gap opens between the master and the slave petals.
C g = Fg 9 actual / Fg 9 ideal
Eq. (21)
The thrust coefficient Cg can also be calculated consistent with the flow coefficient using equation 6 repeated.
Eq. (22)
The gross thrust coefficient compares the actual thrust (measured) with an ideal definition of thrust produced by an ideal
expansion to ambient pressure. Cg will also be a function of nozzle throat to exit area ratio, A8/A9.
2.3.2 ALL STAKE-HOLDERS SHOULD BE AWARE OF THRUST & COEFFICIENT DEFINITIONS
The definition of ideal Cg based fully expanded variable geometry results in a lower numerical value for the calculated
real nozzle system. This is because it is not possible for the nozzle to fully expand to ambient pressures at off-design
conditions. Several experimenters have opted to define the ideal coefficient as the maximum achievable for a given
hardware geometry. This has the advantage, that the thrust respectively velocity coefficient becomes a true measure of
nozzle flow quality. The coefficient defined in such a way is independent from nozzle area ratio A9/A8, pressure ratio
P8/Pamb and nozzle leakage provided a sufficiently detailed mathematical model of the nozzle is employed.
Although other definitions for ideal thrust are possible, they are generally non-standard and should only be used for
sound reasons, related to a particular geometry. All parties should fully understand and use the definitions including
nozzle designers, sub-scale modelers, analytical simulation engineers, test analysis engineers, technical manager/system
evaluators, customers and sometimes suppliers.
An alternate formulation of the thrust equation reported in the literature [97] is shown in equation 21.
Fg9 actual = A9 (Ps9-Pamb)+W9*V9 ideal * C* v9 * Ca * Cf
Eq. (21)
5-67
In this formulation Ca is the angularity coefficient from the Reference [97] Fig. 5-12 that accounts for the losses due to
non-axial exit flow velocity. The coefficient Cf from the Reference [97] Fig. 5.13 accounts for the effect of boundary
layer momentum loss caused by friction in the nozzle. Both these coefficients depend on the nozzle area and the primary
petal angle. The velocity coefficient C* v9 is held to represent a nozzle efficiency. The advantage to define the
coefficient is this way is discussed in the next section.
Note that in the turbofan engine this coefficient accounts mainly for losses due to incomplete mixing of the core flow,
bypass flow and nozzle cooling air and is not the same as the velocity coefficient Cv as defined earlier.
2.4.
Figure 64 - Test result for nozzle velocity coefficient using 1-D geometry model
For high-pressure ratio nozzles with significant cooling flow, the assumptions in treating the cooling flow can change
the nozzle thrust coefficient. Typically cooling air entering upstream of the nozzle throat is treated as part of the primary
flow, even if the injection location does not allow for effective mixing and the momentum of the flow is not considered.
Cooling flow on the expansion flaps is generally not included in the thrust calculation. As cooling becomes more
significant, the modeler may choose to treat these cooling flows as separate streams with their own thrust contribution
and thrust coefficient, rather than try to develop a thrust coefficient representation to cover the nozzle behavior over a
wide range of cooling flow and nozzle pressure ratio levels.
2.4.2 LOW NOZZLE-PRESSURE RATIOS
When the nozzle pressure ratio is less than 3.5 to 4.0 the C*v9 increases rapidly as nozzle pressure-ratio decreases, Figure
20. This occurs because the calculated force on the struts holding the secondary petals becomes negative. In turn, this is
because the pressure inside the divergent section is lower than ambient causing the forces on the divergent master petals
to exert a pull on the struts, see Figure 65. Under these conditions, the slave petals detach from the master petals as
can be clearly seen in Figure 63. Airflows from outside to inside the nozzle, which reduces the effective nozzle exit area
ratio A9/A8. This results in a better match of the effective nozzle area ratio at the low-pressure ratio condition and the
nozzle appears to perform better than calculated with the standard assumption, Cd = 0.995.
5-68
5-69
1,0
1,05
0,98
1,15
1,20
0,9
1,25
1,10
0,95
0,8
Flow detachment @
low NPR improves
thrust !
1,30
1,35
1,40
2,4
1,0
1,5
2,0
2,5
3,0
3,5
4,0
4,5
5,0
5,5
Figure 66 - Thrust gain due to flow detachment at low nozzle pressure ratios
0.71
0.82
0.71
0.60
1.03
0.82
2.23
2.33
2.44
2.55
2.66
2.12
2.01
1.90
1.79
1.68
1.58
1.47
1.36
1.25
1.14
1.03
0.92
0.82
Ps7,wall
Figure 67 - Schematic of divergent nozzle wall section with pressure instrumentation locations
Figure 68 - Correlation of static pressure near the nozzle exit with nozzle pressure ratio
If one hypothesizes the requirement that the Nozzle throat area, A8, always controls the flow, i.e. the effective nozzle
area ratio, Ae9/Ae8, is always greater than 1, an empirical correlation to calculate nozzle exit discharge coefficient, Cd9,
may be defined. In this example the nozzle operates much like a convergent nozzle and expands the flow to ambient
pressure for nozzle pressure ratios below the sonic limit of ~1.8, (Ps9wall/Pamb = ~ 1.0. At nozzle pressure ratios above 1.8
the level of measured Ps9wall/Pamb trends lower. The curve labeled Ps9L/Pamb in Figure 68 is a best-fit curve through the
measured data; it represents the lower limit of the Ps9/Pamb for the given nozzle. Lines of constant divergent nozzle area
ratio calculated from one-dimensional considerations are also shown in the Figure.
5-70
The limiting curve of Ps9L/Pamb = f (P7/Pamb) is used in the simulation in the determination of Cd9 as follows:
1.
2.
If Ps9/Pamb is greater than the limiting value from Figure 68, then Cd9 = 0.995
3.
If Ps9/Pamb is lower than Ps9L/Pamb In Figure 68, then use the theoretical (A9/A8)eff from the limiting curve.
4.
Calculate the required Cd9, and recalculate the exit flow conditions.
This procedure requires that the static pressure at the nozzle exit never be lower than that consistent with the limiting
curve; therefore the simulation calculations may be slightly different than the measured values for Cd. The method
provided good consistency of the thrust coefficient, C*v9, however, and therefore provides a useful correlation for
characterizing nozzle thrust performance, Figure 69.
INLET SYSTEMS
The inlet aerodynamic design characteristics of the powerplant have significant effects on the engine performance and
stability and are of vital concern to the engine designer. Inlets are designed to guide required engine flow into the front
5-71
face of the engine with minimum loss and distortion. For sub-sonic flight velocity this generally means properly faired
inlet surfaces to minimize flow separation. For supersonic flight velocity, the design must minimize losses due to
formation of shock waves. In any case to assure engine stability, the inlet design must provide low flow distortion and,
in the supersonic case, stable shock structures. The inlet design will also have important effects on overall flight vehicle
drag.
The inlet functions as follows in the overall powerplant system:
1.
It accelerates or decelerates ambient air to the engine forward flange station to achieve the entering flow Mach
number that will provide most efficient operation of the engine turbomachinery.
2.
3.
4.
5.
6.
7.
It produces additional air vehicle drag due to scrubbing and pressure forces generated by free stream flow over the
external cowl.
Inlet drag and pressure loss should be faithfully modeled in the engine simulation to calculate accurate average total
pressures for the core stream and the bypass stream at the fan inlet face. Clear definitions are required for thrust-drag
bookkeeping
3.1.
DIFFUSION AND ACCELERATION OF AIRFLOW INTO THE PROPULSION SYSTEM
The inlet is designed to induct air into the powerplant with minimum loss at the aircraft design flight condition. For
subsonic and low supersonic aircraft the design is usually a fixed inlet area with significant external diffusion and
accompanying rise in static pressure, see Figure 70. The energy contribution of the velocity of the airflow to the engine
cycle is depicted in the H-S diagram shown in Figure 71. A typical subsonic aircraft design condition would have the
inlet area approximately 20% larger than the area of the entering stream tube far upstream of the engine where the flow
which will enter the engine is unaffected by air vehicle velocity.
The diffusion process from station 0 far upstream to station 1 at the inlet lip is very nearly isentropic at the design
condition and the static pressure rise is an ideal function of the flight velocity, equation 22 and Figure 71.
An exception to this would be off-design conditions like sea level static where the engine flow required is very large
relative to the inlet area and some of the entering flow comes from behind the inlet and may actually intersect the
ground. Engine test installations minimize this effect by using a bellmouth inlet which is designed to minimize lip
overspeed, Figure 72. Another exception at the aircraft design condition would be a case where external heating of the
flow occurs due to hot gas ingestion from a missile. For supersonic inlets losses due to shock wave formation internal
and external to the inlet must be considered and will be discussed below.
h = V2/2
Eq. (22)
1
12
V flight
BypassStream
2
FLOW
Flo
A0 V
Fan
A1
Core Stream
Core Compressor
5-72
Pt0
Enthalpy
Pt2
Ps2
Ht0=Ht2
Velocity head at Fan Face =Cx2/2
Pamb
0
Entropy Increase due to Inlet Pressure Loss
Entropy
St
ag
na
tio
nS
tre
am
lin
Airflow
5-73
Spillage
Drag
0.8
Inlet Mass Flow Ratio MFR
Figure 74 - Inlet spillage drag is a function of mass flow ratio
Additive Drag + Lip suction + Scrubbing Friction = Spillage Drag
Eq. (23)
e
lag
l
i
Sp
Supersonic
Subsonic
5-74
Normal Shock
Oblique Shocks
Fan Face
A0
Fan Face
W /
M>5.0
Eq. (24)
REFERENCES
5-75
VI
INTRODUCTION
Three important subjects of the physics of air systems have been chosen:
Part 1: Tappings and Preswirl Systems
Part 2: Rotating Holes and Two Phase Flow
Part 3: Labyrinth Seals
In these fields, many new papers are available. Correlations out of the literature for the discharge coefficients, which
include the most important parameters are compared and discussed. For two-phase flow, simple correlations for
engineering purposes are recommended, especially for vent lines of aero engines. It is concluded that much more effort
is required in order to push knowledge beyond the state of current literature that is partly controversial and not
comprehensive enough for computerized engineering.
The aim of this review is to collect the major aspects for air systems calculations. It does not contain all the different
correlations out of the literature. The choice presented in the following is based on a subjective judgement of the authors
based on their experience. Furthermore it is aimed to keep it simple, therefore, the effects of only the most important
parameters are taken into account.
1.1.
A
b
C
Cd
Cp
Cp
d
dh
fs
fs,rec
k
K2
ks
H
f
Cv
DAB
NOMENCLATURE
Area (m)
Fin groove width (mm)
Absolute velocity (m/sec)
Discharge coefficient (-)
Specific heat (KJ/KgK)
Diameter (m)
Hydraulic diameter (m)
Swirl factor (-)
Receiver swirl factor (-)
Hodkinson carry over fact.
Carry over factor (-)
Correct. Factor f. stepped seals
Step height (mm)
Fin height (mm)
Velocity coefficient (-)
Pt1 Ps 2
, Effective pressure ratio (-)
Pt1 Ps1
l
L
m
Ma
W
P
Ps
Pt
Q
R
r
TAPPINGS
m))
Re
s
T
t
tn
Us
U
Downstream
Air is tapped off outwards for several purposes: Cooling air (C/A), engine bleeds i.e. for starting or for pneumatic
purposes and customer bleeds. Inwards the air is only used for cooling purposes.
5-76
In order to minimize the performance impact of the tapped air on the engine cycle, it is important to choose the lowest
possible compressor stage. The purpose for which the air is used dictates a certain pressure level and necessitates
tapping off with a minimum of pressure losses. On the other hand, from a design point-of-view it is required to keep the
number of tappings to a minimum.
To calculate overall engine secondary-air-system characteristics with a computer program, accurate correlations for all
flow elements are necessary. For tappings they are usually presented in form of Cd - values as a function of pressure
ratio.
Results from rig tests or from numerical calculations are usually from three-dimensional models and a transformation
into 1D correlations is necessary. In this process a lot of information is lost, this and further reasons are responsible that
only moderate accuracy can be expected.
2.1.
OUTWARD TAPPINGS
The following geometries are encountered:
Plain
With a step
Radiused
With subsequent diffuser
Between vanes
Plain tappings are the most common ones, but they have relatively high pressure losses.
Eq. (25)
1
1
Pt 1 Abl Ps 2 2 Ps 2
mid =
1
R Tt 1 Pt 1 1 Pt 1
Eq. (26)
Cd = f ( DAB )
Eq. (27)
DAB =
Pt 1 Ps 2
Pt 1 Ps1
Eq. (28)
In order to fit all configurations it is convenient to base the DAB-parameter on total upstream and the static downstream
pressure. Otherwise, if the upstream static pressure is used it becomes difficult in case the downstream static pressure is
higher than the upstream one.
5-77
Pt 1 Pdif ,s 2
Pt 1 Ps1
Eq. (29)
Eq. (30)
Assuming further that the diffuser behaves like an ordinary one, i.e. the distortion of the inlet velocity profile is
neglected, bleed port characteristics with a diffuser can be obtained from the basic DAB with a Cp from diffuser charts
and vice versa. Of course, DAB could be derived from DABdif.
5-78
5-79
PRESWIRL SYSTEMS
Two types, radial and axial pre-swirl systems are in use. There was an opinion, that the two types of systems should
behave considerably different. Rig tests have shown that for practical applications it is fair enough to apply the same
correlations for both. The whole problem can be divided into four parts:
1.
2.
Swirl loss and mixing in of boundary layer fluid in the pre-swirl chamber.
3.
COVERPLATES.
The second point might be different for radial and axial pre-swirl systems, but there are not enough test results available
to quantify the matter. Uncertainties arise from rig results, which are specific to engine configurations and up to now no
telemetry tests results are available. Furthermore, it is recommended to agree on a nomenclature and the basic theory in
order to facilitate comparisons.
The temperature drop into the rotating system is:
Cta = C * cos
Eq. (31)
Eq. (32)
C = f s * Cn
Eq. (33)
and
Cn = Cv * Cid
Eq. (34)
it follows
C = Cv * f s * Cid
Eq. (35)
with
2
Ps ,n ,2
Cid =
R Tt 1
1
Pt ,n ,1
Eq. (36)
Cv depends on nozzle shape, surface roughness and pressure ratio, it is in the order of 0.9 to 0.95. The swirl factor fs
comprises wall friction and turbulence losses and mixing in of boundary layer fluid. It is in the order of 0.8.
5-80
4.1.
PRE-SWIRL NOZZLES
The shape of the nozzles can be optimized by CFD methods without big effort. It is important to get a nearly rectangular
velocity profile, in order to get a good discharge coefficient and, more important, a good velocity coefficient, which is
coupled with the Cd-value. The discharge coefficient in itself is not so important as there is usually enough space
available to have the required amount of pre-swirl nozzles. A low velocity coefficient, however, would reduce the preswirl with the consequence of higher temperatures in the rotating system. An important means of achieving a high Cd
and Cv is a low roughness of the pre-swirl nozzles.
4.2.
PRE-SWIRL CHAMBER
The size of the chamber is not very important, but the distance nozzle exit and receiver entry should at least be larger
than 1.2 dn to avoid an interaction between the two flow fields: that of the nozzle jet and that of the receiver.
Radial nozzles have the advantage that there is no boundary layer due to disk pumping which can mix into the receiver
flow.
The swirl loss results from a complex 3D flow structure including a powerful chamber vortex.
4.3.
RECEIVER
he receiver holes are normally large and consequently the pressure losses small.
If the swirl can be utilized, i.e. if there is a coverplate or an annular space for a free vortex downstream of the receiver
the swirl losses are of importance.
The swirl loss of the receiver is a function of length over diameter ratio of the holes and of the ratio of receiver to nozzle
area. The receiver swirl factor is defined as:
Eq. (37)
equivalent to equation
Eq. (25)
Pt ,rel ,1 Pt ,rel ,2
2
Wrel ,1
2
Eq. (38)
The difficulty is, to define a pressure ratio across the receiver. A pressure loss is a total pressure difference and would
have to be determined by averaging the velocity profiles up- and downstream of the receiver. Normally these profiles
are not available, furthermore the pressure losses are small and the pressure profiles very uneven which would render
these results questionable. In Schmitz (1995) and Popp et al. (1996) the pressure and swirl losses of different receiver
configurations are plotted.
4.4.
COVERPLATES
In Hasan et al. (1997), Liu (1997), and Zimmermann (1989) it is shown, that the free vortex flow between coverplate
and turbine disc can be fairly well calculated by CFD, the same applies for a forced vortex. Many engine applications
have no coverplate, i.e. the swirl is lost by the transition into the disc. The main reason for having a coverplate is to
enable good seals on both sides of the swirl chamber in order to minimize leakages and to achieve a sufficiently high
pressure for a good blade cooling air supply.
4.5.
CONCLUSIONS
LABYRINTH SEALS
Most companies in the turbomachinery industry have a wealth of unpublished test results, which are partly transformed
into company restricted correlations. Therefore, the opinion in this review is necessarily based on the knowledge of the
authors. If this paper initiates more research and exchange of ideas, or even correlations, the authors would meet part of
their targets.
This chapter comprises labyrinth seals. There are several books for these subjects, e.g. Trutnovsky (1981 and 1942 first
edition) and others. Since there is a good basis, progress needs much effort, but there are still some geometrical and
5-81
1 ( Ps / Pt ) 2
=
R(n ln( Ps / Pt )
Qid =
mid Tt
A Pt
Eq. (39)
m = k 2 Cd mid
4500
Eq. (40)
4000
n=1
3500
2
Qid
3000
2500
6
2000
1500
1000
10
500
0
1
1.5
2.5
3.5
4.5
Pt/Ps
5-82
3.5
10
2.5
8
6
4
1.5
n=2
1
0
10
15
20
25
30
35
40
t/s
k=
Eq. (41)
n 1
s/t
1
n s / t + 0,02
n
n 1
k1 =
Eq. (42)
1.45
1.4
k2 = k1 *
1.35
1.3
n 0.5
k1 = ( n-1 )
k1
1.25
1.2
1.15
1.1
1.05
1
1
11
n (number of fins)
k 2 = k1 k
Eq. (43)
Carry-over factors derived from CFD calculations as tabulated in Zimmermann and Wolff (1987) compared with those
from equivalent conditions in Figure 86 and Figure 87, are in the same order, i.e. they are in line with physical
understanding.
5.1.2 COEFFICIENT OF DISCHARGE
The Cd values in Figure 88 and Figure 89 comprise the effect of flow contraction and various corrections, they are
plotted against the stream-wise Re Number (Re = Udh/ with U, the velocity at the fin tip). Above Re = 210
only an effect of s/b for small seal clearances, where friction losses are of larger influence.
there is
5-83
0.7
s/b = 0.23
0.6
0.44
Cd (n = 2)
0.5
0.4
0.83
0.3
0.2
0.1
0
100
1000
10000
100000
Re
Cd (n > 2)
0.7
0.6
s/b = 0.23
0.5
0.44
0.83
0.4
0.3
0.2
0.1
0
100
1000
10000
100000
Re
5.1.3 GROOVES
With the use of coatings on the stator, the design is often such, that running into the liner is allowed under certain
extreme operating conditions. Therefore, it is important to know the performance of a labyrinth seal running in a groove.
Figure 90, derived from Zimmermann et al. (1994), shows a correction factor for Cd values without grooves as a
function of the groove depth related to gap width and with the groove width related to groove depth as parameter. It can
be seen that large increases of the throughflow up to a factor of approximately 2.5 can occur.
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
Re > 5000
bN / b > 2.0
bN / tN > 2.5
bN/tN = 10
5
2.5
1 (extrapolated)
1
(t + s )/s
2.8
Figure 90 - Correction factor for straight-through labyrinth seals with stator grooves
5-84
Honeycomb on the stator layers with three different sizes as a function of labyrinth gap width related to cell size, where
delta m is the increase or decrease of the leakage flow induced by the honeycomb. The correction is derived from
Stocker (1978) and agrees well with company owned data. For big seal clearances the negative delta arises from the
increased roughness of the outer wall. For small seal clearances, the effective gap is enlarged. Figure 91 shows a
correction for honeycomb
100
80
delta_m [%]
60
L = 1/8 inch
40
L = 1/16 inch
20
L = 1/32 inch
0
-20
-40
0
0.2
0.4
0.6
0.8
s/L
1.50
1.45
Standard Orifice
(l/dh 0.1) dh = d
1.40
Cd,r/Cd,se
1.35
1.30
Stepped Labyrinth
dh = 2*s
1.25
1.20
1.15
1.10
1.05
1.00
0.00
0.10
0.20
0.30
0.40
0.50
r/dh
5-85
0.5
0.48
0.46
0.44
t/s = 10
Cd
0.42
0.4
0.38
20
0.36
30
0.34
minimum leakage
0.32
40
0.3
0
0.2
~0.25
0.4
0.6
0.8
1.2
h/t
ROTATION
From Waschka et al. (1991) it can be seen that for Re 10000 the influence of rotation on throughflow is negligible.
But for laminar flow, especially for low Re Numbers the effect is high. As the pressure losses for laminar flow are due
to increasing friction losses with decreasing Re Number they depend on surface area and form. Therefore, a generalized
correlation cannot be very accurate.
5.2.
STEPPED LABYRINTH SEALS
The flow function of the ideal labyrinth in Figure 85 is as well valid for stepped seals.
m = k s Cd mid
Eq. (44)
(1/n)
Cd (single Fin)
(Pt/Ps)
0.9
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.7
b
0.6
=2.5
0.5
0
s/b
Figure 94 - Discharge coefficients for stepped labyrinth seals (single fin)
5.2.1 IDEAL CD-VALUES
Figure 94 contains the Cd values for single fins derived from Snow (1952) to be applied for stepped seals because of the
assumption that there is a one dynamic head loss after every fin. It is furthermore assumed that the pressure ratios are
equal for each fin.
5.2.2 CD CORRECTION FACTOR
As the assumption above is not fully applicable, a correction factor is necessary. Figure 95 shows this factor as a
5-86
function of gap width related to fin width. It represents test results with +/- 5 % accuracy.
5.2.3 CORNER RADIUS
See Figure 92 and the associated section, where the effect of corner rounding for stepped seals is included.
5.2.4 FLOW REVERSAL
From Waschka (1991) it can be seen that the influence of flow reversal on the Cd- value of stepped labyrinth seals is
small. Only for Re 10 there is an effect of approximately 10 to 20%, but that rarely occurs with stepped seals in aero
engines, see Zimmermann et al. (1994).
3
1.5
1.4
1.3
1.2
Pt/Ps = 1.0
1.2
1.6
1.4
1.8
2.0
2.5
KS
1.1
1
0.9
0.8
0.7
0.6
0.5
0
0.5
1.5
2.5
3.5
s/b
ROTATION
Waschka et al. (1991) [129] have shown that for approximately Re 3.102 the influence of rotation on throughflow is
negligible.
5.2.6 STEP HEIGHT
Figure 96 based on unpublished company owned data shows that there is a flat minimum at H/t equal to 0.075. It is
recommended to place practical applications around that value. For decreasing s/b Cd*ks increases, because in the then
relatively longer gap the flow reattaches and thus increases cd.
1.2
s/b = 0.2
0.4
0.5
1
0.8
Cd*K
1.0
1.5
0.6
b
0.4
n = 5 fins
Pt/Ps = 1.2
b = 1.0 mm
(M5:1)
0.2
0
0
0.05
0.1
0.15
0.2
0.25
0.3
H/t
For both, stepped and straight through labyrinth seals the influence of Reynolds Number is only significant for
laminar flow.
It is advantageous to design straight through labyrinth seals with flat chambers. But the thermal stability of the fins
in a rub situation has to be watched.
It is as well advantageous to realize flat steps and small chambers for stepped labyrinth seals. The influence of flow
reversal on the Cd values of stepped seals is small.
5-87
There is a strong need for further research, especially concerning whole models and correlations for all important
parameters.
All correlations in this paper could be refined.
Better or new correlations are required for Reynolds Number effect, influence of rotation, especially at low Re
Numbers, furthermore for the influence of chamber depth (or step height) and for the axial displacement of stepped
seals.
There is a wealth of literature but it is difficult to create complete correlations from individual research projects that
do not fit together.
REFERENCES
97. Oates, Gordon Aircraft Propulsion Systems Technology and Design, AIAA Education Series, 1989
98. Chen, D. K., Chen, Z. H., Zhao, Z.S.and Zhua, N., 1986, The Local Resistance of a Gas- Liquid Two- Phase Flow
through an Orifice, Int. Journal of Heat and Fluid Flow, Vol. Nr. 3, pp. 231-238.
99. Chisholm, D., 1980, Two-Phase Flow in Bends, Int. J. Multiph. Flow, Vol. 6, pp. 363-367.
100. Dittrich, R.T.; Graves, C.C., Discharge Coefficients for Combustor-Liner Air Entry Holes : Circular holes with
parallel flow , NACA-TN3663, 1956.
101. Farthing, P.R., (1989), The Use of Deswirl Nozzles to Reduce the Pressure Drop in a Rotating Cavity With a
Radial Inflow, ASME Paper 89 - GT - 184.
102. Fischer, R., 1995, Calculation of the Discharge Characteristic of a n Orifice for Gas-Liquid Annular Mist Flow,
Int. J. Multiph. Flow, Vol. 21.
103. Friedel, L. ,1978, Druckabfall bei der Strmung von Gas/Dampf-Flssigkeits-Gemischen in Rohren. Chem. Ing.
Technik, Vol.50, pp. 167- 180.
104. Graham, E. J., 1967, The Flow of Air-Water Mixtures through Nozzles, NEL Report, No. 308.
105. Grim, R. E., 1969, Fluid Flow Characteristics Through Orifices in Enclosed Rotating Discs, M.S. Thesis, AFIT.
106. Hasan, J. C. 1997, Flow in a Cover-Plate Pre-Swirl Rotor-Stator System, ASME Paper 97 - GT - 243.
107. Hodkinson, B., 1940, Estimation of the Leakage Through a Labyrinth Gland, Proceedings Inst. Mech. Eng. Vol.
141, pp. 283 - 288.
108. Idelchic, I. E., 1960, Handbook of Hydraulic Resistance, Atomic Energy Com. And Nat. Science Found. AEC TR 6630.
109. Lockart, R. W. , and Martinelli, R., 1949, Proposed correlation of data for isothermal two phase, two component
flow in pipes, Journ. Chem. Eng. Progr., Vol. 45, pp. 39-48.
110. Mc Greehan, W. F. and Schotsch, M. J., 1987, Flow Characteristics of Long Orifices with Rotation and Corner
Radiusing, ASME Paper 87-GT-162.
111. Mc Greehan, W. F., and Ko, S. H., 1989, Power Dissipation in Smooth and Honeycomb Labyrinth Seals, ASME
Paper 89 - GT - 220.
112. Millward, J. A., and Edwards, M. F., 1994, Windage Heating of Air Passing Through Labyrinth Seals, ASME
Paper 94 - GT - 56.
113. Meyfarth, P.F., and Shine, A. J., 1965, Experimental Study of Flow Through Moving Orifices, J. of Basic
Engeneering, pp. 1082- 1083.
114. Mller, M., 1990,EJ 200: Durchflubeiwerte fr Spaltabzweigungen in Verdichtern, Technischer MTUBericht.
115. Liu, X., 1997, Flow in a Corotation Radial Inflow Cavity Between Turbine Disc and Coverplate, ASME Paper 97
- GT - 137.
116. Parker, D. M. and Kercher, D.M., 1991, An Enhanced Method to Compute the Compressible Discharge
Coefficient of Thin and Long Orifices with Inlet Corner Radiusing, HDT- Vol. 183, Heat Transfer in Gas Turbine
Engines, Asme 1991.
117. Popp, O. Kutz, J., Zimmermann, H., CFD-Analysis of Coverplate Receiver Flow, ASME Paper 96 - GT - 357.
118. Reichert, A. W., Birillert, D.and Simon H., 1997, Loss Prediction for Rotating Passages in Secondary Air
Systems, ASNE Paper 97- GT- 215.
119. Rohde, R. E., Richards, H. T. and Metger, G. W., 1969, Discharge Coefficients for Thick Plate Orifices with
Approach Flow Perpendicular and Inclined to the Orifice Axis, NASA TN D- 5467.
120. Samoilowich, G. S., 1957, Coefficients of Flow Through Pressure Equalizing Holes in Turbine Discs, D. S. I. R.
Translation C. T. S. No. 541.
121. Meierhofer, B.; Franklin, C. J., 1981, 'An Investigation of a Preswirled Cooling Airflow to a Turbine Disc by
Measuring the Air Temperature in the Rotating
122. Channels', ASME 81-GT-132.
123. Schmitz, D., 1995, Analyse von Axialen und Radialen Vordralldsensystemen, Technischer MTU Bericht.
124. Snow, E. W., 1952, Diskussionsbeitrag , Proc. Inst. Mech. Engrs., Vol. 166.
125. Stocker, H. L., 1978, Determining and Improving Labyrinth Seal Performance in Current and Advanced High
Performance Gas Turbines, AGARD-CP-273.
126. Storek, H., Brauer, H., 1980, Reibungsdruck-verlust der Adiabaten Gas/Flssigkeits- Strmung in Horizontalen
und Vertikalen Rohren, VDI Forschungsheft Nr. 599.
5-88
127. Theissing, P. Eine allgemeingltige Methode zur Berechnung des Reibungsdruckverlustes der
Mehrphasenstrmung, Chemie-Ing.-Techn. 52, Heft 4 (1980).
128. Trutnovsky, K., and Komotori, K., 1981, Berhrungsfreie Dichtungen, VDI - Verlag
129. Waschka, W., 1991, Zum Einflu der Rotation auf das Durchflu- und Wrmebergangsverhalten in
Labyrinthdichtungen und Wellenduchfhrungen, Dissertation, Universitt Karlsruhe.
130. Trutnovsky, K., and Komotori, K., 1981, Berhrungsfreie Dichtungen, VDI - Verlag.
131. Weissert, I. (1997) Numerische Simulation dreidimensionaler Strmungen in Sekundrluftsystemen von
Gasturbinen unter besonderer Bercksichtigung der Rotation. VDI Fortschrittsbericht, Reihe 7, Nr. 313.
132. Wood, J.D., Dickson, A.N., 1973, Metering of Air-Oil Mixtures with Sharp-Edged Orifices, Dep. Mech. Eng.
Rep., Heriot- Watt University, Ricarton, Edinburgh, U.K.
133. Zimmermann, H., 1989, Some Aerodynamic Aspects of Engine Secondary Air Systems, ASME Paper 89 - GT 209.
134. Zimmermann, H., Kammerer, A., Fischer, R. and Rebhan, D., Two-Phase Flow Correlations in Air/Oil Systems of
Aero Engines, ASME Paper 91-GT- 54.
135. Zimmermann, H., Kammerer, A., Wolff, K. H., 1994, Performance of Worn Labyrinth Seals. ASME Paper 94 GT - 131.
136.
Zimmermann, H., and Wolff, K.H., 1987, Comparison between Empirical and Numerical Labyrinth Flow
Correlations, ASME Paper 87 - GT - 86.
ROTATING HOLES
The aim of this review is to collect the major aspects for systems calculations. It does not contain all different
possibilities out of the literature, thus it is based on a subjective judgement of the author. For simplicity, only the effects
of the most important parameters are taken into account.
Rotating holes are often the bottleneck to sections of the air system network, because the discharge coefficients are
small. Consequently, large areas in rotating pieces are required which is often critical, especially with respect to the
stressing of these parts. For rotating holes the final method of correlation has yet to be found. Therefore, in this chapter
four methods are compared to each other and discussed.
7.1.
SHARP-EDGED HOLES
For sharp edged holes with small l/d all sources quote similar test results. The more it is astonishing that for the other
cases there are so large discrepancies.
7.2.
ROUNDED HOLES
There is only little and contradictory literature available on the influence of the rounding radius and the interaction with
the l/d effect.
Several methods of correlation are in use. The definitions of Cd are the following:
Cd =
m
mid
Eq. (45)
2 Ps
1
Ps
mid = A P
(
) (1 (
)
RT 1 P
P
2
Eq. (46)
P and T may be expressed either in the absolute or in the relative frame of reference. In the absolute frame of reference,
P and T are P0 and T0, whereas in the relative frame of reference Prel and Trel have to be used:
1
Prel = P0 (1 +
Ma 2 ) 1
2
Trel = T0 (1 +
Ma =
1
Ma 2 ) with
2
U
R T0
Eq. (47)
Eq. (48)
Eq. (49)
Samoilowich (1957), Meyfarth and Shine (1965) and McGreehan and Schotsch (1987) plot Cd in the absolute system
against U/Cax,id:
5-89
m
P
2
Cax , id = id =
RT0 (1 ( s )
P0
A
1
Eq. (50)
More recently Weissert (1997) presents Cd in the relative frame of reference versus Cax:
Cd mid
A
Cax =
Eq. (51)
For , the density in the vena contracta has to be used. This necessitates iteration and the first Cd has to be guessed.
Another method would be to plot Cd in the relative system against U/Wid with:
P
2
RTrel (1 ( s )
Prel
1
Wid =
Eq. (52)
Because U/Wid approaches 1 for U approaching infinity, U/Wid is limited between 0 and 1 and not between 0 and
as is the case with U/Cax and U/Cax,id. Therefore, U/Wid is advantageous, especially if limited test data have to be
extrapolated.
In the following diagrams the different correlations are transformed to the same basis, in order to allow a comparison.
The upstream total equals the static condition. From Figure 97 it can be seen that in the absolute frame of reference
Cd,rot,abs may exceed 1. This is the more the case the greater Rk/D is. Cd,rot,rel in the relative frame of reference
1.2
Rk/D
L/D
Weiert
Cd,rot,abs
1.0
0.5
2.6
P0/Ps
McGreehan & Schotsch
0.8
1.1
1.2
1.3
1.5
1.8
2
0.6
Samoilowitsch
0.4
0.2
0.0
0.0
0.5
1.0
1.5
2.0
2.5
U/Cax,id
Rk/D= 0.5
L/D= 2.6
66
P0/Ps
1
Weissert
Cd0.8
,ro
t,r
el 0.6
1.1
1.2
1.3
1.5
1.8
2
0.4
Samoilowitsch
0.2
0
0
0.5
1.5
2.5
U/Cax,id
5-90
U/Cax,id is higher than the static Cd , because of the work done to the distorted flow with strong secondary vortices.
With large L/d the work done is mainly independent of the corner radius, but influences the efficiency, i.e. the
conversion into pressure rise which yields higher Cd values than for shorter L/d. In Figure 97 to Figure 99 the Cd-values
are plotted against U/Cax,id, whereas in Figure 100 against U/Wid. This has the clear advantage that the Cd-values have
to approach different Cd Correlations 0 at U/Wid = 1 on the abscissa. This is advantageous if, due to the limited extent
of available test data, extrapolations become necessary. The difference between the three compared correlations partly
results from the difference in the test results. Only Weissert (1997) has modeled the maximum in the relative system on
the left hand side of the diagrams which is due to the work done to the secondary flows.
1.2
Rk/D= 0.5
L/D=
2.6
666
P0/Ps
1.0
Cd,rot,rel/Cd,0
Weissert
0.8
1.1
1.2
1.3
1.5
1.8
2
0.6
0.4
Samoilowitsch
0.2
0.0
0.0
0.5
1.0
1.5
2.0
2.5
U/Cax,id
Rk/D= 0.5
L/D=
2.6
Weissert
1.0
Cd,rot,rel/Cd,0
P0/Ps
0.8
0.6
Samoilowitsch
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
U/Wid
Rk/D= 0.5
L/D= 2.6
1.0
Cd,rot,rel/Cd,0
Weissert
P0/Ps
0.8
1.1
1.2
1.3
1.5
1.8
2
0.6
Samoilowitsch
0.4
0.2
0.0
0.0
0.5
1.0
U/Cax
1.5
2.0
2.5
5-91
Rk/D= 0.5
L/D=
0.8
2.6
U/Wid
Weissert
P0/Ps
0.6
1.1
1.2
1.3
1.5
1.8
2
Samoilowitsch
0.4
0.2
0
0
0.5
U/Cax 1.5
2.5
Eq. (53)
with:
l/d 0
5-92
It is recommended to plot related Cd values in the relative frame of reference against U/Wid.
If U/Cax is used on the horizontal axis, errors due to the coupling with Cd can occur in case the correlation is used
for extrapolating beyond the test data.
The difference between the various test results is not understood.
New test results are required to clarify the latter point and to refine the correlations.
Only those correlations are chosen for this chapter which are easy to handle for engineering purposes and as there is not
much experience in this field, test data are added to give more confidence into the application.
In Zimmermann et al (1991) [134] a review was published mainly based on water/air test results. Since then, new
correlations have come forward and oil / air test results are available. Because of limited space the formulas or basic
diagrams are not reproduced, they have to be taken from the references. The air/oil test data are taken from an ongoing
research program and a forthcoming thesis from R. Fischer.
8.1.
PIPE FLOW
The Lockart & Martinelli (1949) [109] correlation is still in use, therefore it has been compared with those of Storek &
Brauer (1980) [126], Friedel (1978) [103] and air/oil test results (Figure 103). As already stated in Zimmermann et al
(1991) [134] the correlation of Friedel is recommended.
5-93
18000
16000
14000
[P
a
/m
/
P
12000
10000
8000
6000
4000
2000
0
0
0.2
0.4
0.6
0.8
7000
6000
5000
4000
3000
2000
Chisholm (1980), Friedel (1978)
Chisholm (1980), Storek and Brauer (1980)
air/oil measurements
1000
0
0.000
0.200
0.400
0.600
0.800
1.000
Figure 104 - Pressure loss of a 90 bend Comparison of available correlations with air/oil measurements
8.3.
MITRE BENDS
Figure 104 shows the pressure loss comparison of a 90 mitre bend with correlations from Chisholm (1980) [99] and
from Theissing (1980) [127], it can be seen that the Chisholm correlation fits better but for pure air there is a big
discrepancy. For a 60 mitre bend the situation is very similar, Figure 105.
5-94
20000
18000
90 mitre bend
D=0.02 m
Ps0=2 bar
T=370 K
Chisholm 60 g/s
Chisholm 40 g/s
measurement, 40 g/s
measurement, 60 g/s
Theissing 40 g/s
Theissing 60 g/s
16000
14000
theory
pure air
12000
10000
8000
6000
4000
2000
0
0
0.2
0.4
0.6
0.8
0.12
0.1
0.08
0.06
measurement,
measurement,
measurement,
prediction, x =
prediction, x =
prediction, x =
0.04
0.02
x = 0.4
x = 0.3
x = 0.2
0.4
0.3
0.2
0
0.4
0.5
0.6
0.7
D = 20 mm
d = 10 mm
l / d = 0.5
T = 370 K
Ps0= 4 bar
0.8
0.9
Po A
Po
2
1 Ps ,
R ( 1)
Po
Cd can be obtained from pure air correlations for instance from Parker & Kercher [116].
Eq. (54)
5-95
80
70
60
50
40
30
20
10
0
0.00
0.20
0.40
0.60
0.80
1.00
m x 0.44 T
= 33221
. = Qcrit
Po A
Eq. (55)
Figure 108 summarizes all the test results, which show a scatter band of approximately +/- 10 % which is not bad for
two phase flow.
4000
4500
D = 20 mm
l/d = 0.5
T = 330 - 370 K
+ 10 %
3500
3000
- 10 %
2500
2000
1500
1000
500
0
0
0.2
0.4
0.6
0.8
5-96
40000
30000
20000
10000
0
0.00
0.20
0.40
0.60
0.80
1.00
VII
The standard of fuel system and actuator modeling that is required depends on the objective for the work. When the
objective is to analyze and develop a control-system, then very detailed models of the hardware and software will be
required. When the basic control is functioning satisfactorily, and the objective is to evaluate engine performance and
handling, the control system models can usually be simplified significantly.
Simplification of the control system models for engine handling and performance work is required to obtain acceptable
run times for the model. The detailed control models normally require shorter integration time steps than are acceptable
for engine handling work.
Acceptable control-system models for engine handling work can be developed using simple transfer-functions and the
steady-state characteristics of the control. It may be necessary to change the coefficients of the transfer functions with
engine operating conditions. The information required to define suitable transfer functions for the control systems can
be obtained from the more detailed models that should have been built to facilitate control design.
Where control-system dynamics must be modeled accurately, then all dynamic valves need to be modeled as first or
second-order systems. For all valves, fluid pressure, spring force, friction effects, leakage and the effect of fluid masses
should be considered and included where appropriate. Fluid pressure-drop-to-flow relations and fuel compressibility
should be modeled throughout the system. The effects of flow forces (Bernoulli Forces) should be included in valves
that operate at low force levels.
It is not practicable to model the effects of heat transfer and other thermal effects in a dynamic control-system model.
Where such effects are of interest, a separate model should be built for this purpose, using simplified control modeling.
The effect of manufacturing tolerances can be simulated, in a detailed control-system model, by changing parameters
within the model. It is rarely practical to include terms for all components where tolerances could influence control
performance, so engineering judgement must be used in determining how to model particular scenarios. The same
constraints apply to modeling of failures within the control system.
Modern controls are built around electrohydraulic and electromechanical transducers and actuators. These can often be
modeled satisfactorily using the manufacturer's declared characteristics and response data, but more detailed modeling is
advisable if performance is marginal. It is also advisable to model delays incurred in the signal conditioning associated
with many transducers.
Modeling of the DECU will depend on how detailed the control system models are and what the objectives for the
model are. The engine control laws will normally be modeled, but the functionality associated with transducers and
output devices may be omitted for the simpler models.
Clearly, control-laws, which are implemented digitally, are bespoke and modeling can only exist at a certain level for a
5-97
given logical feature. However, a decision may be made to omit a certain branch of logic depending on the application
of the model.
There will be other features of a fuel system, which require modeling at the appropriate level of detail, examples being
heat-to-fuel, filtration, fuel chemistry changes, pressure losses, and compressibility effects.
1.1.
TYPES OF MODELS
There are two main types of pump used in modern fuel systems: the fixed displacement type and centrifugal flow
pumps. Gear pumps are the most common form of fixed displacement pump and are widely used in modern, mainengine fuel-systems. Centrifugal pumps are used primarily in low-pressure fuel supply systems but are increasingly
finding applications in reheat controls.
In the past, variable stroke piston pumps were used in many main engine fuel systems, and vapor core pumps (inlet
throttled C.F. Pumps) were used for reheat applications. Although these pumps are to be found on many engines still in
operation, they are unlikely to be used on new engine controls and are not described in this report.
1.2.
MODELING THE FIXED DISPLACEMENT PUMP
This type of pump is probably the simplest to model. The steady state characteristic (shown below) is used to determine
pump delivery flow given pump speed and delivery pressure. There are no dynamics in the pump to be represented.
2000
increasing
pressure
Flow
(gph)
0
0
Speed
(%)
100
1600
Delivery
Pressure
(psia)
increasing
speed
0
0
Delivery Flow
(gph)
2000
5-98
To avoid algebraic loops in the model, it may be necessary to include a first order lag as shown in the diagram below.
The lag should not exceed 0.01 seconds.
2P
force = P.area
force
acc =
mass
Qnet
P = K.
.dt
vol
Figure 113 Hydraulic control system equations
1.5.
ELECTRO-HYDRAULIC SERVO VALVES
Modern engine control systems are operated using several types of electrohydraulic servo valve. The most common
types are:
Solenoid;
Single stage servo valve - flapper, jet deflector and jet pipe types;
Two stage servo valve - using one of the above first stages to drive a second stage spool.
Response of these servo valves can normally be ignored for engine handling work, as the bandwidth of the devices is
typically above 100Hz. For more detailed work, the manufacturers response data can be used, and an example of a twostage servo valve is shown later. For detailed analysis of the control systems, it is often necessary to model individual
components within the servo valves, but this is beyond the scope of this document.
The solenoid valve is usually used for on-off operations and can often be treated as a simple switch. Where the response
is critical, the device can be represented as a first order lag with a time constant obtained from the manufacturers data.
A schematic of a single-stage flapper-valve is shown below, the jet deflector and jet pipe devices are similar, although
5-99
the steady state characteristics need to be treated differently when detailed models are required.
Torque
motor
armature
and
flapper
HP
HP
C1
LP
C2
5-100
throttling when a C.F. pump is used. Both direct acting and servo pressure-drop regulators may be used.
The response of the pressure drop regulator is normally much faster than that of the metering valve, and can be ignored
for engine handling work; only the steady state characteristics of the valve need be included. Where fuel system
problems are to be investigated, a more detailed PDR model is required, but this is beyond the scope of this document.
1.5.3 POWER ACTUATORS
These are used to control IGVs, nozzle area and other engine components. Typically the power actuator is a ram, with
the servo supply set by a two-stage servo valve. An LVDT or other transducer measures Ram position, and a digital
computer is used to close the loop. Provided that the control has been correctly designed, the response of the servo valve
will be insignificant in the overall control loop. An acceptable model can be obtained by setting the ram velocity as a
function of DECU drive current and available pressure drop across the servo valve, after allowing for the pressure
required to overcome actuator loads.
servo valve
HP
metering valve
to burners
LP
pressure drop
control
and spill valve
5-101
5-102
s
s +
n
Kf
Lf
where :
n =
Bf
. n
2.Kf
K2.Kw
Kv =
Kf . A
5-103
Figure 121 - Engine and control-system models running together with inter-process communication
It is common practice for engine models to be written in FORTRAN although the emphasis is shifting towards C. As
such, an engine model may be imported into proprietary development environments in the form of a user code block.
Custom interface routines may be required. The interface routines should allow easy routing of additional signals if
required.
Development environments such as MATLAB or MATRIXx have suites of tools (toolboxes) with which the model
must be compatible. It is common practice for all dynamic elements in a system to be numerically integrated by the host
environment. Several integration algorithms are available; the most appropriate for the particular dynamics present, can
be chosen. This said, some engine models might perform their own integration internally. Iterative techniques used in
thermodynamic solutions facilitate implicit integration approaches that may have certain advantages in numerical
5-104
stability. Engine models are also manipulated to generate linear models in state-space form.
In the early stages of a project, it is common to find that some control variables are not modeled as independent program
input variables. For example, a variable mixer may be present in the engine concept, but early performance studies do
not need a geometry-to-thermodynamic-effect calculation. This is because the model can be matched by prescribing
what such a feature has to achieve, rather than predicting the effect of a specified geometry. Clearly, in the case of
control-system design, a geometry model is required, although some simplification may be acceptable at the preliminary
stage.
1.5.7 USING PERFORMANCE (CYCLE-MATCH) MODELS FOR CONTROL-SYSTEM DESIGN AND ANALYSIS
Control engineers are perhaps most familiar with engine models presented in the form shown in Figure 122 below. The
engine is a dynamic simulation, which requires the host environment to perform the numerical integration of state
derivatives. Often explicit integration techniques are used such as Runge-Kutta 4th order. A model presented in this form
is also able to be linearized (see below). Whereas this is the standard model form used in many areas of dynamic
modeling, it is one particular form (of several) in which a performance model may be presented. It can be thought as a
states-matching mode. This reflects the models function of finding a solution (by iteration) to a question posed in terms
of specified values of inputs [u] and states [x].
Figure 122 - A general form of dynamic model using external integration for simulation
Inputs [u] are boundary terms e.g. environmental temps, pressures, fuel-flows and geometry for an engine;
Outputs [y] are parameters of interest e.g. thrust, specific-fuel consumption, surge margin;
States [x] are those parameters whose responses are governed by a differential (dynamic) equation e.g. shaft speed;
State derivatives [xdot] are the time differentials of the state variables.
1.5.7.1 LINEARIZATION
In order to linearize the engine model, it must be presented in this form - i.e. with states set up as program input, and
state derivatives as output. Thus the model may be manipulated (by parametric perturbation of states and inputs) around
a base point to generate the linear (partial-derivative or state-space) engine representation which is necessary for
standard controller design and analysis methods.
1.5.7.2 STEADY-STATE SYNTHESIS (INITIALIZATION)
Consider the engine model presented as shown in Figure 123. In order to generate a steady-state solution for a
prescribed set of boundary conditions (inputs, e.g. fuel flow), the values of the state variables have to be ascertained.
With a performance model, this is achieved by the iteration that is also used for the thermodynamic solution. With other
types of engine model which do not use iteration, the initial values of states can be determined in several ways:
By invoking an external iteration function, trimming, which varies the states to achieve a zero state-derivative (at
fixed inputs).
By running a settling transient where the states are initialized at arbitrary values and the simulation run over a period
of time at constant inputs until steady state is achieved. Thus the initial combination of arbitrary states and
prescribed (fixed) inputs combine to give a non-zero state-derivative. This is numerically integrated to predict the
next time-step value of state variable. The model should eventually achieve steady state where xdot = 0.
By using table look-ups of states vs. inputs for steady-conditions.
By setting the inputs and states at values stored from a previous run (not necessarily a steady-condition), moving the
input variables to the required values over a time-base, and allowing the model to achieve steady-state.
5-105
Figure 124 - A Performance model running in transient mode using internal integration
The base points for linearization are not necessarily at steady-state conditions. Any mid-transient point (generated by
transient mode) can be selected as a base point for linearization.
Generalization:
The three modes are fundamentally distinguished by the iteration matching-scheme used to produce the solution in
each case (which is transparent to the user). In addition, the dynamics routines are not called in steady-state mode.
The steady-state matching-scheme is the simplest and does not allow excess power on any shaft, nor any stored W, P
or T within a volume. This matching scheme is extended for the states-matching mode and for the transient mode,
where an extra matching pair is added for each dynamic equation being considered. Ref. 1 expands on this using a
simple example.
Iterative models are not dependent on the solution of gas dynamics equations for dynamic or transient simulation valid simulation can be achieved by modeling only the shaft dynamics. Modeling of gas dynamics (volume packing)
can be added if required.
When gas dynamics are not modeled, the model sampling frequency is around 5Hz. With gas dynamics, the
bandwidth is increased to 30-40Hz.
5-106
The performance process is largely preoccupied with the prediction and analysis of steady-state performance.
Solution by iteration is essential for this type of work. Performance engineers often regard transient synthesis as an
extension of the steady-state model. Controls engineers often see things the other way around with the steady state
being a particular state of a dynamic system, which is perhaps never truly achievable. However regarded, the
modeling techniques are the same.
1.5.7.4 USING CONTROL-SYSTEM MODELS IN PERFORMANCE AND OPERABILITY ASSESSMENT
Integrated engine and controller models are seldom required in traditional performance activities. However, for
operability studies, such as compressor stability assessment, a control-system model is essential. The emphasis is placed
on the understanding of engine behavior in response to control-system action. Data interface requirements therefore
differ - the performance engineer wants full visibility of internal engine parameters and engine analysis tools. It is also
the case that other sub-systems, for instance the secondary air-system and oil systems, should be represented in the best
possible way.
Clearly, if a controller model has been developed in a controls development environment, it is desirable that this should
also run in the performance environment. Models constructed using Graphical User Interfaces (GUI) can be represented
in pure code forms using proprietary code-writing tools. This facility offers potential for inter-environment use. C and
ADA are common languages for this task.
Figure 125 - Control-system model imported (embedded) into performance modeling environment
Figure 125 shows a controller model derived using picture-to-code tool in a performance development environment.
Several issues arise:
Interface of dissimilar languages (e.g. FORTRAN and C);
Ideally, the control-system and engine model should be separately configurable items;
The integrated model in this environment should be technically equivalent (in simulation terms) to the alternative
scenario of the engine-model imported as a subservient element in the controls environment;
Initialization.
In the case of FORTRAN and C, there is little technological challenge in interfacing. It is more a case of careful
handling of the different conventions used in the two languages. Naturally, the platform must support both languages.
It is possible to avoid embedding the controller within the engine model, by using the inter-process communication
technique shown in Figure 121. For work centered on performance studies, it is desirable that the combined model
should be run from the performance domain.
Performance work often involves the prediction of nominal steady-state performance under the action of the controlsystem, at, say, a particular power demand or rating. For example, as expressed by a certain pilot-lever angle. Ideally,
the full, integrated model should be used to generate these points. However, the computing task can become a burden
because of the requirement to initialize certain dynamic terms in the combined model.
A control-system (or other engine sub-systems) can also be structured to take on the form in Figure 121. Thus a
combined controller and engine model, both exhibiting the form in Figure 121 and running in a development
environment (such as MATRIXx or Matlab) can be initialized in the ways discussed above. Initialization is not so
straightforward when:
1.
2.
Any constituent part of a whole system is not in the standard form (Figure 122);
When constituent parts of the model are spread across different environments (Figure 121);
5-107
3.
When the whole model is running in an environment that does not feature the trimming function (Figure 125).
1 is solved by providing ensuring a standard form, or providing models which are self-initializing.
2 ought not be a fundamental problem if 1) is solved; inter-process communication will be increased for the
initialization phase.
3 becomes an issue when using combined models in the performance environment. Often, a series of steady-state flight
points (for given values of pilot demand) is to be generated. This can require initialization for each point. In such a case,
the settling transient method is an excessive burden on execution time. However, there are ways of alleviating this
overhead. The transients are not of interest, so some dynamic terms can be muted or modified to reduce settling time
(e.g. heat-soakage can be muted and shaft inertia reduced). A problem can arise if this measure is taken to extremes: the
whole-system stability may be compromised and therefore prevent a steady condition being achieved! A trimming
function may be used, however such a facility may not be available in all environments. Also, with an extended iteration
scheme, there is a potential convergence hazard with the number of control-system states that may need initialization.
An alternative approach is to use a subset of the controller: the ratings structure, to establish the level of performance in
terms of the engine rating parameters. An iterative engine model can be run to such parameters (with due regard to
engine limit loops) and converge on a steady-state solution (either on the prescribed rating or on an overriding limit)
without any settling period. From this point, the initial conditions in the controller may be determined either by backcalculation or by including controller terms in the iteration loop. The engine condition thus obtained should be identical
(within iteration tolerance) to that obtained by transient settling. However, there are some situations when this approach
is not feasible. Even with iteration; the time-settling technique is sometimes essential, for instance for engines with
switching bleeds or with unusual peckingorder of limits and ratings. The full model is, of course, the best means of
determining the steady-state performance, if indeed such a condition exists.
Where the model is being used for true transient simulation, the execution time associated with the start point is less
significant and so the transient settling method may be acceptable. It may be beneficial in this case to initialize the
engine model at an arbitrary fuel flow first, thus giving the controller a feedback consistent with the initial engine input.
The combined model can then bootstrap to the required steady-state condition, at which point the controller state
variables (e.g. numerical integrators) are correctly initialized for the forthcoming maneuver of interest. From the users
point of view, the settling method can be made virtually transparent. The settling transient can be hidden and discarded
after steady state has been achieved. The only symptom will be in the execution time.
The issues discussed above extend to other systems such as intakes and secondary air system that interact with the
engine.
REFERENCES
137. Marcus S Horobin May 1998, Cycle-match models used in functional engine design - an overview, Presented at
the RTO Symposium: Design Principles and Methods for Aircraft Gas Turbine Engines
A-1
Appendix
Summary Analysis of AVT-18 Survey
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
TABLE OF CONTENTS
Table of Contents ....................................................................................................................................................... A-1
Introduction................................................................................................................................................................ A-1
Contact Details........................................................................................................................................................... A-1
Outline of Business .................................................................................................................................................... A-2
Guidance Notes for Completing the Questionnaire................................................................................................... A-2
Scope of Modeling Activity in Your Organization ................................................................................................... A-3
Computing Platforms ................................................................................................................................................. A-3
Computing Languages and Tools .............................................................................................................................. A-3
Input and Output ........................................................................................................................................................ A-4
Issue of Engine Models To Customers .................................................................................................................. A-4
Requirements of the Model.................................................................................................................................... A-5
Engine Modeling Technique and Model Type ...................................................................................................... A-6
Modeling of Physical Processes............................................................................................................................. A-7
Associated Capabilities .......................................................................................................................................... A-8
Initialization ........................................................................................................................................................... A-8
Associated Models (e.g. Control-System) ............................................................................................................. A-9
Interfaces with Other Models and Systems and Databases ................................................................................. A-10
Testing and Validation ......................................................................................................................................... A-10
Documentation ..................................................................................................................................................... A-11
Standards .............................................................................................................................................................. A-12
Further Comments................................................................................................................................................ A-13
General Conclusions ............................................................................................................................................ A-13
Questionnaire ...................................................................................................................................................... A-15
INTRODUCTION
In the formative stages of AVT-18 (then WG-29), it was decided to approach the world community who would
potentially have an interest in the type of publication, which was emerging as the remit of AVT-18. This was considered
to be useful on three counts:
To identify current modeling practice;
To identify expectations and future needs for performance predictions;
To guide the efforts of AVT-18 in their preparation of a worthwhile reference document.
Accordingly, a detailed survey was prepared and distributed to users and creators of engine models under the following
banner heading:
PERFORMANCE PREDICTION AND SIMULATION
OF GAS TURBINE ENGINE OPERATION
(whole-engine thermodynamic performance and control-system)
Survey of creators and users of gas-turbine performance models
The survey comprised 18 sections, which are analyzed in order below. In retrospect, a simpler survey would have had
advantages. The complexity of the survey led to a time penalty in respect of the third objective above. As a
consequence, the working group had to anticipate the response in order to focus effort.
The quality of response varied. Some returns were minimal; some contained a high level of detail. Where appropriate,
some contributions are quoted verbatim (anonymously). An assurance was given at the outset that specific responses
would not be cross-referenced to the source in this final report this has been honored. There were a few instances of
odd responses, which explain the fuzzy statistics in some places. In general, the overall response revealed few surprises.
CONTACT DETAILS
38 completed surveys were returned. It is estimated that this represents a rate of return of about 30%. Recipients
included engine and aircraft manufacturers, academic institutions, airlines and supporting industries and agencies.
A-2
Country
Returns
Belgium
Canada
France
Germany
Netherlands
UK
USA
1
2
3
9
7
7
8
TOTAL
37
Comment
The returned sample cannot be considered a scientific sample. However, enough response was gained to
get a reasonable view of current practice.
OUTLINE OF BUSINESS
Industry
Returns
Airline
Academia
Aircraft Mfr.
Engine Mfr.
Independent
Evaluation
Agency
Government
Agency
Supporting
Industry
Military Force
TOTAL
1
6
4
12
6
3
3
2
37
Comment
There may be some distinction in terms of emphasis on certain modeling capability between different institutions.
However, responses summarized are not cross-referenced against source in the interests of brevity. The response from
airlines and overhaul bases was disappointing.
The survey elicited specific responses from Creators of engine models and Users of engine models. Respondents
indicated their role, and whether they used models created in-house or by an external source. The Venn diagram in
Figure 1 summarizes this. There is some inconsistency e.g. one cannot be a user of an in-house model without being a
model creator!
USER
IN-HOUSE
1
EXTERNAL
3
16
10
3
2
CREATOR
USER/CREATOR DISTRIBUTION
A-3
Comment
The emphasis is firmly on aerothermal modeling of the physical processes involved, although there is some use of
database engine models for some applications. The scope of model use from the sample represents a large subset of the
main body document, therefore validating the approach taken by the working group.
COMPUTING PLATFORMS
Mainframe activities being replaced by workstation (UNIX increasingly replacing VMS) and PC networks largely
running Win95 and NT (a few cases of MS-DOS, and LINUX applications)
Portable PCs used for field activities
Real-time execution often uses specialized platforms (e.g. Harris, ADI)
The overall view in terms of platform types mentioned is summarized below:
Platform Type
Workstation
PC
Mainframe
Special
TOTAL
45
35
10
10
100
Comment
The recession of mainframe usage and ascent of PC-based applications was expected.
FORTRAN is widely used most respondents indicated some usage (50% applications)
A-4
C++ is emerging as a GUI-compatible language (to replace FORTRAN) already wide usage (~25% applications)
There was some mention of Visual Basic, Delphi etc.
MATRIXx and Matlab are front-runners in development environments, with some use of Easy5
Some use of automatic code generation (75% No, 25% Yes)
Language
Usage (%)
~60%
~25%
Beacon
~10%
Others
~ 5%
Of those who do not use automatic code generators, 15% plan to do so in future
Comment
FORTRAN remains prevalent but, as expected, there is a strong move toward GUIs, which is leading to the adoption of
compatible languages and code structures. Automatic code generators can be used where appropriate for executing
models on specific target machines.
Some respondents considered their present file-based interfaces satisfactory. User-friendliness may reduce training time,
but is perceived by some as inhibiting the experienced user.
Engine models are being issued in forms to interface with larger system models (~50% of respondents indicated a
requirement of current practice). The interfaces in this case are largely customized according to the specifics of the
application.
Comment
With GUIs and the increase of multi-disciplinary interaction, efficient interfaces between models (whether between
models and users or with other models) are extremely important.
10
The breakdown of these 26 reponses in terms of the supply methods mentioned is given below:
Supply methods
Instances
12
4
15
2
11
Special customer requirements for the way in which the model is supplied:
MS-Excel compatible output
AS 681 interfaces
ARP 4868 interfaces (emerging standard)
Engine models are supplied to:
Aircraft manufacturers
A-5
Control-system manufacturers
Simulator manufacturers and integrators
Academic institutions (research and education)
Propeller manufacturers
Missile manufacturer
Government evaluation dept.
Pipeline operators
Power utility operators
Military operators / maintainers
Propulsion system customers
Design
Performance
Controls
Air-Systems
Testing
Airframes
Propulsion research
Flight simulations
Comment
A wide exposure of engine models with highly skilled users. Adoption of recommended practices and standards is
essential for efficient cross-industry usage.
11
Application area
Operator Pass-Off / Depot Test Stand
Manufacturer Pass-Off Test Stand
Flight Line - Ground Testing
Flight Line - Flight Data
Academic / Educational
Development Testing
Preliminary Engine Design
Engine Development
Control System Design
Control System Development
Engine Selection
Aircraft / Engine Integration
Flight Simulation
Accident Investigation
Diagnostic Capability
Measurement Drift
Research
Engine Condition Monitor
Aircraft Performance Analysis
Thrust in Flight
Instances
5
5
7
7
15
14
13
9
15
13
12
15
12
9
4
1
1
2
1
1
A-6
Responses here varied even amongst similar applications. This could be taken as some having unrealistic expectations of
the type of model being used. The main body of this report should help in this respect. The general response under each
of the headings is summarized below.
11.1
CONSISTENCY
In preliminary design studies, consistency is assured, as the model is the prime source of data. Engine design and
development models will require closure with selected test data. Prediction using the calibrated model may require
better than 1% agreement at high power cases. Transient models may be required to agree within 2-3% of overall
handling times larger tolerances may be acceptable (e.g. 8%).
11.2
BANDWIDTH
Many applications were for steady-state performance, hence bandwidth was not an issue. For non-steady models, 5Hz is
considered adequate for most engine handling and operability studies. Models in this case only generally consider the
heat soakage elements and rotor dynamics. Other studies may involve consideration of the gas dynamic behavior. 0-D or
1-D models will have a bandwidth of up to ~50Hz. 2-D and 3-D models demonstrate higher bandwidth.
11.3
INTEGRATION TIME-STEP
Integration time-step is linked to the dynamic behavior being modeled, and the mathematical stability of the integration
technique being used. Ideally, the integration time-step should be chosen to be consistent with the bandwidth of
simulation required. 1/20th of the wavelength of the fastest dynamic event is the usual rule-of- thumb. Thus, integration
time-steps of around 1ms are consistent with 50Hz modeling methods.
11.4
OUTPUT TIME-STEP
This is normally set at the integration time-step, as this is consistent with the frequency range of interest. However, in
some cases, output might be generated at multiples of the integration step to minimize output when the detail is not
required, or when the integration time-step is set very small for stability purposes.
11.5
RUN TIME
Pilot and Hardware-in-the-loop applications require real-time execution time. For non-real-time work, execution speed
is a matter of user convenience. Typically, 5x real-time is tolerated.
Comment
It is obvious that the model application (rather than the user) should be the prime driver of model requirements. There
appears to be a need for guidance in this area which the main body document should provide.
Modern statistical design and analysis techniques benefit from ultra-fast running (i.e. faster than real-time).
12
The distribution of the techniques relative to application on the basis of occurrence is given below.
Technique
Aerothermal
Piecewise Linear
Partial Derivative
Transfer Function
Neural Network
Fuzzy Logic
Non-Linear
TOTAL %
Steady
State
Non S-S
Real-Time
(HITL)
Embedded
28
6
4
1
3
0
0
17
31
23
2
5
2
3
0
0
14
24
11
6
4
1
1
1
20
38
17
7
8
7
3
1
1
18
11
7
6
3
1
0
0
11
21
11
5
4
4
1
0
0
10
4
2
3
2
1
0
0
5
10
4
2
3
2
1
0
0
5
27
11
8
4
2
53
100
22
7
8
6
3
47
(246 responses)
Fuzzy-logic and non-linear were specific additions by 2 separate respondents.
Points to note
TOTAL (%)
A-7
Comment
The high proportion of database techniques is suspected to have arisen by tradition and the perceived or demonstrable
incompatibility of aerothermal methods (in most cases: iterative) with certain applications.
13
Thermodynamic / Aerothermal
Varying complexity;
Largely 0D or 1D representation;
Some 2D modeling of fan exit conditions;
Shaft torsion dynamics used for helicopter rotor systems (long shafts);
Full range modeling capability fairly common usually for applications which require less thermodynamic detail
(e.g. controls design);
Deterioration modeling includes:
Blade fouling, blade erosion, tip clearance changes.
Surge and stall modeling prevalent with an even spread of response for ingestion and failure (various types)
Others mentioned:
Blow-out
Fuel ingestion
Battle damage
Solution methods
A-8
Global iterations
Comment
This section merely gives a flavor of the technical scope of modeling gas-turbine engines. The main body document
addresses the techniques and modeling areas in detail.
14
ASSOCIATED CAPABILITIES
Other capabilities of engine models in terms of how they can be manipulated are recorded below:
Capability
Instances
22
16
11
16
steady-state - 24
transient - 14
Kalman filtering;
Multi-mode engine operation;
Rubber engine studies;
Optimization of transient behavior;
Optimization of mechanical design;
Optimization of engine design based on different technology levels (aerothermodynamic and material properties) for
a given turbine blade life;
Comment
Differing types of models will exhibit different capabilities. However, this table shows an even spread of the
requirements that a single modeling technique must fulfil.
15
INITIALIZATION
Only the first was foreseen when the survey questions were created. See the comment below. Clearly, 1 must precede 2
for iterative models.
Quote: Initial guess consistency is more important than absolute accuracy
Initialization Method
Automatically at any flight case and power level
At specific flight case and/or power level
Instances
17
15
2
14
A-9
Iteration
Table look-up
Transient stabilization
10
7
17
Comment
Given the two definitions (the author really only had item 2. in mind when setting the survey), there may be some
confusion in the responses. Some entries in the table could therefore be applicable to either initialization scene. It is
clear however, that minimizing execution time, and enabling an ease-of-use is important, therefore efficient and
transparent initialization (of either type) is fundamental.
16
Models associated (in some cases the associated models were described as rudimentary) with engine models were:
Air-system;
Control-system;
Fuel-system;
Sensor systems;
Slave and auxiliary systems;
Rotor;
Propeller;
Intake;
Air-conditioning system;
Flight control-system;
Handling bleed valves;
Dynamometer;
Aircraft drag;
Exhaust;
Air-start;
Combustion;
Gearbox;
Ship hull model;
Lubrication system;
Nacelle;
Emissions;
Noise;
Heat-exchanger.
Some of these models were embedded inside the engine model. Some were separately configurable.
One respondent said that all dynamic models used the standard form:
[y,xdot] = f (x,u)
There were some instances of multi-objective optimization tools being used. Tools within specific development
environments are used for analysis and design.
Separate configuration of subsystem was considered an advantage for the following reasons:
A-10
Comment
Co-execution of different models is essential in the inter-disciplinary arena (typified by the long list of associated
models above). Separately configured models simplify departmental interfaces, and help to ensure the accountability for
subsystem models lies with the model originator.
17
OTHER SYSTEM
Now
Future
Engine geometry
15
8
8
5
8
5
Airframe model
Aircraft systems
Life-cycle cost model
Emissions/observables
Global environment
Diagnostic model
15
7
4
9
1
10
4
3
6
8
2
5
Comments
Ability to predict performance from given geometry
Aid identification of risks that are not evident from
aerothermal cycle models
Potential to reduce full-scale testing
Quote: At this point in time, it is possible that the prospect of greater interaction among groups making up the
engineering infrastructure will increase the suspicion that the intent of this trend is to reduce the number of jobs
necessary to bring the product to market. I believe that this attitude has been responsible for the resistance found
among many engineering managers to the establishment of a single modeling environment that would support both
component and full-engine applications. As with other advances in the state of engineering art, it will require a
conclusive demonstration of the value of modeling to overcome this feeling and foster the realization that effective
numerical testing facilities will promote greater engineering opportunities and control of process.'
Most said that increased model interfacing would involve a wider range of users. One respondent said that most users
would be Performance and Systems engineers (this being the main area of integration activity). Whole-system analysis
tools are used in some cases.
Comment
The table shows the extent of integration, and confirms the emergence of a greater interest in environmental and cost-ofownership issues. The quote encapsulates the symptoms of greater integration of modeling activities.
18
Quote: Validation is a continuing process that starts with checks for gross violations of the basic physical principles.
Consistency:
Direct comparison sometimes automated;
By calibration with test data (Ansyn Analysis by Synthesis);
A-11
Other points:
Must be provided with model when issued;
User often wants a greater level of confidence than the creator;
Extra evidence/information can be requested;
Evidence in some cases is flimsy.
Quote: The model producers have not spent enough time wringing out their model, little validation evidence or
documentation is provided
Comment
Validation documentation is required which is comprehensive and reassuring.
19
DOCUMENTATION
Facility
Results
PC word processor
Mainframe/workstation-based
documentation package
Automatic Documentation
Further Comments
Is the documentation supplied by the
model producer Satisfactory?
A-12
Nil returns
Comment
Model documentation needs to be comprehensive, clear, concise, timely, easy-to-access and browse.
20
STANDARDS
Gas turbine engine steady-state performance presentation for digital computer programs;
AS 755
Gas turbine engine performance station presentation for digital computer programs;
ARP 1210
ARP 1257
ARP 1420
ARP 4148
ARP 4191
Gas turbine engine performance presentation for digital computer programs using FORTRAN 77;
ARP 4868
Application programs requirements for the presentation of gas turbine performance on digital computers;
AIR 4548
A-13
Many creators did not complete this section or provided little information. This could be interpreted that there are those
who are either unaware of the standards being applied or are unaware of any standard being applicable. It may also have
been too time consuming to look them up. Adherence to standards is essential especially with closer integration of
engineering disciplines.
21
FURTHER COMMENTS
One return stated that the Controls community used the models for design and development. Most models were
generated by Controls but lately the Controls dept had integrated subsystem models generated by external and internal
suppliers.
One return stated that a manual on engine simulation theory and practice would be of great interest.
22
GENERAL CONCLUSIONS
The survey, although not particularly rigorous in the analysis, has captured the current practice across a broadly
representative slice of the industry.
It is possible to see the extent of the task in providing a single modeling methodology to satisfy all application areas.
It identifies the issues that need careful consideration before modeling is fully exploited to significantly reduce
component testing.
The survey response validates the foresight of the working group in selecting the content of the main body
document.
A-15
1. Contact details
NAME
POSITION
TELEPHONE
FAX
E-MAIL
COMPANY / ORGANISATION ADDRESS
2. Outline of business
(e.g. engine manufacturer, aircraft simulation etc.)
Do you :
create engine performance models
use engine performance models
use engine models supplied by an external source
Please return completed survey form to :
NAME
TEL
FAX
ADDRESS
representing :
Advisory Group for Aerospace Research and Development
PEP Working Group 29
7, rue Ancelle
92200 Neuilly-sur-Seine
FRANCE
A-16
5. Computing platforms
Describe current and planned systems including operating system, network and user interface details.
(e.g. mainframe, workstation, PC, network cluster, parallel machine etc.)
languages
development environments
Indicate relative usage if you use more than 1 language or toolset (cover current and planned situations)
Are automatic code generation tools used ? if so give details concerning their nature and application
7. Input / Output
Briefly outline the type of i/o facilities currently used in conjunction with engine models.
(e.g. GUI, labelled COMMON blocks, free-format input data, telemetry requirements etc.)
Are the current interfaces satisfactory? if not, describe the type of interface which would better suit your
needs.
Is there a requirement (current or future) for engine models to be supplied in a high-level form (for
example, as a user-code block within a proprietary modelling environment e.g. MATRIXx) rather than in
the general, lower-level form defined by current Aerospace Standards.
A-17
YES / NO
ALL
What is the technical background of those using engine models?
- range of disciplines (e.g. Design, CFD, Performance, Controls, Air-Systems, Testing, Airframes etc.)
- skill level of these users (e.g. engineer, technician - indicate relative usage)
Application area
Operator Pass-Off / Depot Test Stand
Manufacturer Pass-Off Test Stand
Flight Line - Ground Testing
Flight Line - Flight Data
Academic / Educational
Development Testing
Preliminary Engine Design
Engine Development
Control System Design
Control System Development
Engine Selection
Aircraft / Engine Integration
Flight Simulation
Accident Investigation
Other - please specify
A-18
- output timestep
- integration timestep
- run time (relative to real-time)
- other requirements (e.g. diagnostic capability)
aerothermal
(cycle model)
linearpiecewise
partial
derivative
transferfunction
neural net
steady-state
non
steady-state
real-time
(H.I.T.L)
embedded
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
S/A
Synthesis is a general term referring to the prediction of engine performance in steady and non-steady
conditions
Analysis is a general term referring to the evaluation of engine and component performance and
includes diagnostic functions.
Steady-state refers to the ability to model nominal performance as 'single-shot' cases, rather than waiting
for a transient to stabilise (i.e. no time-base)
Non-steady-state : The often vague and confusing distinction between 'transient' and 'dynamic' models
is avoided by the use of 'non-steady-state' which encompasses any model which is run over a time base.
H.I.T.L : Hardware-in-the-loop
Embedded refers to real-time models used in controllers as observers or predictors
The distinction is made between Linear-Piecewise models and Partial-Derivative models to separate
the conventional state-space (matrix) implementation of the former, from more customised forms of
models using partial derivative terms.
A-19
acceleration
twist (e.g. long shafts)
volume dynamics - continuity of :
mass
momentum
energy
low Mach no. assumptions used in volume dynamics
thermal dynamics (heat soakage)
transient corrections to steady-state characteristics
component interactions
steady-state thermodynamics
real gases (i.e. non-ideal)
water/methanol injection
humidity effects
inlet distortion
combustion chemistry options
emissions
deterioration
mechanical interactions (e.g. tip clearance)
partial mixing of gas streams
modelling range
overspeed (beyond normal controlled overshoot)
sub-idle (starting range)
windmilling
modelling complexity
1 dimensional modelling
2 dimensional modelling
3 dimensional modelling
Please expand on any of the items above as appropriate :
11.2 Abnormal engine operation
Are any of the following modelled ?
compressor surge and stall
rain/hail/ice ingestion
control failure
shaft breakage
other abnormality (specify)
further comments :
11.3 Solution methods
notes
A-20
Outline the following, indicating as necessary where and how they are applied.
- integration approach and methods (e.g. explicit / Runge-Kutta 4)
steady-state
transient
13. Initialisation
Are models initialised ....
automatically at any flight case and power level
at specific flight case and/or power level
(in this case the model may be flown to required initial condition from a datum condition
e.g. SLS/ISA/max-rating)
Is automatic initialisation considered a benefit ?
How is initialisation achieved ?
back calculation
iteration
table look-up
transient stabilisation (time dependent)
pick-up of stored initial parameters
(e.g. end-point of previous run)
further comments:
A-21
Now
Future
When?
Why?
engine geometry
(e.g. CAD)
component models
(aero-mechanical - specialist component design area origin)
aircraft model
(detailed - e.g. used for engine/airframe interaction studies)
airframe model
(basic - e.g. for use in mission analysis and flight simulation)
aircraft systems
A-22
17. Documentation
CREATOR
As a creator, what facilities are used to produce model documentation:
notes
PC word processor
automatic documentation
(using tools integrated with modelling environment)
USER
Is the documentation supplied by the model producer satisfactory ? if not, describe the documentation
that would better suit your needs
18. Standards
ALL
List relevant computing/modelling standards and guidelines observed.
(e.g. SAE interface standards and guidelines, Software quality control standards)
G-1
Glossary
Term
Chapter
Description of Term
0-D model
Chapter 2
1-D model
Chapter 2
2-D model
Chapter 2
3-D model
Chapter 2
A8
Accuracy
Adaptive model
Additive drag
Adiabatic
Adjusted
performance
Adverse weather
Chapter 2
Chapter 5
Chapter 2
Chapter 4
Chapter 3
Chapter 2
Aero acoustics
Chapter 4
Aerodynamic
forcing functions
Aero-elastics
Chapter 2
Aerothermodynamics
Afterburning,
Reheat, Post
Combustion,
Augmentation
AIR
Airframe designers
Chapter 2
A model which presents the gas conditions at discrete stations along the engine
and where no physical length is implied
A model which presents the gas conditions at stations along the engine where the
longitudinal location is defined in units of length
A model, which presents the gas, conditions at stations along the engine and
provides profile information either in the radial or circumferential sense (usually
radial).
A model, which presents the gas, conditions at stations along the engine and
provides profile information both radially and circumferentially.
Exhaust nozzle throat area
The measure of a model's ability to replicate the true physical entity
A model which adjusts itself to a set of observations
Pressure force on external stream tube surface in front of inlet
A process which occurs without loss or gain of heat
The performance, which is a result of trimming, engine inputs following an
analysis process - automated or manual.
Phenomena, such as clear air turbulence, thunderstorms, and low altitude wind
shear that may affect safety of flight on each route to be flown and at each airport
to be used.
The study of sound transmission through the air, in terms of the effects of
environmental noise from machinery, vehicles, aircraft.
Force on blade(s) due to aerodynamic flow over the blade; in conjunction with
blade vibration analysis
Coupled motion of solid surfaces due to elasticity of solid materials and
aerodynamic forces.
The analysis of aerodynamic phenomena at high gas speeds incorporating the
essential thermodynamic properties of gas into the examination.
Addition of fuel & combustion after the last turbine to provide additional thrust
Chapter 2
Chapter 4
Chapter 2
Chapter 2
Introduction
Airframe
propulsion
integration
Analog simulation
Chapter 2
Analysis
Chapter 2
Annular cascade
AnSyn
Chapter 4
Chapter 2
ANSYN Factors
(matching)
API
Chapter 3
Application
software
APU (Auxiliary
Power Unit)
ARP
AS
Average Engine
Chapter 2
Chapter 5
Chapter 5
Chapter 3
Chapter 2
Chapter 2
Chapter 3
G-2
Average passage
Chapter 2
Axi-symmetric
model
Bandwidth
Beta lines
Chapter 3
Chapter 4
Chapter 4
Blade geometry
Chapter 4
Blade loading
Chapter 2
Chapter 4
Blade surface
roughness
Blade untwist
Chapter 4
Bleed flows
Chapter 4
Body forces
Chapter 2
Buzz
Chapter 4
Bypass ratio
Chapter 2
CAD
Chapter 5
Calibration
Camber
Casing treatment
Chapter 2
Chapter 4
Chapter 2
Cause-effect
relationship
CFD
Chapter 3
CFD
turbomachinery
Choked nozzle
operation
Class
Chapter 2
Clean inlet
performance
Closed loop
control
Cold flow rigs
Chapter 2
The range of input frequency over which a model gives valid results
Set of external lines draw on a compressor map to aid in the construction of
appropriate tables that describe compressor performance in a numerical simulation
Turbomachinery blade description in terms of blade shape, stagger, lean, inlet and
exit metal angles, camber, thickness, solidity, etc
The work capacity of a stage; Loading is increased as angle of attach is increased
until flow separation occurs
Stacking of steady state blade performance to provide overall compressor
performance
Surface roughness that can effect aerodynamic and thermodynamic performance
over a compressor or turbine blade
Compressor blades are usually twisted from hub to tip to obtain an optimum
angle of attack or loading at all radii; Aerodynamic forces can untwist the blade
while in motion
Airflows to or from a component used for cooling or external air condition within
the aircraft
Any external force that act on a volume element of a body and is proportional to
the volume, such as gravity force
Usually associate with the inlet where the shock is moving back-and-forth in the
throat causing a "buzzing" sound
The ratio of cold stream mass flow-rate to hot stream mass flow rate in a turbofan
engine
Ratio of the amount of air that bypasses the core of the engine to the amount of air
that passes through the core
Computer Aided Design. CAD programs help engineers in the design processes.
For example, CAD programs can help design gas turbine components and check
compatibility of parts in an assembly. CAD tools often have advanced
visualization capabilities to show the geometry of parts and assemblies.
The process of adjusting a model to replicate a specific set of data
The difference in the inlet and outlet blade angles
Refers to compressor or fan tip casing refinements to increase stall margin; may
be in the form of circumferential grooves or cross blade slots
An empirical or computer-generated exchange rate of an input to an output of a
physical process
Computational fluid dynamics - a numerical simulation technique usually solving
3D viscous equations
Computational Fluid Dynamics simulation using Navier-Stokes equations for
turbomachinery applications
An operating point corresponding to the maximum mass flow-rate through the
nozzle.
A class defines the structure of an object in object oriented software code. A
synonym for class would be object type.
No inlet distortion present
Chapter 3
Chapter 4
Combustion
Aerodynamic Load
Combustion
Efficiency
Combustor
blowout
Combustor
primary zone
Combustor relight
Compact engine
Chapter 4
Chapter 2
Chapter 2
Chapter 5
Chapter 3
Chapter 4
Chapter 2
Chapter 2
Chapter 2
Chapter 2
Part of combustor where combustion is stoichiometric (i.e. all fuel is burned with
100@ O2)
Re-ignition of the primary combustor
A simplified engine model
G-3
model
Complexity
Component
characteristics
(CHICS)
Component level
cycle code (CLM)
Component
matching
Compressor axial
gap
Compressor
recovery
Computer platform
Chapter 3
Introduction
Chapter 2
Chapter 2
Chapter 4
The process of integrating engine components such that each component operates
at the appropriate operating point, working line or trajectory
The axial spacing between rotor and stator blades of a compressor
Chapter 2
Chapter 5
The computer platform is the combination of the hardware and software needed
for gas turbine performance calculations. The hardware is the physical part of the
computer; the operating system represents the software required to use the
hardware with application software such as gas turbine simulation programs
Management of a computer program either by software or by a manual process
that sets version control and checks out code to users and checks it back in
without stepping on the work of others
The measure of a model's ability to replicate the reference database
Process by which the engine is controlled; usually imbedded within an on-board
computer
Feedback loops in a control system. Deviation from a desired output is detected,
and an input related to the difference is applied to reduce the deviation
An imaginary boundary encompassing a component which allows it to be
considered as a gross entity
The 'homing' onto a solution by iteration
Type of exhaust nozzle to accelerate the gas flow to supersonic speeds
Configuration
management
Chapter 5
Consistency
Control logic
Chapter 5
Chapter 2
Control loops
Chapter 2
Control volume
Chapter 4
Convergence
ConvergentDivergent nozzle
CORBA
Chapter 5
Chapter 4
Chapter 4
Core flow
Corrected
parameters
(referred
parameters)
Cowl lip
Cradle drag
Customer bleed
Cycle decks
Cycle match model
Data base
Chapter 2
Chapter 3
Data validation
Debugging tools
Chapter 2
Chapter 5
Chapter 2
Chapter 3
Chapter 3
Introduction
Chapter 2
Chapter 5
Deck
Degree of reaction
Chapter 4
Design
Chapter 2
Design and
verification
Chapter 2
Common Object Request Broker Architecture. CORBA provides an objectoriented approach to writing distributed applications. Distributed applications and
CORBA would enable integral gas turbine simulations using simultaneous
execution of different programs simulating separate engine modules, on separate
computers at separate locations
Fractional flow that runs gas generator of the turbofan engine; hot stream flow
Turbomachinery operating point characterizing parameters corrected by inlet
thermodynamic conditions to eliminate the dependence (sometimes referred to as
non dimensional parameters (mass-flow, rotational speed, pressure ratio )
although not truly non-dimensional)
Front part of an inlet; usually where an oblique shock develops
Loads, induced by parasitic airflows, on the engine installation in the test cell
External bleed generally used for aircraft air conditioning
An engine model/program (usually 0D type) using the cycle match technique
A model using iteration to achieve flow compatibility between engine components
A collection of organized data, usually residing in a number of files in a computer
system. A database can be as simple as a shopping list or as complex as a
collection of thousands of sounds, graphics, and related text files.
The process by which input (and output) data is checked for correctness
Computer programs that help find errors in computer program code. Often,
debugging tools are integrated in the development environment. Modern
debugging tools enable computer programmers to monitor the execution of the
program, line-by-line in the program code while being able to query all relevant
program parameter values.
Originally describing the set of punched cards comprising a computer program this term is still used to refer to digital computer models.
A measure of the extent to which the rotor contributes to the overall static
pressure change in a turbomachinery stage
1) An operating point corresponding to the design values of mass flow-rate,
pressure ratio and rotational speed in a turbo-machine
Also known as the Program Definition and Risk Reduction phase;
G-4
Deterioration
Deterministic
Chapter 5
Chapter 3
Development and
validation
Development
environment
Chapter 2
Development
process
Introduction
Development
testers
Introduction
Diagnosis mode
Diagnosis
techniques
Digital simulation
Chapter 5
Chapter 2
Direct numerical
simulation
Discretization
Displacement
pumps
Dissociation
Chapter 4
Distortion
(circumferential
and radial)
Distributed
computing
Chapter 2
DLL (Dynamic
Link Library)
Chapter 5
Dry/wet operation
(Afterburner)
Durability
Chapter 3
Dynamic model
Effective nozzle
area ratio, A9/A8
EGT (Exhaust Gas
Temperature)
Electro hydraulic
servo units
Embedded engine
model
EMD (Engine
Manufacturing
Development)
Introduction
Chapter 4
Emissions index
Empirical model
End wall effects
Chapter 4
Chapter 2
Chapter 4
Engine aging
Engine altitude
facility
Chapter 2
Chapter 2
Chapter 4
Chapter 2
Chapter 3
Chapter 4
Chapter 4
Chapter 5
Chapter 2
Chapter 2
Chapter 4
The flow weighted mean total temperature of the working fluid at a plane
immediately downstream of the last turbine stage
Hydraulically powered servo units that are electrically controlled
Chapter 2
Chapter 2
Life cycle phase that translates the most promising design approach into a stable,
interoperable, producible, supportable, and cost-effective design; validate the
manufacturing or production process; and demonstrate system capabilities through
testing
The ratio of mass of pollutants to unit mass of fuel
A representation of an engine set up in terms of a priori observations
The boundary layer formation effects at the hub and casing of a turbo-machine
causing pressure losses and blockage
cf. deterioration
Wind tunnel whose working section can simulate altitude conditions of pressure,
temperature and humidity,
G-5
Engine anomalies
Engine component
model
Engine condition
monitoring
Engine
configurations
Engine controls
Chapter 2
Introduction
Chapter 2
Engine cycle
Chapter 2
Engine design
process
Engine
deterioration
Engine fleet
management
Engine health
Introduction
Engine model
Engine operators
Engine simulation
Engine to engine
scatter
Chapter 2
Introduction
Introduction
Chapter 2
Environmental
effects
EPR (Engine
Pressure Ratio)
Euler equations
Evaluation testers
Chapter 3
Activity of analyzing operation data for aircraft engines to ensure engine safety by
adapted repair and maintenance planning
Quantified performance engine status described by component characteristics
difference relative to a status
A set of thermodynamic assumptions representing an engine
People or company involved in engine operation / usage
A computer implementation of a model run to give the time response of an engine
Diversity of engine behavior due to:
component characteristics variance, generated by the bill of material tolerance
transducer / controller accuracy causing dispersion of component throttling and
command
Extraneous effects on an engine imposed by environmental conditions.
Chapter 2
The pressure ratio of an engine cycle available to the turbine and nozzle
Chapter 2
Introduction
Event driven
Chapter 5
Exhaust nozzle
external loads
Factors, deltas
(adders, scalars)
FADEC, DECU
FAR (Fuel Air
Ratio)
Fault detection
FHV (Fuel
Heating Value),
LHV
Finite element
model
Chapter 4
Chapter 3
Chapter 2
Chapter 2
Chapter 5
Chapter 2
Chapter 3
Chapter 2
Chapter 2
Chapter 4
Chapter 2
Chapter 3
Chapter 2
Finite rate
chemistry
Flame-holder
Chapter 2
Chapter 2
Flight simulators
Chapter 2
Chapter 4
G-6
Flow coefficient
CD
Flowpath
Fluid dynamic
blockage
Chapter 4
Flutter
Chapter 4
FOD
FORTRAN
Chapter 2
Chapter 2
Free vortex
Chapter 4
Frozen
Chapter 4
Functionality
Fundamental
pressure losses
Gas generator
Gas path analysis
Gas properties
effects
Gas sampling
Chapter 3
Chapter 3
Global iteration
Global system
level analysis
Grid generation
Chapter 3
Chapter 4
Groaning, screech,
organ noise,
chugging, and
growl
Grooves
Gross thrust
Chapter 4
GUI (Graphical
User Interface)
Gutter
Hardware in the
loop
Chapter 2
Chapter 4
Chapter 3
Chapter 2
Chapter 4
Chapter 3
Chapter 5
Chapter 4
Chapter 2
Chapter 4
Chapter 4
Chapter 2
Chapter 2
Chapter 4
Chapter 2
Ideal thrust
Chapter 4
Idle
IGV & VSV
Chapter 2
Chapter 2
Chapter 3
Chapter 3
Chapter 3
Chapter 4
cockpit.
Nozzle discharge coefficient; Ratio of actual flow rate to ideal flow rate
The path that the working fluid follows during the flow through a machine
The available flow area excluding the flow blockage due to boundary layer
growth or separation in a duct, nozzle, and intake or within a compressor. ( Net
flow area / Total area )
A self-induced (flow induced) oscillating motion of improperly designed fan or
rotor blades
Foreign Object Damage, damage caused by ingestion of external material
FORmula TRANslater. One of the earliest third generation programming
languages with origins going back to the 1950s. FORTRAN is the traditional
computer language for the scientific community and the majority of gas turbine
simulation code to date is implemented in FORTRAN
Flow with concentric circles in which there is no change of total energy per unit
weight with radius.
When the time for a change in state of a chemical process is shorter than the
relaxation time, than the gas is said to be frozen at a fixed composition.
Logical process action expected from the system
The stagnation pressure drop in a combustion chamber associated with the rise in
the temperature due to combustion.
Compressor-burner-turbine; the internal gas path power cycle
See Analysis
Effects of gas properties on a thermodynamic process
Specific process of chemical analysis by gas extraction for combustion
efficiency, emissions and temperature assessments
An iteration loop around a complete engine model
Overall characterization of the system described by sole inputs/outputs
relationships
Division of 2 or 3 dimensional flow domain into finite volumes. For CFD, grid
generation is required to divide the space in which the flow is analyzed in small
grid elements: spatial discretization
Colorful ways of describing unsteady gas path behavior; Usually associated with
Combustion
G-7
Chapter 4
Chapter 4
Chapter 2
Chapter 2
Chapter 4
Stall cells that initiate rotating stall; may be multiple cells prior to full stall
The flow of a fluid where the changes in density with other thermodynamic
parameters is negligible (low Mach number flows).
Thrust of an engine while in flight; fully installed
Partial derivatives
Chapter 2
The process of setting up a model such that the initial conditions are as required
(often at steady-state)
Inlet airflow area
The ability of the inlet and the engine to interface for prolonged periods without
interference under prescribed environmental conditions
The ability of the inlet and the engine to interface for prolonged periods without
interference under prescribed enviromental condtions
Intake loses expressed in terms of the ratio of total pressure at the compressor
inlet to that defined in front of the intake
Drag due to more air
The inlet normal shock is not at the minimum area and may be expelled out the
front of the inlet
A parameter which perturbs a system e.g. fuel flow, intake conditions
Technical and commercial activities involving the engine operators, with a
technology to operate satisfactorily the engine population provided by engine
manufacturers
Engine performance with all external bleeds and power extraction activated
Chapter 5
Chapter 3
The entry duct into an engine, which may be used to induce compression
The process of estimating the time-dependent behavior of state variables
Chapter 4
Chapter 3
Airflow pathways between the rotors and disks used for transferring cooling air
between compressor and turbines
Chapter 2
Chapter 3
Chapter 2
Chapter 4
Chapter 2
Chapter 4
Chapter 2
Input
In-service support
Chapter 2
Chapter 2
Chapter 4
Installed
performance
Intakes
Integration
(implicit &
explicit)
Integration time
step
Internal air system
(secondary air
system)
Iron bird
Isentropic
IT&E
IT&E
Chapter 2
Iteration
Chapter 5
Jacobian matrix
Kalman filtering
Chapter 3
Chapter 2
Kernel
Labyrinth seals
Labyrinth seals
windage
Large eddy
simulation
Legacy system or
code
Chapter 4
Chapter 4
Chapter 4
Level of detail
Life assessment
Life cycle
Lifing model
Chapter 3
Chapter 2
Chapter 2
Chapter 2
Lifing model
Chapter 2
Linear cascade
Chapter 4
Chapter 4
Chapter 5
G-8
Linearization
Chapter 2
Liner cooling
Local equilibrium
Chapter 4
Chapter 4
Chapter 3
in rig test
The process by which the partial derivatives which characterize the dynamic
behavior of a system are derived
Airflow used to cool combustor or augmentor liner metal temperature
The condition of having thermal, mechanical and chemical equilibrium at a
specified point in a thermodynamic system.
Type of combustion process; fuel is premixed and already vaporized; very lean
Chapter 2
Chapter 3
Chapter 4
Chapter 2
Manufacturers
Introduction
Manufacturing
tolerances
Maps (CHICS)
MAR (Moving
Actuator Ring)
Mass flow
function
Mass, momentum,
energy
conservation
Master/Slave
petals
Mathematical
engine model
Max AB
Mean line , rowrow model
Measurement
uncertainty
Mil power
Mil specification
Min AB
Minimum engine
Chapter 2
Pipes, pumps, and controls for the oil lubrication process and the fuel delivery
process
Linear Variable Differential Transformer - a type of displacement transducer
Human-piloted action for generating demands to the hardware by visual
monitoring of states engine/plane indicators connected to the simulation loop
company in charge of a product (from design through development to
manufacturing and certification) with specified technical use
Range of acceptable characteristics described in the bill of materials
Chapter 2
Chapter 4
Chapter 4
Chapter 4
Chapter 4
Introduction
Chapter 2
Chapter 2
Chapter 3
Chapter 2
Chapter 3
Chapter 2
Chapter 3
Mixer
Chapter 4
Mixer efficiency
Model
assumptions
Model creators
Chapter 4
Chapter 3
Model fidelity
Model user
Chapter 2
Introduction
Monte Carlo
Chapter 2
Moore's law
Chapter 5
Multi disciplinary
Multi disciplinary
Multi stream
model
Chapter 2
Chapter 2
Chapter 3
Introduction
G-9
Navier-Stokes
equations
Net thrust
NGV
NH, N2, XNH
NL, N1, XNL
Non-recoverable
stall
Nox
Nozzle area ratio
NPR
Chapter 2
Chapter 2
Chapter 2
Chapter 2
Chapter 2
Chapter 2
Chapter 4
Chapter 4
Chapter 2
NPSS
Chapter 5
Numerical
optimization
Numerical stability
Object
Chapter 2
Chapter 3
Chapter 5
Object oriented
Introduction
Off design
On-board engine
performance
One stream model
Chapter 2
Chapter 2
Chapter 3
Chapter 2
Operating Point
Chapter 4
Operating System
Chapter 5
Operation
Introduction
Output
Over under
expansion
Chapter 4
Chapter 3
Parallel
compressor
Parallel computing
Chapter 2
PC-based
Introduction
PDF (Probability
Density Function)
Perfect gas
Chapter 4
Performance
Performance
seeking control
Physical model
Chapter 3
Chapter 2
Chapter 2
Chapter 4
Chapter 4
G-10
Piecewise linear
PLA
Plausibility check
Polytropic,
isentropic
efficiency
Chapter 2
Chapter 2
Chapter 5
Chapter 4
Portability
Chapter 5
Post certification
Chapter 2
post stall
Power balance
Chapter 2
Chapter 2
Power off-take
Power turbine
Chapter 2
Chapter 4
Preliminary design
Chapter 2
Pressure loss
Pre-swirl system
Chapter 3
Chapter 4
Profile losses
Programming
languages
Chapter 4
Chapter 5
Chapter 3
Chapter 2
Ram drag
Reactive chemistry
Chapter 2
Chapter 4
Real gas
Chapter 4
Real-time engine
models
Relaxation
Research
Introduction
Reverse flow
combustor
Reverse flow
compressor
Reynolds number
Chapter 2
Chapter 4
Rig-engine effects
Chapter 2
Risk reduction
Chapter 2
RNI (Reynolds
numbers Index)
Robustness
Chapter 3
Rossby number
Rotating distortion
Rotating holes
Rotating stall
Chapter 4
Chapter 4
Chapter 4
Chapter 2
Chapter 3
Introduction
Chapter 2
Chapter 2
G-11
Rotor moments of
inertia
RQL (Rich
Quench Lean)
Rumble
Chapter 2
Chapter 3
Safe operational
limits
Sauter mean
diameter
Schedule
Chapter 3
Screech
Chapter 4
Scrubbing
Secondary air
system
Chapter 4
Chapter 4
Sensors
Chapter 3
Separation
Chapter 4
Shell
Chapter 5
Shock boundary
layer interactions
Chapter 4
Shock wave
Chapter 2
Signal noise
Chapter 2
Simulation mode
Chapter 5
Chapter 2
Chapter 2
Specific fuel
consumption
Specific thrust
Splitters
Spool down
Stage by stage
Stagger
Stall
Chapter 2
Chapter 2
Chapter 3
Chapter 2
Chapter 2
Chapter 4
Introduction
Stall line
Chapter 2
Stall margin
Chapter 2
Stalling pressure
rise
Standard
atmosphere
Chapter 4
Chapter 4
Chapter 4
Chapter 2
Chapter 5
Chapter 3
The ratio of the sum of all volumes of droplets in a spray to the total surface area
of all droplets.
Refers to a variation of a parameter or hardware (such as fuel flow or variable
vane position) as a function of some measure of flight condition (such as burner
pressure for fuel flow and rotor speed for variable vane position)
Description of a combustion instability which sounds like a "screech" due to its
high frequency (400 Hz) content pressure perturbation
Cleaning of exhaust gases
Airflow system used to bring relatively cool air to hotter parts of the engine such
as turbine blades; Airflow system uses cavities within the rotor shafts and bladed
disks to transfer cool air to hotter parts.
Instrumentation to provide information to control logic for proper engine
operation
A situation occurring when the streamlines cannot follow the contour of the body
because of the adverse pressure gradients.
A computer program that provides an external interface for another program. This
may be a operating system shell (such as the various UNIX shells) that provide
enhanced user interfaces, or shells that provide a GUI around a particular
application program.
Transonic or supersonic flow interacting with the boundary layer resulting in
shocks and boundary layer separations and causing losses in intakes nozzles or
turbomachinery.
A pressure wave passing through a fluid medium in which the pressure, density
and particle velocity undergo drastic changes
Parasitic measurement perturbation induced by measurement system (acquisition
electronics)
A mode of operation of a model where measured data is used as input. Frequently
relates to models that can describe transient behavior
Ambient conditions represented by air at a temperature of 15 C (59 F), a
barometric pressure of 1.01325 bar (14.696 psia) corresponding to average sea
level atmospheric conditions at middling latitudes
The flow weighted mean total temperature of the working fluid at a plane
immediately upstream of the first stage turbine rotor blades. (See TET)
A measure of the ability of a model to resolve differences in properties in terms of
physical proximity
Fuel consumption per unit thrust or per unit specific work output
Net thrust per unit mass flow-rate of air inflow
Generally a flow splitter between the fan and the high pressure compressor
The deceleration of a shaft rotor usually due to combustor extinction
Modeling each stage of turbomachinery as a separate thermodynamic process.
Angle between the blade camber line and the axial direction
Compressor instability; Sometimes synonymous with Surge. May be rotating stall
or non-recoverable stall
Locus of stability points for all speeds beyond which the compressor will stall or
surge
Measured at a constant airflow rate; the amount of pressure rise available between
the stall line and the operating line: stall margin can be affected by inlet distortion,
engine-to-engine tolerances, deterioration, and clearances
The pressure rise @ constant corrected airflow that will cause the compressor to
stall
Ambient conditions represented by air at a temperature of 15 C (59 F), a
barometric pressure of 1.01325 bar (14.696 psia)
2).ISA corresponds to average values of temperature and pressure at middling
G-12
Standards
State space model
State variables
Steady-state
Stepped labyrinth
seals
Straight model
Chapter 5
Chapter 2
Chapter 2
Introduction
Chapter 4
Stratification
Stream line
curvature code
Sub idle
Surface cooling
Chapter 4
Chapter 2
Chapter 2
Chapter 4
Surge
Introduction
Surge Cycle
Surge line
Chapter 3
Chapter 2
Swirl losses
Swirling flow
Synthesis
System
identification
T&E (test and
evaluation)
Tappings (bleed
ports)
Temporal
resolution
Test cell
Chapter 4
Chapter 4
Chapter 2
Chapter 2
Chapter 2
Thermal
management
system
Thermodynamic
parameters
Through flow code
Thrust
Chapter 3
Chapter 2
Chapter 2
Chapter 4
Chapter 5
Chapter 2
Chapter 2
Introduction
Chapter 2
Chapter 2
Thrust coefficient
Cg
Thrust vectoring
Time average
equations
Time between
overhauls
Chapter 4
Time lag
Tip clearances
Chapter 3
Chapter 4
Chapter 2
Chapter 2
Chapter 4
Chapter 3
Chapter 3
latitudes.
Technical requirements expressed by certifying organizations
A linear representation of a dynamic system
Parameters in a model whose rate of change is defined by a dynamic equation
The description of a system where there are no unbalanced forces or energies
A type of seal design
The model that satisfy the engine matching equations for a given component
characteristics
Separation into layers
A type of 2D turbomachinery code which uses blade geometry, correlations of
blade loss and flow exit deviation to determine blade steady performance
Engine power at which the system is in start up mode
Associated with cooling across a metal surface by convection using cooler airflow
from some source within the engine
Compressor instability; Violent reversing of flow within the compressor; Usually
3-to-15 Hz
Reoccurring compressor surge with periods of recovery; Cyclic in nature 3-15 Hz
Locus of stability points for all speeds beyond which the compressor will stall or
surge
Pressure losses in a duct or turbo-machine due to the swirl component of flow.
Flow that has a circular motion on top of its principal direction
The generation of a prediction based on a collection of component assumptions
A process by which the transfer function of a dynamic process can be derived by
observing its outputs
Strong coupling between modeling and simulation technology with experimental
data during the development process
Holes or slots in the casing for extraction or introduction of air from some other
source
Dealing with time stepping or time domain
Installation embedding the engine or component to reproduce their actual
theoretical environment for characterization of global and detailed behaviors
1) The flow weighted mean total temperature of the working fluid at a plane
immediately downstream of the last stage turbine rotor blades. 2) The flow
weighted mean total temperature of the working fluid at a plane immediately
upstream of the first stage turbine rotor blades.
The ratio of the net power output to the heat consumption based on the lower
heating value of the fuel
A design that controls the temperature of a component or region
Parameters defining the state of the working fluid during the engine cycle, such as
temperature, pressure, enthalpy, entropy etc.
Another name for streamline curvature or meridional type codes
Unbalanced force caused by the pressure forces across and the difference in the
momentum of air entering and the exhaust gasses leaving a gas turbine engine
Ratio of actual thrust to ideal thrust
Off centerline axis thrust produced by a vectoring exhaust nozzle
Equations that do not consider fluctuating pressure perturbations - mean flow type
equations
Time spent on wing between required major planned maintenance action on
engine component to restore hot parts temperature margin - relative to the
declared red-line
A time-based delay or skew
Physical distance between the rotor tip and the casing; generally the larger the
clearances the worse the performance
The flow weighted mean total temperature of the working fluid at a plane
immediately upstream of the first stage rotor blades.
Static Conditions refer to thermodynamic properties not considering the flow
velocity. Total (or Stagnation) Conditions refer to thermodynamic conditions
G-13
Transfer function
Transient
Trending
Trim setting
Trimming
Turbine flow
capacity method
Turbine nozzle
Chapter 2
Introduction
Chapter 2
Chapter 3
Chapter 4
Chapter 3
Turbofan
Chapter 2
Turbojet
Chapter 2
Turbomachinery
matching
Turboshaft
Chapter 3
Chapter 2
Turbulence model
Chapter 4
Chapter 4
Two-spool or
Dual-Spool
Unsteady
User environment
Chapter 2
Introduction
Chapter 2
User friendly
Chapter 2
Utilities
Chapter 5
Variable area
turbine
Variable cycle
engine
Variable geometry
Chapter 4
Variable nozzle
Chapter 4
Velocity
coefficient, Cv
Velocity triangles
Chapter 4
Vitiated air
Chapter 3
Volume dynamics
Warm flow rigs
Chapter 3
Chapter 4
Weak extinction
WFB burner fuel
flow
Wind tunnel
Chapter 4
Chapter 2
Chapter 2
Any of family of devices in which fluid is pumped through duct to flow past
Chapter 4
Chapter 2
Chapter 2
Chapter 4
G-14
Windage losses
Windmilling
Chapter 4
Chapter 2
Chapter 2
Chapter 2
2. Originators References
3. Further Reference
RTO-TR-044
AC/323(AVT-018)TP/29
5. Originator
ISBN 92-837-1083-5
4. Security Classification
of Document
UNCLASSIFIED/
UNLIMITED
6. Title
9. Date
Multiple
April 2002
11. Pages
Multiple
12. Distribution Statement
356
There are no restrictions on the distribution of this document.
Information about the availability of this and other RTO
unclassified publications is given on the back cover.
13. Keywords/Descriptors
Air flow
Aircraft engines
Axial flow compressors
Calibrating
Computer systems
Computerized simulation
Design
Exhaust nozzles
Gas turbine engines
Maintenance
Manuals
Mathematical models
Methodology
Model validation
Model verification
Models
Performance modeling
Questionnaires
Reviews
14. Abstract
A Technical Team of the NATO RTO has created a manual on aircraft gas turbine simulation,
ranging from applications to latest methodology of modelling techniques. Application is from
design to operation and maintenance. Twenty two examples of models are reviewed, including
educational examples and completed engine models. Present computer platforms in use for such
models are reviewed, and an outlook on development is given. Executable models are included.
On the theoretical side recent and advanced developments of the modelling of components are
included. A survey of model users and a glossary are provided. The manual aims at increasing
the use and the value of engine computer simulations in NATO Nations and NATOs design
and use of engines.
BP 25 7 RUE ANCELLE
F-92201 NEUILLY-SUR-SEINE CEDEX FRANCE
Tel
ecopie
LOrganisation pour la recherche et la technologie de lOTAN (RTO), detient un stock limite de certaines de ses publications recentes, ainsi
que de celles de lancien AGARD (Groupe consultatif pour la recherche et les realisations aerospatiales de lOTAN). Celles-ci pourront
e ventuellement e tre obtenues sous forme de copie papier. Pour de plus amples renseignements concernant lachat de ces ouvrages,
adressez-vous par lettre ou par telecopie a` ladresse indiquee ci-dessus. Veuillez ne pas telephoner.
Des exemplaires supplementaires peuvent parfois e tre obtenus aupr`es des centres nationaux de distribution indiques ci-dessous. Si vous
souhaitez recevoir toutes les publications de la RTO, ou simplement celles qui concernent certains Panels, vous pouvez demander detre
inclus sur la liste denvoi de lun de ces centres.
Les publications de la RTO et de lAGARD sont en vente aupr`es des agences de vente indiquees ci-dessous, sous forme de photocopie ou
de microfiche. Certains originaux peuvent e galement e tre obtenus aupr`es de CASI.
FRANCE
O.N.E.R.A. (ISP)
29, Avenue de la Division Leclerc
BP 72, 92322 Chatillon Cedex
BELGIQUE
Etat-Major de la Defense
Departement dEtat-Major Strategie
ACOS-STRAT-STE Coord. RTO
Quartier Reine Elisabeth
Rue dEv`ere, B-1140 Bruxelles
CANADA
Services dinformation scientifique
pour la defense (SISD)
R et D pour la defense Canada
Minist`ere de la Defense nationale
Ottawa, Ontario K1A 0K2
DANEMARK
Danish Defence Research Establishment
Ryvangs Alle 1, P.O. Box 2715
DK-2100 Copenhagen
ESPAGNE
INTA (RTO/AGARD Publications)
Carretera de Torrejon a Ajalvir, Pk.4
28850 Torrejon de Ardoz - Madrid
ETATS-UNIS
NASA Center for AeroSpace
Information (CASI)
Parkway Center
7121 Standard Drive
Hanover, MD 21076-1320
PAYS-BAS
NDRCC
DGM/DWOO
P.O. Box 20701
2500 ES Den Haag
GRECE (Correspondant)
Hellenic Ministry of National
Defence
Defence Industry Research &
Technology General Directorate
Technological R&D Directorate
D.Soutsou 40, GR-11521, Athens
POLOGNE
Chief of International Cooperation
Division
Research & Development Department
218 Niepodleglosci Av.
00-911 Warsaw
HONGRIE
Department for Scientific
Analysis
Institute of Military Technology
Ministry of Defence
H-1525 Budapest P O Box 26
PORTUGAL
Estado Maior da Forca Aerea
SDFA - Centro de Documentaca o
Alfragide
P-2720 Amadora
ISLANDE
Director of Aviation
c/o Flugrad
Reykjavik
REPUBLIQUE TCHEQUE
DIC Czech Republic-NATO RTO
a PVO Praha
VTUL
Mladoboleslavska ul.
Praha 9, 197 06, Ceska republika
ITALIE
Centro di Documentazione
Tecnico-Scientifica della Difesa
Via XX Settembre 123a
00187 Roma
ROYAUME-UNI
Dstl Knowledge Services
Kentigern House, Room 2246
65 Brown Street
Glasgow G2 8EX
LUXEMBOURG
Voir Belgique
TURQUIE
Mill Savunma Bas,kanli i (MSB)
ARGE Dairesi Bas,kanli i (MSB)
06650 Bakanliklar - Ankara
NORVEGE
Norwegian Defence Research
Establishment
Attn: Biblioteket
P.O. Box 25, NO-2007 Kjeller
AGENCES DE VENTE
NASA Center for AeroSpace
Information (CASI)
Parkway Center
7121 Standard Drive
Hanover, MD 21076-1320
Etats-Unis
Les demandes de documents RTO ou AGARD doivent comporter la denomination RTO ou AGARD selon le cas, suivie du
numero de serie (par exemple AGARD-AG-315). Des informations analogues, telles que le titre et la date de publication sont
souhaitables. Des references bibliographiques compl`etes ainsi que des resumes des publications RTO et AGARD figurent dans les
journaux suivants:
Scientific and Technical Aerospace Reports (STAR)
Government Reports Announcements & Index (GRA&I)
STAR peut e tre consulte en ligne au localisateur de
publie par le National Technical Information Service
ressources uniformes (URL) suivant:
Springfield
http://www.sti.nasa.gov/Pubs/star/Star.html
Virginia 2216
STAR est e dite par CASI dans le cadre du programme
Etats-Unis
NASA dinformation scientifique et technique (STI)
(accessible e galement en mode interactif dans la base de
donnees bibliographiques en ligne du NTIS, et sur CD-ROM)
STI Program Office, MS 157A
NASA Langley Research Center
Hampton, Virginia 23681-0001
Etats-Unis
BP 25 7 RUE ANCELLE
F-92201 NEUILLY-SUR-SEINE CEDEX FRANCE
Telefax 0(1)55.61.22.99 E-mail mailbox@rta.nato.int
DISTRIBUTION OF UNCLASSIFIED
RTO PUBLICATIONS
NATOs Research and Technology Organisation (RTO) holds limited quantities of some of its recent publications and those of the former
AGARD (Advisory Group for Aerospace Research & Development of NATO), and these may be available for purchase in hard copy form.
For more information, write or send a telefax to the address given above. Please do not telephone.
Further copies are sometimes available from the National Distribution Centres listed below. If you wish to receive all RTO publications, or
just those relating to one or more specific RTO Panels, they may be willing to include you (or your organisation) in their distribution.
RTO and AGARD publications may be purchased from the Sales Agencies listed below, in photocopy or microfiche form. Original copies
of some publications may be available from CASI.
POLAND
Chief of International Cooperation
Division
Research & Development
Department
218 Niepodleglosci Av.
00-911 Warsaw
HUNGARY
Department for Scientific
Analysis
Institute of Military Technology
Ministry of Defence
H-1525 Budapest P O Box 26
PORTUGAL
Estado Maior da Forca Aerea
SDFA - Centro de Documentaca o
Alfragide
P-2720 Amadora
CZECH REPUBLIC
DIC Czech Republic-NATO RTO
a PVO Praha
VTUL
Mladoboleslavska ul.
Praha 9, 197 06, Ceska republika
ICELAND
Director of Aviation
c/o Flugrad
Reykjavik
DENMARK
Danish Defence Research
Establishment
Ryvangs Alle 1, P.O. Box 2715
DK-2100 Copenhagen
ITALY
Centro di Documentazione
Tecnico-Scientifica della Difesa
Via XX Settembre 123a
00187 Roma
FRANCE
O.N.E.R.A. (ISP)
29 Avenue de la Division Leclerc
BP 72, 92322 Chatillon Cedex
LUXEMBOURG
See Belgium
GERMANY
Streitkrafteamt / Abteilung III
Fachinformationszentrum der
Bundeswehr, (FIZBw)
Friedrich-Ebert-Allee 34
D-53113 Bonn
NETHERLANDS
NDRCC
DGM/DWOO
P.O. Box 20701
2500 ES Den Haag
NORWAY
Norwegian Defence Research
Establishment
Attn: Biblioteket
P.O. Box 25, NO-2007 Kjeller
SPAIN
INTA (RTO/AGARD Publications)
Carretera de Torrejon a Ajalvir, Pk.4
28850 Torrejon de Ardoz - Madrid
TURKEY
Mill Savunma Bas,kanli i (MSB)
ARGE Dairesi Bas,kanli i (MSB)
06650 Bakanliklar - Ankara
UNITED KINGDOM
Dstl Knowledge Services
Kentigern House, Room 2246
65 Brown Street
Glasgow G2 8EX
UNITED STATES
NASA Center for AeroSpace
Information (CASI)
Parkway Center
7121 Standard Drive
Hanover, MD 21076-1320
SALES AGENCIES
NASA Center for AeroSpace
The British Library Document
Canada Institute for Scientific and
Information (CASI)
Supply Centre
Technical Information (CISTI)
Parkway Center
Boston Spa, Wetherby
National Research Council
7121 Standard Drive
West Yorkshire LS23 7BQ
Acquisitions
Hanover, MD 21076-1320
United Kingdom
Montreal Road, Building M-55
United States
Ottawa K1A 0S2, Canada
Requests for RTO or AGARD documents should include the word RTO or AGARD, as appropriate, followed by the serial
number (for example AGARD-AG-315). Collateral information such as title and publication date is desirable. Full bibliographical
references and abstracts of RTO and AGARD publications are given in the following journals:
Scientific and Technical Aerospace Reports (STAR)
Government Reports Announcements & Index (GRA&I)
STAR is available on-line at the following uniform
published by the National Technical Information Service
resource locator:
Springfield
http://www.sti.nasa.gov/Pubs/star/Star.html
Virginia 22161
STAR is published by CASI for the NASA Scientific
United States
and Technical Information (STI) Program
(also available online in the NTIS Bibliographic
STI Program Office, MS 157A
Database or on CD-ROM)
NASA Langley Research Center
Hampton, Virginia 23681-0001
United States