THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS

345 E. 47th St, New York. N.Y. 10017

98-GT-343
The Sodely shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the Sodety or of its Divisions or
Sections, or printed in its publications. Discussion is [minted only if the paper Is published in an ASME Journal. Authorization to photocopy
for internal or personal use is granted to libraries and other users registered with the Copyright Clearance Center (CCC) provided
53/article or 54/page is paid to CCC, 222 Rosewood Dr., Danvers, MA 01923. Requests for special permission or bulk reproduction
should be addressed to the ASME Technical Publishing Department.

Copyright 0 1998 by ASME AD Rights Reserved Printed in U.SA

Gas Turbine Cycle Design Methodology:

A Comparison of Parameter Variation with Numerical Optimization

1111111111111111111111
BREAK

Joachim Kurzke
Engine Performance Department
Mill Munchen GmbH
MOnchen, Germany

ABSTRACT routines. This can be done for example with the Windows-based gas turbine
In gas turbine performance simulations often the question arises: What performance program GasTurb developed by the author.
is the best thermodynamic cycle design point? This is an optimization task Certainly it is not sufficient, to get a single cycle as the best solution
which can be attacked in two ways: One can do a series of parameter from the computer program. There arc two main reasons for that:
variations and pick from the resulting graphs the best solution or one can Firstly, the numerical optimization algorithm will find the optimum of
employ numerical optimization algorithms that produce a single cycle which the mathematical model only as opposed to the "true" optimum. An exotic
fulfills all constraints. cycle as result of an optimization run is mostly a hint to a deficiency of the
The conventional parameter study builds strongly on the engineering model. In such a case most probably a design constraint has been
judgement and gives useful information over a range of parameter overlooked when defining the problem.
selections. However, when values for more than a few variables have to be Secondly, it is always of interest to know about the neighborhood of the
determined while several constraints are existing, then numerical optimum solution. From a parameter study limited to the region of interest
optimization routines can help to find the mathematical optimum faster and it becomes obvious which design variables and constraints have the biggest
more accurately. Sometimes even an outstanding solution is found which impact on the result.
was overlooked while doing a preliminary parameter study. One of the advantages of numerical optimization is, that the region
For any simulation task a sophisticated graphical user interface is of where parameter studies should be performed is narrowed down
great benefit This is especially true for automated numerical optimizations. significantly.
It is quite helpful to see on the screen of a PC how the variables are
changing and which constraints are limiting the design. A quick and clear 2. PARAMETER STUDIES
graphical representation of trade studies is also of great advantage. The The results of parameter studies are normally presented as graphics. In
paper describes how numerical optimization and parameter studies are a single graph one can show for given ranges of two parameters the results
implemented in a Windows - based PC program. for several dependant quantities. In a cycle study for a single spool turbojet
Man example, the cycle selection of a derivative turbofan engine with engine, for example, with compressor pressure ratio and burner exit
a given core shows the merits of numerical optimization. The parameter temperature as design parameters, one can plot the specific fuel
variation is best suited for presenting the sensitivity of the result in the consumption over specific thrust. In the resulting carpet one can additionally
neighborhood of the optimum cycle design point. show lines for other calculated parameters as for example the turbine
pressure ratio and the turbine exit temperature, see figure I.
1. INTRODUCTION Let us assume, for example, that the design aim is a low cost turbojet
The traditional way to select the thermodynamic cycle of a new gas with a single stage turbine and an uncooled turbine exit casing. The feasible
turbine employs extensive parameter variations. For a complex engine with region of design parameter combinations can be marked easily in the carpet
many design variables this is a time-consuming task. One looks for the since it is limited by the following constraints:
optimum solution in a certain respect.
Instead of screening a wide range for the design variables with turbine pressure ratio < 3.5 single stage turbine
systematic parameter variations it is also possible to do an automatic search turbine exit temperature < 1200K uncooled turbine exit casing
for the optimum engine design with the help of numerical optimization

Presented at the International Gas Turbine & Aeroengine Congress & Exhibition
Stockholm, Sweden — June 2—June 6, 1998
This paper has been accepted for publication in the Transactions of the ASME
Discussion of it will be accepted at ASME Headquarters until September 30, 1998
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 From figure 1 one can read that with these constraints the cycle with maximum of the figure of merit) faster and more accurately. As shown
the highest specific thrust has a burner exit temperature of 1600K and a above with the turbojet example, in a parameter study with only two
compressor pressure ratio of around 16.5. variables it is easy to find an optimum solution. If there are three variables
the situation is not so clear. With more than three variables the picture may
get obscure. In complex studies the true optimum may never be found with
Compr.
32.5 I M° 7.5 Pressure the conventional parameter study.
Ratio There are many numerical optimization algorithms known from
cotale 1603
e
10 literature. They can be divided basically into two major groups: Methods
— 30
es 03CtISCO 12.5 that use gradient information and others. In the program GasTurb there is
JC
is one method from each group implemented. A short explanation how these
-"
s- 275
-
1Y-lt‘uco
algorithms work is given in the following chapters.
0 No)
25
frstolellfr
Turbine 3.1 Gradient Method
22.5 Pressure moo nl wre Them is a good example for the optimization task: A mountaineer shall
3 Ratio climb the highest peak in a certain region. He has no map and the weather
7
or. 20 005`ne °de is foggy. His only tool is an altimeter. What is he going to do? He will
to
certainly check his surroundings first and then go in the direction of the
17.5 steepest ascent In the end he will come to the top of a mountain. This is a
12 14 16 18 20 place where each step leads downwards.
Net Thrust NM
The steepest ascent may, however, lead toward a border (which is either
Figure 1: Result of a parameter study the lower or upper limit of a design variable) of the region. Then our
mountaineer will walk along the border until he reaches the place where
each step leads downwards or out of the allowed region.
However, most gas turbine cycles are much more complex than the Is that the end of the story? Not necessarily. There might be several
turbojet example discussed above. A two-spool turbofan has the following summits within the region. Our mountaineer may have found the highest
five cycle design variables: peak by chance, but he cannot be sure of that. He has to check other parts
of the region. In mathematical terms there might be "local" optimums
low spool pressure ratio besides the "global" optimum.
high spool pressure ratio Up to now we have not spoken of constraints. They are like fences. A
burner exit temperature part of the region is forbidden to our mountaineer. His task is made more
bypass ratio difficult because on his way to the summit he may have to walk downwards
fan pressure ratio for a while to avoid a forbidden region. The fences (the constraints) often
exclude the summit (where each step leads downwards) as an acceptable
There will also be more constraints than with the turbojet example: solution. They create local optima that would not exist without fences.
Constraints make the task of optimization difficult.
low pressure turbine inlet temperature < limit Let us turn to the mathematical algorithm now. The mountaineer who
uncooted low pressure turbine first makes test steps in several directions uses the "gradient strategy" as a
compressor exit temperature < limit search method. With the test steps he is looking for the partial derivatives
material limit for compressor disk avav, For each optimization variable he must do one test step before he
fan pressure ratio < limit can start his way in the "right" direction.
single stage fan After the first step uphill the local gradients will be different. The test
high pressure turbine pressure ratio < limit steps could now be repeated to find the new direction. Test steps take time,
single stage turbine however, and it is therefore better to go on in the same direction as long as
the altitude increases. Reaching a fence (violating a constraint) could be
In a more detailed study there will be even more design variables as for another reason for stopping the climb. Only then will new gradients be
example the stage numbers for the high and the low pressure turbine. It is sought. The new direction will eventually take you along a fence.
obvious that with a parameter study it will be very difficult and time The gradient search algorithm implemented in GasTurb was derived
consuming to find the optimum values for the design variables. from ref. Pl. The principle is the following (see figure 2). We begin at the
point marked "Start I", looking for the direction of the steepest gradient
3. NUMERICALOPTIMIZATION ("Direction 1"). Following this direction we walk to the highest point. Then
By the way how is the optimum defined in a mathematical sense? In a we change the direction by 90° (orthogonal). This can be done without
parameter study that question must not be answered a priori. In a numerical evaluating the local gradient. We again go for the highest point here. To
optimization, however, a figure of merit must be clearly defined before the define the third direction we use the experience from the first two directions.
calculation can commence. The figure of merit might be the specific fuel We connect the point "Start 1" with the optimum point found along
consumption of a turbofan at cruise which is to be minimized. For a fighter "Direction 2". We follow this direction again as long as altitude increases.
engine it might be that the specific thrust shall be maximized. One can also This procedure can be applied repeatedly until the search steps or the
think of a weighted combination of these parameters. changes in the "figure of merit" become very small. There is also a
When values for more than a few variables have to be determined while maximum limit for the number of optimization steps. In the example of the
several constraints are existing, then numerical optimization routines can figure the optimum is found along search direction 6 (not marked in the
help to find the mathematical optimum (i.e., the minimum respectively figure, perpendicular to direction 5).

2
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 The dashed line in the figure shows how optimization would go on, if 4. CYCLE SELECTION FOR A DERIVATIVE TURBOFAN
only local gradient information is used. With this simple strategy, the search A very common design task is to adapt an existing engine for a new
direction would change very often. application. It is quite obvious that in this case there are more constraints
than during the design of a brand new engine. In this chapter at first the
basic engine will be described and then the design variables, the constraints
and the figure of merit for the numerical optimization of a derivative engine.
Stitt 2

4.1 Description of the basic engine

Let us assume that we can start from an existing unmixed flow turbofan
engine for a business jet. This type of engine has a rather low overall
pressure ratio and a moderate burner exit temperature compared to the big
turbofim engines used on commercial airliners. The main cycle parameters
are shown in the table below.

Flight Condition 11 km / Mach 0.8

Max Climb
installed

Thrust 3.64 kN

SFC 19.7 g/kN's

Figure 2: Optimization strategy . Bypass Ratio 4.5

Up to now we have only dealt with optimization without constraints. In Burner Exit Temperature 1350 K
the figure there is a shaded zone which suggests a forbidden region. If we
use the strategy just described the search for the optimum will end at the
Overall Pressure Ratio 17.82
point "A" along "Direction 5". We cannot find the global optimum if we
Core Pressure Ratio 12
begin at "Start 1". If we begin at "Start 2", however, we will be at the top of
the hill very quickly. ISA Corrected Mass Flow 60 kg/s

3,2 Random Search

The engine configuration and the nomenclature are shown in figure 3
The second optimization strategy offered by CrasTurb is based on ref.
[3]. In an adaptive random search, random numbers are used for the
optimization variables that are concentrated around the best solution found
before. The algorithm is

with
new value for optimization variable
value of V; producing the best figure of merit
search region for variable V,
range reduction coefficient (positive integer)
distribution coefficient (positive odd integer)
random number between zero and one

To start an adaptive random search one should have a variable

Figure 3: Configuration of the basic engine
combination which ftdfills all of the constraints. At the start of the search
is 10 and lc, is 1. In one search run, the program tries (40 times the number
of Optimization Variables) random engine cycles. When all cycles have 4.2 Design variables, constraints and fiaure of merit
becnSculatSthenlwiIl be duplicated and lçwill be increased by 2. The
search region will get smaller. Mother (40 times the number of Design Variables
Optimization Variables) cycles will be calculated and then k R will be Besides the pressure ratios of the new booster and the fan there will be
duplicated again and Ic will be further increased by 2. This procedure will the bypass ratio and the burner exit temperature among the design
be repeated until all cycles for Ic R=80 have been tried. Cycles that do not variables of the growth engine. A new low pressure turbine will be required
fulfill the constraints will be ignored. while the gas generator remains unchanged. The configuration of the
growth engine will be as shown in figure 4.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The core compressor of the new engine must not necessarily be lp turbine inlet temp To <1150K
operated at the same operating point as in the basic engine. In fact that compressor exit temp T 3 < 750K
might even be impossible because doing that would require an increase in core spool speed < reference + 5%
the mechanical spool speed beyond the limits of the original design. Thus fan tip diameter <0.75m
we get as two further design variables for the derivative engine the core Max Climb thrust >4,5 kN
compressor mass flow and its pressure ratio.
It is standard practice, not to read a compressor map with given mass Figure of Merit
flow and pressure ratio, but with given corrected speed and a value for an The specific fuel consumption (SFC) for Max Climb rating is the figure
auxiliary coordinate (here called beta), see for example ref. 4. In the list of of merit which is to be minimized. This will automatically result in a low
the design variables we get instead of the compressor mass flow and its fuel consumption for cruise.
pressure ratio the two equivalent variables corrected speed and map
coordinate beta. 4.3 Mathematical model of the engine
Altogether there are six design variables for the derivative engine. A mathematical model of the growth engine requires a mixture of
design and off-design calculations. The components on the low pressure
spool will be newly designed while the core components will be operated at
some off-design condition compared to the design point of the basic engine.
We select as the cycle design point for the growth engine the Max
Climb rating at altitude. For this flight conditions the optimum values for the
design variables will be found.
The mathematical model of the engine must take into account, that the
design point efficiencies of the fan, the booster and the low pressure turbine
will change with the aerodynamic loading. For axial compressors an
appropriate conflation been published by Glassman (ref 5) and for the low
pressure turbine one can use a simplified version of the preliminary turbine
design routine from Warner (ref 6).
The efficiency and the surge margin of the core compressor will be read
from the map dependent from the values for the design variables core
compressor corrected speed and map coordinate beta.
Note that the temperature limits for T, and T o in the list of constraints
are not applicable to the Max Climb rating, but for the flight case with the
highest temperatures encountered in the flight envelope. That means, that
the numerical model of the engine must be capable to simulate both the
Figure 4: Configuration of the growth engine Max Climb flight case at altitude (as a cycle design point) and the Take Off
rating for the "hot day" (LSA+15K) at sea level, Mach 0.2 (as an off-design
condition).

5. OPTIMIZING THE GROWTH ENGINE

Constraints
There are several constraints for the new engine design to be observed 5.1Ranaes for the design variables
The common core with the basic engine requires that both high pressure Another argument for setting the range of a design variable is, that
turbines have practically the same flow capacity. We want the Mach either the lower or the upper limit represents a true limit for the engine
number at the core exit also to be nearly the same and that has the design. In our example this is the case for the pressure ratio of the single
consequence, that the flow capacity of the low pressure turbine must also stage fan which is introduced with an upper limit of 1.9. In the table below
be very similar between both engines. As a consequence there will be the ranges for all six design variables are given.
practically no difference between both engines with respect to the high Before the numerical optimization algorithm can start, we need to
pressure turbine pressure ratio. define a range for the design variables. On one side this range should be as
A further constraint is, that the low pressure turbine inlet temperature narrow as possible because then the search for the optimum will take less
To must be below say 1150K which allows to design an uncooled low effort. However, when the range is too narrow, then the true optimum
pressure turbine from inexpensive materials. might be excluded from the search unintentionally.
The core compressor will eventually cause several constraints for the
design. There might be a temperature limit when the last stage is made from
titanium, for example. Also, a mechanical speed limit may exist And last min value max value
but not least the minimum surge margin requirements must be met.
fan P/P 1.1 1.9
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. booster P/P 1.4 2.3
In our example the task is, to increase the Max Climb thrust by 25%.
Engine designs with less thrust than required will not be acceptable and bypass ratio 4 6
therefore the thrust is a design constraint for the growth engine. In
summary, the design constraints for the derivative engine are: burner exit temp 1300 K 1600 K

relative core compr. corr. speed 0.9 1.02

hp turbine flow capacity reference value ±5%
lp turbine flow capacity reference 5% core compr. map coord. beta 0.3 0.8

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The optimization can be stopped at any time which allows to check the
52 Starting point best solution found in more detail than possible from the values for design
Many numerical optimization algorithms require, that a set of design variable and constraints alone. For both the engine design point (Max climb
variables which fulfills all constraints must be known before the calculation at altitude) and the offdesign condition (SL Take Off ISA+151C Mach 0.2)
can commence. The cycle of the basic engine is within the ranges of all there are all details accessible. This includes even graphs with the low
design variables, however, obviously it does not fulfill the minimum thrust pressum turbine design (figure 7) and the operating points in the component
constraint. maps at off-design.
How can we get a valid cycle to start with? One possible approach
would be, to do a rough parameter study which has only the aim to find a 5.4 Local and global optima
feasible solution, but not the best solution for the problem. However, this As explained in the chapter about the gradient search strategy with
parameter study takes more effort than necessary. We can redefine the numerical optimization there is always the danger, that the algorithm finds
figure of merit for the moment and do a slave optimization with the aim of only a local optimum but not the global optimum within the parameter
maximizing the Max Climb thrust. The minimum thrust constraint is range. When there are several local optima within the feasible region, then
dropped for that preliminary exercise which makes the cycle of the basic it depends on the starting point of the algorithm which local optimum will
engine valid as a start point. be found. Therefore one should repeat the optimization run several times
While the slave optimization is running, one can observe on the and pick from all local optima the best one.
computer screen the progress. As soon as a cycle is found which has more One can find easily a new starting point for the optimization by
Max Climb thrust than required (and fulfills all constraints) we can stop. redefining the search direction. Instead of minimizing the specific fuel
Now we redefine the figure of merit as specific fuel consumption and consumption one looks during a restart run for the cycle with the maximum
introduce the minimum thrust constraint. The final optimization can SFC. The random adaptive search will lead for each restart run to a different
commence now. starting point even when it commences several times from the same
optimum.
In figure 5 one can see from the graphics for the figure of merit, that
•
twice a restart has happened. This was caused by the algorithm called
1.1 Oen Fan Prmun Rao II I t st. 12 HP714 Cca iksv WI1Rga 1.41 "Endless Random Search" which restarts automatically after the algorithm
10 IP CA:ammo Pruitt.* Rabe 2$ SUP I 113 LPT IN Corr Flaw W4151hut 52
has homed into an optimum. The best solution found will be stored in
■ ■1116110
11 memory and can be restored as soon as the calculation is stopped.
Onign Bypass Roo 6 0 Ur bin Tempraturs T15 119)
Some important data for our growth engine example is summarized and
tX0 Duns En Tenpenturt t 0 HPC Ext Impress 73 750 compared to the basic cycle in the table below.
00 lei NWsprIr25.1100 102 OP Design R HP Sped Sped IFS

02 Auxlmy Coctdinso DM1 05 05 LPC lip Dimon 075

Basic Engine Growth Engine

Max Climb Hot Day Max Climb Hot Day

11 ilk 41P11. Take Off Take Off
Ape ell Molt
VINO* Thrust 3.61 13.10 4.50 16.94
(kFll
Ow Figure el art
91,11071
SFC 19.86 14.23 18.93 13.33
Cycle No 1691
(g/(kff s)]
ete 9.

Figure 5: Graphical user interface of GasTurb Bypass 4.5 4.65 5.06 5.23
Ratio

Fan 1.775 1.62 1.73 1.60

5.3 Graphical user Interface P13/P2
Figure 5 shows the optimization window of GasTurb for the example
of this paper with six horizontal gauges for the design variables on the left Ideal Jet 0.761 0.886 0.726 0.839
VS. Ratio
and seven gauges for the constraints in the upper right part. The gauges are
continuously updated while the optimization is running. In the lower part of Booster 1.5 1.33 1.80 1.61
the SCIECTI the figure of merit is shown both as numbers and as graphic with P24/P2
a dot for every valid solution.
One can immediately see from the gauges when a variable or a HPC 12 11.36 12.3 11.67
constraint is driven toward a range boundary respectively limit When the P3/P25
range boundary of a design variable happens to be not a true limit for the
74 [KJ 1350 1479 1393 1530
engine design then one can stop the calculation and redefine the range for
the coni,vonding design variable. W41 Rsid 1,35 1,35 1,31 1,31
In practice it happens quite often, that during the rust attempts the
optimization problem is not formulated correctly. In such a case the W45Rstd 4,98 4,96 5,01 5,03
numerical algorithm drives the mathematical model in a direction which is
obviously nonsense because a constraint was forgotten, for example. 731Iq 610 708 649 750
Therefore an easy to survey graphical user interface is very helpful for
74511 973 1076 1030 1108
avoiding a waste of computing time.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 6. DISCUSSION OF THE RESULTS 7. SUMMARY
The optimum growth engine is influenced by three of the design With a conventional parameter study it is very difficult to find the
constraints. It has a fan diameter of 0,75m, i.e., it uses the largest fan which optimum solution for a problem as soon as four or more design variables
was allowed in this exercise. The second constraint which had an impact on and several constraints are involved. With the help of numerical
the design of the growth engine is the compressor exit temperature which optimization algorithms one can easily find the mathematical correct
was limited to 750K for the hot day take off case. The third constraint was solution to the problem. Extensive parameter studies around the solution
the minimum high pressure turbine flow capacity W41Rstd. will help to understand why this combination of design variables is the best
All design variables did remain within the predefined range during the choice and how sensitive the figure of merit is to small deviations from the
optimization. The thrust increase for Max Climb rating at altitude is 25% optimum.
and at Take Off even 29%. Note that both engines run during Take Off
speed with 7% more mechanical high pressure spool speed than at Max
Climb in this example. The specific fuel consumption at altitude is nearly 5/5 19.75
5% better for the growth engine. 9.5
In the table there is also a row for the ideal jet velocity ratio. From 19.25

Lo
5
theoretical considerations one can derive, that this ratio should be equal to cc
the product of fan and low pressure turbine efficiency when an unmixed 5
flow turbofan is to be optimized for SEC. Note that the numerical °
al t Thalicksi,
optimization algorithm has automatically found a cycle for which the jet
WI
C.44 8141
1kkS\
%%.141 4‘
11%1114911111111
; 8.75
velocity ratio is near to its theoretically best value. TOC.
4
00 *frit: 4407
470,0411,441
5sF
1925
19

27.5 eat ;4375 -II.

.4113agtf
a 20 le
.75

25
Ara
10
3.5 3.75 4 4.*et 4.5 t p475 5 5.25 5.5

3 20 Figure 8: Sensitivity for Bypass Ratio and Burner Exit temperature

u• 17.5'

ie 15
wx 19.6
12.5 38 re1.4 4728a4141
3 1 02
10
3.8 4 4.2 4.4Net Ttlitt 4.8 5 5.2 5.4 t 19.4 5-
all/I
111111■119 ISO
Figure 6: Sensitivity for Booster and Outer Fan Pressure Ratio 0

5 19.2
annm.
WlNMTryj 45 g

It is quite useful to do a parameter variation in the neighborhood of the

optimum solution. This gives an insight to the sensitivity of the result.
Figure 6 shows the influence of "Outer Fan Pressure Ratio" and "Booster
Pressure Ratio" on thrust and specific fuel consumption. Note that the ci.
0
19 elligt if 3 3

parameter on the vertical axis was mainly selected because it spreads the Trust 4.5 IN
carpet nicely. The corrected flow at the low pressure turbine exit is not a 18.84 35
4.4 4.5 4.55
very important engine design parameter. Net4.45
Thrust NM
Actually the limiting line for the fan diameter (0.75m) is identical to the Figure 7: Sensitivity for HPC Map Operating Point
line for the minimum high pressure turbine flow capacity. To the left of the
line for booster pressure ratio 1.79 the lower limit for the high pressure
turbine flow capacity is violated and to the right of this line the fan diameter As an example, the cycle selection of a derivative turbofan engine with
a given core shows the merits of numerical optimization. The parameter
is too big. In this graph only one point is a valid solution: the square which
variation is best suited for presenting the sensitivity of the result in the
fulfills the thrust requirement. One can see, that the optimum solution is
neighborhood of the optimum cycle design point. Sometimes this leads to
pretty much boxed in by the design constraints.
a redefinition of the figure of merit or the constraints imposed on the
In figure 7 two more design variables were systematically varied around
solution. In rare cases even an outstanding solution is found which was
the optimum solution. Again the square marks the only point which fulfills
overlooked while doing a preliminary parameter study.
all constraints.
Figure 8 deals with the operating point of the high pressure compressor.
High values for the auxiliary coordinate beta go with a low surge margin.
The square marks the point with lowest SFC in the region of feasible
designs.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 8. REFERENCES
I. Kurzke Joachim, 1998 "Manual GasTurb 8.0 for Windows - A
Program to Calculate Design and Off-Design Performance of Gas
Turbines", Available from the author
2. Heinrich G. Jacob, 1982. "RechncrgestOtzte Optimierung stalischer
und dynamischer Systeme", Fachberichte Messen - Steuem - Regain,
Springer Verlag
3. R. C. Kelahan, J. L. Gaddy, 1978. "Application of the adaptive
random search to discrete and mixed integer optimization." International
Journal for Numerical Methods in Engineering, Vol. 12, 289-298 (1978)
4. Kualce Joachim, 1996 "How to Get Component Maps for Aircraft
Gas Turbine Performance Calculations", ASME Paper 96-GT-164
5. Arthur J. Glassman, 1992 "Users Manual for Updated Computer
Code for Axial-Row Compressor Conceptual Design" NASA Contractor
Report 189171
6. Warner L. Stewart, 1961 "A Study of Axial-Row Turbine Efficiency
Characteristics in Terms of Velocity Diagram Parameters", ASME Paper
6l-WA-37
7. H.Cohen, G.F.C. Rogers, H.LH. Saravartarnuttoo "Gas Turbine
Theory", 4th Edition, Addison-Wesley Longman London 1996

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 03/04/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use