Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

System optimization of turbofan engines using

genetic algorithms
A. Homaifar

Department of Electrical Engineering, North Carolina A&T State University, Greensboro, NC,
USA

H. Y. Lai

Department of Mechanical Engineering, North Carolina A&T State University, Greensboro,


NC, USA

E. McCormick

Research Triangle Institute, Centerfor Systems Engineering, Research Triangle Park, NC, USA

This paper presents an application of genetic algorithms to the system optimization of turbofan engines.
Genetic algorithms are relatively new general-purpose optimization algorithms that apply the rules of natural
genetics to explore a given search space. In order to characterize the many measures of aircraft engine
performance, two d&erent criteria are chosen for evaluation. These criteria are thrust per unit mass flow rate
and overall eficiency. These criteria are optimized using four key parameters including Mach number,
compressor pressure ratio, fan pressure ratio, and bypass ratio. After observing how each parameter inJEuences
objective functions independently, the two objective functions are combined to examine their interaction in a
multiobjective function optimization. Numerical results indicate that genetic algorithms are capable of
optimizing a complex system quickly. The resultant parameter values agree well with previous studies.

Keywords: genetic algorithms, optimization

Introduction added benefits. A turbofan engine uses a turbojet with a


larger fan area that divides the incoming airflow into two
Modern aircraft engine design is an exceptionally com- streams. The “hot” stream passes through the engine and
plex task. Today’s engines contain literally tens of thou- is burned with fuel. The “cold” stream bypasses the
sands of parts. Hill and Peterson,’ Zucrow,’ Sheperd,3 engine and is compressed in the bypass stage. These two
Bathie,4 and Hale5 have described many different criteria streams result in two different exit velocities, U: and
used to judge the performance of an engine, and these U%.
criteria are subject to optimization. This paper uses The turbofan engine consists of seven fundamental
genetic algorithms (GA) to assist in the optimization of stages,’ as shown in Figure I. These stages include:
this design process for a turbofan engine.
The premise behind the turbofan engine comes from l-2 Inlet diffuser and ducting stage-decreases air velo-
the observation that thrust increases with airflow rate city as air is carried to compressor inlet.
for a given fuel consumption rate. An uneconomical 2-3 Compressor stage-compresses air.
method for increasing airflow rate would be simply to 3-4 Combustion stage-heats air through mixing and
increase the engine size. This, however, would result in burning of fuel.
an unnecessarily heavy engine that might offset any 4-5 Turbine stage--expands air to obtain power to
drive the compressor.
5-6 Afterburner stage-additional heating of air
Address reprint requests to Dr. Homaifar at the Dept. of Electrical through more mixing and burning of fuel.
Engineering, North Carolina A&T State University, Greensboro, NC 6-l Exhaust nozzle stage-accelerates and exhausts air.
27411, USA.
2-8 Fan bypass stage-bypasses some air around the
Received 6 October 1992; revised 5 July 1993; accepted 28 July 1993 engine’s internal operating parts.

72 Appl. Math. Modelling, 1994, Vol. 18, February 0 1994 Butterworth-Heinemann


System optimization using genetic algorithms: A. Homaifar et al.

while the strings with low fitnesses receive a lower


number of copies or even none at all. Crossover refers
to splitting a string into two parts at a randomly gen-
erated crossover point and recombining it with another
string that has been split at the same crossover point.
This procedure serves to promote changes in the best
strings that will give them even higher fitnesses. Mutation
is the random alteration of a bit in the string. This assists
a 1 2 3 4 5 6 7 in keeping diversity in the population.
In explaining the mechanisms of GA, a few terms need
Figure 1. Stages of a turbofan engine. to be defined. Because we are dealing with binary strings,
a notation must be developed to denote similarity subsets
(schemata). A schema is a similarity subset with strings
Genetic algorithms are used in this study because they that contain similarities at some bit positions. We can
1

are able to quickly optimize the objective functions expand this format even further with the introduction of
involving subfunctions of multivariates. The resultant a wild card character, *, in addition to the binary set
parameter values are accurate and can be used for design (0, l}. For example, the set {0001,0101, OOll} can be
optimization in the early engine design stages. The details described by the similarity template O**l. Using this
of GA are presented in this context. notation, we can now specify a schema’s order and
deJining length. For a given schema, h, its order o(h) is
defined as the number of fixed bit positions within that
Genetic algorithms schema. The defining length of a schema, 6(h), is the
Genetic algorithms are general-purpose optimization al- distance between the outermost fixed positions of a
gorithms with a probabilistic component. They provide schema. As an example, the schema Ol****O has order
a means to search poorly understood, irregular spaces. 3, defining length 5, and can represent 16 different
Genetic algorithms have successfully been applied to a individuals.
variety of function optimization, parameter searches, and With these definitions we can now present the funda-
machine learning applications. John Holland6 originally mental theorem of genetic algorithms, the schema theor-
developed GA and provided its theoretical foundations. em.’ The schema theorem enables us to calculate a lower
GA was developed to simulate some of the processes bound on the expected number of a particular schema,
observed in natural evolution, a process that operates on h, following reproduction, crossover, and mutation. This
chromosomes (organic devices for encoding the structure theorem is stated as
of living beings) rather than on living beings. Natural
selection links chromosomes with the performance of m(h, t + 1) 2 m(h, t) 50 6(h) - (1)
their decoded structures. The processes of natural selec- 5 cE p,o(h) 1
tion cause those chromosomes that encode successful
structures to reproduce more often than those that do where m is the expected number of schemata, t is the
not. Recombination processes create different chromo- generation index, 1 is the overall string length, c(h) is the
somes in children by combining material from the chro- average fitness of those strings representing the subset h,
mosomes of the two parents. Mutation may cause the 5 is the average fitness of the entire population, pc and
chromosomes of children to be different from those of P,,, are, respectively, the crossover and mutation probabil-
their parents. ities. Examining the schema theorem, we see that it
GA appropriately incorporates these features of na- states that a schema will grow when it is short, has low
tural evolution in computer algorithms to solve difficult order, and has above-average fitness.
problems in the way that nature has done-through Genetic algorithms have many advantages over other
evolution. GA requires the problem to be maximized (or methods. Currently most literature defines three main
minimized) in the form of an objective (cost) function. In types of search methods: calculus-based, enumerative,
GA, a set of variables for a given problem is encoded and random.8 Calculus-based methods can be divided
into a binary string or any other coding structure, into two classes: indirect and direct. In indirect methods,
analogous to a chromosome in nature. These strings are local extrema are determined by finding where the gra-
converted to a numerical value and then linearly mapped dient of the objective function is equal to zero. Direct
over the range allowed for the variable. This value is then methods follow along the objective function in a direc-
used to evaluate the objective function, yielding a “fit- tion related to the local gradient. Both classes share two
ness.” GA selects parents from a pool of strings (popula- main disadvantages that greatly limit their usefulness.
tion) according to the basic criteria of survival of the First, both are local in scope; that is, if a function has
fittest. It reproduces new strings by recombining the multiple local maxima, these methods may drive toward
parts of the selected parents in a random manner. one of these values without ever approaching the global
The repopulation of the next generation is done using maxima. The second deficiency of calculus-based me-
three methods: reproduction, crossover, and mutation. thods is their dependence on the existence of derivatives.
Reproduction means simply that strings with high fitness However, many “real world” functions are discontinuous
should receive multiple copies in the next generation and noisy, and therefore do not work well with a method

Appl. Math. Modelling, 1994, Vol. 18, February 73


System optimization using genetic algorithms: A. Homaifar et al.

that prefers smooth, continuous functions. Enumerative would be possible to show bias toward one function if
methods offer an attractive advantage-simplicity-but one is considered to be more important than the other.
this advantage carries a high cost.
Enumerative schemes basically take a discretized
Algorithm performance
search space and examine the objective function at every
point. Although simple in technique, this brute force The main purpose of this computational experiment is to
method is quite inefficient, and its execution times be- examine the ability of GA to perform a multiparameter,
come too long as the search space becomes larger. multiobjective function optimization on a real-world
The last method is the random search method. In problem. Although there are many different measure-
many respects, this method can perform as poorly as ments of engine performance, in this study the thrust per
enumerative methods because of its inefficiency. Ran- unit mass flow rate (TMA) and the overall efficiency, q,
domly searching through a space and saving the best are the two functions under investigation. TMA is an
results can become time-consuming as the space becomes important consideration in turbofan engine design be-
large. cause size constraints limit an engine’s mass flow rate.
The primary advantage of GA is its robustness. GA Therefore, we seek to get as much thrust as possible for
works through function evaluation, not through differ- this mass flow rate. Overall efficiency is also important
entiation or other such means. Although GA begins to engine design. Overall efficiency is the product of two
with a randomly generated set of points, it exploits the other efficiencies-propulsion efficiency and thermal
information contained in those points to drive it through efficiency. Propulsion efficiency is the ratio of an engine’s
the search space. Because GA is based on function thrust power to the rate of production of propellant
evaluation, it can be applied to all types of optimization kinetic energy. Thermal efficiency is the ratio of the rate
problems as long as they are properly coded. of addition of kinetic energy to the propellant to the total
Because of its robustness, GA has been used in opti- energy consumption rate. Overall efficiency is an impor-
mization problems as diverse as image analysis by tant part of engine design because flight range is directly
Grefenstette and Fitzpatrick9 and gaming strategy by proportional to overall efficiency. To obtain these two
Axelrod.” Another practical engineering example of functions, we must examine each stage of the engine and
GA’s application is Goldberg’s study” of a system of 10 derive expressions for the total (stagnation) temperature
pipes and 10 pumping stations. Also, expert systems can and pressure ratios across these stages. Using the con-
be improved upon by using GA, as shown by Davis and servation equations for mass, momentum, and energy,
Coombs.iz we derived equations governing the behavior of each
The software used for this paper was developed by engine stage. These equations are given in the appendix.
Goldberg.8 The simple genetic algorithm (SGA) pro- Using these equations, we can evaluate expressions for
gram, written in Pascal, is modified to allow the evalu- TMA and overall efficiency. The resultant equations are
ation of multiple parameters and multiple objective func-
tions. Multiple parameters are incorporated by letting
‘T = (1 + f)ub + pz.4,”- (1 + #l?)u (3)
each string be divided into substrings that represent the &a
different variables. For example a string of length 18
could be used to represent three substrings of length 6, (ii, - u)u
each of which represents a different parameter. In this
ro = (1 + PI fQR
(4)
evaluation, the string length is set at 36 and the number
of substrings is four. The precision for each substring is When evaluating these functions, engine performance
determined from the equation is considered to be a four-parameter problem. The
parameters used in evaluating these functions are: Mach
u m=x - Umin number, M; compressor pressure ratio, P,,/P,,; fan
7[ =

2”’ - 1
(2) pressure ratio, P,,/P,,, and bypass ratio, p. It is obvious
from the equations derived in the appendix that there
where rr is the precision of the parameter, U,,, and are many other possible parameters to add to this list,
Urnin are the maximum and minimum of the allowed but these variables are chosen because there are data
value of the parameter, and A1 is the length of the readily available for comparison. Reasonable values are
substring. Obviously, using multiple parameters requires therefore assumed for such parameters as free-stream
increasingly longer string lengths if each variable is to pressure and temperature and the stage efficiencies. The
have an acceptable precision. Multiple objective func- initial ranges for the parameters, their individual string
tions are incorporated into the program by initially lengths, and their respective accuracies as shown in Table
evaluating the population using the first objective func- 1. The ranges for the two pressure ratios are typical upper
tion. These fitnesses are then normalized by dividing by and lower bounds for each. The limits on Mach number
the maximum fitness in the population for that objective were made to ensure a reasonable stall speed and to
function. In the same manner, the second objective prevent shock formation. Finally, the upper limit on
function is evaluated, normalized, and added to the bypass ratio was set partially by the increase in weight
normalized result from evaluation of the first objective penalty for larger fan and cowling sizes. The lower limit
function. After dividing by two, we have a fitness that on bypass ratio is the lowest reasonable value for an
varies from 0.0 to 1.0. Also, by choosing this method it engine to be considered a true turbofan. The string

74 Appl. Math. Modelling, 1994, Vol. 18, February


System optimization using genetic algorithms: A. Homaifar et al.

Table 1. Variables to be optimized particular parameter’s value will affect the other param-
eters. In one experiment, P,,/P,, was allowed to vary
String and To4 was held constant at 1089 K. Also, two different
Parameter u max u I?%” length Accuracy Mach numbers were used, 0.66 and 0. For M = 0.66, a
maximum TMA of approximately 400 N-s/kg was ob-
M 0.9 0.3 8 0.0024 tained with P,,fP,, z 6. For M = 0, a maximum TMA
25.0 5.0 10 0.0196
PO31 PO2
of approximately 60 N-s/kg was obtained with Po3/
P&I Pea 3.0 1.0 9 0.0039
6.0 0.1 9 0.0115 P,, E 8. In both cases, TMA behaved quadratically in
B
nature on a log scale, increasing to a maximum and then
decreasing. This supports the fact that GA drove M as
low as it could.
lengths were then chosen to give the best possible resolu- Pinkel and Karp’s second experiment kept M con-
tion while trying to minimize computation time. stant at 0.66 and used three different To4 values: 811 K,
The variables that had to be defined for GA are 1089 K, and 1367 K. For 811 K, TMA was a maximum
population size, crossover probability, mutation prob- of approximately 245 N-s/kg with P,,jP,, E 3. For
ability, and maximum number of generations. Their 1089 K, TMA was a maximum of 400 N-s/kg with Po3/
respective values are 300, 0.65, 0.001, and 1000. Popula- P,, E 6. For 1367 K, TMA was a maximum of approx-
tion sizing is an important requirement in GA. Popula- imately 610 N-s/kg with P,,/P,, E 9. As before, TMA
tions must be large enough to provide adequate diversity. behaved quadratically on a log scale, but, as x4 in-
However, the larger the population, the greater the creased, the peak became flatter giving an indication as
number of calculations. The population used here is to why GA fluctuated between 16 and 18 for its maxi-
based on Goldberg’s study on minimum population mum TMA. Given the combustor temperature used in
size. 13 our work of 1555 K, and the lower M that GA chose,
the value of P,,/P,, E 17 becomes a reasonable solution.
To see what effect crossover and mutation probability
Analysis had on convergence, an additional run was made using
With the proper variables set and the range for the crossover and mutation probabilities of 0.5 and 0.005,
parameters determined, four different cases were ana- respectively. The results, shown in Figures 3(a)-3(e)
lyzed. Case 1 was done using TMA as the objective closely match those in Figures 2(a)-2(e). However, there
function with the Mach number required to be subsonic. appears to be greater fluctuation in the parameter values
Case 2 examined overall efficiency with Mach number of the most fit individual from one generation to the next,
subsonic. Case 3 examined the two functions combined accompanied by a lower convergence value for TMA.
and the Mach number subsonic. Case 4 reexamined This can be explained by the fact that the reduced
overall efficiency and TMA when Mach number was crossover probability is not allowing GA to piece to-
allowed to be supersonic. gether good copies often enough, thereby decreasing the
convergence value. While this action is taking place, the
increased mutation probability is creating more random
Case 1 bit changes, assisting in the fluctuation from one gen-
The initial run was made with TMA as the objective eration to the next.
function. The results for each parameter and the objec-
tive function are shown in Figures 2(a)-2(e). The figures
Case 2
show the parameter values for the most fit individual of
each generation. As the figures show, GA tended to The second case is made with overall efficiency as the
converge to its final answer within approximately 300 objective function. The results are shown in Figures
generations for each parameter. With a population of 4(a)d(e). Again, GA converged rather quickly to its
300, this equates to 90,000 function evaluations to con- final solution. In examining the equation for thermal
vergence. On a 25MHz 80386 PC with a 80387 math efficiency, differentiation with respect to u would imply
coprocessor, each generation took approximately 5 sec. that thermal efficiency is maximized when u z u,/2. For
This yields a computer time to convergence of approx- the flight conditions specified, this would entail driving
imately 25 min. The only exception with respect to con- Mach number as high as possible, which is what hap-
vergence is the compressor pressure ratio, P,,/P,,. Al- pened with GA. This fact also tended to drive the other
though this value tended to fluctuate between 16 and 18, parameters to their final values. One should also notice
the smoothness of the objective function shows that this that the highest overall efficiency obtained is approx-
small variation has little influence on the final result. imately 0.22. Because this value tended to peak as Mach
Work done by Pinkel and Karpi4 examined the influence number peaked, it suggests that even higher Mach num-
of Mach number, compressor pressure ratio, and com- bers could lead to higher overall efficiencies.
bustor temperature on TMA. Although their experi-
ments examined only one variable at a time with other
parameters being fixed, they provided us with confirma- Case 3
tion of GA’s strength. Because they examined more than The third run is made while optimizing the two functions
one value of each parameter, we can infer how each together. The two functions were combined into a single

Appl. Math. Modelling, 1994, Vol. 18, February 75


System optimization using genetic algorithms: A. Homaifar et al.

O-
0 IM 200 SW 4QQ 600 2QO 7QQ E4 000 1000 2Qo 2OQ 4QQ 203 MO 700 MQ QQQ IQOQ

2a Gene&ion Generation
Figure 2(a). Bypass ratio vs. generation where thrust/mass flow Figure 2(b). Mach number vs. generation where thrust/mass
rate is objective function. flow rate is objective function.

L4.-

0 IQ 2uaQQ4m6mMQ7QobQQ nQQ IQQQ

2c Generatin 26 Generation

Figure 2(c). Po3/Po2 where thrust/mass flow rate is objective Figure 2(d). Po8/Poa where thrust/mass flow rate is objective
function. function.

2 IOQQ
i:c, nm.’
2cQ-

7QQ-
Q 100 2M 500 100 MD 203 7OQ 2QQ QQO IQM

2e Generation

Figure 2(e). Optimization of thrust/mass flow rate.

76 Appl. Math. Modelling, 1994, Vol. 18, February


System optimization using genetic algorithms: A. Homaifar et al.

3a Generation 3b Generation

Figure 3(a). Bypass ratio using different PC and Pm Figure 3(b). Mach number using different PC and Pm.

3c Generation 3d Generation

Figure 3(c). P&PO2 using different PC and Pm Figure 3(d). Po8/Poa using different PC and Pm

3e Generation

Figure 3(e). Thrust/mass flow rate using different PC and Pm.

Appl. Math. Modelling, 1994, Vol. 18, February 77


System optimization using genetic algorithms: A. Homaifar et al.

4m

Generation 4b Gmeration

Figure 4(a). Bypass ratio where overall efficiency is objective Figure 4(b). Mach number where overall efficiency is objective
function. function.

0.6

to
L-5
E 2

1.6. - ’ 7

17
0 100 wowo4oownwo7ooow ooo moo

4c Generation 4d Generation

fo3/Po2 where overall efficiency is objective func Figure 4(d). f,,s/P,, where overall efficiency is objective function.
Figure 4(c).
tion.

4e Generation

Figure 4(e). Optimization of overall efficiency.

78 Appl. Math. Modelling, 1994, Vol. 18, February


System optimization using generic algorithms: A. Homaifar et al.

ii4
a

L 2..
al

l-

0+ A
0 I00 ZOO 500 400 500 MO 700 5w 900 IWO

5a Generation 5b Generation
Figure 5(a). Bypass ratio using combined objective function Figure 5(b). Mach number using combined objective function.

5c Generation 5d Generation
Figure 5(c). P,,JPo2 using combined objective function Figure 5(d). Pos/P,,a using combined objective function.

1000
2
2950 0.2
2.
&Oo
IA
E 0 18
0

‘C
Isso
In E 0.16
=KJo =
5
2 750
s 0.14
2 700 0.12 -

i-’
650 0.17 , 1 , , , , , ,
0 100 200 300 400 500 600 700 600 900 loo0 0 100 200 300 400 600 600 700 600 900 1000

5e Generation
5f Generation

Figure 5(e). Thrust/mass flow rate using combined objective Figure 5(f). Overall efficiency using combined objective func-
function. tion.

Appl. Math. Modelling, 1994, Vol. 18, February 79


System oprimizarion using genetic algorithms: A. Homaifar et al.

k I.4
&
b-i
9:
5 12

z ’
J”
Em
6
I&
r: M.
1.
K4-
Jb-dk___-
k--X
6;i IO0 ZOO 300 700 I00 OM 1wO .
yo&wtl 4m wo MO 700 Ma SO0 1000

6iY0 ‘Oa z00 spo Gmeration

Figure 6(a). Bypass ratio where overall efficiency is objective Figure 6(b). Mach number where overall efficiency is objective
function. function.

6c GeneIation 6d Generation

Figure 6(c). Po3/Po2 where overall efficiency is objective func- Figure 6(d). P&/P,, where overall efficiency is objective func-
tion. tion.

6e Generation
Figure 6(e). Optimization of overall efficiency.

80 Appl. Math. Modelling, 1994, Vol. 18, February


System oprimizarion using generic algorithms: A. Homaifar er a/.

,aO; 100 200 660 ytaa~on~~ 700 6W 662 IWO

7b Generation

Figure 7(a). Bypass ratio where thrust/mass flow rate is objec- Figure 7(b). Mach number where thrust/mass flow rate is objec-

tive function. tive function.

106 200 yI(I 400 6cQ 666 700 630 6m 1666 7d”
I
‘O” 600

czeneratiin Generation
7c
Figure 7(c). Po3/Po2 where thrust/mass flow rate is objective Figure 7(d). Pos/Poa where thrust/mass flow rate is objective
function. function.

7e Cenerdiin

Figure 7(e). Optimization of thrust/mass flow rate.

Appl. Math. Modelling, 1994, Vol. 18, February 81


System optimization using genetic algori. 7s: A. Homaifar et al.

function as described earlier in the sectic .3n genetic tion, Figures 7(a)-7(e) bear this fact out. With this in
algorithms. Output of this run is show in Figures mind, we must conclude that for the results obtained for
5(a)-5(f). Analysis of this run gives an ,&%cation of overall efficiency shown in Figures 6(a)d(e), the loca-
which parameters have the greatest influence on each tion of a shock wave on the engine inlet has a drastic
individual objective function. Although Mach number influence on the parameters we used. When one considers
needed to be low for optimum TMA and high for the physical consequences of a shock wave, the results
optimum overall efficiency, it can be seen that if both tend to make more sense. The occurrence of a normal
functions are considered equally important, the require- shock leads to an increase in stagnation pressure down-
ment for overall efficiency wins out and drives the Mach stream of the shock. With this natural increase in pres-
number higher. To a lesser extent, the same is true for sure, one can see that higher compressor pressure ratios
the bypass ratio and TMA. Although overall efficiency are no longer needed. The shock also leads to the
dictates a moderate value for bypass ratio, TMA be- negation of the effect of bypass ratio. Because the effect
comes greater as bypass ratio becomes greater. This of bypass ratio becomes negligible, this in turn eliminates
necessity remains after the two functions are combined. the influence of fan pressure ratio, explaining why the
Also, compressor ratio appears to be more relevant to erratic results for this parameter have no effect on overall
TMA than overall efficiency. Notice that the fan pressure efficiency.
ratio tends to level out to the same value for both
objective functions. This shows that while the fan pres-
Conclusion
sure ratio has a bearing on both, the effects of the “hot”
stream dominate both functions. When examining Fig- In this study, genetic algorithms have been applied to a
ures 5(e) and S(f), one can surmise that while a low complex turbofan engine design problem. Although un-
Mach number may be essential for a high overall effic- der most circumstances a multiparameter, multiobjective
iency, it does indeed have a detrimental effect on TMA, function optimization problem would be considered a
which resulted in a nearly 40% drop from the TMA difficult task, GA handled the task quite successfully.
obtained when it was the sole objective function. In the Although the model used here to represent a turbofan
same way, the changes in bypass ratio and compressor engine is a relatively simple one, the procedure would be
ratio led to a 10% drop in overall efficiency when exactly the same with a more elaborate model. Results
compared with the value obtained when optimizing of assorted runs agreed with experimental and single-
overall efficiency alone. These losses, especially the dra- parameter optimization results. Of course, realistic de-
matic change for TMA, show the delicate balancing GA sign would include many more parameters, but for GA
performed to obtain the best overall solution. The best this would only mean increasing the string length to
overall solution, however, may not lead to the most incorporate these additional parameters. The ability of
well-balanced solution, and therefore the two functions GA to handle a problem of this nature could make it a
may need to be weighted so each bears an equal burden very important tool in the future of engine design.
of loss.
Acknowledgment
Case 4 This work is partially supported by grants from the
As noticed earlier, overall efficiency is apparently limited NASA Center of Research Excellence under grant num-
by the constraint on the upper bound of the Mach ber NAGW-2924 and Honeywell Inc. under grant
number. To see if this is true, an additional run was made number 48057. The authors wish to thank them for their
with an upper bound on Mach number that exceeded support.
unity. The new upper bound on Mach number was set
to 1.8. If the Mach number exceeded unity, the program
Nomenclature
was modified to assume a normal shock wave as per
Anderson” across the engine inlet instead of allowing CPC constant-pressure specific heat (cold), kJ/kg-K
the possibility of the engine swallowing the shock. The ‘ph constant-pressure specific heat (hot), kJ/kg-K
results for this run are shown in Figures 6(a)d(e). fuel to air ratio
Overall efficiency did indeed increase slightly, but the L flight Mach number
effect on the parameters is more pronounced. Optimum atmospheric pressure, Pa
compressor ratio decreased by approximately one third, stagnation pressure at stage *, Pa
while bypass ratio changed from being as high as possible fuel heat of reaction, kJ/kg
to being as low as possible. Also, the optimum fan universal gas constant, kJ/kg-K
pressure ratio became highly erratic and did not con- atmospheric temperature, K
verge after 1000 generations. stagnation temperature at stage *, K
To validate that the algorithm for supersonic Mach flight velocity, m/s
numbers was correct, another run was made with TMA bypass ratio
as the objective function. Because TMA requires Mach Y ratio of specific heats (cold)
number to be as low as possible, there should be no Y1 ratio of specific heats (hot)
difference between this run and Case 1. Given slight qb burner efficiency
changes due to the probabilities of crossover and muta- % compressor efficiency

82 Appl. Math. Modelling, 1994, Vol. 18, February


System optimization using genetic algorithms: A. Homaifar et al.

qd diffuser efficiency
fan efficiency 642)
Ylr
%I nozzle efficiency
% turbine efficiency
643)

PO1= PO, 644)


References P, = P, WY
1 Hill, P. G. and Peterson, C. R. Mechanics and Thermodynamics of
r,, = T,, = T,, WI
Propulsion, 3rd ed. Addison-Wesley Publishing Co., Reading, MA,
1970
2 Zucrow, M. J. Principles of Jet Propulsion and Gas Turbines. John 647)
Wiley & Sons, New York, 1948
3 Sheperd, D. G. Aerospace Propulsion. American Elsevier Publish-
ing Company, New York, 1972
648)
4 Bathie, W. W. Fundamentals of Gas Turbines. John Wiley & Sons,
New York 1984
5 Hale, F. J. Introduction to Aircraft Performance, Selection, and
Design. John Wiley & Sons, New York, 1984 649)
6 Holland, J. H. Adaptation in Natural and Artificial Systems. The
University of Michigan Press, Ann Arbor, MI, 1975
T04
I Goldberg, D. E. Genetic algorithms and Walsh functions: Part I, -- 1
A gentle introduction. Complex Systems 1989b, 3, 129-152
8 Goldberg, D. E. Genetic Algorithms in Search, Optimization, & +z3 T (AlO)
Machine Learning. Addison-Wesley Publishing Co., Reading, MA,
1989
x-2
9 Grefenstette, J. J. and Fitzpatrick, J. M. Genetic search with
approximate function evaluations. Proceedings of an International P 05 ‘
Conference on Genetic Algorithms and Their Applications, ed. J. J. -=
Grefenstette, Pittsburgh, 1985 P 04
10 Axelrod, R. The evolution of strategies in the iterated prisoner’s
dilemma. Genetic Algorithms and Simulated Annealing, ed. L.
Davis. Pitman, London, and Morgan Kaufmann Publishers, 1987
]+;[(5$‘:‘-
11 Goldberg, D. E. Computer-aided gas pipeline operation using

(-I )
genetic algorithms and rule learning. Ph.D. Thesis, Technical
T04 CR,
Publication BER-84001SP, University of Alabama, 1986 (1 + f)’ c Ir
12 Davis, L. and Coombs, S. Optimizing network link sizes with T03 p,
genetic algorithms. Conference on Computer Simulation and
Modeling, Tucson, AZ, 1987b (Al 1)
13 Goldberg, D. E. Sizing populations for serial and parallel genetic
algorithms. University of Alabama, Tuscaloosa, The Clearing-
house for Genetic Algorithms, TCGA Report No. 88004, 1988
6412)
14 Pinkel, B. and Karp, 1. M. A thermodynamic study of the turbojet
engine. NACA Report No. 891, 1947 6413)
15 Anderson, J. D. Modern Compressible Flow. McGraw-Hill Book
Company, New York, 1982
6414)

Appendix (A13
The following is a list of the equations used in the
formulation of the thrust per unit mass flow and overall
efficiency objective functions. They describe the relation-
ship among pressure ratios, temperature ratios, and other
properties for the turbofan engine, as given in Figure I. (A171
The derivation of these equations is given in Hill and
Peterson.’

Appl. Math. Modelling, 1994, Vol. 18, February 83

You might also like