Você está na página 1de 8

IMK-14 Research & Developement 18(2012)4, EN115-122

UDK 621 ISSN 0354-6829

EN115

Modified Ant Colony Algorithm for Solving Engineering Optimization


Problems
1

Grkovi Vladan1,* - Bulatovi Radovan1


The Faculty of Mechanical and Civil Engineering in Kraljevo

This paper presents ant colony optimization - ACO. After presenting the biological basis of ASO, it explains the method
itself and then the main algorithms of the ASO method as well as their mathematical models. The "Modified Ant Colony
Algorithm" (MACA) is presented in detail as well as the manner of its operation and it application to optimization
examples of engineering problems, such as: optimization of spring, pressure vessel, welded beam and speed reducer. At the
end, the results obtained by the MASM method are compared to the results previously obtained by other methods.
Key words: Optimization, ant colony, engineering optimization problems
0. INTRODUCTION
The ant colony optimization algorithm studies behaviour
of real ant colonies used for solving optimization
problems, most frequently for finding the shortest path in
the given graph.
The term ant colony optimization was introduced by
Marco Dorigo in his doctoral dissertation in 1992, as a
result of research on the approach to combinatorial
optimization. Initially, the algorithm was applied to the
traveling salesman problem and the quadratic assignment
problem. Since 1995 Dorigo et al. have been working on
various extended versions of the original idea [1] .
All this was preceded by a significant experiment carried
out by Deneubourg in 1990. In order to examine the
behaviour of ants, Deneubourg et al. [2] conducted a series
of experiments using a bridge with two branches placed
between the anthill and the food. During the their
experiments, they first used the same lengths of the
branches, and then different lengths of bridge branches.
They noticed that ants would always find the shortest path
between the food source and their colony. It is in the
nature of ants to deposit the trail of pheromone behind
themselves when they find food, both on their way from
the anthill to the food and on their way back to the anthill.
Ants can detect pheromone, so while deciding which path
to choose they make decision on the basis of the intensity
of pheromone trail they detect and they will decide to take
the direction of a more intense pheromone trail. There are
two reasons why the intensity of pheromone can be
different. The first reason is pheromone evaporation, i.e.
the more time passes since depositing pheromone, the less
intense the trail is, and the second reason is the number of
ants that have deposited the pheromone. If more ants have
previously decided for a path, the intensity of pheromone
trail will be higher and the probability that the path is good
wil be greater so that the other ants will also take it.
Toksari [3] described ant colony optimization for finding
the global minimum. His algorithm is based on the fact
that each ant searches only around the best solution of the
previous iteration.

This paper will apply the modified algorithm to solving


engineering problems.
The first engineering problem that will be solved here is
spring optimization. This problem was described by Arora
[6] and Belegundu [7], and it consists of minimization of
the weight of spring which has certain constraints on
minimum deflection, shear stress, surge frequency, limits
on the outside diameter and design variables.
The second problem is pressure vessel optimization which
was previously analyzed by Sandgren [9], who was the
first to propose this problem. The problem was later
described by Kannan and Kramer [8]. The aim was to
reduce the total costs, including the costs of material,
forming and welding.
The third problem is welded beam optimization in which it
is necessary to optimize the miminum cost subject to
constraints on shear stresses, bending stresses in the beam,
buckling load, end deflection of the beam and other side
contraints described by Rao [10].
The last problem is speed reducer optimization in which it
is necessary to find the minimum weight of the speed
reducer so that the conditions of constraints on bending
stress of the gear teeth, surface stresses, transverse
deflections of the shafts and and other constraints analyzed
by Mezura and Coello could be satisfied [11].
H. Eskandar et al. [5] obtained, by applying the new
Water cycle algorithm for solving these problems, one of
the best minima for optimization of spring, pressure vessel
and speed reducer design.
By applying Improving proposed algorithm harmony
search (IPHS) for solving these problems, Jaberipour and
Khorram [4] obtained significant minima, especially in
optimization of welded beam design.
Carlos A. Coello Coello [12] used Genetic algorithms
(GA) for solving numerical examples.
Zhao et al. [13] applied Hybrid GA with flexible allowance
technique (GAFAT) algorithm to solving these engineering
problems and obtained significant minima, particularly in
problems of speed reducer optimization.

*Kontakt adresa autora: Fakultet za mainstvo i graevinarstvo, Dositejeva 19, 3600 Kraljevo, Srbija, grkovic.v@mfkv.kg.ac.rs

EN116

IMK-14 Research & Developement

max

1 MAIN ANT COLONY OPTIMIZATION


ALGORITHMS

ij ij (1 ) + ij best

Several Ant Colony Optimization (ACO) algorithms have


been proposed in literature. The original Ant System and
two successful variants: Max-min Ant System and Ant
Colony System will be presented here. In order to illustrate
the difference between these three algorithms, the
traveling salesman problem will be used as a concrete
example.
1.1 Ant system (AS)
The Ant System is the first ACO algorithm. Its main
characteristic is that, at each iteration, the pheromone
values are updated by all the m ants that have built a
solutions in the iteration itself. The quantity of pheromone
ij which is deposited on the edge that connects the cities i
and j is updated in the following way:
m

ij ij (1 ) + ij

(1)

k =1

where is the evaporation rate, m is the number of


ants, and ij is the quantity of pheromone on the edge (i,
j) by the ant k:

ij k

(2)

In the construction of their solution, ants select the


following city to be visited through a stochastic
mechanism. When the ant k is in the city i and if a partial
solution s p has been constructed so far, the probability of
going to the city j is given by:
p

il

( ) is the set of feasible components, i.e. edges


p

(i,l), where l is a city which has not been visited by the


ant k yet. The parameters and control the relative
importance of the pheromone value versus the heuristic
information ij , which is obtained in the following way:
ij =

is defined as:

a if x > a,

=
[ x ] b if x < b,
0 otherwise.

and ij best is
a

(6)

1 , if (i , j ) belongs to the best tour,

(7)
ij = Lbest
0, otherwise.

where Lbest is the length of tour of the best ant. It may be


best

either the best tour found in the current iteration or


iteration-best, Lib , or the best solution found since the
start of the algorithm or best-so-far, Lbs , or a combination
of both.
As for the upper and lower bounds of pheromone values,
min and max , they are commonly obtained empirically.
and max on the basis on analytical considerations.

The most interesting contribution of ACS is the


introduction of a local pheromone update in addition to the
pheromone update performed at the end of the
construction process (it is called offline pheromone
update). The local pheromone update is performed by all
the ants after each construction step. Each ant applies it
only to the last edge covered:

ij ij (1 ) + 0
where

(3)

where N s

[ x]

1.3 Ant Colony System (ACS)

where Q is a constant, and Lk is the length of the tour


constructed by the ant k.

il

where max and min are respectively the upper and the
lower bounds imposed on the pheromone. The operator

Nonetheless, there are some guidelines for defining min

Q , if ant k used the edge i , j

=
L
0, otherwise.

ij ij
, if c N ( s

pijk = il il
c N s
( )
0, otherwise.

(5)

min

(4)

d ij

where d ij is the distance between the cities i and j.


1.2 MAX MIN Ant System (MMAS)
This algorithm is an improvement over the original AS. Its
main characteristic is that only the best ant updates the
pheromone trails. The difference in the pheromone value
is realized in the following way:

( 0,1]

is

the

(8)
coefficient

of

pheromone

evaporation, and 0 is the initial value of pheromone.


The main goal of the local pheromone update is to
diversify the search performed by subsequent ants during
an iteration: by decreasing the pheromone concentration
on the traversed edges, the subsequent ants are encouraged
to choose other edges, and, consequently, to produce
different solutions. It provides a smaller probability that
several ants will produce identical solutions during one
iteration.
The offline pheromone update is similar to MMAS and it is
applied at the end of each iteration by only one ant, which
can be either iteration-best or the best-so-far. However,
the update formula is slightly different:

{ ij (1 ) + ij , if ( i , j ) belongs to the best tour,


(9)
ij
ij , otherwise.
As in MACA, ij best =
1 / Lbest where Lbest can be either
Lib or Lbs .

Grkovi, V. - Bulatovi, R.

EN117

IMK-14 Research & Developement

Another important difference between ACS and AS is in


the decision rule used by the ants during the construction
process. In ACS, the following rule is used: the probability
for an ant to move from the city i to the city j depends on a
random variable q uniformly distributed over [0, 1], and a
parameter q0 , if q q0 , then j = arg max c N
il

(s

},
){

il

il

otherwise the following equation is used:

ij ij
, if cil N ( s p )

pij = il il
c N s
( )
0, otherwise.
p

x ( i )best - the previous best solution,

h - the arbitrary value of

, h ] for the

iteration h , it is chosen by the option rand

h - the value for the correction of solution for the


iteration h , in the initial iteration this value is
given in the code and its value is chosen in relation
to the range of possible solutions of the variables

( x1 , x2 , x3 , ...) , and for each following iteration it is


(10)

il

corrected (pheromone evaporates) in the following


way:

=
0, 2 h 1
h

2. MODIFIED ANT COLONY ALGORITHM (MACA)

(12)

After obtaining new solutions (for each ant), the values of


the functions f ( x ) are determined for new values and
it is evaluated whether they are better than the previous
best solution f ( x )min . If the new value of the function is
new

The modified ant colony algorithm is based on the


improved MAX MIN Ant System algorithm (MMAS)
(1.2). Its main modification is that only the best ant
updates the path with the pheromone trail, i.e. in this case
only the solution of the best ant is taken, and then all other
ants vary around that solution.
In the initial phase it is necessary to define the number of
ants m , which represents a very important parameter. The
quality of the initial solution as well as the number of
iterations I , which represents the number of corrections in
the second loop, depend on this parameter. When the
algorithm starts, the same (fixed) quantity of pheromone
= 1 is placed on the whole function f(x) so that each ant
could arbitrarily (randomly) choose one location on the
function. In the first step (loop), all ants ( m ) randomly
choose the location they will reach (which values of the
variables x1 , x2 , x3 , ... they will take). When each ant finds
its location, then the value of the function f(x) which is
optimized for the values of variables ( x1 , x2 , x3 ,...) chosen
by the ant is determined. When the values of the function are
obtained for each ant, it is evaluated which ant has chosen
the best solution, and the value of that solution is recorded
f(x) min . The best soluton should be better than the
approximate solution (the approximate solution must be
given in the code), and if it is not better then the loop must
be passed through again. When a solution better than the
approximate one is obtained, it is saved as well as the
values of the variables ( x1 , x2 , x3 ,...) and then it is possible
to pass to the second step. In the second step (loop), all ants
take the location of the best solution from the first step
f ( x )min and vary around that value for the value which
represents an arbitrary value of [ , ] . New solutions
obtained by varying around the previous best one are
obtained by the equation:

x (=
i )h
x ( i )best h
new

(11)

where:

i - the number of variables,


h - the number of iteration,

x ( i )h

new

- the new solution for the iteration h ,

better than the previous best solution f ( x ) < f ( x )min ,


then the best solution will be the new value
new

f ( x )min = f ( x )

new

. Each better solution is saved as well

as the values of its variables ( x1 , x2 , x3 , ...) .


After that pheromone is updated by the equation:

( h )= ( h 1) + 0, 01 f ( x )min

(13)

This procedure described by the second step is one


iteration inside the second loop and it will be repeated up
to the maximum given number of iterations (correction of
solutions). If a solution better than the required one (the
required solution must be defined in the code) is not
obtained even after the given number of iterations, then the
whole loop must be passed through again. When the
solution which is better than the required one is obtained,
then that solution and the values of its variables are
recorded in a new file.
Relying on the described principles, it is now possible to
present a simple optimization algorithm (code 1).
repeat until the end repeat for each ant
create an arbitrary solution evaluate the solution
choose the ant with the best solution
determine the best solution
repeat the end
repeat for each ant
create a new solution (varying around the best
solution)
evaluate the solution in relation to the currently
best one
repeat the end
update pheromone trails
end of repetition

Code 1. Modified ant colony algorithm


3. EXPERIMENTAL ENGINEERING EXAMPLES FOR
OPTIMIZATION
The main problem with these examples is to find the
optimum solution (in this case minimum) which must
satisfy a series of given constraints. Optimization with

Modifikovani algoritam kolonije mrava za reavanje inenjerskih optimizacionih problema

EN118

IMK-14 Research & Developement

constraints has always been a field of great interest.


Generally speaking, optimization problems with a series of
constraints can be formulated in the following way:

subject
to
the
Minimize:
f ( x ) , x Rn

x23 x3
0
g1 ( x ) =
1
71785 x14
=
g2 ( x )

following constraints:

g i ( x ) 0, 1 i < l

h j ( x=
) 0, l + 1 j m

(14)

T
x = [ x1 , x2 , x3 ,..., xn ] denotes

vector,
l denotes the number of inequalities in constraints,
m l denotes the number of equations
in constraints.

n th

dimensional

This chapter will present certain examples of engineering


problems, such as: optimization of spring, pressure vessel,
welded beam and speed reducer. The basis of the problem,
the objective function, variable parameters that should be
found as well as the constraints that should be respected
will be shown. Then the results obtained by the MACA
method (Chapter 2) will be presented and they will be
compared to the optimum results for these four examples
obtained and published so far. Analysis and obtaining of
results by MACA were performed in the code written in
MatLAB 2009.
3.1 Optimization of the tension/compression spring
This problem consists of minimization of spring weight
(Figure 1) subject to constraints on minimum deflection,
shear stress, surge frequency, limits on the outside
diameter and design variables. The variables which should
be optimized are

12566 x x x
3
2 1

4
1

1
5108 x12

140, 45 x1
g3 ( x ) =
1
0
x22 x3
g 4=
( x)

where:

4 x22 x2 x1

x1 + x2
1,5

1 0

0, 05 x1 2 ; 0, 25 x2 1,3 ; 2 x3 15
By applying MACA to solving this problem, the number
of ants m=10000 was used and a slightly better solution
0.012650712535 was obtained.
A detailed presentation of the results obtained by the
MACA method and comparison of several best results
obtained by using other algorithms are given in Table 1.
3.2 Optimization of the pressure vessel
This problem consists of a cylindrical vessel which is
capped at both ends by hemispherical heads as it is shown
in Figure 2. The goal is to reduce the total costs, including
the costs of material, forming and welding. There are four
variables that should be optimized:

the shell thickness T s (x 1 ),


the thickness on the head T h (x 2 ),
the inner radius R(x 3 ), and
the length of the cylindrical section of the vessel,
excluding the heads L(x 4 ).

The problem is reduced to minimization of the function:


0, 6224 x1 x3 x4 + 1, 7781x2 x32 + 3,1661x12 x4 + 19,84 x12 x3
subject to the following constraints:
g1 ( x ) =
x1 + 0, 0193 x3 0

the coil diameter D(x 2 ),


the wire diameter d(x 1 ) and
the number of active coils N(x 3 ).

g2 ( x ) =
x2 + 0, 00954 x3 0
g3 ( x ) =
x3 x4
2

4
3

x3 + 1296000 0
3

g 4 ( x ) =x4 240 0
g 5 ( x ) = 1,1 x1 0
g 6 ( x ) = 0, 6 x2 0
0 x1 100 ; 0 x2 100
10 x3 200 ; 10 x4 200

Fig. 1. Schematic view of the spring with variable


parameters
Formally, the problem can be expressed as minimization
of the function ( x3 + 2 ) x2 x12 subject to the following
constraints:

Fig. 2. Schematic view of the pressure vessel with variable


parameters
Grkovi, V. - Bulatovi, R.

IMK-14 Research & Developement

EN119

6 PL
4 PL3
( x) =
, ( x)
.
By applying MACA to solving this problem, the number of =
2
3
x
x
Ex
x
4
3
3
4
ants m=100000 was used and a solution slightly worse
than the best one, but still considerably better than certain
x2 x6
previous algorithms, was obtained and it is
4, 013E 3 4
x3
E
36
5964.505957036.
=
Pc ( x )
1
,
2
L
2 L 4G

A detailed presentation of the results obtained by the


MACA method and comparison of several best results
P
= 6000lb ; L
= 14in ; E
= 30 106 psi
obtained by using other algorithms are given in Table 2.
G=
12 106 psi ; max =
13600 psi
=
max 30000
=
psi ; max 0, 25in.

3.3 Optimization of the welded beam

The welded beam structure consists of the beam A and the


weld required to hold the beam to member B (Figure 3). It
is necessary to optimize the welded beam for the minimum
cost subject to constraints on shear stresses ( g1 ) ,
bending stress in the beam ( g ) , buckling load P ( g ) ,
2

end deflection of the beam ( g 6 ) and side constraints.


The four variables that should be optimized which are
shown in Figure 3 are:

the size of the weld h(x 1 ),


the length of the welded section of the beam l(x 2 ),
the beam width t(x 3 ),
the beam thickness b(x 4 ).

The problem consists of minimization of the function


2
1,10471x1 x2 + 0, 04811x3 x4 (14 + x2 ) subject to the following
constraints:
g1 ( x ) = ( x ) max 0
g 2 ( x ) = ( x ) max 0

Fig. 3. Schematic view of the welded beam with variable


parameters
By applying MACA to solving this problem, the number of
ants m=10000 was used and a solution slightly worse than
[4] was obtained, but it was still considerably better than
certain
previous
algorithms;
that
solution
is
1.729335567988.
A detailed presentation of the results obtained by the
MACA method and comparison of several best results
obtained by using other algorithms are given in Table 3.

g 3 ( x ) = x1 x4 0
=
g 4 ( x ) 0,10471x1 + 0, 04811x3 x4 (14 + x2 ) 5 0
2

g 5 ( x=
) 0,125 x1 0
g 6 ( x ) = ( x ) max 0

In this optimization problem it is necessary to find the


minimum weight of the speed reducer so that the
conditions of constraints on bending stress of the gear
teeth, surface stress, transverse deflections of the shafts
and other constraints could be satisfied. There are seven
variables (Figure 4):

g7 ( x ) =
P Pc ( x ) 0
0.125 x1 10 ; 0.1 x2 10
0.1 x3 10 ; 0.1 x4 5

where:

( x ) =( ' ) + 2 ' ''


2

=
'

''
,=

2 x1 x2

=
R

x22
2

2R

+ ( '' )

M P L +
,=

x1 + x3
,
2

x2

the width between the shafts b(x 1 ),


the module of the teeth m(x 2 ),
the number of teeth in the pinion z(x 3 ),
the length of the first shaft between the bearings
l 1 (x 4 ),
the length of the second shaft between the
bearings l 2 (x 5 ),
the diameter of the first shaft d 1 (x 6 ),
the diameter of the second shaft d 2 (x 7 ).

2
x1 + x3
,
12 2

x22

=
J 2 2 x1 x2

MR

x2

3.4 Optimization of the speed reducer

This is an example of a mixed integer programming


problem. The third variable z ( x3 ) (the number of teeth) is
an integer. The problem can be expressed as minimization
of the function

Modifikovani algoritam kolonije mrava za reavanje inenjerskih optimizacionih problema

EN120

IMK-14 Research & Developement

0, 7854 x1 x2 3, 3333 x3 + 14, 933 x3 43, 0934 1, 508 x1 x6 + x7


2

+7, 4777 x6 + x7 + 0, 7854 x4 x6 + x5 x7


3

subject to the following constraints:


g1 =
( x)
g2 =
( x)

27

1 0

x1 x2 x3
397, 5
2

1 0

x1 x2 x3

5 x2

g 9 ( x=
)

x1

1 0

x1
1 0

12 x2

=
g11 ( x )

1,1x7 + 1, 9

=
g10 ( x )

1,5 x6 + 1,9

1, 93 x4

g4=
( x)

1, 93 x5

1 0

x4

1 0

2, 6 x1 3, 6 ; 0, 7 x2 0,8 ; 17 x3 28

x2 x6 x3

7, 3 x4 8, 3 ; 7, 3 x5 8, 3 ; 2, 9 x6 3, 9

1 0

x5

g3=
( x)

1 0

5, 0 x7 5, 5

x2 x7 x3

By applying MACA in solving this problem the number of


ants m=100000 was used and an approximate solution,
which was slightly worse than [16] and [5], but still
considerably better than certain previous algorithms, was
obtained and it is 3009.6692386349.

g5 ( x )

g 8 ( x=
)

745 x4
6

+ 16,9 10
x2 x3
110 x63

745 x
x x + 157, 5 10

1 0

g6 ( x )
g 7 ( x=
)

A detailed presentation of the results obtained by the


MACA method and comparison of several best results
obtained by using other algorithms are given in Table 4.

1 0

85 x7
x2 x3

1 0

40

Fig. 4. Schematic view of the speed reducer with variable parameters


Table 1. Comparison of results for optimization of the spring
Variable values

x1
x2
x3
g 1 (x)
g 2 (x)
g 3 (x)
g 4 (x)
f(x)

Belegundu [7]
0.050000
0.315900
14.25000
- 0.000014
- 0.003782
- 3.938302
- 0.756067
0.0128334375

Arora [6]
0.053396
0.399180
9.185400
0.000019
- 0.000018
- 4.123832
- 0.698283
0.0127302737

Coello [12]
0.051989
0.363965
10.890522
- 0.00126
- 0.000025
- 4.061337
- 0.722697
0.012681

Grkovi, V. - Bulatovi, R.

Eskandar [5]
0.051680
0.356522
11.300410
-1.65-13
- 7.9-14
- 4.053399
- 0.727864
0.012665

MACA
0.052310747184
0.372242379221
10.414160013342
680.655755e-6
835.529214e-6
- 4.091394977807
-1.075446873594
0.012650712535

EN121

IMK-14 Research & Developement

Table 2. Comparison of results for optimization of the pressure vessel


Variable values

Sandgren [9]

x1
x2
x3
x4
g 1 (x)
g 2 (x)
g 3 (x)
g 4 (x)
f(x)

1.125
0.625
47.700
117.701
- 0.204390
- 0.169942
54.226012
- 122.299000
8129.1036

Kannan,
Kramer [8]
1.125
0.625
58.291
43.690
0.000016
- 0.068904
- 21.220104
- 196.310000
7198.0428

Coello [12]

Eskandar [5]

MACA

0.8125
0.4375
40.3239
200.0000
- 0.034324
- 0.052847
- 27.105845
- 40.00000
6288.7445

0.7781
0.3846
40.3196
- 200.0000
-2.95e-11
- 7.15e-11
- 1.35E-06
- 40.0000
5885.3327

0.82217025865521
0.40638075763737
42.6028471508185
170.484922696647
64.691355586e-6
50.404181438e-6
227.89696231e-6
- 69.515077303352
5964.5059570367

Coello [12]

Eskandar [5]

MACA

0.2088
3.4205
8.9975
0.2100
- 0.337812
- 353.9026
- 0.00120
- 3.411865
- 0.08380
- 0.235649
- 363.23238
1.74830941

0.205728
3.470522
9.036620
0.205729
- 0.034128
- 3.49e-05
- 1.19e-06
- 3.432980
- 0.080728
- 0.235540
- 0.013503
1.724856

0.202611802838422
3.538184790851369
9.038385657692625
0.205720941513918
4.79880374405e-3
-10.4259994343302
-3.1091386754e-3
-3.42682278102344
77.6118028384e-3
-235.548165281e-3
-7.86546916333e-3
1.729335567988

Table 3. Comparison of results for optimization of the welded beam


Variable values

x1
x2
x3
x4
g 1 (x)
g 2 (x)
g 3 (x)
g 4 (x)
g 5 (x)
g 6 (x)
g 7 (x)
f(x)

Zhao [13]
0.20572964
3.47048867
9.03662391
0.020572964
1.724852

Jaberipour,
Khorram [4]
0.20573
3.47049
9.03662
0.20573
0.0
0.0
5.55e-17
- 3.4329
- 0.0807
- 0.2355
- 9.09e-13
1.7248

Tabele 4. Comparison of results for optimization of the speed reducer


Variable values

x1
x2
x3
x4
x5
x6
x7
f(x)

Zhao [13]
3.50000000
0.70000000
17.0000000
7.30000000
7.71531991
3.35021466
5.28665446
2994.4710

Jaberipour, Khorram [4]


3.500
0.700
17
7.3
7.715332338339
3.350215109256
5.286664035454
2994.4

4. CONCLUSION
This paper describes application of a metaheuristic
algorithm, the modified ant colony algorithm MACA, in
optimization of examples of engineering problems with
the constant number of variables. The use of this algorithm
results in obtaining optimum or approximate results, which
is shown by examples. In this case one result which is
better than the previous results is obtained, and the other
ones are approximate but optimum. Further research and
development of this algorithm may modify and improve
this method for the purpose of obtaining even better and
more optimum results.

[3]

[4]

[5]

5. REFERENCES
[1] Marco Dorigo & Thomas Sttzle: Ant Colony
Optimization, 2006.
[2] Denebourg, J. L., Aron, S., Goss, S., & Pasteels, J.
M.: The self-organizing exploratory pattern of the

[6]
[7]

Eskandar [5]
3.50000
0.70000
17
7.30000
7.715319
3.350214
5.286654
2994.471066

MACA
3.4999999994707
0.6999999999999
17
8.0517801843653
8.0846770710370
3.3516892746758
5.2867821117062
3009.6692386349

Argentine ant. Journal of Insect Behaviour, 3, 1990,


159-168.
M. Duran Toksari: Ant colony optimization for
finding the global minimum. Applied Mathematics
and Computation 176 (2006) 308316.
Majid Jaberipour, Esmaile Khorram: Two improved
harmony search algorithms for solving engineering
optimization problems. Commun Nonlinear Sci
Numer Simulat 15 (2010) 33163331
H. Eskandar, A. Sadollah, A. Bahreininejad, M.
Hamdi: Water cycle algorithm A novel
metaheuristic optimization method for solving
constrained engineering optimization problems.
Computers and Structures 2012.
J.S. Arora: Introduction to Optimum Design,
McGraw-Hill, New York, 1989.
A.D. Belegundu: A Study of Mathematical
Programming Methods for Structural Optimization.

Modifikovani algoritam kolonije mrava za reavanje inenjerskih optimizacionih problema

EN122

IMK-14 Research & Developement

Dept. of Civil and Environmental Engineering,


University of Iowa, Iowa, IA,1982.
[8] B.K. Kannan, S.N. Kramer: An augmented Lagrange
multiplier based method for mixed integer discrete
continuous optimization and its applications to
mechanical design. Journal of Mechanical Design,
Transactions of the ASME 116_1994. 318320.
[9] Sandgren E.: Nonlinear integer and discrete
programming in mechanical design optimization. J
Mech Des ASME 1990;112(43):2239.
[10] S. S. Rao: Engineering Optimization, 3rd edn.,
JohnWiley and Sons, 1996.

[11] Mezura-Montes E, Coello CAC.: Useful infeasible


solutions in engineering optimization with
evolutionary algorithms, MICAI 2005. Lect Notes
Artif Int 2005; 3789 (2005): 65262.
[12] Carlos A. Coello Coello: Constraint-handling in
genetic algorithms through the use of dominancebased tournament selection. Advanced Engineering
Informatics, 16 (2002): 193203.
[13] Jia-qing Zhao, Ling Wang, Pan Zeng, Wen-hui Fan:
An effective hybrid genetic algorithm with flexible
allowance technique for constrained engineering
design optimization.
Expert
Systems
with
Applications 39 (2012) 60416051.

Grkovi, V. - Bulatovi, R.

Você também pode gostar