Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/271452623
READS
10
4 AUTHORS, INCLUDING:
Mir Mohammad Ettefagh
Hasan Biglari
University of Tabriz
University of Tabriz
11 PUBLICATIONS 54 CITATIONS
SEE PROFILE
SEE PROFILE
Milad Azvar
Sabanci University
1 PUBLICATION 0 CITATIONS
SEE PROFILE
H. Emdadi
Department of Computer
Science
University of Tabriz
Tabriz, Iran
habib.emdadi@yahoo.com
I.
Beam;
Modal
INTRODUCTION
II.
+Q16L
PROBLEM FORMULATION
A. Analytical model
Fig. 1 shows an analytical model of a laminated
composite beam composed of several layers.
+C22L cos 4
Fig. 1.
C11 =
I1
2 w0
t w
4 w0
I 3 2 0 2 + EI eq
=0
2
t
x t
x 4
where:
B2
EI eq = b D11 11
A11
(1)
C12 =
E1
E2
, C22 =
1 12 21
1 12 21
E1 ( 21 )
1 12 21
, C44 = G12
A11 = b Q11*L ( z L z L 1 )
L =1
z 2 zL2 1
B11 = b Q L
2
L =1
3
3
n
z
L 1
D11 = b Q11*L L
3
L =1
*L
11
Det
( sin ( l ) + sinh ( l ) )
( 2 cos ( l ) + 2 cosh ( l ) )
( 2 cos ( l ) + 2 cosh ( l ) )
3 sin ( l ) 3 sinh ( l )
=0
(2)
where
2
= q + q + 4 pr
2p
q + q2 + 4 pr
=
2p
, p = EIeq ,
q = I3 2 , r = I1 2
(3)
f obj = (i i ) 2
i =1
(4)
5 GPa E1 30 GPa
Now, it is optimization algorithms duty to minimize the
object function and identify the elastic constant.
III. METAHEURISTIC ALGORITHMS
A. ABC Algorithm
The artificial bee colony (ABC) algorithm is a new
population-based stochastic heuristic approach,
introduced for the first time by Karaboga in 2005 [8].
This algorithm is very simple and flexible, which does
not require external parameters like crossover rates,
especially suitable for engineering application. The
colony of artificial bees consists of three groups of bees
to search foods, which includes employed bees,
onlookers and scouts.
Accelerometer
Force Transducer
Composite Beam
Shaker
Fig. 2.
C. Design objective
Before defining design objective, definition of
frequency error must be presented, which is defined as
the square of deviation between measured frequency
( i ) and analytically calculated frequency ( i ), as
described in Section B and A, respectively. Therefore,
the objective function is written as a mathematical
equation as follows:
B. PSO Algorithm
The PSO method introduced for the first time in
1995 by Kennedy and Eberhart [9]. This method is
based on social behavior of organisms such as fish
schooling and bird flocking. Because of its simple
concept and quick convergence, PSO can be applied to
various applications in different fields. The approach
uses the concept of population and a measure of
performance like the fitness value used with other
evolutionary algorithms. Similar to other evolutionary
algorithms, the system is initialized with a population of
random solutions, called particles. Each particle
TABLE I.
E2
G 12
12
5.21GPa
1.83GPa
0.36
TABLE II.
PSO
ABC
TS
SA
Error =
( f1 268)
+ ( f 2 736 ) + ( f3 1440 )
2
(5)
CASE STUDY
where the values of 268, 736 and 1440 (in Hz) are the
natural frequencies of the composite beam, derived by
experimental modal analysis as described in section II.
In Fig. 3, the frequency response function of the beam
has been shown. In this figure the mentioned natural
frequency of the beam can be observed.
TABLE V.
st
Experiment
f1
f2
f3
Identified
Elastic constant
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
Average
Best
Worst
262.92
735.93
1441.8
23539499969.6303
5.381
11.117229
262.76
735.61
1441.3
24273528274.6193
5.4097
11.231237
262.92
736.09
1442.3
32866790080.2181
5.558
11.026652
262.92
735.93
1441.8
31575589535.0553
5.381
11.023517
262.76
735.45
1441
12709107167.0193
5.3556
11.061685
262.76
735.45
1440.8
34132330047.2828
5.3285
11.177567
262.61
735.3
1440.5
27418342727.0149
5.4641
11.065933
262.76
735.61
1441.1
10769611050.2733
5.3735
11.088365
262.61
735.3
1440.5
21270177829.81
5.4641
11.069687
262.92
735.93
1441.8
28312693572.6635
5.381
11.009096
24686767025
5.4096
11.0871
10769611050
5.3285
11.0091
34132330047
5.558
11.23124
Error
CPU
Time
nd
2 Frequency (736 Hz)
Fig. 3.
Frequency response function of the beam, extracted by
experimental modal analysis
Experiment
f1
f2
f3
Identified
Elastic
constant
Error
CPU
Time
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
Average
Best
Worst
262.76
735.45
1440.8
25080361099.8693
5.3285
30.911357
262.76
735.45
1440.8
22762260251.5533
5.3285
30.534592
262.76
735.45
1440.8
26307411152.9975
5.3285
31.294821
262.76
735.45
1440.8
41361733546.9444
5.3285
31.468497
262.76
735.45
1440.8
25133369513.629
5.3285
29.856366
262.76
735.45
1440.8
46006136592.5867
5.3285
31.999353
262.76
735.45
1440.8
25082300588.272
5.3285
30.068815
262.76
735.45
1440.8
35582371137.5373
5.3285
31.508160
262.76
735.45
1440.8
25121571605.5497
5.3285
30.762982
262.76
735.45
1440.8
20702609477.3485
5.3285
30.735458
TABLE IV.
29314012497
5.3285
30.91404
20702609477
5.3285
29.85637
46006136593
5.3285
31.99935
TABLE VI.
Experiment
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
Average
Best
Worst
f1
f2
f3
Identified
Elastic
constant
Error
CPU
Time
263.4
737.52
1445
23721197124.5346
6.9389
12.282198
262.29
734.34
1438.8
27172203067.5831
6.0764
12.255190
263.72
738.16
1446.4
10244320741.3003
7.9969
12.410953
263.24
736.89
1443.9
25224710983.6245
6.1866
12.441645
262.13
733.86
1437.8
14976271892.8799
6.6225
12.380435
263.24
736.73
1443.5
30500827746.0295
5.972
12.518850
12.423675
262.45
734.98
1439.9
21661556749.5209
5.6482
262.76
735.61
1441.1
30879303453.1958
5.3735
12.181088
262.76
735.61
1441.3
22483529978.2795
5.4097
12.340560
262.29
734.18
1438.3
13116927031.4515
6.236
12.418477
21998084877
6.24607
12.36531
10244320741
5.3735
12.18109
30879303453
7.9969
12.51885
Experiment
f1
f2
f3
Identified
Elastic
constant
Error
CPU
Time
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
Average
Best
Worst
262.76
735.45
1440.8
25111420602.7902
5.3285
21.725442
262.76
735.45
1440.8
25111019341.6118
5.3285
21.798221
262.76
735.45
1440.8
25110049454.8997
5.3285
21.773744
262.76
735.45
1440.8
25110491037.4499
5.3285
21.700789
262.76
735.45
1440.8
25111125973.0664
5.3285
21.615837
262.76
735.45
1440.8
25110083759.024
5.3285
21.657098
262.76
735.45
1440.8
25110944378.4238
5.3285
21.633943
262.76
735.45
1440.8
25109279966.035
5.3285
21.587906
262.76
735.45
1440.8
25109435893.2731
5.3285
21.998107
262.76
735.45
1440.8
25110854295.7699
5.3285
21.475473
25110470470
5.3285
21.69666
25109279966
5.3285
21.47547
25111420603
5.3285
21.99811
Fig. 4.
Identified parameter (elastic constant) for different
algorithms
Fig. 5.
Fig. 7.
Error values variation versus 100 runs for different
algotrithms
Fig. 6.
Cost value variation versus 100 runs for different
algotrithms
Fig. 8.
Computational time versus generations for different
algorithms
[2]
[3]
[4]
[5]
[6]
Fig. 9.
Convergence plot versus generations for different
algorithms
[7]
[8]
[9]
[10]
[11]
V.
CONCLISION
[12]
[13]
[14] S.H. Sishaj and P. Simon, Artificial bee colony algorithm for
economic load dispatch proble with non-smooth cost
functions, Electric Power Components and Systems, vol.
38(7), pp. 786-803, 2010.
[15] F. Kang, J. Li, H. Li, Z. Ma and Q. Xu, An improved artificial
bee colony algorithm, IEEE 2nd International Workshop on
Intelligent Systems and Applications (ISA), 2010.
[16] H. Emdadi, M.Yazdanian, M. M. Ettefagh and M. R. FeiziDerakhshi, Double four-bar crank-slider mechanism dynamic
balancing by meta-heuristic algorithms, International Journal
of Artificial Intelligence & Applications (IJAIA), vol. 4(5),
September 2013.
[17] M. M. Etttefagh, M. Abbasi and H. Emdadi, , Path synthesis
of the four-bar mechanism using ABC algorithm and
comparing with BGA, proceeding of IEEE Internationa
Symposium on Innovations in Intelligent Systems and
Applications 2013 (INISTA 2013),.