Professional Documents
Culture Documents
Science
Science
Composite Structures
journal homepage: www.elsevier.com/locate/compstruct
a r t i c l e
i n f o
Article history:
Available online 17 May 2008
Keywords:
Multiobjective optimization
Genetic algorithms
Composites laminated structures
Finite element analysis
a b s t r a c t
A technique for the design optimization of composite laminated structures is presented in this work. The
optimization process is performed using a genetic algorithm (GA), associated with the nite element
method (FEM) for the structural analysis. The GA is adapted with special operators and variables codication for the specic case of composite laminated structures optimization. Some numerical examples are
presented to show the exibility of this tool to solve different kinds of problems. Two cases of multiobjective optimization of plates under transverse or in-plane load are studied. In these examples the minimization of two objectives, such as weight and deection or weight and cost, are simultaneously
performed and a pareto-optimal set is obtained by shifting the optimization emphasis using a weighting
factor. The stiffness maximization of a composite shell under pressure load is presented in the last example, where the geometrically nonlinear behavior of the structure is considered. Some aspects of the optimization performance, such as the apparent reliability and the computational cost, are investigated in
each application.
2008 Elsevier Ltd. All rights reserved.
1. Introduction
In the recent decades, structural applications with composite
laminated materials have been growing, requiring great effort on
the development of analysis and design techniques. An advantage
of using ber-reinforced composites over conventional materials is
that the former can be tailored to specic requirements of certain
applications [1]. However, the large number of design variables
and the complex mechanical behavior associated with such materials turn the structural design much more difcult and laborious
than those involving conventional materials. These characteristics
have motivated the use of optimization methods in the sense of
turn the composite material structural design a more systematic
and well dened task, becoming less dependent to the designer
sensitivity and achieving the maximum material performance [2].
The earlier works in the eld of composite structures optimization employed the same methods already used to optimize conventional material structures. These methods are based on gradients of
the objective and constraints functions with respect to the design
variables, which are considered to be continuous in the design
space. Such works resulted in limited success because composite
laminate design falls on a discrete optimization problem, since in
practice the variables are restricted to few values imposed by the
manufacturing process. Moreover, the composite optimization
* Corresponding author. Tel.: +55 51 3308 3587; fax: +55 51 3308 3999.
E-mail address: amawruch@ufrgs.br (A.M. Awruch).
0263-8223/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2008.05.004
444
The organism tness is obtained directly from an objective function using simple structure information and gradient evaluations
are not required.
In real designs cases, when the structural geometry is usually
complex and the prediction of the structural behavior must be
accurate, it is necessary to use numerical tools, such as the FEM,
for the structural analysis. These methods are computationally
expensive and may turn the optimization processes with GA
impracticable when a large amount of analyses is required. Many
researchers have proposed modications to the classical GA structure to take advantage of composite laminates characteristics and
minimize the computational cost. Some of these new strategies
are applied in this work, consisting essentially in a GA restructuring of the variable codication and the genetic operators.
When more than one objective is handled in the design process,
a multiobjective optimization problem may be solved considering
all the objectives simultaneously, providing a set of optimum designs (pareto-optimal set), depending of the emphasis given to
each one of the objectives. The pareto-optimal set may be very useful when the critical objective is not known a priori. Various
researchers have studied the problem of multiobjective optimization of laminated structures, but the use of this approach together
with GA and the nite element method (FEM) have not been
widely explored.
In the present work, two examples of multiobjective optimization of composite laminate plates using GA and FEM are studied. A
third application deals with the single optimization of a semicylindrical shell considering the geometrically nonlinear behavior
of the structure.
very common constraint adopted in most of optimization problems, and it is used in this work too. The TsaiWu failure criterion
[7] is used for the failure prediction in a ply. A safety factor against
failure kf can be obtained in linear problems with the TsaiWu failure function using the material strength parameter for traction,
compression and shearing at each of the principal material axes.
In geometrically nonlinear analysis the structural failure is veried
at each load step of the incremental solution method, which is
stopped if material failure is detected.
445
more of its characteristics. This may lead to a less random operation, which is supposed to be more efcient in guiding the evolution towards optimization objectives. New operators named
orientation alteration, material alteration, ply addition and ply
deletion, are introduced to replace the classical mutation in the
present GA [8,9]. A description of these new operators is presented
below, followed by an example of their applications as it is shown
in Fig. 3.
Orientation and material alterations are implemented similarly
to the classical mutation, but are independently applied to orientation and material chromosomes, respectively. Different orientation
and material operator probabilities (poa and pma) may be adopted,
which is useful, since each chromosome may converge at different
velocities in most optimization processes. Additionally, these operators are not applied to genes with empty stack code, when it is
present in the chromosome.
Two distinct operators are used to vary the number of layers in
the laminate. The rst one, named ply addition, acts in the chromosomal strings, at a given probability ppa, introducing a new layer
close to the laminate mid-plane (end of chromosomes), and
removing an existing pair of genes with empty stack code. The
genes that represent the new layer are randomly created, assuming
any value contained in each respective alphabet codication. In an
opposite way, the ply deletion operator acts taking out the last
pair of genes of the organism chromosomes (innermost layer)
and adding a pair of genes with empty stack code at the outermost
position. This operator is applied with a given probability ppd. Both
operators manipulate at the innermost laminate layer because it
has little effect on bend or twist behavior of the plate or shell, causing no abrupt changes in the design.
Fig. 3. (a) Orientation and material alteration; (b) ply addition; (c) ply deletion.
446
PN
An
i
i1 X g P
where X ig is the total number of generations analyzed in the ith optimization procedure. In GA optimization procedures it is very common to occur repeated analyses because one specic design may
appear in many generations during the process. When a memory
containing information about the performed analyses is used, associated to the GA, these repeated analyses can be avoided, resulting
in an important reduction of the computational cost. In these cases,
another measurement of the computational cost can be stated tacking the average number of the effectively performed analyses (Ar),
obtained dividing the number of analyses that were effectively carried out by the number of GA executions (N).
The criterion to stop the optimization process, which was used
in all examples presented here, is based in two parameters: the
upper limit of the number of generations (NLG) and the maximum
number of generations with no improvement of the best design
(NSD). Once one of these limits is reached, the optimization process
is stopped and the best laminate of the last generation is taken as
the optimization result. NLG and NSD are dened in each optimization procedure, depending on the complexity of a specic problem.
4.1. Weight and deection minimization of a composite laminated
plate under transverse load
This example deals with the design of a composite laminated
square plate, subjected to a uniform pressure load on its surface.
Minimization of the structural weight and deection are the design
objectives. The two objectives must be considered in conjunction
with the constraints imposed by material failure and maximum
values of contiguous plies thickness with the same ber orientation. These are two opposite objectives, since improvements in
one of them leads to depreciation of the other, and they must be
considered at the same time in the optimization, requiring a multiobjective approach. In its formulation, the objective function
must contain both objectives, which are weighted by a factor
which controls the emphasis given to each one of the objectives
in the optimization. As a result of the variation of the weighting
factor used in the objective function, this problem has a set of optimal solutions (pareto-optimal set) instead of a single solution.
The structure geometry, boundary conditions and the mechanical properties of the composite material are presented in Fig. 5.
The elastic constants are the Youngs modulus in the ber direction
(E1) and transverse to the ber direction (E2), the shear modulus
(G12) and the Poissons ratio (m12), respectively. Strength parameters for traction and compression for longitudinal and transversal
directions are given by F1t, F1c, F2t, and F2c, respectively. The
remainder parameters are the shear strength (F6) and the specic
weight (q). The structure must support a design pressure load of
0.1 MPa with no material failure (the TsaiWu failure function
must be lower than 1.0 for the whole plate) and thickness of contiguous plies with the same ber orientation must not be greater
than 2 mm or less than 0.75 mm.
In the plate design, the laminate is restricted to be symmetric
with 8 layers, being represented in the GA by a pair of chromosomes, with 4 genes each one. The ber orientation angle and
the thickness of each layer are the optimization variables. They
must assume one of the discrete values given in Table 1, where
the codication adopted in the GA is also given. The number of
genes combined with the number of possible values of each variable leads the size of the design space (SDS) to be equal to 65536.
In this example the tness evaluation FIT must consider both,
weight and deection reduction, at the same time. It is done by
Eq. (2a), where the tness value is taken as the inverse of the objective function, and the weighting factor a is introduced to allow the
447
Table 1
Genes alphabet and possible discrete variable values
Orientation gene alphabet
Code
Orientation angle
Code
1
2
3
4
1
1
1
1
1
2
3
4
0.75
1.00
1.50
2.00
ply
ply
ply
ply
at
at
at
at
0
45
+45
90
if FF 6 1;
1
;
FFTv1aW 1aD
if FF > 1
Tv is referred to the violation of the limit of contiguous plies thickness with the same ber orientation. Tv is equal to the exceeding
value violating the limit xed to the thickness of contiguous plies
with the same ber orientation. As an example, if the thickness of
each one of two contiguous plies with the same ber orientation
is equal to 1.5 mm, the exceeding value violating the limit (i.e.
Tv) is 1.0, since this work adopts a limit of 2.0 mm. The constant
1.0 is added to Tv in order to avoid a division by zero in Eq. (2a)
when Tv = 0.
To allow the analysis of the optimization performance, GA is
executed 50 times for each a, which is taken varying from 0.0 to
1.0 with increments equal to 0.05. A population size P = 50 and
the elitist scheme parameter Ne = 5 are adopted (see Section 3.3).
The genetic operators are used with the probabilities poa = 4%,
pma = 2%, and pgs = 80%, while the probability of ply addition (ppa)
and ply deletion (ppd) operations are set to zero, since the number
of layers must remain xed during the optimization (see Sections
3.2.2 and 3.2.3). The parameters used as a criterion to stop the process are NLG = 250 and NSD = 125.
The distribution of weight and central displacement of all the
feasible designs in the problem is shown in Fig. 6. Points AP in this
2a
where
W
W W min
1;
W max W min
D
D Dmin
1:
Dmax Dmin
2b
448
gure are the designs that form the pareto-optimal set, which must
be obtained by the GA, according to the emphasis given to each of
the objectives. Details of the pareto-optimal set are presented in
Table 2.
Good reliability levels were obtained for the optimization with
most of the a values, as can be observed in Table 3, where the column r represents the standard deviation of the apparent reliability
R. The loss of reliability in the optimizations using a equal to 0.55
occurs because the GA nds many times the design identied by L
instead of the design identied by J, which is the correct solution in
this case. It happens because both designs J and L have practically
the same tness value for a equal to 0.55. The same occurs for a
equal to 0.70, when the GA nds the design identied by O in many
optimization processes instead of the design identied by N, which
is the correct solution. In optimizations where a is taken as being
equal to 0.80, 0.85 and 0.95, the low level of reliability occurs because the designs obtained in these cases have values of the tness
function very close to the optimal designs identied by O and P.
The average number of analyses required (An), the average
number of analyses effectively performed (Ar) and the ratio between Ar and the size of the design space (SDS), which is equal
to 65,536, are also shown in Table 3 for each value of a. An expressive reduction of the computational cost is observed if the analysis
Table 2
Pareto-optimal designs
Optimal
design
Laminate
Weight
(N)
Deection
(mm)
Weighting
factor a
A
B
C
D
E
F
G
H
I
J
L
M
N
O
P
219.7
211.9
204.0
196.2
188.4
180.5
172.7
164.8
157.0
149.1
141.3
133.4
125.6
117.7
109.9
7.9
8.8
9.6
10.8
11.8
13.2
15.0
17.3
19.9
23.3
27.2
32.1
38.5
46.6
59.5
0.00.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.750.85
0.901.0
Table 3
Optimization results with GA for the square plate
Optimal designs
R (%)
r (%)
An
Ar
Ar/SDS (%)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
A
A
A
A
A
C
E
F
G
H
I
J
L
M
N
O
O
O
P
P
P
100
100
100
100
100
98
100
100
100
100
100
84
98
100
80
94
78
68
92
86
92
0.0
0.0
0.0
0.0
0.0
2.0
0.0
0.0
0.0
0.0
0.0
5.2
2.0
0.0
5.7
3.4
5.9
6.6
3.8
4.9
3.8
7401
7780
7759
7640
7587
8659
8299
7681
7949
7694
7750
8386
8035
8146
7830
8205
7805
7798
8559
8067
8001
3641
3781
3852
3842
3923
4645
4463
4014
4242
4299
4392
4472
4095
4146
3747
3568
3175
3161
3404
3379
3347
5.56
5.77
5.88
5.86
5.99
7.09
6.81
6.12
6.47
6.56
6.70
6.82
6.25
6.33
5.72
5.44
4.84
4.82
5.19
5.16
5.11
449
Fig. 7. Differences of the tness values of points B and D with respect to their neighbor points. (a) When FIT(B) > FIT(A), then FIT(C) > FIT(B); (b) when FIT(D) > FIT(C), then
FIT(E) > FIT(D).
the unfeasible designs, driving the GA to feasible areas of the design space. A small bonus, proportional to the safety factor, is
incorporated to the objective function in the case of feasible
450
Table 4
Designs of the pareto-optimal set
Table 5
Optimization results with GA for the plate with in-plane load
Optimal
design
Laminate
Weight
(N)
Cost
(uc)
kb
kf
A
B
C
D
E
F
ge
ge
ge
45ge ; 90ge
2 ; 02 ; 45 ; 02 S
ge
ge
ke
90ge
2 ; 02 ; 452 ; 02 S
ge
ge
ke
90ge
2 ; 02 ; 45 ; 452 S
ke
ke
;
0
;
45
; 90ke
45ge ; 90ge
2
2 S
2
ke
;
0
45ge ; 45ke ; 90ke
4
2 S
ke
45ke
3 ; 904 S
24.49
25.44
26.39
27.34
28.30
29.25
73.46
64.62
55.77
46.93
38.09
29.25
1.04
1.30
1.50
1.64
1.56
1.30
55.12
30.93
31.84
18.00
17.16
16.04
designs. This bonus promotes an additional goal in the optimization process, seeking to maximize the value of the safety factor. As
a consequence of this formulation, the result of the optimization
process is supposed to be the safest design having the best
weighted combination of weight and cost.
Parameters W* and C*, used in Eq. (3a), are the dimensionless total weight and cost of the plate, respectively. They are given by Eq.
(3b) using the maximum and minimum possible values of weight
and cost, which are easily obtained by the extreme combination
of materials and number of layers. Instead of working directly with
a linear combination of the dimensionless weight and cost, the
implemented objective function uses the square of these variables,
already weighted by the factor a. This alternative formulation is
necessary because the results of optimizations that uses a linear
combination of W* and C* as objective function is always one of
the extreme points A or F (see Fig. 9). It occurs because the pareto-optimal set is arranged in a straight line in the weight-cost
plane. The new formulation provides a curved distribution of optimal points, allowing the GA to nd them. The optimal solution is
dened as the point which is located at the closest distance from
the origin of the weight-cost reference system, being this disq
tance given by aW 2 2 1 aC 2 2 .
8
q1
>
>
>
OBJ
aW 2 2 1 aC 2 2
106 k ; if k P 1;
<
q1
>
>
>
: OBJ k 2
aW 2 2 1 aC 2 2
;
if k < 1;
3a
where
W
W W min
1;
W max W min
C
C C min
1:
C max C min
3b
To study the GA performance, 25 optimization processes were executed for each a, which is taken varying from 0.0 to 1.0 with increments equal to 0.1. The GA is used with a population size P = 30
and a elitist parameter Ne = 4, together with the following probabilities for the genetic operators: poa = 4%, pma = 2%, ppa = 4%, ppd = 8%
and pgs = 80%. The parameters of the criteria used to stop the process
are NLG = 300 and NSD = 100. The results of the optimizations performed here show that the GA can obtain one optimal design for
every tested a, but not all of the 25 GA executions performed with
each value of a are successful. This happens because GA has difculty to search for the optimal design in regions where the objective
function has low gradients, as a consequence of the introduction of
the bonus proportional to the safety factor. There are many points
(designs) with tness values very similar to the optimal design tness value, having the same weight and cost of the optimal design,
but with a smaller safety factor. Table 5 shows the apparent reliability (R) and its standard deviation (r), obtained from GA the 25 optimizations for each value of a, considering two situations. In the rst
column of Table 5 optimizations founding only one optimal design
are computed as successful, while in the second column all designs
with the same weight and cost of those included in the rst column,
Pareto-optimal
set
Optimal
designs
R (%)
r (%)
R (%)
r (%)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F
F
F
E
E
D
C
B
A
A
A
100
100
100
100
100
40
20
8
84
92
88
0.0
0.0
0.0
0.0
0.0
9.8
8.0
5.4
7.3
5.4
6.5
100
100
100
100
100
100
100
100
100
100
100
0
0
0
0
0
0
0
0
0
0
0
Near
optimal
designs
An
Ar
Ar/SDS
(%)
3829
3839
3862
3750
3732
4805
4498
5999
4172
4740
4622
1238
1714
1709
1683
1723
2209
2336
2868
2037
2410
2219
2.21
3.06
3.05
3.01
3.08
3.95
4.18
5.13
3.64
4.31
3.97
FIT
!
NC crit NC 2max
:
U max V nlc 1
451
Fig. 11. Critical load level and maximum displacement of feasible designs.
ering the number of genes in the chromosomal string and the number of possible values for each gene, the size of the design space in
this example is 2187. The whole design space has been analyzed to
allow the evaluation of the GA, and the feasible designs are shown
in Fig. 11, where the optimal design is indicated.
The optimal design, indicated in Fig. 11, is dened by the stacking sequence [(904, 45)2, 902]S and the following value of the
parameters were obtained: NCcrit = 0.563, Umax = 0.0272 m and
FIT = 20.698. As expected, the optimal design has most of its plies
with bers oriented at 90, providing the best bending properties.
The 45 plies are introduced in order to satisfy the constraint related to the number of contiguous plies with the same ber orientation. The curve load displacement at point A for the optimal
design is presented in Fig. 12, together with the same curve for a
design conguration with a smaller tness value. In this case the
stacking sequence is [453, 04, 452]S, NCcrit = 0.228, Umax =
0.0307 m and FIT = 7.427. In the analyses executed for all the
points of the designs space the basic load increment ki in the GDCM
was taken equal to 0.05.
Nine combinations of the GA parameters P and NLG, presented in
Table 6, are used to study the inuence of these parameters in the
optimization performance. Considering that the maximum number
of analyses (Amax) to be executed in a optimization process is determined by the product (NLG + 1) P, it is possible to impose an
upper limit to An by adjusting these two parameters. The values
of P and NLG are obtained by working with ranges of Amax/SDS equal
to 12.5%, 25% and 50%. Using the expression for Amax stated before
Fig. 10. Shallow shell geometry, boundary conditions and material properties.
452
Fig. 12. Central displacement load level for the optimal design and for a low tness value.
Table 6
Combination of the parameters P and NLG
Amax/SDS = 12.5%
Amax/SDS = 25.0%
Amax/SDS = 50.0%
Comb
P (Ne)
NLG
(NSD)
Comb
P (Ne)
NLG
(NSD)
Comb
P (Ne)
NLG
(NSD)
1
2
3
7 (1)
14 (2)
21 (3)
38 (30)
19 (15)
12 (10)
4
5
6
7 (1)
14 (2)
21 (3)
77 (44)
38 (22)
25 (15)
7
8
9
7 (1)
14 (2)
21 (3)
155 (55)
77 (27)
51 (18)
and adopting the population size P equal to 7, 14 and 21 (proportional to 1, 2 and 3 times the number of genes in the chromosome),
it is possible to determine the value of NLG for each range of Amax/
SDS. The values of the parameters Ne and NSD are conveniently adjusted in each of the nine combinations, but these parameters are
also decisive in the efciency of the process. The genetic operators
probabilities adopted here are poa = 4% and pgs = 80%, being the
remainder set of probabilities equal to zero because they are not
used in this problem. The GA is executed 50 times for each of the
nine combinations of parameters in order to study their inuence
in the apparent reliability (R) and in the computational cost of
the optimization process.
Fig. 13 shows the performance of the GA considering each one
of the combinations of P and NLG tested in this work. In combina-
parameter ki. If the adopted value is too big, the nonlinear solution
is inaccurate or even may not converge. In the other hand, if the value is smaller than the necessary, the analysis process may became
very slow, which is very undesirable when an optimization by GA
is used. The challenge in this case is to nd a value for ki that provide accurate analyses for all the designs with the minimum computational cost.
In this example a strategy for the automatic renement of the
parameter ki is implemented to reduce the cost of the analyses
and to guarantee the convergence and the accuracy of the nonlinear solution. The rst step in the implementation is carried out
monitoring the structural response during the nonlinear solution
and identifying anomalous behaviors that can be interpreted as
solution instability or error. When this situation occurs, the process is stopped and restarted using a smaller value for ki. The
way the structural response is monitored and the factor of reduction of the parameter ki are problem dependent and can be implemented in different forms. In the present example the error is
identied when the increment in the displacement of the central
point of the shell in a specic load step is greater than 10 times
the maximum displacement veried in the previous load step.
The renement of the parameter ki is accomplished by taking one
half of its current value. A new structural analysis in the optimization process using GA to look for a new design is always started
using the value of ki dened by the user, even if the renement
process was used in the previous analysis.
The results of four optimizations executed for this case are presented in Table 7. The optimizations 1 and 2 were executed using
the xed value ki = 0.025, while the optimizations 3 and 4 were
executed using the value ki = 0.05, together with the renement
strategy. In the optimizations 1, 3 and 4 the optimal solution for
the problem is given by the laminate [904, 45, 902]S, with the following values for the critical load and maximum displacement:
NCcrit = 0.451 and Umax = 0.02895 m. This stacking sequence is almost similar to that obtained for the best design of the previous
example. It contains many plies with ber oriented at 90, in order
to maximize the shell stiffness, with a 45 ply introduced to attend
the problem constraint. The best design found by the optimization
2 is the laminate [902, 452, 90, 45, 90]S; however, the nonlinear
analysis of this structure, using ki = 0.025, results in a wrong solution, invalidating this optimization. The load level central displacement curves for the designs found in optimizations 1, 3 and
4, together with the curve referred to the right solution (using
ki = 0.0125) of the design found in optimization 2, are shown in
Fig. 14. In this gure an additional region with snap-back and
snap-through points is observed in the curve corresponding to
the best design of optimization 2. This behavior occurs in many designs explored during the optimization process, making the nonlinear analysis of such structures much more difcult since these
behaviors are correctly reproduced only when small values of the
parameter ki are used.
The values adopted for the GA parameters resulted in a good
performance in terms of computational cost, considering the small
size of the design space in this problem. Furthermore, a great
reduction in the number of analyses to be executed in each optimization is obtained avoiding repeated analyses using a memory
Table 7
Optimization performance with or without the renement strategy
Optimization
1
2
3
4
ki
0.025
0.05
With ren.
a
Ttotal and Tan
optimization 1.
An
Ar
Ar/SDS (%)
Ttotala
Tana
483
441
504
651
293
297
259
325
13.4
13.6
11.8
14.8
1.000
0.996
0569
0.682
1.000
0.982
0.643
0.616
dimensionless
values
obtained
with
reference
to
the
453
Fig. 14. Load level central displacement for the best design obtained in optimizations 1, 2, 3 and 4.
Acknowledgement
The authors wish to thank the Brazilian agencies CNPq and
CAPES for its nancial support.
454
References
[1] Walker M, Smith RE. A technique for the multiobjective optimization of
laminated composite structures using genetic algorithms and nite element
analysis. Compos Struct 2003;62:1238.
[2] Grdal Z, Haftka RT, Hajela P. Design and optimization of laminated composite
materials. Wiley & Sons; 1999.
[3] Soremekun GAE, Grdal Z, Haftka RT, Watson LT. Composite laminate design
optimization by genetic algorithm with generalized elitist selection. Comput
Struct 2001;79:13143.
[4] Goldberg DE. Genetic algorithms in search, optimization, and machine
learning. Addison-Wesley Publishing Company; 1989.
[5] Bathe KJ, Ho L. A simple and effective element for analysis of general shell
structures. Comput Struct 1981;13:67381.
[6] Yang Y, Shieh M. Solution method for nonlinear problems with multiple
critical points. AIAA J 1990;28(12):21106.
[7] Daniel IM, Ishai O. Engineering mechanics of composite materials. Oxford
University Press; 1994.
[8] Soremekun GAE. Genetic algorithms for composite laminate design and
optimization. M.Sc. thesis (Master of Science in Engineering Mechanics).
Polytechnic Institute and State University, Blacksbourg, Virginia, USA; 1997.
[9] Nagendra S, Jestin D, Grdal Z, Haftka RT, Watson LT. Improved genetic
algorithm for the design of stiffened composite panels. Comput Struct
1996;58(3):54355.
[10] Almeida FS. Laminated composite material structures optimization with
genetic algorithms. M.Sc. thesis PPGEC/UFRGS, Porto Alegre, Rio Grande do
Sul, Brazil; 2006 [in Portuguese].