Вы находитесь на странице: 1из 8

Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

www.springerlink.com/content/1738-494x

DOI 10.1007/s12206-014-0319-8

A new topology optimization scheme for nonlinear structures


Young-Sup Eom1 and Seog-Young Han2,*
1

Department of Mechanical Engineering, Graduate School, Hanyang University, 17 Haengdang-Dong, Seongdong-Gu, Seoul, 133-791, Korea
2
School of Mechanical Engineering, Hanyang University, 17 Haengdang-Dong, Seongdong-Gu, Seoul, 133-791, Korea
(Manuscript Received October 18, 2012; Revised June 25, 2013; Accepted January 15, 2014)

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Abstract
A new topology optimization algorithm based on artificial bee colony algorithm (ABCA) was developed and applied to geometrically
nonlinear structures. A finite element method and the Newton-Raphson technique were adopted for the nonlinear topology optimization.
The distribution of material is expressed by the density of each element and a filter scheme was implemented to prevent a checkerboard
pattern in the optimized layouts. In the application of ABCA for long structures or structures with small volume constraints, optimized
topologies may be obtained differently for the same problem at each trial. The calculation speed is also very slow since topology optimization based on the roulette-wheel method requires many finite element analyses. To improve the calculation speed and stability of
ABCA, a rank-based method was used. By optimizing several examples, it was verified that the developed topology scheme based on
ABCA is very effective and applicable in geometrically nonlinear topology optimization problems.
Keywords: Artificial bee colony algorithm; Nonlinear finite element analysis; Rank-based method; Sensitivity number; Topology optimization
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

1. Introduction
To date, topology optimization has been applied for various
linear structural problems [1-9]. However, when a very large
load is applied or the structural deformation is very large,
geometric nonlinearity may occur. In this case, the nonlinearity must be considered in the analysis and design in order to
obtain a more stable and robust structural topology. Stiffness
designs of geometrically nonlinear structures using topology
optimization based on the SIMP (solid isotropic material with
penalization) method have been performed [10]. It was found
that minimization of complementary elastic work rather than
compliance as an objective function, should be used for geometrically nonlinear topology optimization. In addition, it was
verified that there is no difference between the obtained topologies of four- and nine-node rectangular elements when a
mesh-independent filtering scheme is employed, while the
computation speed of the four-node element is much faster
than that of the nine-node element. Huang and Xie [11, 12]
also carried out nonlinear topology optimization with load and
displacement constraints using a bidirectional evolutionary
structural optimization (BESO) method.
It is known that the SIMP and BESO methods widely used
in topology optimization may provide different topologies
under the same constraints due to the parameters used in each
*

Corresponding author. Tel.: +82 2 2220 0456, Fax.: +82 2 2220 2299
E-mail address: syhan@hanyang.ac.kr

Recommended by Associate Editor Jongsoo Lee


KSME & Springer 2014

method. For example, even in the case of linear topology optimization in the SIMP and BESO methods, the obtained optimized topologies may be different due to parameters of the
filtering scheme or the element removal ratio [1, 3]. It is expected that these phenomena may be severe in nonlinear topology optimization. Therefore, a new nonlinear topology
optimization scheme is needed to obtain a stable and robust
optimal topology regardless of the filtering parameter or removal ratio.
Nature-inspired computation methods have been applied for
optimum designs using natural or physical phenomena with an
evolutionary optimization process. As opposed to local search
methods, these techniques are close to global search methods.
In general, stochastic and probabilistic methods are implemented. The artificial bee colony algorithm (ABCA) suggested by Karaboga [13] is one of the nature-inspired computation methods, which imitates bee colony behavior of gathering nectar. Recently, it has been applied to determine a variable of concrete dam [14], and in linear topology optimization
for static [15] and dynamic stiffness problems [16].
Since topology optimization methods are based on an iterative optimization process, the computing time is very important. It is known that ABCA is a very fast and accurate algorithm among evolutionary algorithm (EA), particle swarm
optimization (PSO) and differential evolution (DE) approaches for multi-dimensional numeric problems [17], and
the genetic algorithm (GA), particle swarm algorithm (PSO)
and particle swarm-inspired evolutionary algorithm (PS-EA)

2780

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

in finding a global minimum for multivariable functions [17,


18]. It was also verified that a topology optimization method
based on ABCA is faster than SIMP for static stiffness problems [15] and BESO for dynamic stiffness problems [16].
Furthermore, a topology optimization method based on
ABCA has not been applied to topology optimization for
geometrically nonlinear structures. For these reasons, ABCA
was adopted as a topology optimization algorithm in this study.
In this study, a new topology optimization algorithm based
on ABCA is proposed for nonlinear structural problems. The
algorithm can be applied to both linear and nonlinear structures. The distribution of material is expressed by the density
of each element in the application of ABCA. In order to prevent a checkerboard pattern in optimized layouts, a filter
scheme was implemented. In order to improve the convergence rate and stability of the optimization process, a rankbased method was employed. The optimized topologies were
compared with those of the SIMP [19] and soft-killed BESO
methods [12] to evaluate the effectiveness and applicability of
the proposed algorithm.

2.1 Linear topology optimization


Topology optimization is used to obtain material distribution to minimize an objective function. In order to find the
material distribution, the design domain must be divided into
finite elements. Commonly, topology optimization for a structure is performed to obtain the stiffest structure under a prescribed volume. Hence, the compliance, C , is employed as
an objective function and the topology optimization can be
formulated as follows.
Minimize

: C.

If the discrete function is applied to the stiffness matrix, the


following is obtained.
K = c ( x)K 0

where K 0 is the stiffness matrix in the design domain.

2.2.1 Formulation for geometrically nonlinear topology optimization


In geometrically nonlinear topology optimization under a
load constraint, the complementary work, W C , shown in Fig.
1 as an objective function, provides a better result than the
compliance [10]. Nonlinear topology optimization used complementary work as an objective function can be formulated
as follows:
Minimize

where complementary work can be expressed as the follows.

(2)

where F is the load vector, U is the displacement vector,


and K is the stiffness matrix.
If a method using the density distribution such as the SIMP
method is applied to topology optimization, the optimized
topology naturally includes a grey region. Thus, it is very difficult to apply the optimal topology to a real structure. In order
to produce the material distribution without a grey region, a
discrete function, c , is introduced [12].
1 if x is a solid element
0 if x is an empty element .

c ( x) =

(5)

(1)

W C = lim
n

1 T
1
F U = U T K -1U
2
2

: f ( x) = W C .

Subjected to : equilibrium and Vs Vs .

Here, Vs is the target volume and Vs is the volume for


the present topology. The compliance is defined as follows:
C=

(4)

2.2 Geometrically nonlinear topology optimization

2. Topology optimization

Subjected to : equilibrium and Vs Vs .

Fig. 1. Load-deflection curves in nonlinear finite element analysis.

(3)

1 n
( FiT - Fi-T1 ) (U i + U i-1 ) .
2 i=1

(6)

2.2.2 Sensitivity number for load constraint


The sensitivity number, a e , indicates the effectiveness of
the i-th element on the objective function. The sensitivity
number for a load constraint, a e , in geometrically nonlinear
topology optimization can be expressed as the following equation [12].
An adjoint equation is introduced by adding a series of

ae =

U U i-1
f ( x)
1 n
= lim ( Fi T - Fi-T1 ) i +

n 2
xe
xe
i =1

xe

(7)

vectors of Lagrangian multipliers into the objective function


as follows:

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

1 n
( Fi T - Fi -T1 ) (U i + U i -1 )

n 2
i =1

3. A new topology algorithm

f ( x ) = lim

(8)

+ li ( Ri + Ri -1 )

where Ri and Ri-1 are residual forces at the i and i - 1 ,


respectively, defined as follows.
Ri + R i-1 = Fi - Fi int + Fi-1 - Fi-int1 = 0.

(9)

The sensitivity number for the modified objective function


of Eq. (8) can be expressed as follows.
a e = lim
n

1 n T
U i-1
T U i
+
( Fi - Fi-1 )

2 i=1
xe
xe

R U i Ri-1 U i-1 ( Ri + Ri-1 )


+li i
+
+
.
xe
U i xe U i-1 xe

(10)

3.1 Artificial bee colony algorithm (ABCA)


The fitness, fiti, plays the role of the objective function in
ABCA. As the fitness value increases, the candidate solution
improves. The fitness of the i-th element is defined using the
sensitivity number of each element as follows [15, 16].
1
ai 0

fiti = 1 + a i
1 + abs (a ) a < 0
i
i

a i : sensitivity number of the i -th element .

xij + fij ( xij - xkj )


xij

If the increment of the load is small enough, the relationship


between the load and displacement can be assumed to be linear, where Ri / U i and Ri-1 / U i-1 can be expressed using the tangential stiffness matrix, K i , at the i-th increment as
follows.
Ri
R
= i-1 = - K it .
U i U i-1

(11)

The sensitivity number for the objective function can be


written using the following equation.

U i U i-1
1 n T
T
t
+
( Fi - Fi-1 - li K i )

2 i=1
xe
xe

( Ri + Ri-1 )
+l
.
xe

(12)

T
i

In order to remove the ( U i / xe + U i-1 / xe ) term, the


Lagrangian multiplier is calculated as follows.
li K it = Fi T - Fi-T1 = K it (U i - U i-1 )

(13)

li = U i - U i-1 .

(14)

Using Eqs. (13) and (14), the sensitivity number can be expressed as follows.
a e = lim
n

1 T ( Ri + Ri-1 )
li x
2 i=1
e

F int F int
1 n
= - lim (U iT - U iT-1 ) i + i-1
n 2
xe
i =1

xe

if xij xkj
otherwise

(17)

fij : random number in the range [-1,1] .

In the above equation, we can ignore the index j , because


we can express the two-dimensional index number as an integer in MATLAB programming. Therefore, the variables vi ,
xi and xk indicate a candidate solution and two arbitrary
initialized solutions, respectively. The subscripts i and
k indicate the index of each discretized element randomly
chosen in the full design domain. The temporary candidate
solution, vi , is obtained using the random search method
shown in Eq. (17). Then, the positions of the employed bees
are determined by comparing the fitness values corresponding
to vi , xi and xk .
An onlooker bee chooses a food source depending on the
probability, pi , associated with the i-th element. pi is determined by transforming the i-th elements fitness into a
probability, ranging between 0 and 1. The procedure for
searching for a candidate solution, vi , is performed by choosing xi and xk only in the case of satisfying the condition,
with a random number [0,1] < pi for each element in the full
design domain. That is, an improved candidate solution is
searched by raising the probability of the employed bees
movement to a food source with a larger fitness by increasing
the number of random trials (roulette-wheel method [20]). A
candidate solution, vi , from the old solutions, xi and xk , can
be generated as follows [17, 18].
pi =

(16)

All employed bees, that leave the hive to search for promising flower patches, are initialized randomly over the full design domain as shown in the following equations [14].
n ij =

a e = lim

2781

fiti
n

fit

j =1

(15)

n i = xi + fi ( xi - xk )
fi : random number in the range [-1,1] .

(18)

After each candidate solution is produced and evaluated by

2782

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

the artificial bee model, its performance is compared to that of


its old solution. If the new solution has equal or better quality
than the old solution, the old one is replaced by the new one.
Otherwise, the old solution is retained.
The topology optimization procedure using the suggested
ABCA proceeds as follows [15].
(1) Establish the design domain and parameters of topology
optimization using ABCA.
(2) Calculate the fitness values for the initial design domain
using finite element analysis.
(3) Perform the employed bee phase. Determine the positions of the xi and xk elements and the randomly chosen
temporary candidate solution (vi ) . The positions are determined by applying the following equations.
xi = INT [rand [0,1] (number of all food source)]
xk = INT [rand [0,1] (number of all food source)]

(19)

n i = xi + INT [rand [0,1]( xi - xk )]


INT = interger value, rand = random number .

(4) Perform the onlooker bee phase. Search for the new employed bee position determined by an onlooker bee. Calculate
the probability of each element and select the temporary candidate solution.
pi =

fiti

(20a)

Fig. 2. A flowchart of the bee colony algorithm [16].

N'

fit j

C
k -i +1

(W

j =1

(20b)

n i = xi + INT [rand [0,1]( xi - xk )] .

error =

- WkC- N ' -i +1 )

i =1

t .

N'

(22)

C
k -i +1

W
(5) Perform the scout bee phase. Search occupied elements
(employed bee colony) and abandoned elements (scout bee
colony). Compare the fitness of employed bee colony with
that of the scout bee colony.
(6) Obtain a candidate solution satisfying all constraints.
(7) Calculate the total strain energy for the updated design
domain using finite element analysis. In this step, a developed
filter scheme [21] is implemented to prevent a checkerboard
pattern which can occur when the finite element method is
used for topology optimization.

i =1

Here, W is the complementary work, t is an allowable


convergence error, k is the present iteration number, and N
is the integral number which results in a stable complementary
work in at least ten successive iterations.
If the updated topology is not converged, calculate fitness
values for the updated design domain with the density of entire elements updated by the density updating scheme using
finite element analysis. Return to step 3 and repeat the above
steps until an optimized topology is obtained. A schematic
diagram of the suggested ABCA is described in Fig. 2.

w(r )a
ij

ai =

j =1
M

n
j

, w(rij ) = rmin - rij

( j = 1, 2,...M ) .

(21)

w(r )
ij

3.2 A new topology algorithm for geometrically nonlinear


problems

j =1

Here, M is the total number of nodes in the circular subdomain, w(rij ) is the linear weight factor, a j n is the nodal
sensitivity number of the j-th node, rmin is the length scale
parameter, and rij is the distance between the center of the
element i and the j-th node.
Check whether the updated topology converged to the optimized topology using the following convergence criterion
[21].

A new topology algorithm for both linear and nonlinear


problems is proposed based on ABCA in the previous section.
The main characteristics of ABCA were modified because of
the following reasons. When topology optimization based on
ABCA is carried out, the effects of the topology and density at
the previous iteration must be considered.
Since onlooker bees search for nectars probabilistically using the roulette-wheel method [20] in step 4 of the ABCA,
they may be located at bad positions. This results in providing

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

2783

Fig. 4. A clamped long beam with a given load.


(a) Design domain

(b) The SIMP method

(c) The BESO method

rank-based method is implemented with ABCA, the scout bee


phase for escaping the local minimum in step 5 can be omitted
since the probability is not used to search for a new solution in
the proposed algorithm.
The density reflection ratio, Cr , is introduced to make the
fitness values stable. If the density reflection ratio is not properly chosen, the fitness values are severely varied, and as a
result, the topology may not converge. The density update rule
is modified to Eq. (23) using the density reflection ratio, Cr ,
and the densities at the previous and present iterations.
r k +1 = Cr r k -1 + (1 - Cr ) r k .

(d) The proposed method with the roulette-wheel method

(23)

Here, r k is the density at the present iteration, and r k -1


is the density at the previous iteration. It is found that Cr is
appropriate in the range of 0.85~0.9 for a fine mesh size for
both linear and geometrically nonlinear problems from numerical experiments like those shown in Tables 1 and 2.

4. Numerical examples
(e) The proposed method with rank-based method
Fig. 3. Design domain and optimal layouts for various topology optimization methods.

a different topology at each trial. In the case of a short or simple structure, the bees at bad positions easily move to good
positions using the roulette-wheel method. As a result, an
optimal topology can be obtained. However, in the case of
linear or nonlinear topology optimization for a long or complex structure, since the effects of the topology and density at
the previous iteration are very severe, the optimized topology
is very differently obtained at each trial. In order to overcome
this problem, the rank-based method [20] is implemented
instead of the roulette-wheel method. The rank-based method
ranks a temporary candidate solution by the probability pi
calculated by each elements fitness value. It is assumed that
the result of the rank-based method would be the same as that
obtained when the roulette-wheel method is infinitely performed.
Fig. 3 shows the results of topology optimizations for a
long cantilever beam using various techniques. When the roulette-wheel method was used, a different optimal topology
was obtained at each trial. It is far from the optimized topology obtained from the SIMP and BESO methods. When the
rank-based method was employed, the optimized topology
coincided with that of the SIMP and BESO methods. If the

4.1 Example 1: a clamped long beam under a given load


A clamped beam having dimensions of 0.2 m1.6 m 0.01
m was subjected to 30N at the center of the bottom surface, as
shown in Fig. 4. When a filter scheme was applied to topology
optimization, there was no difference of the final results between four- and nine-node rectangular elements, while the
computing time decreased by a factor of 16 when the fournode rectangular element was used [10]. Thus, the design
domain was divided by the four-node rectangular element in
this study. The material was assumed to have a Youngs
modulus of 30 MPa and a Poissons ratio of 0.3. The allowable convergence error, t , was set at 0.001. The objective
was to obtain the stiffest structure under a volume constraint
of 20% of the original volume.
When the proposed method is applied for nonlinear topology optimization, the effects of the density reflection ratio, C ,
the filtering parameter, rmin , and the size of the finite element
must be considered. In order to examine these effects, the
values of C used were 0.75, 0.8, 0.85, 0.9, and 0.95, the values of rmin used were 1.5, 2, and 2.5, and the element sizes
evaluated were 12015, 24030, and 36045. The linear and
nonlinear topology optimization results of each combination
are shown in Tables 1 and 2, where - indicates that the optimal topology for the case was not obtained.
By comparing the results of the proposed method with those
[19] of the SIMP method shown in Fig. 5, it can be seen that

2784

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

Table 1. Results of example 1 for the linear case.


Element size

rmin

C = 0.75

C = 0.80

C = 0.85

C = 0.90

C = 0.95

120

15

1.5

2.0

2.5

1.5

2.0

2.5

C = 0.80

C = 0.85

C = 0.90

C = 0.95

240

30

1.5
360

45

2.0

2.5

Table 2. Results of example 1 for the geometrically nonlinear case.


Element size

rmin

C = 0.75

1.5
120

15

2.0

2.5

1.5
240

30

2.0
2.5
1.5

360

45

2.0
2.5

the optimal topologies for linear and geometrically nonlinear


cases are very similar to those of SIMP. The complementary
works were calculated to be 0.23~0.24J for the linear case and
0.22~0.24J for the geometrically nonlinear case. It was also
found that the appropriate density reflection ratio, C, was 0.95
for the mesh size 24030 and 0.85~0.9 for the mesh size
36045 for the geometrically nonlinear case.
Consequently, the appropriate density reflection ratio, C, is
0.85~0.9 for the mesh size of 36045 for geometrically

nonlinear cases. In other words, the density reflection ratio, C,


must be in the range of 0.85~0.9 and the mesh size must be
fine enough in order to obtain an optimized topology regardless of the filtering parameter, rmin .
4.2 Example 2: a simply supported beam under a given load
A simply supported beam with dimensions of 0.8 m 0.2
m0.001 m was subjected to 200N at the center of the top

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

2785

5. Conclusions
(a) Linear case

(b) Geometrically nonlinear case


Fig. 5. Optimal topology using the SIMP method [19].

Fig. 6. Design domain of example 2.

(a) Linear case

(b) Geometrically nonlinear case


Fig. 7. Optimized topologies using the SIMP method [19] and the
proposed method.

surface as shown in Fig. 6. The design domain was divided


into a mesh size of 16040 by the four-node rectangular element, considering the results of Sec. 4.1. The material is assumed to have a Youngs modulus of 1 GPa and a Poissons
ratio of 0.3. The allowable convergence error, t , was set at
0.001. The objective was to obtain the stiffest structure under
a volume constraint of 20% of the original volume. The density reflection ratio, C , and the filtering parameter, rmin were
set at 0.85 and 2, respectively. The results are shown in Fig. 7.
The allowable convergence error, t , was set at 0.001. The
objective was to obtain the stiffest structure under a volume
constraint of 20% of the original volume.
From comparisons of the results of the proposed method
with those [19] of the SIMP method shown in Fig. 7 for the
linear and geometrically nonlinear cases, it is seen.

In this study, a new topology optimization algorithm base


on ABCA was proposed. Topology optimization using the
proposed algorithm was performed for linear, and geometrically nonlinear problems. From the results, the following conclusions can be made.
(1) The proposed method provides a clear optimized topology consisting of only solid elements, whereas those from the
SIMP consist of grey and solid elements.
(2) It was verified that the proposed algorithm can be effectively applied for topology optimization based on a comparison of the optimized topologies of the SIMP method.
(3) It was found that the density reflection ratio, C, must be
in the range of 0.85~0.9 and the mesh size must be fine
enough in order to obtain an optimal topology regardless of
filtering parameter ( rmin ).

References
[1] M. P. Bendsoe and N. Kikuchi, Generating optimal topologies in structural design using a homogenization method,
Comput. Meth. Appl. Mech. Eng., 71 (28) (1988) 197-224.
[2] H. P. Mlejek and R. Schirmacher, An engineer's approach to
optimal material distribution & shape finding, Comput. Meth.
Appl. Mech. Eng., 106 (1993) 1-6.
[3] Y. M. Xie and G. P. Steven, A simple evolutionary procedure for structural optimization, Comput. Struct., (1993)
885-896.
[4] O. M. Querin, G. P. Steven and Y. M. Xie, Evolutionary
structural optimization (ESO) using a bidirectional algorithm,
Eng. Comput., 15 (1998) 1031-1048.
[5] Q. Q. Liang and G. P. Steven, A performance-based optimization method for topology design of continuum structures
with mean compliance constraints, Comput. Meth. Appl.
Mech. Eng., 191 (2002) 1471-1489.
[6] Q. Q. Liang and G. P. Steven, Performance-based optimization of structures: Theory and applications, Spon press, Taylor and Francis Group, London, New York (2005).
[7] J. Sethian and A. Wiegmann, Structural boundary design via
level set and immersed interface methods, J. Comput. Phys.,
163 (2000) 489-528.
[8] T. Belytschko, S. P. Xiao and C. Parimi, Topology optimization with implicit functions and regularization, Int. J. Numer.
Meth. Engng., 57 (2003) 1177-1196.
[9] K. X. Zou, C. M. Chan, G. Li and Q. Wang, Multiobjective
Optimization for Performance-Based Design of Reinforced
Concrete Frames, J. Struct. Eng., 133 (2007) 1462-1474.
[10] T. Buhl and C. B. W. Pedersen and O. Sigmund, Stiffness
design of geometrically nonlinear structures using topology
optimization, Struct. Multidisc. Optim., 19 (2000) 93-104.
[11] X. Huang and Y. M. Xie, Topology optimization of nonlinear structures under displacement loading, Eng. Struct., 30
(2008) 2057-2068.
[12] X. Huang and Y. M. Xie, Evolutionary topology optimization of Continuum Structures, John Wiley & Sons. Ltd.,

2786

Y.-S. Eom and S.-Y. Han / Journal of Mechanical Science and Technology 28 (7) (2014) 2779~2786

(2010) 121-150.
[13] D. Karaboga, An idea based on honey bee swarm for numerical optimization, Engineering Faculty, Computer Engineering Department, Erciyes University, Technical reportTR06, (2005).
[14] F. Kang, J. Li and Q. Xu, Structural inverse analysis by
hybrid simplex artificial bee colony algorithms, Comput.
Struct., 87 (2009) 861-870.
[15] J. Y. Park and S. Y. Han, Swarm intelligence topology
optimization based on artificial bee colony algorithm, Int. J.
Precis. Eng, Man., 14 (1) 2013) 115-121.
[16] J. Y. Park and S. Y. Han, Application of artificial bee colony algorithm to topology optimization for dynamic stiffness
problems, Comput. Math. with Appl. (2013) article in press.
[17] D. Karaboga and B. Basturk, On the performance of artificial bee colony (ABC) algorithm, Appl. Soft Comput., 8
(2008) 687-697.
[18] D. Karaboga and B. Basturk, A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm, J. of Glob. Optim., 39 (2010) 459-471.

[19] D. Jung and H. C. Gea, Topology optimization of nonlinear


structures, Finite Elem. Anal. Des., 40 (2004) 1417-1427.
[20] Z. Gu^^rdal, R. T. Haftka and P. Hajela, Design and optimization of laminated composite materials, John Wiley &
Sons, Inc. (1999) 191-201.
[21] X. Huang and Y. M. Xie, Convergent and meshindependent solutions for the bi-directional evolutionary
structural optimization method, Finite Elem. Anal. Des., 43
(2007) 1039-1049.

Seog-Young Han received his M.S. and


Doctor Degree in Mechanical Engineering from Oregon State University, Corvallis, USA, in 1984 and 1989, respectively. Now he is a professor in School
of Mechanical Engineering of Hanyang
University, Korea. His reseach area is
topology and shape optimization, nature-inspired optimization algorithm and structural strength
analysis.

Вам также может понравиться