Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

ARTICLE IN PRESS

Finite Elements in Analysis and Design 46 (2010) 229–237

Contents lists available at ScienceDirect

Finite Elements in Analysis and Design


journal homepage: www.elsevier.de/locate/finel

Topology design for multiple loading conditions of continuum structures


using isolines and isosurfaces
Mariano Victoria a, Osvaldo M. Querin b,, Pascual Martı́ a
a
Structures and Construction Department, Technical University of Cartagena, Campus Muralla del Mar, 30202 Cartagena (Murcia), Spain
b
School of Mechanical Engineering, The University of Leeds, Leeds LS2 9JT, UK

a r t i c l e in fo abstract

Article history: Isolines topology design (ITD) is an iterative algorithm for the topological design of two-dimensional
Received 28 July 2008 continuum structures using isolines. This paper presents an extension to this algorithm for topology
Received in revised form design under multiple load cases of two/three-dimensional continuum structures. The topology and the
3 August 2009
shape of the design depend on an iterative algorithm, which continually adds and removes material
Accepted 28 September 2009
depending on the shape and distribution of the contour isolines/isosurfaces of the required structural
Available online 2 November 2009
behaviour. In this study the von Mises stress was investigated. Several examples are presented to show
Keywords: the effectiveness of the algorithm, which provides very detailed contours without the need to interpret
Topology design the topology in order to obtain a final design. The ITD algorithm demonstrates how the use the multiple
Multiple load cases
loading conditions can produces more stable and realistic designs with a little additional complexity.
Three-dimensional continuum
& 2009 Elsevier B.V. All rights reserved.
Isosurfaces
Isolines
Fixed grid

1. Introduction method can be used to achieve a multiple criterion design for a


structure in thermal problems. Akgün et al. [8] presented a study
Structures used in real environments are subject to multiple whose main objective was to demonstrate the efficiency of the
loading conditions and in many cases need to be modelled in adjoint method under MLC with and without constraints lumping.
three-dimensions. Therefore, to obtain more realistic structures, Bendsøe and Sigmund [9] extended solid isotropic microstructures
these should be designed for multiple load cases, be three- with penalty (SIMP) method to MLC, where the objective function
dimensional, and be feasible, among other aspects. is formulated as a minimization of a weighted average of the
Diaz and Bendsøe [1] and Allaire, et al. [2] extended the compliances for each of the load cases. Allaire and Jouve [10]
Homogenization method to multiple load cases (MLC). Xie and extended the level set (LS) method for shape and topology
Steven [3,4] took into account the highest element stress level over optimization for 2D and 3D problems to new objective functions
all load cases for identifying the extreme efficiency of material such as eigenfrequencies and multiple loads. Cervera and Trevelyan
usage. Those elements with the least efficiency under all load [11] presented and ESO approach based on boundary element
conditions are progressively removed until the remaining elements method (BEM) and non-uniform rational B-splines (NURBS). To
become relativity efficient for at least one of all the load cases. Chu solve multiple load cases problem, a logical AND/OR algorithm is
et al. [5] presented a simple evolutionary procedure based on finite used, where the material is selected for the removal process only if
element analysis (FEA) to minimize the weight of structures while it is low-stressed in all load cases, and it is selected for the
satisfying stiffness requirements. A wide range of problems additional process if it is high-stressed in any of the load cases.
including those with multiple displacements conditions, moving Aguilar et al. [12] developed a computational model for multi-
loads, and multiple load cases are considered. Young et al. [6] objective optimization problems and it is based on genetic
successfully extended bi-directional evolutionary structural opti- algorithms (GA). Zhou and Li [13] presented a method to optimize
mization (BESO) to MLC for two-dimensional (2D) and three- the topology of structures under multiple load cases with stress
dimensional (3D) problems. Li et al. [7] showed how the ESO constraints. Fiber-reinforced orthotropic composite is employed as
the material model to simulate the constitutive relation of truss-
like continua.
 Corresponding author. The first work where the homogenization method was modified
E-mail addresses: mariano.victoria@upct.es (M. Victoria), to account for 3D effects can be found in Bendsøe [14] and Diaz and
ozz@mech-eng.leeds.ac.uk (O.M. Querin), pascual.marti@upct.es (P. Martı́). Lipton [15]. Olhoff et al. [16] used optimum 3D microstructures for

0168-874X/$ - see front matter & 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.finel.2009.09.003
ARTICLE IN PRESS
230 M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237

topology optimization of linearly elastic 3D continuum structures The criteria value at the ith node of an element ðsni Þ is
subject to a single case of static loading. To visualize topology determined by (3).
designs, a penalization technique was applied. Borrvall and PN e
Petersson [17] considered large-scale topology optimization of e¼1 sni
sni ¼ ð3Þ
elastic continua in 3D using the regularized intermediate density N
control. In order to deal with problems of large size, parallel where i is the ith node number of an element; seni is the nodal
computing is used in combination with domain decomposition. criterion value at node i for each element surrounding that node,
Hsu et al. [18] presented an automated process for interpreting 3D N is the number of elements connected to that node.
topology optimization results into a smooth CAD model. The The nodal value is determined from the criteria values at each
topology optimization is solved by the SIMP method. Koguchi and Gauss point extrapolated to the nodes using the shape functions
Kikuchi [19] developed a surface reconstruction algorithm, which of the element.
consisted of three parts: (1) an enclosed isosurface geometry from
which the topologically optimized model was generated, (2)
features detected, and (3) the parametric CAD solid model 3. Design using isosurfaces
reconstructed as bi-quartic surfaces splines.
The novelty of this work is that it extends the isolines topology The use of isolines/isosurfaces in two/three-dimensions re-
design (ITD) algorithm [20] to problems experiencing multiple spectively to obtain the optimum design of a structure have been
load cases and which are modelled with three dimensional finite used in some recent studies, e.g. Woon et al. [31], Cui et al. [32],
elements. The method of determining the isosurfaces within De Ruiter and Van Keulen [33], Hsu and Hsu [34], Koguchi et al.
fixed-grid finite elements (FG-FE) is given, together with a brief [19], etc.
explanation of the fixed-grid finite elements analysis method The aim of this work has been to extend the ITD algorithm [20]
(FG-FEA). Several examples are presented to show the effective- to problems experiencing multiple load cases and which are
ness of the algorithm. The usefulness of the algorithm lies in that modelled with three dimensional finite elements.
it is capable of providing quality solutions with very detailed The ITD is an iterative algorithm which redistributes (adds and
contours, without the need to interpret the topology or of using a removes) material inside of a design domain until it reaches a
tuning process in order to obtain a final design. desired volume fraction. The redistribution process consists of the
following four steps: (1) Obtain the design criteria distribution
within the design domain; (2) Determine the minimum criteria
2. Fixed grid finite element analysis level (MCL), where its intersection with the design criteria
distribution produces the new structural boundary, shown for a
The fixed grid method was first introduced by Garcia and 2D continuum in Fig. 1; (3) Eliminate all regions from the design
Steven [21] as a tool for numerical estimation of two-dimensional domain where the criteria distribution is lower than the MCL; (4)
elasticity problems. The application of FG-FEA to two-dimensional This design modification requires the re-evaluation of the
problems has been a research topic during the last years [22–25]. remaining structure in order to recalculate the design criteria
A lot of works have been also done with 3D structures like Suzuki distribution.
et al. [26], Garcia et al. [27,28], etc. The benefits of using FG-FEA The MCL is calculated at each iteration and depends on the
over conventional FEA in this work are that: (1) FG does not need a distribution of the design criterion and on the volume of the
fitted mesh to discretize the analysis domain; (2) the boundary of design domain in that iteration, given by (4).
the design is disassociated from the mesh [21]; (3) designs using  
FE-FEA do not contain checkerboard patterns, making the design ni  i i
Vi ¼ V0 þ Vf ð4Þ
more reliable for manufacture [29]; (4) solution time is sig- ni ni
nificantly reduced [30].
In FG-FEA, the elements are in a fixed position and have the where i is the ith iteration, V0 is the initial volume of the design
real design superimposed on them. This means that there are domain; Vf is the final volume desired for the design; ni is the total
elements which lie Inside (I), Outside (O), or on the Boundary (B) number of iterations to use for the ITD to design the structure.
of the design. Once the criterion has been calculated for each element in the
The elemental stiffness matrix ðKe Þ is given by (1). design domain, these are arranged in decreasing order of criterion
value. An element by element volume summation of the ordered
8 e list is carried out until a volume is reached which is as close as
< KI
> if x ¼ 1
e e possible to the target volume given by (4), where the level of error
K ¼ KO if x ¼ 0 ð1Þ
>
: e e e between the summed and target volume depends on the size of
KB ¼ KI x þ ð1  x ÞKO if 0 o x o 1
the elements. The criteria value of the next element in the ordered
list is then used as the value for the MCL. For multiple load cases,
ðeÞ
where x is the design fraction inside the element, KI element the same process is applied for each individual load case. The final
stiffness matrix for an element inside, KO is the element stiffness design then consists of the superposition of the design for each
matrix for an element outside, and KB for an element boundary. load case.
Normally KO r104  KI .
In this work, the value of the criteria in each element (se) is
3.1. Criterion selection
calculated using (2).

PnG The design criterion used in this work was the von Mises
wk sek
se ¼ Pk n¼G 1 ð2Þ stress, which for a three-dimensional continuum is given by (5).
k ¼ 1 wk qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
sVM ¼ pffiffiffi ðs1  s2 Þ2 þ ðs2  s3 Þ2 þ ðs1  s3 Þ2 ð5Þ
2
where nG is the number of the Gauss points in the elements, and w
is the weighting factor for each Gauss point. where, s1, s2, and s3 are the principal stresses.
ARTICLE IN PRESS
M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237 231

Fig. 1. Structural boundary is defined by the intersection of the MCL with the criteria distribution.

Fig. 2. Look-up table for the MC algorithm showing the 16 different topologic states.

3.2. Combination of individual load case designs to produce the BESO like methods [3,4,6,7,11]. This is achieved by the 9 steps
global design given below:

The multiple load case global design is obtained by the 1. Determine the volume ðVi Þ of the design domain in the current
superposition of the designs obtained for each individual load (ith) iteration using (4).
case, which is analogous to the AND/OR method applied to ESO/ 2. For each load case calculate the average (or mean) criterion ðsmean
l Þ.
ARTICLE IN PRESS
232 M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237

3. Calculate the required volume for every load case design cell. The basic assumption of this algorithm is that a contour (MCL
domain ðVil Þ using (6). isosurface) can only pass through a finite element in a limited
b V number of ways. This algorithm requires the value of the MCL as
Vil ¼ PlNLC i ð6Þ well as the value of the criteria at each node, and consists of two
l¼1 bl
basic steps:
where i is the ith iteration, l is the lth load case, NLC is the total
number of load cases, bl is the volume control weighting factor.
1. Identify from Fig. 2 the topological state of each element;
It’s calculated using the ratio of the average criterion for the lth
2. Determine the shape of the contour of the MCL isosurface
load case with the maximum average criterion amongst all
through each element.
load cases maxl ¼ 1;NLC fsmean
l g and is given by (7).
smean
l
bl ¼ ð7Þ The interaction of an isosurface through a cubic element can
maxl ¼ 1;NLC fsmean
l
g
have a maximum of 256 different topological states. But, since a
4. Use the individual load case design volume from (6) to cube has double symmetry, the maximum number of states can
calculate the MCL for each load case ðsMCLl Þ. be reduced to 16, Fig. 2. This shows all different states where a
5. For each load case, calculate shape of the design domain using black circle at a node means that the value of the criteria at that
that load case MCL ðsMCLl Þ. node is less than the MCL (i.e. outside of the design).
6. Use the shape of the design domain to calculate the volume When only one of the nodes in an edge of an element is marked
e
fraction of each finite element for all of the load cases ðxl Þ. with a black circle it indicates that the MCL isosurface intersects
e
7. The volume fraction for an element ðx Þ is then given by that edge, which is the case for topological states 2–15. To find
the maximum volume for that element from all of the load that intersection point, linear interpolation can be used. To
cases (8). construct facets from the intersection points, Delanauy triangula-
xe ¼ max fxel g ð8Þ tion [40] can be used. The shape of the MCL isosurface through the
l ¼ 1;NLC element is then obtained by connecting these facets as shown in
8. Calculate the superimposed volume of the design domain Fig. 2.
using (9).
X
N
e 3.4. Structural boundary stabilization
Visuper ¼ Ve  x ð9Þ
e¼1

where e is the eth finite element number, Ve is the volume of When the MCL is modified, the structural boundary changes
the eth element. and this affects the criterion distribution. Therefore, before the
9. Calculate the volume error between the target volume ðVi Þ and new iteration is started, an iterative process of reanalysis and
the superimposed volume ðVisuper Þ using (10). material redistribution is carried out until the change in the
domain volume between successive boundary adjustments is less
DV ¼ Visuper  Vi ð10Þ than a minimum volume change limit ðDV%Þ. Typical value are
DV% ¼ 0:5–2.5%.
If the absolute value of the volume error is less than the Vi  Vi1
volume of an element ðjDVj rVe Þ, then the process can be DV% ¼  100 ð13Þ
Vi1
terminated.
If the volume error is negative and greater than the volume of This iterative process only requires a few iterations, although the
an element, then (11) is used to modify the required volume for exact number depends on the value of the volume of design
every load case design domain ðVil Þ and steps 4 through 9 are domain at the ith iteration ðVi Þ determined by (4).
repeated.
Vil ¼ Vil þ bl  Ve ð11Þ
4. The topology design algorithm for multiple load cases
If the volume error is positive and greater than the volume of
an element, then (12) is used to modify the required volume
The procedure for implementing the ITD method for multiple
for every load case design domain ðVil Þ and steps 4 through 9
load cases is as follows:
are repeated.
Vil ¼ Vil  bl  Ve ð12Þ
1. Define the design and non-design domains, supports, loads,
Note that since this process has been implemented using a and material properties.
fixed grid, then all elements have the same volume (Ve). 2. Specify the size of fixed grid mesh.
3. Specify the final design volume Vf, the total number of
iterations ni, and the minimum volume change limit DV%.
3.3. Minimum criteria level extraction
4. Specify the design criterion to use: von Mises stress, etc.
5. Carry out a FG-FEA of the design domain for each load case.
There are several approaches to generation of a 3D surface, e.g.
6. Using the method from Section 3.2 determine the structural
Keppel [35], Herman et al. [36], Farrell [37], Shen and Jhonson
boundary for the design.
[38], Koguchi et al. [19], etc.
7. If the percentage volume change is greater than the minimum
The procedure to generate the structural boundary in 3D
volume change limit ðDV%Þ, go to step 5, else go to step 8.
designs depends on the determination of the MCL isosurface. In
8. If the total number of iterations ni has been reached, go to step
order to determine the line segments that produce the profile of
9, else increment the iteration number i by 1 and go to step 5.
the boundary, the contouring algorithm called Marching Cubes
9. Stop the design process.
[39] was implemented.
The Marching Cubes (MC) method uses a divide and conquer
approach, treating each finite element independently as a cube This process can be viewed in the flow chart of Fig. 3.
ARTICLE IN PRESS
M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237 233

Start

Specify:
Design and FG domain,
non-design region, loads and
boundary condition,
topology parameters

FG-FEA for each load case

Determine
the target volume Vi

Calculate
the MCL for each load case

Calculate shape of design


domain for each load case

Update target volume for Calculate elemental volume Update target volume for
each load case Vil fraction for each load case each load case Vil

Calculate superimposed
volume of design domain

No Superimposed and
Target volume equal

Stabilization Process
Yes

Stop design No
process

Yes
End

Fig. 3. Flow chart showing the ITD process for structures with multiple load cases.

Table 1
ITD parameters.
F1 = 10 kN F2 = 10 kN
Example ni Vf =V0 ðDV%Þ NLC
F2 = 10 kN F1 = 10 kN
Square under torsion 50 0.300 1.5 2
Beam with a roller support 50 0.200 1.5 3
Short cantilever 100 0.150 1.5 2
Electric mast 50 0.025 1.5 2

0.2 m 1m

5. Examples

To illustrate the ITD algorithm, four structures were studied and F2 = 10 kN


F1 = 10 kN Thickness : 0.1 m
are presented here: (1) A square under torsion, (2) a beam with a
roller support, (3) a short cantilever, and (4) an electric mast.
F2 = 10 kN F1 = 10 kN
For all the examples the elastic modulus of I elements is 210 GPa,
the Poisson’s ratio is 0.3, and elastic modulus of O elements is
0.0210 MPa. The FE used for the 2D examples is the four-node plane Fig. 4. Square under torsion.
ARTICLE IN PRESS
234 M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237

Fig. 5. Final designs for the square under torsion: (a) ITD algorithm; (b) Diaz and Bendsøe [1].

10m

3m

Thickness : 0.1 m

F1 = 10 kN F2 = 10 kN F3 = 10 kN

2.5m 2.5m 2.5m 2.5m

Fig. 6. Beam with a roller support.

Fig. 8. Optimum topologies [9]: (a) all loads in one load case; (b) multiple loading
cases.

1m

F1 = 10 kN

Fig. 7. Final designs for the beam with a roller support: (a) all loads act
simultaneously; (b) multiple load cases.
1m

stress quadrilateral element with four Gauss integration points [41],


and for the 3D example is the eight-node isoparametric with eight
Gauss integration points [41]. The design criteria used was the von
Mises stress. Table 1 shows the ITD parameters used.
Thickness : 0.1 m
5.1. Square under torsion
F2 = 10, 50, 100 kN
The design domain is a square area of 1 1 m with a centrally
located square hole where all displacements are restricted. The Fig. 9. Short cantilever.
ARTICLE IN PRESS
M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237 235

Fig. 10. Designs for the short cantilever beam: (a) F1 =F2 = 10 kN; (b) F1 = 10 kN, F2 = 50 kN; (c) F1 = 10 kN, F2 = 100 kN.

2.5 m
2.5 m

5m
5m
F2 = 10 kN

F1 = 10 kN

5m
Fig. 11. Designs for the short cantilever beam. Volume fraction: Vf =V0 ¼ 0:35:
(a) ITD algorithm; (b) Li et al. [42].

F1 = 10 kN

F2 = 10 kN

20 m

10 m

10 m

10 m
5m
Fig. 12. Design for the short cantilever beam with F1 = 10 kN and F2 = 55.5 kN.
Fig. 13. Electric mast.

mesh used has 250  250 elements. The structure is subject to two Similar solutions to this example (Fig. 8) were obtained by
different load cases, represented by loads F1 and F2, Fig. 4. Bendsøe and Sigmund [9]. It is interesting to note that the solution
The designs generated with ITD algorithm (Fig. 5a) reveals an for a single load case (Figs. 7a and 8a) in which all three loads acts
excellent agreement with optimum topology obtained (Fig. 5b) simultaneously produces a truss-like structure that is not rigid.
using homogenization method [1]. Hence for this problem, the single load case arrangement
produces an unstable structure based on squared frames
whereas the multiple load cases problem produces (Figs. 7b
5.2. Beam with a roller support and 8b) a stable structure based on triangular frames.

The design domain is a rectangular area of 10  3 m. The mesh


used has 160  48 elements. It has a fixed support in the bottom 5.3. Short cantilever
left-hand corner and a roller support in the bottom right-hand
corner. Three identical vertical downward loads F1, F2, and F3 are The dimensions of the cantilever beam are 1 1 m. The design
applied at the bottom edge, Fig. 6. domain was discretized using a mesh of 200  200 FE, fully
The resulting designs for multiple load cases, where F1, F2, and clamped along the left edge. Two vertical loads are applied,
F3 represent each load case, and the single load case, where F1, F2, F1 ¼ 10 kN at the upper right-hand corner and F2 at lower free end
and F3 act simultaneously, are given in Fig. 7. with three different values 10, 50, and 100 kN, Fig. 9.
ARTICLE IN PRESS
236 M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237

Fig. 14. Electric mast. Final designs for: (a) single load case; (b) multiple load case.

The resulting MLC designs for different ratios between F1 and design criteria may be local, such as the von Mises stress; (2) the
F2 are given in Fig. 10. For the case of F1 ¼ F2 the resulting design is produced structure has smooth boundaries and needs no further
symmetric (Fig. 10a) and when F1 aF2 the design is asymmetric, interpretation, enhancement or processing; (3) the external shape
Fig. 10b and c. Fig. 11a (for a volume fraction of 0.35) shows great of the designs are not dependent on the fixed-grid FE mesh
agreement with the work of Li et al. [42] with the ESO method, density, although the internal features of the topology are.
Fig. 11b. Therefore, if more internal features are desired to be present in
Fig. 12 shows the design for the ratio between forces of the final design, a progressively denser FE mesh must be used.
F2 =F1 ¼ 5:5. For ratios higher than this, the optimal topology In practise, most of designs are usually subjected to multiple
always converged to the design of Fig. 10c. load cases, and in many cases, structures designed for multiple
loads are more stable and robust than designs subject to a single
5.4. Electric mast load case.
Four examples of optimum topology design of 2D/3D con-
The design domain is the T-shaped box of Fig. 13. Two tinuum structures subject to multiple load cases were presented
symmetric vertical loads F1 and F2 are applied in the middle of to demonstrate the applicability and effectiveness of ITD algo-
the lower edges of the horizontal part of the T-section and rithm in this work.
represent the loads exerted by the wires on the mast. Fixed The main conclusion of this work is that the ITD algorithm is
supported boundary conditions are applied at the corners of the useful design method for 2D and 3D structures with single and
base of the T-shaped box. A one-quarter FEA model using a multiple loading conditions.
90  150  15 mesh and symmetry conditions was used to
generate the design. Acknowledgements
Fig. 14a is the case where all loads act simultaneously and
Fig. 14b shows the resulting design for multiple load cases where This work has been partially supported by the CARM
F1 and F2 represent each load case. (Consejeria de Educación, Ciencia e Investigación de la Región
The ITD algorithm produces a truss-like design that evolves de Murcia) and the Technical University of Cartagena. Its support
topology of real electric masts. The number of generated bars or is greatly appreciated. Travelling funds for the second named
truss elements which emerge depends on the mesh density and author were provided by the School of Mechanical Engineering at
size of the fixed grid domain. A full scale real industrial the University of Leeds.
application would require a much finer mesh, and a larger domain
in the vertical direction. References

[1] A.R. Diaz, M.P. Bendsøe, Shape optimization of structures for multiple loading
6. Conclusions
conditions using a homogenization method, Structural Optimization 4 (1992)
17–22.
This paper presents an enhancement to the ITD algorithm, [2] G. Allaire, Z. Belhachmi, F. Jouve, The homogenization method for topology
and shape optimization. Single and multiple loads case, European Journal of
where the multiple load cases are considered in design process. By Finite Elements 5 (1996) 649–672.
using the fixed-grid FEA method of analysis, changes in the [3] Y.M. Xie, G.P. Steven, Optimal design of multiple load case structures using an
topology of the structure are easily and efficiently handled and evolutionary procedure, Engineering Computations 11 (1994) 295–305.
[4] Y.M. Xie, G.P. Steven, Evolutionary Structural Optimization, Springer, Berlin,
stress or other criteria isosurfaces are easily generated and used to
Heidelberg, New York, 1997.
generate the emerging structural topology. [5] D.N. Chu, Y.M. Xie, A. Hira, G.P. Steven, Evolutionary structural optimization
The ITD is an iterative algorithm where the generation of new for problems with stiffness constraints, Finite Elements in Analysis and
contours allows for the removal and redistribution of material. Design 21 (1996) 239–251.
[6] V. Young, O.M. Querin, G.P. Steven, 3D and multiple load case bi-directional
The use of the isosurfaces of the desired structural performance evolutionary structural optimization (BESO), Structural Optimization 18
has several benefits: (1) the process works globally although the (1999) 183–192.
ARTICLE IN PRESS
M. Victoria et al. / Finite Elements in Analysis and Design 46 (2010) 229–237 237

[7] Q. Li, G.P. Steven, O.M. Querin, Structural topology design with multiple [26] K. Suzuki, H. Ohtsubo, K. Terada, The analysis of 3D solid using multi-scale
thermal criteria, Engineering Computations 17 (6) (2000) 715–734. voxel data, in: Proceedings of the IV World Congress on Computational
[8] M.A. Akgün, R.T. Haftka, K. Chauncey, J.L. Walsh, Efficient structural Mechanics, Buenos Aires, Argentina, 1998.
optimization for multiple load cases using adjoint sensitivities, AIAA Journal [27] M.J. Garcia, O.E. Ruiz, L.M. Ruiz, O.M. Querin, Fixed grid finite element
39 (3) (2001) 511–516. analysis for 3D linear elastic structures, Computational Mechanics (WCCM
[9] M.P. Bendsøe, O. Sigmund, Topology Optimization. Theory, Methods and VI), China, 2004.
Applications, Springer -Verlag, New York, 2003. [28] M.J. Garcia, M. Henao, O.E. Ruiz, Fixed grid finite element analysis for 3D
[10] G. Allaire, F. Jouve, A level-set method for vibration and multiple loads structural problems, International Journal of Computational Methods 2 (4)
structural optimization, Computer Methods Applied Mechanics and Engi- (2005) 569–585.
neering 194 (2005) 3269–3290. [29] F.S. Maan, O.M. Querin, D.C. Barton, Extension of the fixed grid finite element
[11] E. Cervera, J. Trevelyan, Evolutionary structural optimisation based on method to eigenvalue problems, Advances in Engineering Software 38 (2007)
boundary representation of NURBS. Part I: 2D algorithms, Computers and 607–617.
Structures 83 (2005) 1902–1916. [30] M.J. Garcı́a, G.P. Steven, Fixed grid finite element analysis in structural design
[12] J.F. Aguilar, H. Rodrigues, H. Pina, Multi-objective optimization of structures and optimisation, in: Second ISSMO/AIAA Internet Conference on Approx-
topology by genetic algorithms, Advances in Engineering Software 36 (2005) imations and Fast Reanalysis in Engineering Optimization, May 25–June 2,
21–28. 2000.
[13] K. Zhou, X. Li, Topology optimization of structures under multiple load cases [31] S.Y. Woon, L. Tong, O.M. Querin, G.P. Steven, Optimising topologies through a
using a fiber-reinforced composite material model, Computational Mechanics multi-GA System, in: The Fifth World Congress of Structural and
38 (2006) 163–170. Multidisciplinary Optimization (WCSMO 5), Lido di Jesolo, Italy, 2003,
[14] M.P. Bendsøe, Optimal of Structural Topology, Shape, and Material, Springer, pp. 229–230.
New York, 1995. [32] C. Cui, H. Ohmori, M. Sasaki, Computational morphogenesis of 3D structures
[15] A.R. Diaz, R. Lipton, Optimal material layout for 3D elastic structures, by extended ESO method, Journal of IASS 44 (141) (2003) 51–61.
Structural Optimization 13 (1997) 60–64. [33] M.J. De Ruiter, F. Van Keulen, Topology optimization using a topology
[16] N. Olhoff, E. Rønholt, J. Scheel, Topology optimization of three-dimensional description function, Structural and Multidisciplinary Optimization 26 (2004)
structures using optimum microstructures, Structural Optimization 16 (1998) 406–416.
1–18. [34] M.H. Hsu, Y.L. Hsu, Interpreting three-dimensional structural topology
[17] T. Borrvall, J. Petersson, Large-scale topology optimization in 3D using parallel optimization results, Computer and Structures 83 (2005) 327–337.
computing, Computer Methods Applied Mechanics and Engineering 190 [35] E. Keppel, Approximating complex surfaces by triangulation of contour lines,
(2001) 6201–6229. IBM Journal of Research and Development 19 (1975) 2–11.
[18] Y.L. Hsu, M.S. Hsu, C.T. Chen, Interpreting results from topology optimization [36] G.T. Herman, J.K. Udupa, Display of 3D digital images: computational
using density contours, Computer and Structures 79 (2001) 1049–1058. foundations and medical applications, IEEE Computer Graphics and Applica-
[19] A. Koguchi, N. Kikuchi, A surface reconstruction algorithm for topology tion 3 (5) (1983) 39–46.
optimization, Engineering with Computers 20 (2006) 1–10. [37] E.J. Farrell, Color display and interactive interpretation of three-
[20] M. Victoria, P. Martı́, O.M. Querin, Topology design of two-dimensional continuum dimensional data, IBM Journal of Research and Development 27 (4) (1983)
structures using isolines, Computer and Structures 87 (2009) 101–109. 356–366.
[21] M.J. Garcia, G.P. Steven, Fixed grid finite elements in elasticity problems, [38] H.W. Shen, C.R. Jhonson, Seeping simplifies: a fast isosurface extraction
Engineering Computations 16 (2) (1999) 145–164. algorithm for unstructured grids, in: Proceedings of Visualization, 1995, pp.
[22] H. Kim, M.J. Garcia, O.M. Querin, G.P. Steven, Y.M. Xie, Introduction of fixed 143–151.
grid in evolutionary structural optimization, Engineering Computations 17 [39] W.E. Lorensen, H.E. Cline, Marching Cubes: a high resolution 3D surface
(4) (2000) 427–439. construction algorithm, Computer Graphics 21 (4) (1987) 163–169.
[23] M.J. Garcia, C.A. Gonzalez, Shape optimization of continuum structures via [40] B.N. Delaunay, Sur la sphe re vide. Izvestia Akademii Nauk SSSR, Otdelenie
evolution strategies and fixed grid finite element analysis, Structural and Matematicheskikh i Estestvennykh Nauk 7 (1934) 793–800.
Multidisciplinary Optimization 26 (2004) 92–98. [41] O.C. Zienkiewicz, R.L. Taylor, J.Z. Zhu, The Finite Element Method, Its Basis &
[24] Y. Liu, F. Jin, Q. Li, S. Zhou, A fixed-grid bidirectional evolutionary structural Fundamentals, Elsevier Butterworth Heinemann, 2005.
optimization method and its applications in tunneling engineering, Interna- [42] Q. Li, G.P. Steven, Y.M. Xie, On equivalence between stress criterion and
tional Journal for Numerical Methods in Engineering 73 (2008) 1788–1810. stiffness criterion in evolutionary structural optimization, Structural Optimi-
[25] Y. Liu, J. Feng, L. Qing, A strength-based multiple cutout optimization in zation 18 (1999) 67–73.
composite plates using fixed grid finite element method, Composite
Structures 73 (2006) 403–412.

You might also like