Академический Документы
Профессиональный Документы
Культура Документы
1. Introduction
Topology optimization is one of the most important types of structural optimization. The purpose of
topology optimization is to find the optimal layout of a structure so that given applied loads are
transferred to desired constraints in a specified domain using a given amount of material while
equilibrium and design constraints are satisfied. Two kinds of methods are used in topology
optimization; homogenization method and density method (Bendsoe et al., 2002). Bendsoe et al.
(1988) proposed the homogenization method. The density method was suggested by Bendsoe
(1989). The density of the finite element is used as a design variable in the density method. The
modulus of each finite element is calculated based on the density value. If the density is close to
zero, the corresponding finite element is eliminated as it is considered empty. The density method is
based on isotropic material (Park, 2007). The commercial software NASTRAN is based on the
density method. The commercial software NASTRAN is used for topology optimization (MD
Nastran, 2008). Gea (1996) proposed a design domain method which maintains the advantages from
both the homogenization method and density function approach. Sigmund (1997) proposed the
design of a compliant mechanism using topology optimization. Luo et al. (1998, 1998) studied the
optimal stiffer design of 3D shell/plate structures.
Most of the work mentioned above is based on linear finite element analysis to find the structural
responses. It means that this work is based on the assumption of small deformation during the
optimization process. In the cases of compliant mechanism and energy absorbing structure
applications, the assumption of small deformation does not stand any more. In order to obtain a
more reasonable design, appropriate nonlinearities must be considered. Yuge et al. (1999) modified
the homogenization method for nonlinear deformation. Buhl et al. (2000) studied topology
optimization with geometrically nonlinearity and compared the results of linear analysis and
nonlinear analysis. Mayer et al. (1996) studied the application of topological optimization
techniques to a crashworthiness problem using the elastoplastic material model. Gea et al. (2001)
studied the stiffness optimization problems with geometrical nonlinearities. Bruns et al. (2001)
studied the topology optimization of structures by considering the material and geometrical
nonlinearities. Yoon et al. (2005, 2007) suggested the element connectivity parameterization
method for topology optimization of material nonlinear continuum structures and geometrically
nonlinear structures. Jung et al. (2004) studied the topology optimization of geometrically and
materially nonlinear structures.
Linear response structural optimization is very easy and convenient because the sensitivity
information can be easily calculated. However, structural optimization using nonlinear analysis is
comparatively expensive. It takes a long time to evaluate the design sensitivity of the objective
function or constraints. Therefore the equivalent loads method has been developed to solve the
difficulty of calculating the sensitivity information. Kang et al. (2001) proposed the equivalent
static loads method for dynamic response optimization originally. Shin et al. (2007) modified the
equivalent loads method for nonlinear response size and shape optimization. The main idea is that a
nonlinear response optimization problem is converted to a linear response optimization with
equivalent loads. Equivalent loads are defined as the loads for linear analysis, which generate the
same response fields as those of nonlinear analysis. The method was originally proposed for size
and shape optimization.
A preliminary study was done for topology optimization with equivalent static loads to check the
proposed method (Lee et al., 2010). The overview of the NROEL was published (Park, 2011). The
concept of NROEL for nonlinear static response topology optimization was published (Lee et al.,
2012). In this research, we solved few more case studies to verify the nonlinear static response
topology optimization using equivalent static loads. The results strongly support the proposed
method.
2. Background Theory
2.1.
The response is directly proportional to the load in the linear analysis. Linearity may represent the
real response of the structure or may be the expected response based on assumptions made for
analysis purposes. In linear analysis, we consider that the displacements and rotations are small,
stress is directly proportional to the strain, and also the loads maintain their unique direction when
the structure is deformed. Many practical problems can be solved by the above assumptions made
for linear analysis. But, sometimes the above assumptions are not valid. Sometimes the elastic
material becomes plastic or the material does not have the linear stress-strain relationship, and it is
also possible that due to the material failure and buckling, the structure may lose its stiffness.
Displacement and rotations may be so large that it becomes necessary to write the equilibrium
equations for the deformed structure rather than the original structure. Some parts are in contact
with each other. The contact area is changed when the load is changed. In all the above cases, the
structure shows nonlinear response. Therefore, nonlinearity should be considered in order to obtain
accurate results. In structural mechanics, three kinds of nonlinearity are considered
In material nonlinearity, the material properties are the functions of the state of stress or strain
(Cook, 2002). The nonlinear material response may be nonlinear elastic or plastic. Material
response that is nonlinear and maintains a permanent strain or returns slowly to an unstrained state
on complete unloading is called as nonlinear elastic. Inelastic behavior of materials that maintains a
permanent set on complete unloading is called plasticity.
In geometric nonlinearity, the deformation is large enough that it is required to write the equilibrium
equations of the deformed structural geometry. The loads also change direction, when deformation
increases (Cook, 2002). Contact nonlinearity, a contact problem is a type of a geometrically
nonlinear problem that takes place when different structures, or different parts of the same structure,
either come into contact, separate, or slide into each other. All kinds of nonlinearity are considered
for solving the case studies in this research.
2.2.
Topology Optimization
The concept of optimization is applied to the structural design in structural optimization. Structural
optimization has been employed in various fields, especially in aerospace engineering to reduce the
weight of airplane structures.
Optimization is aggressively applied to various structural designs due to many reasons. First,
structural design process and optimization process are almost similar. The reduction of weight in
structural design is similar to minimization of the objective function in optimization. The structural
design conditions can be transformed easily into optimization constraints. Second, the finite element
method (FEM) can be converted in structural optimization. The objective function and constraints
can be estimated accurately by defining it into definite forms by FEM. FEM is one of the most
developed computational methods for the analysis of system.
b
mean compliance
K(b) z - f = 0
veb V
(1)
3.1.
The equivalent loads are calculated by the following below steps. The equilibrium equation of the
structural nonlinear behavior can be represented as
b, z N z N f
(2)
where K is the stiffness matrix, b is the design variable vector, zN is the nonlinear nodal
displacement vector (responses) and f is the external load vector. The subscript N represents the
nonlinear analysis.
Equivalent loads are obtained by multiplying the linear stiffness matrix (KL) and nonlinear nodal
displacement vector (zN) from Eq. (2).
f eq K L z N
(3)
where feq is the equivalent load vector. The subscript L represents the linear analysis. Eq. (4) is the
equation of linear analysis with the equivalent loads.
L b z L f eq
(4)
where zL is the linear displacement vector. When the linear analysis is performed with the
equivalent loads from Eq. (3), the linear nodal displacement vector (zL) has same values as those of
nonlinear nodal displacement vector (zN) as shown in Eq. (3) and Eq. (4).
3.2.
The topology optimization process using equivalent loads method is illustrated in flowchart as
shown in Figure 2. The steps of the topology optimization process are as follows:
Step 1: Set the initial cycle value k=0, and design variable b(k) = b(0)
Step 2: Perform the nonlinear response analysis in Eq. (2) with b(k)
Step 3: Calculate the equivalent loads in Eq. (3)
Step 4: Perform the linear static topology optimization in Eq. (1) with equivalent loads.
Step 5: The design variables values are transposed into transformation variables by using Eq. (5)
0
i
1
when
bi 1
when
bi 1
(5)
where, i is a transformation variable of the ith design variable and 1 is the separation parameter.
The transformation variables are temporary variables and the total number of transformation
variables is equal to the total number of design variables. The design variables in linear response
topology optimization do not have exact values, but have continuous values between bmin and bmax
as given in Eq. (1). But few elements have value very close to the bmin. Therefore these low density
finite elements create mesh distortion that results in non-reliable optimization results. Therefore, the
separation parameter is needed in order to decide whether this finite element should be included or
eliminated for optimization in the next cycle.
Step 6: Update the design. The elements having transformation variables of value 0 are eliminated
in the finite element model and elements corresponding to the transformation variables having value
1 remain in the FE model. Therefore, the number of elements in the present cycle is less than the
previous cycle.
Step 7: If k=0, then go to the next step. Otherwise check the convergence criterion.
countif b i
bi
k 1
2 n 3
(6)
Where 2 is a small value close to 0, 3 is the percentage and n is the total number of design
variables. The number is counted if the difference of the design variables of the present cycle (k)
and previous cycle (k-1) is more than the specified value (2), and the counted number is smaller
than the percentage of the total design variables. Then the optimization process is terminated
because the convergence criterion is satisfied. Otherwise go to the next cycle.
4. Case Studies
Five case studies are solved by using the equivalent loads method and linear response optimization
method. First two case studies; beam is solved by linear response optimization and equivalent loads
method considering geometric nonlinearity. The optimum results are compared with the numerical
results. Third and fourth case studies; Michell structure and spacer grid are solved by linear
optimization method and equivalent loads method by considering the material and geometrical
nonlinearities. The equivalent loads optimization results are compared with the linear response
optimization result. Fifth case study, a flange with rectangular hole in the mid and a rectangular
beam passing through it is solved with a contact condition. Different combinations of the
nonlinearities are considered for it. The elimination method is used as the cycle update method.
NASTRAN is used for nonlinear response analysis and optimization. Following terminologies are
used in the paper:
MNL: material nonlinearity
GNL: Geometric nonlinearity
CNL: contact nonlinearity
4.1.
Case Study 1
A long beam 800 mm long, 200 mm width, and 10 mm thickness is fixed at the midpoint of the
both ends as shown in Figure 3. A 200 N concentrated force is applied at the center of the top edge.
The number of design variables is 1600, minimizing the strain energy is the objective function, and
the mass constraint is 20 % of the total mass. The Youngs modulus is 1 GPa and the Poissons
ratio is 0.3.
The nonlinear response topology optimization using the equivalent load is formulated as follows:
Find
to minimize
subject to
bi
i = 1,., 1600
strain energy
KL zL feq = 0
mass V mass total x 20 %
(7)
4.2.
Case Study 2
In order to check the consistency in the optimization method and results, the same design problem
as in case study 1 is considered with different loading condition, boundary condition and material
properties as shown in Figure 5. The formulation of the problem is same as given in Eq. (7). The
beam is fixed on both ends and three concentrated forces 200N, 400N and 200N are applied on the
bottom edge of the beam in the downward direction. The material used has Youngs modulus 100
MPa and the Poissons ratio is 0.3.
Linear response topology optimization and nonlinear response topology optimization using
equivalent loads are performed and optimization results are compared with the numerical results
(Gea et al., 2001) as shown in Figure 6. The optimum results are very close to the numerical results
in this case study too. This shows the consistency in the optimization method and results and
justifies the equivalent loads method.
4.3.
Case Study 3
A cantilever plate 550 mm long and 400 mm high is fixed at one end as shown in Figure 7. A 2000
N force is applied at the center point of the free edge. The optimization problem is solved according
to the Eq. (8). The number of design variables is 8800 and objective function is the minimizing the
strain energy. The mass constraint for the topology optimization is 50 % of the total mass. The
material has bilinear elastoplastic properties. The Youngs modulus is 210 GPa, the hardening
modulus is 105 GPa and Poissons ratio is 0.3.
Find
to minimize
subject to
bi
i = 1,., 8800
strain energy
KL zL feq = 0
mass V mass total x 50 %
(8)
Case Study 4
The spacer grid set is one of the main components of a nuclear fuel assembly. There are four spacer
grid in a unit spacer grid set. The fuel rod is inserted into the unit spacer grid set as shown in Figure
9. The function of the spacer grid set is; to support the fuel rod, to provide the path to a cooling
fluent in order to increase the heat transfer from the hot fuel rod to the coolant. The spring and
dimples are the actual supporting parts in the unit spacer grid (Shin et al., 2008). The optimization
of the spring part is preferred.
When the fuel rod is inserted into the unit spacer grid, the spring is deformed by about 3 mm. The
fuel rod is in contact with the spring at the central line. So, the force is applied at the central line of
the spring. Nonlinear analysis is performed using the condition that the spring is moved by 0.3 mm
after the contact of the spring and the fuel rod. The central part of the spacer grid is used as the
design domain in topology optimization as shown in Figure 10.
The formulation to perform topology optimization on the spacer grid is given in Eq. (9). The
number of design variables is 7016 and objective function is the minimizing the strain energy. The
mass constraint for the topology optimization is 70 % of the total mass. The separation parameter
used is 0.001.
Find
to minimize
subject to
bi
i = 1,., 7016
strain energy
KL zL feq = 0
mass V mass total x 70 %
(9)
Case Study 5
The piece of block consist of rectangular hole at the mid of the block. A rectangular beam is passed
through this hole. This block supports the beam and acts like a flange. The dimensions of the block,
hole and beam are: (100x100x100), (30x30x30) and (300x30x30) respectfully. The finite element
model of the block with rectangular beam is shown in Figure 13 (a) and this model is used for
analysis. The finite element model of the block as shown in Figure 13 (b) is used for the
optimization. 100 N of distributed force is applied on each side of the rectangular beam. The
material properties; Youngs modulus is 210 GPa, the hardening modulus is 105 GPa and Poissons
ratio is 0.3.
The formulation of the topology optimization as given in Eq. (10) is used to perform the topology
optimization. The number of design variables is 7280 and objective function is the minimizing the
strain energy. The mass constraint for the topology optimization is 40 % of the total mass.
Find
to minimize
subject to
bi
i = 1,., 7280
strain energy
KL zL feq = 0
mass V mass total x 40 %
(10)
Figure 14. The linear response topology optimization cannot be performed because there is contact
nonlinearity.
5. Conclusions
The equivalent loads method is a valuable tool applied to the nonlinear response topology
optimization. The problem of mesh distortion due to the low density finite elements has been solved
by introducing the transformation variable and new update method. A separation parameter is used
to define the transformation variables. This separation parameter defines the density value to
remove the finite elements having density less than this value. The transformation variables of the
finite elements with a value of 0 are removed from the finite element model for the next cycle. The
updating of the finite element material becomes easy after the removal of low density finite
elements. The equivalent loads method require only few nonlinear analysis and problem is
converged to the optimum.
Five different kinds of case studies are solved by linear response topology optimization and
nonlinear response topology optimization using equivalent loads considering different nonlinearities.
The first two case studies; beam has been solved by considering the geometric nonlinearity and
results have been compared to the numerical results. The optimum results are very close to the
numerical results that validate the proposed method. The consistency in the proposed method has
been verified. The third and fourth case studies; Michell structure and spacer grid have been solved
by considering the material and geometric nonlinearities. The optimum results are obtained. The
nonlinear analysis is performed on the linear response optimum result and nonlinear response
optimum result with the same force and boundary conditions as applied on the initial model to
calculate the strain energy values. The objective of topology optimization is to maximize the
stiffness. Lower the value of the strain energy higher will be the stiffness. The strain energy values
of the equivalent loads optimization results are less than the strain energy values of the linear
response optimization result. This shows that the equivalent loads optimization result is better
compared to the linear topology optimization result. The fifth case study a flange with contact
condition is also solved and optimum results are obtained. The linear response topology
optimization has not been solved due to the contact condition.
Nonlinearity should be considered in the design process, if necessary. Therefore, the equivalent
loads method is an excellent tool to get the optimization results.
References
Bendsoe, M.P. (1989), Optimal shape design as a material distribution problem, Structural
Optimization, Vol. 1, No. 4, pp. 193-202.
Bendsoe, M.P. and Kikuchi, N. (1988), Generating optimal topologies in structural design using a
homogenization method, Computer Method in Applied Mechanics and Engineering, Vol. 71, Issue
2, pp. 197-224.
Bendsoe, M.P. and Sigmund, O. (2002), Topology Optimization: Theory, Methods and Application,
Springer, Germany.
Bruns, T.E. and Tortorelli, D.A. (2001), Topology optimization of non-linear elastic structures and
compliant mechanism, Computer Methods in Applied Mechanics and Engineering, Vol. 190,
Issues 26-27, pp. 3443-3459.
Buhl, T., Pedersen, C.B.W. and Sigmund, O. (2000), Stiffness design of geometrically nonlinear
structures using topology optimization, Structural and Multidisciplinary Optimization, Vol. 19, No.
2, pp. 93-104.
Cook, R. D., Malkus, D. S., Plesha, M. E. and Witt, R. J. (2002), Concepts and Applications of
Finite Element Analysis, 4th edition, pp. 595-638.
Gea, H.C. (1996), Topology optimization: A new microstructure based design domain method,
Computers and Structures, Vol. 61, Issue 5, pp. 781-788.
Gea, H.C. and Luo, J.H. (2001), Topology optimization of structures with geometrical
nonlinearities, Computers and Structures, Vol. 79, No. 20, pp. 1977-1985.
Jung, D.Y. and Gea, H.C. (2004), Topology optimization of nonlinear structures, Finite Elements
in Analysis and Design, Vol. 40, Issue 11, pp. 1417-1427.
Kang, B.S., Choi, W.S. and Park, G.J. (2001), Structural optimization under equivalent static loads
transformed from dynamic loads based on displacement, Computers and Structures, Vol. 79, Issue
2, pp. 145-154.
Lee, H. A., Ahmad, Z. and Park, G. J. (2010), Preliminary study on nonlinear static response
topology optimization using equivalent load, Transactions of the Korean society of Mechanical
Engineers- A, Volume 34, Issue 12, pp.1811-1820 (In Korean)
Lee, H.A. and Park, G.J. (2012), Topology optimization for structures with nonlinear behavior
using the equivalent static loads method, Journal of Mechanical Design, Vol. 134, pp. 031004-14.
Luo, J.H. and Gea, H.C. (1998), Optimal based orientation of 3D shell/plate structures, Finite
Elements in Analysis and Design, Vol. 31, Issue 1, pp. 55-71.
Luo, J.H. and Gea, H.C. (1998), A systematic topology optimization approach for optimal stiffer
design, Structural Optimization, Vol. 16, No. 4, pp. 280-288.
Mayer, R.R., Kikuchi, N. and Scott, R.A. (1996) Application of topological optimization
techniques to structural crashworthiness, International Journal for Numerical Methods in
Engineering, Vol. 39, Issue 8, pp. 1383-1403.
MD Nastran (2008), R3 Quick Reference Guide, MSC. Software Corporation.
Park, G.J. (2007), Analytic Methods for Design Practice, Springer, Germany, pp. 237-243.
Park, G. J., (2011), Technical overview of the equivalent static loads method for non-linear static
response structural optimization, Structural and Multidisciplinary Optimization, Vol. 43, Issue 3,
pp. 319-337.
Shin, M.K., Park, K.J. and Park, G.J. (2007), Optimization of structures with nonlinear behavior
using equivalent loads, Computers Methods in Applied Mechanics and Engineering, Vol. 196,
Issue 4-6, pp. 1154-1167.
Shin, M.K., Lee, H.A., Lee, J.J., Song, K.N. and Park, G.J. (2008), Optimization of a nuclear fuel
spacer grid spring using homology constraints, Nuclear Engineering and Design, Vol. 238, Issue
10, pp. 624-2634.
Sigmund, O. (1997), On the design of compliant mechanism using topology optimization,
Mechanics of Structures and Machines, Vol. 25, Issue 4, pp. 493-524.
Yoon, G.H. and Kim, Y.Y. (2005), Element connectivity parameterization for topology
optimization of geometrically nonlinear structures, International Journal of Solids and Structures,
Vol. 42, Issue. 7, pp. 1983-2009.
Yoon, G.H. and Kim, Y.Y. (2007), Topology optimization of material nonlinear continuum
structures by the element connectivity parameterization, International Journal for Numerical
Methods in Engineering, Vol. 69, Issue 10, pp. 2196-2218.
Yuge, K., Iwai, N. and Kikuchi, N. (1999), Optimization of 2-D structures subjected to nonlinear
deformations using the homogenization method, Structural Optimization, Vol. 17, No. 4, pp. 286299.
Analysis domain
Design domain
Update design
Nonlinear
static
analysis
Linear static
topology
optimization
Start
k=0
End
k=k+1
No
Yes
Nonlinear analysis
Convergence
Calculate equivalent loads
Update design
200 mm
800 mm
Numerical method
(a) Linear
(b) GNL
200 N
200 N
400 N
Numerical method
(a) Linear
(b) GNL
550 mm
400 mm
2000 N
(a) Linear
(b) MNL
(c) GNL
Dimple
Nuclear
fuel rod
Spring
y
z
z
x
(a) Design domain for the topology optimization
16.0 mm
12.85 mm
z
x
(b) Boundary conditions (top view)
0.98 mm
0.91 mm
(a) Linear
(b) MNL
(c) GNL
(a)
(b)
Side view
(c) CNL
Table 1 Strain energy values (Nm) of the Michell structure optimum results
MNL
GNL
171.82
75.35
172.00
163.02
74.80
160.67
Table 2 Strain energy values (Nm) of the spacer grid optimum results
MNL
GNL
10.64
15.45
13.62
9.82
14.02
11.89