You are on page 1of 6

Bat Algorithm for Topology Optimization in

Microelectronic Applications
Xin-She Yang, Mehmet Karamanoglu

Simon Fong

School of Science and Technology


Middlesex University
Hendon Campus
London NW4 4BT, UK

Department of Computer and Information Science


Faculty of Science and Technology
University of Macau
Taipa, Macau

AbstractIn many design applications, designers often have


to find the best geometrical configurations so as to achieve
certain objectives with the minimum amount of materials used.
Such shape or topology optimization problems are usually much
harder to solve than nonlinear optimization problems in a
fixed domain. In this paper, we use the recently developed bat
algorithm to solve topology optimization problems. Results show
that the distribution of different topological characteristics such
as materials can be achieved efficiently. We have also tested the
bat algorithm by solving nonlinear design benchmarks. Results
suggest that bat algorithm is very efficient for solving nonlinear
global optimization problems as well as topology optimization.
Index TermsBat algorithm, inverse problem, optimization
algorithm, topology optimization, metaheuristics.

I. I NTRODUCTION
In many engineering design and industrial applications, one
of the design objectives is to design optimal geometrical
configurations to suit for a particular purpose. For example,
in structural engineering, engineers often have to design the
best way that the support beams are placed in the right places
so as to carry the maximum loads, while the bending and
shear stresses are within the limits imposed by stringent design
codes. In many cases, design domains are not fixed, and
materials can be placed anywhere so that the final structure can
have an optimal shape. The geometry or shape of the structure
is one of the main design objectives [1], [18]. For example, in
microelectronic applications, heat transfer can be an important
issue. Even with the best designs from the electronic point of
view, the layout of materials for microdevices can be very
challenging so that designs are also optimal for heat management. Miniaturization of electronic devices in communications
makes it more important to find optimal shape and distribution
of materials, or more generally the topology of the designs.
Such shape or topology optimization is the main interest
of this paper. In general, topological shape optimization is
much harder to solve than nonlinear problems in a fixed
domain. The geometrical domain of a topology design problem
can be considered fixed, and they can typically be a regular
region (such as a continuous, rectangular region). However,
their design space or search space of the variables can be
significantly larger. For example, we may want to fill a
rectangular region with two materials, though the geometrical
domain is fixed and can be discretized or subdivided into many
978-1-4673-5861-3/12/$31.00 2012 IEEE

150

small subregions, say, 10 10 subregions in a 1 cm 1 cm


square. The number of ways to fill each region or not can
have 2100 different combinations. Thus, it is impossible to
search every combination in practice. Obviously, the geometry
configuration in a continuous domain can often increase the
number of degrees of freedom dramatically, which is also
coupled with nonlinearity and complexity of the problem
itself. Consequently, such problems become very challenging
to solve.
Furthermore, the design objective of an optimization problem has to be properly formulated. Inappropriate formulation
may lead to solve a different problem, or even a wrong
problem, even with most efficient implementation of optimization algorithms. In addition, the evaluation of a given
design with a set of design variables may be time-consuming
if the objective cannot be expressed in any analytical form.
In practice, evaluations of the objective often require computationally extensive tools such as finite-element simulation
software. Conventionally, such problems were either solved by
specific methods such as gradient-based methods together with
finite-element analysis [1], [2], [14]. For example, in structural
topology optimization, these methods can be very efficient in
producing good designs [14]. However, in the context of heat
transfer, such methods can be modified in principle to solve
similar problems, but they tend to be slow.
In different areas such as geophysical applications, inverse
problems are more important, and various specific techniques
exist for inverse parameter estimation [11], [13]. Another
way of solving topology optimization problems is to consider
it as an inverse problem. In an inverse problem, the main
aim is to minimize the differences between observations and
predictions, which is, in fact, an optimization problem. For
topology optimization, we can parameterize the unknown
domain by an array or matrix of indices where 1 means
the region is solid, while 0 means that region is hollow (no
materials). This way, the aim is to use optimization algorithms
to find the index matrix. Obviously, the main objective and
all design constraints have to be satisfied simultaneously. In
general, we have a multiobjective optimization problem. Even
in the simplest case, we still have to deal with a nonlinear
least-squares problem.
In the framework of optimization, we can in principle

solve a shape optimization problem by using optimization


algorithms. However, there are many challenging issues such
as well-posedness, incomplete data, non-unique solutions, uncertainty, and improper problem formulation. Mathematically
speaking, only well-posed problems with sufficient data will
lead to unique solutions. In addition, uncertainty or noise in
data can lead to solutions with large uncertainties. In reality,
data are often incomplete, and the number of degrees of
freedom is large, some extra knowledge of the problem is often
needed so that some proper constraints can be added into the
formulation. Ideally, we can formulate such problems in such
a way that we can solve them using optimization techniques
[1], [7], [15], [17].
Another challenging issue is that traditional optimization
methods do not work well for such highly nonlinear problems.
One of the reasons is that most traditional algorithms such as
gradient-based methods requires derivative information, which
could be difficult to compute. In addition, almost all traditional
optimization algorithms are local search algorithms. That is,
they tend to converge quickly to a local optimum, often not the
global optimum. Therefore, these disadvantages of traditional
algorithms and the challenging issues associated with inverse
and topology problems necessitate to try new methods such
as nature-inspired metaheuristic algorithms.
Among metaheuristic algorithms, swarm-intelligence-based
algorithms have become especially popular. For example,
particle swarm optimization and firefly algorithm are among
the widely used algorithms [9], [10], [4], [17], [18]. Bat
algorithm is another optimization algorithm recently developed
by Yang [19], which has good convergence and performance.
In many engineering applications, we may want to find good
solutions quickly, and bat algorithm is particularly suitable for
such problems as we will demonstrate later in this paper.
The remainder of the paper continues as follows: In Section
II, we will discuss briefly topology optimization in the context
of microelectronic applications, while in Section III, we introduce the bat algorithm in detail. Then, we use bat algorithm
to solve two design problems as benchmarking and validation
in Section IV. Then, we proceed to solve the layout problem
of two different materials for heat transfer management of
a microdevice in Section V. Finally, we discuss some issues
related to topology optimization and topics for further research
in Section VI.
II. T OPOLOGY O PTIMIZATION
Topology optimization, also called shape optimization, can
be in general written as an integral form in terms of an index
variable u. The objective is to
Minimize

(q)d,

(1)

subject to a set of nonlinear constraints or differential equations [1], [14]. The whole domain can be discretized by a
discrete index q {0, 1} which only takes 0 (no material)
and 1 (filled with a given material). This is often true in
151

B
T =1

T =0

Fig. 1. Topology optimization with the maximization of |TA TB |.

structural shape optimization. In many applications, the objective function does not have any explicit dependence on the
design variables, and thus the objective cannot be written in
simple analytical form. In this case, its dependence is implicit
or black-box type. This is especially true for heat management
in microdevices.
Heat management, basically heat transfer modelling and design optimization, is very important for many microelectronic
applications, especially those using large-scale integrated circuits and miniaturization. In fact, nanoscale heat transfer is an
interesting area, and topological optimization for the design
of a nanoscale device is even more challenging [22], [16]. For
example, Evgrafov et al. proposed a topology optimization
benchmark for a nanoscale heat-conducting system with a
size of 150 nm by 150 nm [5]. In real-world applications,
heat transfer can occur at many different scales, and heat
management at smaller scales may be more difficult to control.
Now we extend this benchmark to a standard unit area of 1 mm
by 1 mm, and the aim is to distribute two different materials so
as to maximize the temperature difference |TA TB | at these
two points A and B under the boundary conditions given in
[5] (see Fig. 1).
Two materials used in the design of the unit area have heat
diffussivities of K1 and K2 , respectively. In addition, K1
K2 . For example, Si and Mg2 Si, K1 /K2 10. The domain
is continuous under heat flux conditions at the top and bottom
boudaries and the objective is to distribute the two materials
such that the difference |TA TB | is as large as possible.
In order to formulate this shape optimization properly [7],
[8], [11], [12], we can write
Maximize |TA(q) TB (q)|,
(2)
subject to
hj (x, y, q) = 0, (j = 1, ..., J),
(3)
gk (x, y, q) 0, (k = 1, ..., K),
(4)
where the parameter/index vector q will be determined by
optimization, subject to J equalities and K inequalities. In

addition, (x, y) are the coordinates in the domain .


The whole shape domain in 2D can be subdivided into N by
N small regions/blocks. Inside each block, we use a material
index q so that 1 means this block is filled, and 0 means this
block is empty. In essence, this means that the whole domain
can be parameterized by using a material index array q,
corresponding to N N smaller subregions. In principle, such
optimization can be solved using any efficient optimization
algorithm. However, as the number of free parameters tends
to be very large, and as the problem is often highly nonlinear
and possible multimodal, conventional algorithms such as
hill-climbing usually do not work well. More sophisticated
metaheuristic algorithms have the potential to provide better
solution strategies [9], [4], [17], [20], [21]. In the rest of
this paper, we will use bat algorithm to solve two topology
optimization problems.
III. BAT A LGORITHM
A new bat-inspired algorithm was formulated in 2010 by
Xin-She Yang [19], which were based on the fascinating
characteristics of echolocation of microbats. In order to mimic
the behavour of microbats in the simplest way, some approximations are necessary, and we use the following three
approximations/assumptions:
For sensing the distance ahead of their flight paths,
microbats use echolocation, and they have their unique
instinct to distinguish a target of food/prey from the
background barriers.
During hunting, bats may travel in a random manner at a
velocity vi at position xi with a combination of sensing
frequency fmin , varying wavelength and loudness A0
to hunt for prey. The frequency (or wavelength) of their
emitted pulses can be automatically adjusted, and the
pulse emission rate r [0, 1] can also be fine-tuned
automatically, given the current proximity of their target.
The loudness of echoes by a microbat can be assumed to
decay over time from a large and positive amplitude A0
to a small constant value Amin .
Another sensible simplification made here is that we do
not use any ray tracing process for computing the time delay
and estimating the three dimensional topology. Ray tracing
could have been a useful feature in applications that involve
computational geometry. In our work here, ray tracing is not
adopted due to its very high demand in computational complexity especially for multidimensional cases. The following
approximations are also used, for the sake of simplicity, in
addition to the forementioned. In general the frequency f in
a range [fmin , fmax] corresponds to a range of wavelengths
[min , max ].
A. Bat Motion

For the bats in simulation, the following rules are defined


for regulating how their positions xi and velocities vi are
updated in a d-dimensional search space. The new solutions
xti and velocities v ti at time step t are given by
fi = fmin + (fmax fmin ),
(5)
152

= v ti + (xti x )fi ,
v t+1
i

(6)

(7)
where [0, 1] is a vector randomly drawn from a uniform
distribution. After comparing all the solutions among all the
n bats at each iteration t, a current best solution x that
represents the global best solution can be obtained. This
current best solution corresponds to the best value of the
objective at the current iteration t. As the velocity increment
is defined by the product i fi , fi (or i ) is used to refine
the velocity change while keeping the other factor i (or fi )
constant. The configuration can be readjusted depends on the
domain of the problem of interest. The settings of fmin = 0
and fmax = O(1), are used in our experimentation, and again
it depends on the domain size of the problem of interest. A
random value is assigned as the frequency of each bat, which
is taken uniformly from [fmin , fmax].
In the local search, a new solution is drawn locally by using
a random walk around the current best solution, and it is then
tested to see if it is the best among all the solutions. The
random walk is defined as
xnew = xbest + At ,
(8)
where is a number vector randomly drawn from [1, 1].
At =<Ati > is defined as the average loudness of all the bats
at this time step.
The process of the updating the velocities and the locations
of bats are somewhat similar to that of the standard particle
swarm optimization. The pace and range of the movement
are basically controlled by fi , just like the movement of the
swarming particles. To certain extent, the bat algorithm is
deemed to be a balanced combination of swarm optimization
and the intensive local search, that are governed by the
frequency tuning ability and the variables of loudness and
pulse rate.
Nevertheless, these factors that influence the balance, the
loudness Ai and the pulse emission rate ri , are subject to
iterative updates in the optimization process. The value for
loudness can be chosen arbitrarily by the user at will, as long
as the rules are adhered the loudness subsides and the pulse
emission rate escalates, when a bat is approaching near to its
prey. For simplicity sake, we apply A0 = 1 and Amin = 0.
Letting Amin = 0 assumes that a bat has just found the prey
hence temporarily pause on emitting any sound. Now we have
the following parameter control [19]
At+1
= Ati ,
rit = ri0 [1 exp(t)],
(9)
i
where and are constants.
Actually, has the effect of the cooling factor in a cooling
schedule as in the simulated annealing algorithm. For any 0 <
< 1 and > 0, we have
(10)
Ati 0, rit ri0 , as t .
For a simplest solution, = can be used. In our simulations,
we let = = 0.95.
= xti + v t+1
,
xt+1
i
i

IV. D ESIGN O PTIMIZATION


= 0.7854bh2(3.3333z 2+14.9334z43.0934)1.508b(d21+d22 )
Before we proceed to solve shape optimization problems,
let us first validate our bat algorithm by solving two design
+7.4777(d31 + d32 ) + 0.7854(L1d21 + L2 d22 ),
(16)
benchmarks. In all our simulations, we have used the population size n = 25 (unless given specifically), = = 0.95, subject to
and the total number of iterations as 1000, which gives about
25,000 functional evaluations for each case study.
g1 = bh27z 1 0,
2

A. Design of a Tensional and Compressional Spring

Tensional and/or compressional springs are used widely in


engineering. There are three design variables in a standard
spring design problem: the wire diameter w, the mean coil
diameter d, and the length (or number of coils) L.
The objective is to minimize the weight of the spring,
subject to various constraints such as maximum shear stress,
minimum deflection, and geometrical limits. For detailed
description, please refer to earlier studies [1], [3], [6]. This
problem can be written compactly as
min f (x) = (L + 2)w2 d,
(11)
subject to
3

g1 (x) = 1

d L
71785w 4

0,

g2 (x) = 1

140.45w
d2 L

0,

g3 (x) =

2(w+d)
3

g4 (x) =

d(4dw)
w 3 (12566dw)

(12)

1 0,
+

1
5108w 2

g3 =

1.93L31
hzd41

1 0,
1 0,

1.93L32
hzd42

1 0,
q
1
1 2
6
g5 = 110d
( 745L
3
hz ) + 16.9 10 1 0,
1
q
1
2 2
6
( 745L
g6 = 85d
3
hz ) + 157.5 10 1 0,
g4 =

(17)

g7 =

hz
40

1 0,

g8 =

5h
b

1 0,

g9 =

b
12h

1 0,

g10 =

1.5d1 +1.9
L1

1 0,

g11 =

1.1d2 +1.9
L2

1 0.

2.6 b 3.6,

(13)
Using BA with n = 25, = = 0.95 and 1000 iterations,
we have obtained the same or slightly better solutions than the
best solution obtained by Cagnina et al. [3]
f = 0.012665
(14)
with
x = (0.051690, 0.356750, 11.287126),
(15)
but bat algorithm uses significantly fewer evaluations.
0.25 d 1.3,

397.5
bh2 z 2

In addition, the simple bounds are

1 0,

with the following limits


0.05 w 2.0,

g2 =

0.7 h 0.8,

2.0 L 15.0.

B. Speed Reducer Design

17 z 28,

7.3 L1 8.3,

2.9 d1 3.9,

7.8 L2 8.3,

5.0 d2 5.5.

(18)
(19)
(20)

Furthermore, z must be integers. This is essentially a mixed


variable optimization problem as z takes only integer values,
and all the other variables are continuous.
The best solutions obtained by bat algorithm with n = 25
after 5000 iterations are

b = 3.5, h = 0.7, z = 17, L1 = 7.3, L2 = 7.8,


Optimal design of a speed reducer or a gearbox is another
benchmark design problem with seven design variables [3],
[6], including the face width (b), module of the teeth (h), the
d1 = 3.34336445, d2 = 5.285350625,
(21)
number of teeth on pinion (z ), the length (L1 ) of the first shaft
between bearing, the length (L2) of the second shaft between with
between bearings, the diameter (d1 ) of the first shaft, and the
fmin = 2993.7495888,
(22)
diameter (d2 ) of the second shaft.
The main objective is to minimize the total weight of which are better than f = 2996.348165 obtained by others

the speed reducer, subject to 11 constraints such as bending in


the
literature
[3],
[6].
stress, deflection and various limits on stresses in shafts. This
We have seen that, for both test problems, BA has found
optimization problem can be written as [6]
the optimal solutions which are either better than or the same
as the solutions found so far in the literature.
min f (x) = f (b, h, z, L1, L2 , d1 , d2 )

153

1
1

0.8
0.8

0.6
0.6

0.4
0.4

0.2
0.2

0
0

0.2

0.4

0.6

0.8

0
0

Fig. 2. Initial distribution of two materials: Si (left) and Mg (right).

0.2

0.4

0.6

0.8

Fig. 3. Optimal topology and distribution of two different materials.

V. T OPOLOGY O PTIMIZATION FOR H EAT T RANSFER IN A


M ICRODEVICE
We now solve two topology optimization problems discussed earlier for designing microelectric devices. The main
objective is to distribute two materials (Si and Mg2 Si, with
heat diffusivities of K1 and K2 , respectively) in the unit area
so that the difference |TA TB | can be as large as possible
under the condition that K1 K2 or K1 /K2 10.
By dividing the domain into 40 40 small grids, we can
initialize two materials in an initial distribution as shown in
Fig. 2.
For each configuration generated during the search process,
the temperature distribution is estimated using the finitedifference method by solving the heat conduction equation
with varied indices for material conductivities so that the
temperature difference at the two fixed points should be as
large as possible. The final shape after 1000 iterations is shown
in Fig. 3.
By using BA with n = 40 to search for the possible design
solutions, we have found an optimal shape and distribution of
materials shown in Fig. 3 where Si is shown in cyan (or light
gray) and Mg2 Si is shown in red (or heavy gray in black and
white) in the middle of the domain (reversed S-shape stripe).
The objective here is to maximize the temperature difference
of the two points that are placed vertically on the middle axis.
From Fig. 4, we can see that the temperature difference is
indeed is the largest. The design layout also makes sense that
low conductive materials should be used in the middle so that
heat insulation effect is higher in the middle region.
On the other hand, if we change the boundary conditions
and try to maximize the temperature difference in a different
setting where the two points are placed horizontally on the horizontal middle axis. Then, we can have very different designs
as shown in Fig. 5, which leads to a circular distribution of
less conductive materials around the two points in the middle.
Obviously, the design solutions can depend on the objective
in a very sensible manner as demonstrated in this section. The
154

0.8

0.6

0.4

0.2

0
0

0.2

0.4

0.6

0.8

Fig. 4. Temperature distribution with the maximum |TA TB |.

0.8

0.6

0.4

0.2

0
0

0.2

0.4

0.6

0.8

Fig. 5. Optimal topology and distribution of two different materials.

configuration and boundary conditions are the same, but the


final distribution will naturally depend on the objective we use.
VI. D ISCUSSIONS AND F URTHER T OPICS
By formulating a topology optimization problem in terms
of an inverse problem and optimization, we have demonstrated
that bat algorithm is efficient in finding the optimal shape to
heat transfer design problems in microelectronic applications.
From our simulations, there are two important issues that can
be further improved.
First, the correct formulation of a problem is important so
that it becomes well-posed with unique solutions. This often
requires the incorporation of problem-specific knowledge into
the formulation in terms of proper constraints. Otherwise,
multiple solutions are possible, and sometime there may be
no solution at all. Therefore, care should be taken to ensure
the proper formulation before starting to solve the optimization
problem.
Another important issue is that we have used N N grids
to approach a 2D domain. At the moment, we used N = 40,
which gives 402 = 1600 design variables (indices). However,
if N = 100 is used, then the scale of the problem will
increase significantly to 1002 = 1, 000, 000. Therefore, a naive
approach to increase N to get higher accuracy in approximating a desired topology will increase the computational
costs tremendously. This may suggest that the approach of
approximating the shape is not the most suitable. In fact, in
aerospace engineering, other approaches such as B-splines and
Bezier curves are used as a more efficient way to parameterize
wing shapes. This can reduce the number of degrees of
freedom significantly, which may form an important topic for
further research in shape optimization.
Furthermore, the problems we have solved are in 2D
only. When extended to 3D, the design space becomes more
complex. The number of grids will be O(N 3 ). Even with a
good parameterized approach, we can have spline surfaces
in addition to curves, which still makes it much difficult to
handle. Future research can focus on new ways to formulate
shape optimization problems so that the number of design
variables can become minimal, which itself is an optimization
problem. Obviously, any progress in these areas will enable
us to design better products in a diverse range of applications.
R EFERENCES
[1] M. P. Bendse, Optimization of Structural Topology, Shape and Material, Springer, (1995).
[2] M. P. Bendse and O. Sigmund, Topology Optimization: Theory, Methods and Applications, Springer, Berlin, (2003).
[3] L. C. Cagnina, S. C. Esquivel, and C. A. Coello, Solving engineering
optimization problems with the simple constrained particle swarm
optimizer, Informatica, 32, 319-326 (2008).
[4] R. C. Eberhart and Y. Shi, Y., Comparing inertia weigthts and constriction factors in particle swarm optimization. Proceedings of the IEEE
Congress on Evolutionary Computation (CEC 2000), San Diego, CA.
pp. 84-88 (2000).
[5] A. Evgrafov, K. Maute, R. G. Yang and M. L. Dunn, Topology
optimization for nano-scale heat transfer, Int. J. Num. Methods in
Engrg., 77 (2), 285-300 (2009).

155

[6] A. H. Gandomi, X. S. Yang and A. H. Alavi, Cuckoo search algorithm:


a metaheuristic approach to solve structural optimization problems,
Engineering with Computers, 27, article DOI 10.1007/s00366-0110241-y, (2011).
[7] S. A. Greenhalgh, B. Zhou, and A. Green, Solutions, algorithms and
inter-relations for local minimization search geophysical inversion, J.
Geophys. Eng., 3, 101-113 (2006).
[8] C. L. Kar, I. Yakushin, K. Nicolosi, Solving inverse initial-value,
boundary-value problems via genetic algorithms, Engineering Applications of Artificial Intelligence, 13, 625-633 (2000).
[9] J. Kennedy and R. C. Eberhart, Particle swarm optimization, in: Proc.
of IEEE International Conference on Neural Networks, Piscataway, NJ.
pp. 19421948 (1995).
[10] J. Kennedy, R. C. Eberhart and Y. Shi, Swarm intelligence San
Francisco: Morgan Kaufmann Publishers, (2011).
[11] M. Sambridge, Geophysical inversion with a neighbourhood algorithm
I. Search a parameter space, Geophys. J. Int., 138, 479-494 (1999).
[12] M. Sambridge and K. Mosegaard, Monte Carlo methods in geophysical
inverse problems, Reviews of Geophysics, 40, 3-1-29 (2002).
[13] J. A. Scales, M. L. Smith, and S. Treitel, Introductory Geophysical
Inverse Theory, Samizdat Press, (2001).
[14] O. Sigmund, A 99 line topology optimization code written in Matlab,
Structural and Multidisciplinary Optimization, 21(2), 120-127 (2001).
[15] E. G. Talbi, Metaheuristics: From Design to Implementation, John
Wiley & Sons, (2009).
[16] X. S. Yang, Modelling heat transfer of carbon nanotubes, Modelling
Simul. Mater. Sci. Eng., 13, 893-902 (2005).
[17] X. S. Yang, Nature-Inspired Metaheuristic Algorithms, Lunver Press,
UK, (2008).
[18] X. S. Yang, Engineering Optimization: An Introduction with Metaheuristic Applications, John Wiley and Sons, USA (2010).
[19] X. S. Yang, A new metaheuristic bat-inspired algorithm, in: Nature
Inspired Cooperative Strategies for Optimization (NICSO 2010) (Eds.
J. R. Gonzalez et al.), Springer, SCI Vol. 284, 65-74 (2010).
[20] X. S. Yang and S. Deb, Cuckoo search via Levy flights, in: Proc. of
World Congress on Nature & Biologically Inspired Computing (NaBic
2009), IEEE Publications, USA, pp. 210214 (2009).
[21] X. S. Yang and S. Deb, Engineering optimization by cuckoo search,
Int. J. Math. Modelling Num. Optimisation, 1(4), 330343 (2010).
[22] V. V. Zhirnov, R. K. Cavin, J. A. Hutchby, G. I. Bourianoff, Limits
to binary logic switch scaling - a gedanken model, Proc. of the IEEE,
91(11), 1934-1939 (2003).

You might also like