Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Mechanical Engineering Department, Indian Institute of Technology Bombay, Mumbai 400076, India
Aerospace Engineering Department, Indian Institute of Technology Bombay, Mumbai 400076, India
a r t i c l e
i n f o
Article history:
Received 6 June 2015
Revised 11 December 2015
Accepted 3 January 2016
Keywords:
Micro electro mechanical systems
Pull-in instability
Particle swarm optimization
Hybrid simulated annealing
Electrostatic microbeams
Stable travel range
a b s t r a c t
This paper examines the enhancement of static and dynamic travel range of electrostatically driven microbeams using shape optimization approach. Continuous functions of width and thickness are used for
optimizing the geometry of both cantilever and xedxed microbeams. RayleighRitz energy method is
employed to compute the static and dynamic pull-in parameters. Particle swarm optimization and hybrid simulated annealing are used for shape optimization of microbeams. Constraints on design variables are imposed using penalty approach. Enhanced pull-in parameters obtained for variable geometry microbeams have been validated using 3-D nite element analysis. Optimized shapes of microbeams
show signicant improvement in static and dynamic travel range. Pull-in displacement is increased up to
54.92% for cantilever microbeam and 40.79% for xedxed microbeam with hybrid simulated annealing.
Effectiveness of particle swarm optimization is brought out through representative test cases. The convergence of the particle swarm optimization is approximately ve times faster as compared to the hybrid
simulated annealing, while maintaining the same level of accuracy.
2016 Elsevier Ltd. All rights reserved.
1. Introduction
reduced order model can be used in the present analysis and the
RayleighRitz energy based method as a solver. ROM is specially
preferred for design optimization problems where repetitive analysis has to be carried out by an optimization algorithm until the
objective function is extremized.
optimization of microbeams. Dielectric breakdown voltage is considered as upper limit of the rise in voltage. For prismatic microbeams, pull-in parameters are compared with experimental and
numerical ndings. Estimated pull-in parameters of variable geometry optimized microbeams are compared with 3-D nite element
simulation package COMSOLTM [29].
(1 )
narrow beam ((b < 5h ) ) [9]. Dimensions of the deformable electrode are, length L, width b (x) and thickness h (x) at co-ordinate
x along length and the corresponding dimensions of the referential prismatic microbeam are length L, width b p and thickness h p .
The distance from mid-plane of the deformable electrode to rigid
electrode is kept constant and it is denoted as gm . I(x) represents
area moment of inertia at x. When a step voltage is applied between xed and deformable electrode, attractive electrostatic force
causes deection of microbeam. Beyond a critical value of voltage,
deformable electrode snaps the xed electrode, resulting in collapse of the device. Value of voltage and the deformation at this
point is calculated using RayleighRitz energy method. For analysis, potential energy of the system consist of following terms:
(1) Strain energy due to bending of microbeam.
(2) Strain energy considering von Karman nonlinearity for midplane stretching, is applicable to xedxed microbeam only
[3].
(3) Electrostatic potential considering Palmers fringing eld
model [30].
In the present investigation EulerBernoulli beam theory is applied for modeling. Total potential energy of the system can be expressed as [26],
L
x
d2 w
= E
I(x)
2 0
dx2
2
dx
L
x
dw
E
+
N + A(x)
dx
0
8L
2
dx
2
dx
x
dw
dx
V 2 L
b (x)
2
b (x)
+
1 + ln
dx
2 0
(x)
(x)
g(x) w
g(x) w
(1)
(x) represents
Here, A (x) represents area of cross-section and w
is the axial force due to residual stress. Width
displacement at x. N
variation b (x) in terms of prismatic microbeam width b p and thickness variation h (x) in the form of h p for both the microbeams is as
follows:
b (x) = b p b(x )
(2)
h (x) = h p h(x )
(3)
Fig. 1. Schematic view of proposed variable geometry microbeams with xedxed and xed-free end conditions.
Here, functions b(x) and h(x) describe scaling of b p and h p respectively. Thickness variation causes variation in initial gap, represented by the following equation.
g(x) = g p +
h p
h (x)
2
2
(4)
Here, g p represents initial gap between referential prismatic microbeam and substrate. For taking common reference for nondimensional form of the equation, the distance of mid plane of microbeam from the xed electrode is taken as base, mathematically
represented as
gm = g p +
h p
2
the degree of geometric nonlinearity, represents fringe parameter, t represents dimensionless time and w(x, t ) is estimation of
deection of microbeam in dynamic mode. In this paper, effect of
residual stresses is neglected. Using above mentioned nondimensional terms, potential energy equation in dimensionless form is
=
1
2
3
+
4 2
(5)
x=
x
,
L
1 =
A p g2m
, T =
2Ip
b( x ) =
N=
(x)
w
,
gm
w (x ) =
A p L4
E Ip
b p
,
gm
h p
,
gm
E
Ee f f
N L2
,
E Ip
2 =
0 L4
g2p E Ip
b p h 3p
12
, w(x, t ) =
V2
2
dx
dw(x )
A (x )
dx
dw(x )
dx
d 2 w (x )
d x2
2
dx
2
dx
b( x )
2
+
1 + ln
g( x ) w ( x )
b( x )
g( x ) w ( x )
dx
(6)
b (x)
g(x )
h (x)
A (x)
, g( x ) =
, h (x ) =
, A (x ) =
,
gm
A p
b p
h p
,
I(x)
A p = b p h p , I (x ) =
,
Ip
Ip =
I (x )
In this section, pull-in parameters for static and dynamic operating conditions are estimated numerically.
V2 =
b p L4V 2
2E Ip g3m
(x, t)
w
t
, t=
T
gm
w ( x ) = as ( x )
(7)
are as follows:
dw(x )
dx
w(x )|x=0 = 0,
= 0,
x=0
d2 w(x )
d x2
1
Kb =
2
d3 w(x )
d x3
= 0,
x=1
1
0
d 2 d ( x )
I (x )
d x2
=0
Us (ad ) =
= 0,
dw(x )
dx
=0
w(x )|x=0 = 0,
dx
(ii) Strain energy considering von Karman nonlinearity for midplane stretching [3]
x=1
dw(x )
dx
x=0
N
1
d (x )
+
A ( x ) ad
2
4
dx
d (x )
ad
dx
2
dx
2
dx
From the available resources, shape functions chosen for cantilever microbeam is [26,31],
Ks 4 Kn 2
a +
a ,
4 d
2 d
2 1
2
1
d (x )
d (x )
Ks = 1
A (x )
dx
dx
dx
dx
0
0
Kn = N
w(x )|x=1 = 0,
Us (ad ) =
x=1
where, Dc =
cos + cosh
sin + sinh
(8)
(9)
where D f =
cos cosh
sin sinh
(11)
d2
=0
da2s
(12)
w(x, t ) = ad (t ) (x )
(13)
here, ad (t) indicates time parameter and (x) is spatial function dened in Section 3.1. Potential energy equation in nondimensional
form for dynamic operating condition is rewritten using following
terms
1
Ub (ad ) =
2
Ub (ad ) =
1
0
1
0
a2d Kb
2
d2 w(x, t )
I (x )
d x2
dx
1
0
ad
0
F ( ) ddx
KB 2 KN 2 KS 4
a +
a +
a 2V 2
2 d
2 d
4 d
d 2 d ( x )
I ( x ) ad
d x2
dx
2
dx
0
ad
0
F ( ) ddx
w(x, 0 ) = 0,
w(x, 0 )
=0
t
(14)
w ( 0, t ) = 0,
w ( 0, 1 )
w ( 1, t )
= 0, w ( 1, t ) = 0,
=0
x
x
Cantilever microbeam
w ( 0, t ) = 0,
=
1
Ub (ad ) =
2
d (x )
dx
U f (ad ) = 2V 2
(10)
d
= 0,
d as
w ( 0, t )
2 w ( 1, t )
3 w ( 1, t )
= 0,
=
0
=0
x
x2
x3
Response of the electrostatically driven MEMS devices up to dynamic pull-in voltage is periodic [26,32] and at pull-in point, velocity becomes zero, beyond pull-in response becomes non-periodic.
For extracting dynamic pull-in parameters, conditions expressed in
Eq. (15) hold true,
= 0,
d
=0
d ad
(15)
Table 1
Comparison of the estimates of static and dynamic pull-in voltage of prismatic microbeam obtained from energy based method with the data available
from the literature.
Case
Type of
Mode of
No.
beam
actuation
b
(m)
Pull-in
Relative
Present
work
Corresponding
value from ref.
error(%) in
Pull-in
voltage
37.84
0.09
(Chowdhury et al.) [6]
2.27
0
(Pamidighantam et al.) [34]
2.25
0.44
(Chaterjee and Pohit) [35]
2.61
0.99
(Durieu et al.) [36]
27.95
1.24
(Tilmans and Legtenberg) [14]
3.8
0.26
(Ijntema and Tilmans) [37]
26.4
1.79
(Zand and Ahmedian) [38]
41.68
0.00
( Krylov) [39]
h
(m)
g p
(m)
E
(GPa)
displacement
Present
work
169
0.06
0.45
37.87
Cantilever
Static
100
50
Cantilever
Static
300
50
1.00
2.5
77
0.33
1.12
2.27
Cantilever
Dynamic
80
10
0.5
0.7
169
0.33
0.46
2.24
Cantilever
Dynamic
300
50
77
0.06
0.65
2.58
Fixedxed
Static
210
100
1.5
1.18
166
0.3
0.5116
Fixedxed
Static
500
100
1.5
175
0.3
0.42
3.79
Fixedxed
Dynamic
210
100
1.5
1.18
166.00
0.30
0.73
26.88
Fixedxed
Dynamic
300
20
2.0
169.00
0.28
0.43
41.68
28.3
Table 2
Comparison of estimated static pull-in parameters using energy method with those obtained
using 3-D nite element simulations for prismatic and variable geometry microbeams.
Type of beam
Cantilever
Fixedxed
Beam geometry
Prismatic
Nonprismatic
(Test case S9)
Prismatic
Nonprismatic
(Test case S9)
Pull-in parameters
by proposed method
Pull-in parameters
using 3-D FEA (COMSOL)
Displacement
(m)
Voltage
(Volts)
Displacement
(m)
Voltage
(Volts)
0.9017
1.09
6.45
11.46
0.90
1.078
6.48
11.225
0.9142
1.082
44.29
65.56
0.909
1.094
44.70
66.005
bc ( x ) = y1 ( 1 y2 xy3 )
(16)
hc ( x ) = y5 ( 1 y6 xy7 )
y8
b f ( x ) = y1 ( y2 + ( 1 y2 )[ ( 1 2x ) ] )y4
(17)
2 y7 y8
h f ( x ) = y5 ( y6 + ( 1 y6 ) ( ( 1 2x ) ) )
vi. The term Phmax represents the degree of violation of maximum thickness constraint, represented as,
Subject to :
b min
b p
The term hmax (x ) represents maximum thickness of the variable geometry microbeam conguration at ith iteration of
optimization process. hmax represents the nondimensional
upper bound of thickness.
vii. The term Pvbd shows that if the actuation voltage is greater
than break down voltage of dielectric medium, cost of
penalty Cvbd is imposed,
0 y2 1
0 y3
0 y4
h min
h max
y5
h p
h p
0 y6 1
0 y8
1
1
b(x )dx = 0
1
V Vbd
(18)
In the Eq. (18), f represents objective function with design vector Y and w ps is the pull-in displacement in static mode, w pd is
the pull-in displacement in dynamic mode, to be maximized. Term
P and C with different subscripts show penalties and cost for violating constraints respectively. For example, term Parea and Carea
represents penalty and cost for violating constant area of overlap.
Expression for penalty and cost of violation of different constraints
are mathematically represented as follows:
i. The term Parea species the penalty for violation of constant
area of overlap of the variable geometry microbeam.
Parea = Carea 1
b(x )dx
(19)
ii. The term Pvol quanties the degree of violation of constant volume constraint,volume of the proposed microbeam
should be same as referential prismatic microbeam.
Pvol = Cvol 1
1
0
(20)
iii. Penalty for minimum width constraint violation Pbmin , is expressed as,
2
(21)
bmin (x )
where,
represents minimum width of the microbeam
dened by design parameters y1 , y2 , y3 , y4 at ith iteration.
bmin represents lower bound of the normalized width.
iv. The term Pbmax represents the degree to which maximum
width constraint is violated. Mathematically represented as,
Pbmax =
(x ) bmax ] )
(22)
bmax (x )
The term
represents maximum width of the variable geometry microbeam at ith iteration. bmax represents
the non dimensional upper bound of the width.
v. Minimum thickness violation constraint is penalized using
Phmin , mathematically represented as,
(25)
0 y7
(24)
2
(23)
The term
represents minimum thickness of the proposed microbeam for any combination of design parameters,
y5 , y6 , y7 , y8 at ith iteration. hmin represents the non dimensional lower bound of the thickness.
S (i + 1 ) = S (i ) + V el (i + 1 )
V el (i + 1 ) = uV el (i ) + c1 1 (Pbest S (i ) ) + c2 2 (Gbest S (i ) )
(26)
Where u(i) is called inertia of the particle, is the coecient
determining the effect of previous velocity rate on the current velocity, deciding self condence of the particle. The coecients c1 ,
1 and c2 , 2 decide the impact of Pbest and Gbest respectively. 1
and 2 are random numbers between 0 and 1, c1 and c2 are known
as individual and social factors respectively. In the present study
variable inertia model of PSO is adopted and change in inertia
value is governed by,
u (i ) =
(us u f )
(i 1 ) + u f
(l 1 )
(27)
Algorithm 1 describes the pseudocode of PSO, code is developed in MATLABTM [46] for this algorithm. PSO has been used in
design optimization of various devices. For example, Lake et al.
[25] successfully used PSO for design of slotted MEMS resonators.
= Q
(28)
Q=
ln(Ps )
ln(Pf )
1/(N1)
(29)
Case No.
bmax /L
bmin /L
hmax /L
hmin /L
S1, D1
S2, D2
S3, D3
S4, D4
S5, D5
S6, D6
S7, D7
S8, D8
S9, D9
0.2
0.3
0.4
0.2
0.3
0.4
0.2
0.3
0.4
0.02
0.02
0.02
0.04
0.04
0.04
0.06
0.06
0.06
0.008
0.009
0.01
0.008
0.009
0.01
0.008
0.009
0.01
0.003
0.003
0.003
0.004
0.004
0.004
0.005
0.005
0.005
p=
f
(30)
where, f is the difference of objective function value of two consecutive iterations. It is successfully employed by Trivedi et al.
[50] for shape optimization of electrostatically driven cantilever
microbeam.
In Section 4, the results obtained by implementing particle
swarm optimization (PSO) and hybrid simulated annealing (HSA)
to the objective function are presented and discussed.
4. Results and discussion
In this section, results obtained using particle swarm optimization and hybrid simulated annealing for optimizing pull-in displacement using Eq. (18) are discussed. For generalization purpose,
nine representative test cases for xedxed and cantilever microbeam are considered. Upper and lower limit of geometrical dimensions of microbeam for each test case is shown in Table 3.
Width and thickness constraints in normalized form are chosen from the prismatic beam dimensions available in various research articles [6,9,26]. Nomenclature for various test cases of
microbeam is as follows, S indicates static mode and D represents
dynamic mode of actuation. Numerals represents row number of
the Table 3. Here row number indicates geometric constraints for
the respective test case, for example, S1 represents static mode of
actuation having geometric constraints dened in row number 1 of
Table 3. Similarly, D3 represents dynamic mode of actuation having
constraints corresponding to row number 3 of Table 3.
Fig. 3. Comparison of optimized width and thickness proles of cantilever microbeam using hybrid simulated annealing and particle swarm optimization for dynamic mode
of actuation.
Normalized value of width and thickness for referential prismatic beam is chosen as
[6,9,26].
b pr
L
= 0.15,
h pr
L
= 0.00667, g pr = 0.00667
ii.
iii.
iv.
v.
vi.
vii.
viii.
ix.
Maximum iterations: 60 0 0.
Swarm size : 30.
Initial swarm : Random initialization.
Initial velocity: Randomly initialized.
Social factor : 2.1.
Cognitive factor : 2.1.
Inertia weight: [0.9, 0.4] using Eq. (27).
Function tolerance: 107 .
Fig. 4. Comparison of optimized width and thickness proles of xedxed microbeam using hybrid SA and PSO for dynamic mode of actuation.
The swarm size dependency shown in Fig. 2 clearly indicates that the maximum pull-in displacement value obtained using swarm size 10, 15, 20 and 25 can be improved by increasing the swarm size. If the swarm size is increased beyond 30, not
much improvement is observed in the maximum pull-in displacement value. Therefore, in order to save computational effort without compromising the accuracy of results, swarm size is decided
as 30.
Pull-in displacement is optimized in static and dynamic operating conditions of cantilever and xedxed microbeam discussed
separately in the following sections.
10
Fig. 5. Convergence graphs of cantilever and xedxed microbeam using PSO and hybrid SA for representative test case S9.
ment as compared to referential prismatic microbeam. While implementing hybrid simulated annealing, the results show that maximum rise in static pull-in displacement is 54.92% for test case S3.
4.3.2. Comparison of optimized design variables and pull-in
parameters obtained using PSO and HSA for cantilever microbeam in
dynamic operating condition
Table 5 presents optimized design variables and dynamic pullin parameter for test cases D1 to D9 for cantilever microbeam.
As presented in Table 5, while employing particle swarm optimization, maximum rise in dynamic pull-in displacement is 49.90%
for test case D3. On parallel lines, using hybrid simulated annealing, increment in dynamic travel range is 49.81% for test case
D2. Test cases having maximum range for thickness and width
have shown signicant improvement in the pull-in displacement
in static and dynamic operating conditions. Width and thickness
proles of cantilever microbeams, optimized in static and dynamic
operating conditions show striking similarity in most of the test
cases. Optimized width and thickness proles of cantilever microbeam in dynamic operating condition using PSO and HSA are
shown in Fig. 3.
Convergence criteria for the optimization problem is set as
follows:
11
Pull-in parameters
Case
No.
y1
y2
y3
y4
y5
y6
y7
y8
Normalized
voltage (Vps )
Normalized
displacement (w ps )
% Rise in
w ps
S1
S1
S2
S2
S3
S3
S4
S4
S5
S5
S6
S6
S7
S7
S8
S8
S9
S9
1.333
1.333
1.824
1.928
1.831
2.164
1.021
1.333
1.468
1.835
1.495
1.588
1.303
1.333
1.375
1.598
1.363
1.444
0.045
0.045
0.233
0.233
0.190
0.140
0.164
0.212
0.131
0.781
0.422
0.333
0.623
0.042
0.128
0.149
0.328
0.233
4.440
4.444
1.961
1.786
1.974
1.550
45.375
3.143
2.395
0.995
2.046
1.835
2.100
2.485
2.181
1.451
2.089
1.815
50.0 0 0
50.0 0 0
9.884
10.087
12.437
18.425
7.501
6.739
12.176
1.270
3.147
4.405
1.209
28.316
9.005
8.579
3.089
4.839
1.190
1.180
1.219
1.306
1.319
1.285
1.142
1.190
1.199
1.326
1.303
1.301
1.159
1.185
1.192
1.239
1.190
1.233
0.032
0.032
0.043
0.085
0.091
0.059
0.980
0.188
0.308
0.373
0.201
0.170
0.097
0.330
0.171
0.084
0.189
0.057
2.161
2.246
1.545
1.180
1.174
1.177
22.563
1.571
1.346
0.827
1.009
0.989
1.261
0.984
1.002
0.803
1.014
0.867
30.0 0 0
30.0 0 0
22.752
11.973
11.330
17.404
18.350
3.286
1.882
1.701
3.461
4.147
4.267
1.142
2.468
5.721
2.206
8.496
2.801
2.794
3.644
3.970
3.896
5.149
1.848
2.539
2.728
3.127
2.918
3.020
2.174
2.254
2.306
2.517
2.284
2.407
0.441
0.441
0.461
0.464
0.464
0.466
0.374
0.401
0.405
0.405
0.406
0.408
0.365
0.367
0.367
0.368
0.367
0.368
46.67
46.57
53.33
54.46
54.32
54.92
24.55
33.43
34.60
34.83
35.16
35.60
21.29
22.02
22.19
22.46
22.02
22.42
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
12
Pull-in parameters
Case
No.
y1
y2
y3
y4
y5
y6
y7
y8
Normalized
voltage (Vpd )
Normalized
displacement (w pd )
% Rise in
w pd
D1 (PSO)
D1(HSA)
D2 (PSO)
D2 (HSA)
D3 (PSO)
D3 (HSA)
D4 (PSO)
D4 (HSA)
D5 (PSO)
D5 (HSA)
D6 (PSO)
D6 (HSA)
D7 (PSO)
D7 (HSA)
D8 (PSO)
D8 (HSA)
D9 (PSO)
D9 (HSA)
1.333
1.2900
1.909
1.9755
1.914
1.8228
1.323
1.3333
1.488
1.6472
1.540
1.7119
1.295
1.3137
1.388
1.4140
1.352
1.4182
0.163
0.9200
0.215
0.2593
0.283
0.1916
0.033
0.0320
0.110
0.5666
0.306
0.5065
0.338
0.4481
0.146
0.2977
0.393
0.5286
4.305
2.8191
1.825
1.7032
1.776
1.9885
3.312
3.3118
2.322
1.5038
1.994
1.4407
2.512
2.2572
2.103
1.8826
2.086
1.6661
12.946
0.8988
10.990
8.9821
7.993
12.2963
49.566
49.5660
14.765
2.1779
4.793
2.6329
2.844
2.0 0 09
7.906
3.5738
2.443
1.6832
1.190
1.1642
1.273
1.3369
1.377
1.3279
1.190
1.1900
1.245
1.3081
1.284
1.3529
1.157
1.1678
1.182
1.1726
1.181
1.2315
0.036
0.3783
0.086
0.0981
0.043
0.0920
0.014
0.0136
0.122
0.4653
0.297
0.2131
0.058
0.4115
0.190
0.1230
0.124
0.2025
2.163
2.3464
1.280
1.0871
1.040
1.1534
1.650
1.6502
1.217
0.8444
1.002
0.8383
1.282
1.0451
1.033
1.1006
1.085
0.8408
26.197
1.9996
11.584
10.5439
25.294
11.2152
49.980
49.9796
5.600
1.2449
2.156
3.3936
7.202
0.8351
2.155
3.4036
3.443
2.1918
2.532
2.289
3.491
3.672
3.688
3.516
2.304
2.314
2.576
2.727
2.655
2.861
1.600
2.046
2.088
2.093
2.051
2.138
0.632
0.606
0.623
0.657
0.657
0.656
0.577
0.578
0.583
0.581
0.583
0.583
0.506
0.529
0.529
0.529
0.528
0.529
44.22
43.21
49.53
49.81
49.90
49.71
31.69
31.92
32.92
32.65
32.90
33.11
20.33
20.65
20.69
20.60
20.53
20.63
If the difference of the objective function value of two consecutive iterations is less than or equal to 107 (value of function
tolerance), the optimization algorithm will stop iterations and the
minimum value of the objective function is reported as optimized
value. Convergence graphs for test case S9 is shown in Fig. 5a for
PSO and for HSA is shown in Fig. 5c for cantilever. It clearly indicates that solution converges around 50 0 0 iterations using PSO,
which is almost 5 times faster than those obtained using HSA. It is
depicted from convergence graph that hybrid simulated annealing
is one of the alternatives among the optimization algorithms for
the problem under investigation. Solution using HSA is converged
in around 320 0 0380 0 0 iterations. Width and thickness proles of
cantilever microbeam show striking similarity in most of the test
cases.
4.4. Comparison of the design parameters and normalized pull-in
parameters using PSO and HSA of xed-xed microbeam
In this section, for xedxed microbeam, the optimized pullin parameters using PSO and HSA are compared for 9 different test
cases in static and dynamic operating conditions.
4.4.1. Comparison of optimized design variables and pull-in
parameters obtained using PSO and HSA for xed-xed microbeam in
static operating condition
Results of xedxed microbeam for static operating condition
is shown in Table 6. It shows that using PSO, maximum pull-in displacement is observed in test case S3 showing 34.20% rise as compared to referential prismatic xed-xed microbeam. On parallel
lines, shape optimization using hybrid simulated annealing indicate that static travel range is maximum in test case S2 and it is
increased by 40.79% compared to prismatic xedxed microbeam.
4.4.2. Comparison of optimized design variables and pull-in
parameters obtained using PSO and HSA for xedxed microbeam in
dynamic operating condition
Table 7 shows the comparison of the optimized design and pullin parameters for xedxed microbeam using particle swarm optimization and hybrid simulated annealing optimization techniques
in dynamic operating condition. In dynamic mode of actuation,
shape optimization using PSO show 39.15% rise in travel range for
the test case D2. While shape optimization using hybrid simulated
13
Table 6
Comparison of optimized design parameters of xedxed microbeam for maximizing static pull-in displacement using particle swarm optimization (PSO) and hybrid simulated annealing (HSA).
Design variables
Pull-in parameters
Case
No.
y1
y2
y3
y4
y5
y6
y7
y8
Normalized
voltage (Vps )
Normalized
displacement (w ps )
% Rise in
w ps
S1
S1
S2
S2
S3
S3
S4
S4
S5
S5
S6
S6
S7
S7
S8
S8
S9
S9
1.333
1.333
1.997
1.834
1.633
1.609
1.328
1.333
2.0 0 0
1.913
1.676
1.704
1.108
1.333
1.964
1.799
1.789
1.685
0.053
0.056
0.019
0.010
0.045
0.042
0.139
0.151
0.076
0.063
0.049
0.035
0.046
0.017
0.066
0.048
0.016
0.018
0.231
0.227
0.777
0.762
0.429
0.420
0.252
0.249
0.795
0.774
0.712
0.715
0.174
0.600
1.111
1.003
1.265
1.091
0.783
0.800
0.684
0.569
0.807
0.788
0.814
0.850
0.782
0.711
0.556
0.553
0.332
0.297
0.577
0.496
0.362
0.360
1.200
1.199
1.349
1.349
1.500
1.489
1.200
1.190
1.350
1.349
1.350
1.490
1.200
1.190
1.343
1.349
1.490
1.490
0.008
0.010
0.105
0.010
0.385
0.067
0.010
0.010
0.318
0.206
0.058
0.048
0.081
0.010
0.501
0.445
0.409
0.418
0.670
0.641
1.608
1.417
0.677
1.091
0.927
0.892
1.212
1.015
1.271
1.824
0.900
1.456
1.305
1.284
1.813
1.706
0.204
0.213
0.314
0.238
1.260
0.442
0.151
0.149
0.597
0.513
0.285
0.300
0.175
0.100
0.843
0.726
0.769
0.787
6.991
6.980
10.838
10.768
10.793
10.788
6.916
6.859
10.309
10.331
9.463
10.781
6.020
6.793
9.484
9.112
9.923
9.579
0.375
0.374
0.403
0.429
0.409
0.414
0.370
0.369
0.385
0.397
0.392
0.394
0.343
0.361
0.361
0.363
0.359
0.359
23.04
22.74
32.10
40.79
34.20
35.74
21.53
21.20
26.29
26.92
28.78
29.14
12.67
13.48
18.54
19.04
17.92
17.76
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
Table 7
Optimized design parameters of xedxed microbeam for maximizing dynamic pull-in displacement using particle swarm optimization.
Design variables
Pull-in parameters
Case
No.
y1
y2
y3
y4
y5
y6
y7
y8
Normalized
voltage (Vpd )
Normalized
displacement (w pd )
% Rise in
w pd
D1
D1
D2
D2
D3
D3
D4
D4
D5
D5
D6
D6
D7
D7
D8
D8
D9
D9
1.333
1.333
1.867
1.831
1.637
1.647
1.333
1.333
2.0 0 0
1.910
1.695
1.697
1.327
1.327
1.955
1.984
1.496
2.097
0.010
0.075
0.010
0.010
0.046
0.048
0.143
0.151
0.075
0.062
0.037
0.041
0.034
0.019
0.067
0.065
0.010
0.034
0.340
0.221
0.788
0.761
0.432
0.430
0.253
0.249
0.797
0.773
0.723
0.711
0.507
0.583
1.111
1.112
0.939
1.954
0.500
0.843
0.573
0.567
0.808
0.827
0.827
0.850
0.779
0.709
0.545
0.560
0.356
0.302
0.574
0.586
0.287
0.405
1.190
1.200
1.340
1.350
1.500
1.490
1.200
1.190
1.350
1.350
1.500
1.350
1.188
1.200
1.350
1.350
1.171
1.497
0.010
0.058
0.01
0.010
0.383
0.382
0.009
0.010
0.313
0.206
0.049
0.059
0.013
0.010
0.495
0.503
0.640
0.501
0.618
0.471
1.409
1.415
0.679
0.671
0.949
0.892
0.928
1.017
1.840
1.292
1.367
1.521
1.328
1.555
0.500
1.903
0.211
0.317
0.237
0.239
1.254
1.244
0.148
0.149
0.698
0.513
0.303
0.285
0.106
0.102
0.837
0.798
1.0 0 0
1.0 0 0
6.325
6.364
8.142
9.795
9.853
9.825
6.339
6.268
9.633
9.395
9.826
8.729
6.169
6.244
8.631
8.577
6.481
9.346
0.524
0.523
0.601
0.600
0.578
0.578
0.521
0.518
0.558
0.558
0.554
0.554
0.506
0.507
0.510
0.504
0.505
0.492
21.22
20.96
39.15
38.76
33.69
33.78
20.43
19.94
29.11
29.08
28.16
28.27
17.01
17.33
18.00
16.50
16.94
16.81
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
(PSO)
(HSA)
14
Fig. 8. Time history diagram of microbeams for representative test case D9.
Various test cases are presented to demonstrate the applicability of particle swarm optimization and hybrid simulated annealing method for shape optimization of electrostatically actuated microbeams.
5. Concluding remarks
In this paper, novel design of electrostatically driven microbeams is presented for enhancement of travel range. Shape
optimization of microbeams for enhanced travel range is carried
out successfully using PSO and hybrid SA. To the best of authors
knowledge, this is the rst instance of PSO being used for optimizing a width and a thickness proles for travel range enhancement of electrostatically driven microbeams. From the presented
results,following conclusions can be drawn:
15
16
[42] Schaffert RM. Electrophotography. Focal Press; 1975. ISBN 10: 0240507819
[43] Kennedy J, Eberhart RC. Particle swarm optimization. In: Proceedings of the
IEEE international conference on neural netwroks, vol. 4; 1995. p. 19428.
doi:10.1109/ICNN.1995.488968.
[44] Hu X, Eberhart RC, Shi Y. Swarm intelligence for permutation optimization:
A case study on n-queens problem. In: Proceedings of the 2003 IEEE swarm
intelligence symposium; 2003. p. 2436.
[45] Rosendo M, Pozo A. Applying a discrete particle swarm optimization algorithm
to combinatorial problems. In: Proceedings of the 2010 eleventh Brazilian symposium on neural networks; 2010. p. 23541. doi:10.1109/SBRN.2010.48.
[46] MATLAB. Optimization toolbox users guide (for use with MATLAB). Natick, MA:
MathWorks, Inc.; 2001.
[47] Kirkpatrick S, Gelatt CD, Vecchi MP. Optimization by simulated annealing. Science 1983;220(4598):67180. doi:10.1126/science.220.4598.671.
[48] Trivedi RR, Bhashan A, Joglekar MM, Pawaskar DN, Shimpi RP. Enhancement of
static and dynamic travel range of electrostatically actuated microbeams using
hybrid simulated annealing. Int J Mech Sci 2015;98(4):93110.
[49] Balling RJ. Optimal steel frame design by simulated annealing. J Struct Eng
1991;6(117):178095.
[50] Trivedi RR, Joglekar MM, Shimpi RP, Pawaskar DN. Shape optimization of electrostatically actuated micro cantilever beam with extended travel range using
simulated annealing. In: Proceedings of the 2011 world congress on engineering (WCE 2011), vol. III; 2011. p. 20427.
[51] Najar F, Choura S, El-Borgi S, Abdel-Rahman EM, Nayfeh AH. Modeling and design of variable-geometry electrostatic microactuators. J Micromech Microeng
2005;15(3):41929. doi:10.1088/0960-1317/15/3/001.
[52] Joglekar MM, Pawaskar DN. Pull-in dynamics of variable-width electrostatic
microactuators. In: Proceedings of the ninth biennial conference on engineering systems design and analysis, Haifa, Israel, vol. 4; 2009. p. 32735.