Vous êtes sur la page 1sur 12

Design of Compliant Mechanisms

for Minimizing Input Power

in Dynamic Applications
In this paper, we investigate power flow in compliant mechanisms that are employed in
dynamic applications. More specifically, we identify various elements of the energy stor-
Tanakorn Tantanawat age and transfer between the input, external load, and strain energy stored within the
e-mail: tanakorn@umich.edu compliant transmission. The goal is to design compliant mechanisms for dynamic appli-
cations by exploiting the inherent energy storage capability of compliant mechanisms in
Sridhar Kota the most effective manner. We present a detailed case study on a flapping mechanism, in
e-mail: kota@umich.edu which we compare the peak input power requirement in a rigid-body mechanism with
attached springs versus a distributed compliant mechanism. Through this case study, we
Department of Mechanical Engineering, present two approaches: (1) generative-load exploitation and (2) reactance cancellation,
The University of Michigan, to describe the role of stored elastic energy in reducing the peak input power require-
Ann Arbor, MI 48109 ment. We propose a compliant flapping mechanism and its evaluation using nonlinear
transient analysis. The input power needed to drive the proposed compliant flapping
mechanism is found to be 50% less than a rigid-link four-bar flapping mechanism without
a spring, and 15% less than the one with a spring. This reduction of peak input power is
primarily due to the exploitation of elasticity in compliant members. The results show
that a compliant mechanism can be a better alternative to a rigid-body mechanism with
attached springs. 关DOI: 10.1115/1.2756086兴

Keywords: compliant mechanisms, flapping mechanisms, flapping-wing air vehicles, in-

put power reduction

Introduction ant mechanisms may be efficient when they are operated under a
quasi-static condition, but not a dynamic one. Analysis and syn-
When a compliant mechanism is deformed to transmit force
thesis tools based on a dynamic condition are, therefore, necessary
and motion, some of the input energy is stored in the mechanism
to extend the applications of compliant mechanisms. While the
in the form of strain energy, while the rest is transferred to the
purpose of this ongoing research is to develop a systematic
output load. This stored energy is commonly perceived as energy
method of designing compliant mechanisms based on dynamic
loss, preventing full energy transfer from the input port to the
performance, this paper presents the results, as a preliminary step,
output port. One of the approaches to design efficient compliant of an attempt to understand the benefits of elasticity in dynamic
mechanisms for a quasi-static condition, therefore, attempts to systems. This understanding will be fundamental for the develop-
minimize this stored strain energy. The scenario is completely ment of an efficient and robust design tool during the course of the
different if a compliant mechanism is operated under a dynamic research.
condition, in which strain energy is stored and released during a Due to its multifunctional structure that combines the function
cycle of operation. Most of the stored strain energy, if not all, can of a mechanism and an energy storage component together, a
be recycled. An approach used to design a compliant mechanism compliant mechanism provides a lightweight and compact system.
for a dynamic condition should, therefore, be different from the Its monolithic structure leads to scalability, no wear, and low
one used for a quasi-static condition. Khatait et al. 关1兴 suggest the manufacturing cost. In addition, its ability to incorporate uncon-
use of compliant mechanisms to reduce input torque requirement ventional actuators makes it very attractive to several applications.
in a flapping mechanism. The study is, however, limited to the Nonetheless, if a distributed compliant mechanism is used, further
case of unloaded lumped compliant mechanisms. In addition, benefit can be gained. Since stress and strain are more uniformly
Madangopal et al. 关2兴 optimize a four-bar flapping mechanism distributed than those in lumped compliant mechanisms, there is
with attached springs in an attempt to minimize the input torque less stress concentration but higher energy storage capacity in
requirement of a flapping-wing micro air vehicle. However, how distributed compliant mechanisms. Because of all these benefits, a
the springs enable input torque reduction has not been clearly compliant mechanism is expected to be an excellent alternative to
explained. These two examples have demonstrated the benefits of a rigid-body mechanism, especially for autonomous robots, whose
elasticity in reducing input torque requirement, and lead us to size and weight are critical. In this study, a compliant flapping
believe that a distributed compliant mechanism holds a potential mechanism, which can be applied to flapping-wing micro air ve-
to be an alternative for dynamic systems. hicles developed by other researchers, will be used as a case study
Since most of the early research and development in analysis during the development of the design method.
and synthesis tools for compliant mechanisms were based on a
quasi-static assumption, the performance of the designed compli- Related Work
The use of compliant mechanisms to transmit force and motion
Contributed by the Mechanisms and Robotics Committee of ASME for publica- has been studied for several decades. This type of mechanism
tion in the JOURNAL OF MECHANICAL DESIGN. Manuscript received July 13, 2006; final evolved from work in robotics, where manipulator arms must per-
manuscript received October 25, 2006. Review conducted by Larry L. Howell. Paper
presented at the ASME 2006 Design Engineering Technical Conferences and Com-
form various tasks on workpieces whose locations and dimensions
puters and Information in Engineering Conference 共DETC2006兲, Philadelphia, PA, are not precisely known 关3兴. In this application, the use of com-
September 10–13, 2006. pliance is one of the approaches to accommodate such uncertain-

1064 / Vol. 129, OCTOBER 2007 Copyright © 2007 by ASME Transactions of the ASME

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
ties. Work in the development of compliant mechanisms for ro-
botics applications can be found at least as early as in the 1980s
关3–6兴. In 1983, Holl et al. 关7兴 proposed a compliant mechanism
for joints in prosthetic devices. Holl noted the self-stabilizing
characteristics and other benefits of compliance including reduc-
tion in wear, material weight, and manufacturing cost. Another
work in compliant joints was reported by Trease et al. 关8兴 in 2005,
where several designs of highly effective and kinematically well-
behaved compliant joints were proposed. Since compliant mecha-
nisms were studied extensively, various applications of compliant
mechanisms have been investigated, including bistable compliant
mechanisms 关9兴, compliant micromechanisms and structures
关10,11兴, micromanipulators and microrobots 关12兴, actuator lever- Fig. 1 Four-bar flapping mechanism adapted from Madango-
age 关13,14兴, structural shape morphing in smart structures 关15兴, pal et al. †2‡
surgical tools 关16兴, active flow control 关17兴, and vibration isolation
Various methods of analysis and synthesis of compliant mecha-
nisms can be found in the literature including Pseudo-Rigid-Body Note that the links’ cross sections are adjusted so that their
Models 共PRBMs兲 and continuum mechanics methods for topol- masses match those in Madangopal’s model. The operating fre-
ogy, geometry, and size optimization. Much of the reported work quency is adjusted to match the peak values of inertial torque.
on analysis and design assumes a quasi-static condition. It was not Finally, a damping constant and a constant force component are
until recently that researchers began to incorporate dynamic be- adjusted to match the peak value of the lift. Since the weights of
haviors of compliant mechanisms, such as Du et al. 关19兴, Nishi- the links are small compared to the aerodynamic force on the
waki et al. 关20兴, and Maddisetty and Frecker 关21兴. Li and Kota wing, the gravity effect is ignored in this study to simplify the
关22兴 proposed a systematic method for dynamic analysis of com- analysis. Once the basic understanding is gained, more accurate
pliant mechanisms. Dimensional synthesis based on desired mode results can be obtained by including the gravity.
shapes is presented in 关23兴 by Lai and Anathasuresh. Boyle et al. The input power can be either positive or negative during a
关24兴, Handley 关25兴 et al., and Yu et al. 关26兴 presented the devel- flapping cycle. A positive input power indicates that the motor is
opment in dynamic modeling of compliant mechanisms based on supplying energy into the system, while a negative input power
PRBMs. Khatait et al. 关1兴 used dynamic PRBMs to investigate the indicates that the motor is absorbing energy from the system.
reduction of torque requirement in a flapping mechanism. None of Since the focus of this study is to capture the instantaneous input
the previous work, however, directly addressed the issue of input power experienced by the motor, which affects the design of mo-
power requirement. Since it is the power requirement that governs tors and electrical components, we ignore the energy loss in the
various component designs and affects the overall system’s per- motor and mechanism. In other words, we assume that all energy
formance, we use the power requirement as one of the criteria in absorbed by the motor can be fully recovered. In addition, we
the design method. Even though a design approach based on assume that instantaneous power experienced by the motor is
torque requirement is equivalent to a design approach based on equivalent in both directions 共supplying and absorbing energy兲.
power requirement when the operating frequency is constant, it is The quantity being investigated as an objective function to be
not the case when the frequency changes during the operation, or minimized in this study is, therefore, the maximum of the absolute
when the frequency is one of the design variables during a system value of instantaneous input power, or a peak input power.
design. A design approach based on the power requirement will The values of the spring stiffness and free length are optimized
still be applicable to these situations. using a Generalized Reduced Gradient 共GRG兲 algorithm provided
by ADAMS. The optimal values of the spring’s properties, along
with some other parameters, are shown in Table 2. Plots of inertial
Power Analysis of a Four-Bar Flapping Mechanism torque and lift obtained from this simplified model are shown in
Figs. 2 and 3, respectively. Plots of input powers for the mecha-
To understand the behavior of input power reduction, we first nism with and without a spring are shown in Fig. 4. Various power
perform power analysis on a four-bar flapping mechanism opti- components for the mechanism with a spring are shown in Fig. 5.
mized by Madangopal et al. 关2兴. In their work, the aerodynamic Figure 4 shows that adding a spring reduces peak input power
model is based on quasi-steady blade element analysis. The model by 42% from 1.00 N m / s to 0.58 N m / s. Figure 5 indicates that
includes unsteady wake effects, camber and partial leading edge energy flow in and out of the link’s mass 共kinetic energy rate
suction effects, and post-stall behavior. The grounds for the 共KER兲, spring potential energy rate 共PER兲, and aerodynamic force
mechanism are assumed inertially fixed. For a given set of param- 共Pout兲兲 all significantly contribute to the required input power
eters, the optimal values of spring stiffness k, free length l0, and 共Pin兲. The results themselves, however, do not provide deeper in-
location of attachment a were obtained. We model this mechanism sight into the role of a spring and various components of energy
in ADAMS, which is dynamic mechanical system simulation soft-
ware. The aerodynamic load Fo共t兲 is simplified using a point force
acting perpendicular to the wing at its center. The model for this
study is shown in Fig. 1. Table 1 Links’ dimensions and operating frequency
The input displacement ␾in共t兲 at the operating frequency f can l1 = 6 mm l2 = 20 mm l3 = 330 mm f = 4 Hz
be written as x = 15 mm y = 21 mm a = 50 mm b = 165 mm
␾in共t兲 = 2␲ ft 共1兲
The simplified aerodynamic force, parametrized by a damping Table 2 System parameters and optimal values of spring
constant co, constant force component f o, and normal velocity at properties used in the analysis of the simplified four-bar flap-
the center of the wing vo can be written as: ping mechanism

Fo共t兲 = − covo共t兲 + f o 共2兲 mAB = 0.25 g co = 0.207 N s / m k = 38 N / m

mBC = 0.50 g f o = 0.27 N l0 = 50 mm
The values of the links’ dimensions and operating frequency are mCD = 2.75 g
shown in Table 1.

Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1065

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
Fig. 2 Variation of inertial torque over a cycle of flapping Fig. 5 Various power components of the mechanism with a
motion spring, including kinetic energy rate „KER…, potential energy
rate „PER…, input power „Pin…, and output power „Pout…

exchanged within the system components. We, therefore, investi-

gate the effect of a spring when the system is subject to three 共Fig. 7兲, the constant output force, as being upward, produces
different cases: positive work into the mechanism during the upstroke. To main-
共I兲 with constant output force only tain the steady-state motion, the motor supplies negative work into
共II兲 with inertial force only the mechanism. In other words, it has to absorb energy flowing
共III兲 with damping output force only. from the output. During the downstroke, the output force is still
upward but the motion of the wing is downward. The output force
In each case, the values of spring stiffness and free length are is producing negative work into the system, or extracting the en-
optimized for minimum peak input power. The resulting power ergy from the system. When a spring is added into the system
variations in the mechanism without a spring and with a spring are 共Fig. 8兲, the output force still produces positive work into the
then compared. Based on these results, we attempt to visualize the mechanism during the upstroke. However, instead of having the
path of the energy flow within the system. motor absorb this work from the output, the spring stores this
Case I. The mechanism subject to constant output force only The
values of parameters are shown in Table 3. In this case, we set the
mass of all links and the damping component of output force to be Table 3 Parameters for Case I „constant output force only…
zero to investigate the interaction of the spring and the constant mAB = 0 g co = 0 N s / m k = 0 or 35 N / m
component of output force. The results are shown in Figs. 6–8. mBC = 0 g f o = 0.27 N l0 = 41 mm
From Fig. 6, adding a spring reduces the peak input power by mCD = 0 g
84% from 0.45 N m / s to 0.07 N m / s. To understand this power
reduction, we trace power components of the system over a flap-
ping cycle for each case. For the mechanism without a spring

Fig. 6 For Case I „constant output force only…, adding a spring

reduces peak input power by 84%

Fig. 3 Variation of lift over a cycle of flapping motion

Fig. 7 For Case I „constant output force only… without a

spring, the motor has to supply and absorb the full amount of
Fig. 4 Input powers of the mechanism with and without a energy to balance the energy extracted and injected by the out-
spring. Adding a spring reduces peak input power by 42%. put force

1066 / Vol. 129, OCTOBER 2007 Transactions of the ASME

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
Fig. 8 For Case I „constant output force only… with a spring, Fig. 10 For Case II „inertial force only… without a spring, the
the presence of a spring helps the motor supply and absorb motor has to supply and absorb the full amount of energy
energy extracted and injected by the output force, thus reduc- stored and released by links’ masses
ing the effort needed by the motor to balance the system

From Fig. 12, adding a spring into the system increases the
positive work as elastic energy. The motor, then, needs less effort required peak input power by 78% from 0.55 N m / s to
in balancing the steady-state motion. During the downstroke, the 0.98 N m / s. For the mechanism without a spring 共Fig. 13兲, the
output force extracts energy out of the system. During this period, damping output force extracts energy from the system both during
the spring releases the stored elastic energy to the output. In this upstroke and downstroke. The motor always has to supply this
situation, adding a spring to the system reduces the peak input amount of energy to maintain the steady-state motion. The peak
power requirement. input power is, therefore, equal to the peak output power. When a
Case II. The mechanism subject to inertial force only spring is added into the system 共Fig. 14兲, the motor has to supply
The values of parameters are shown in Table 4. In this case, the energy both to the output and to the spring during the upstroke
links’ masses are nonzero. The output force is set to be zero for and, therefore, requires more power than it does without a spring.
both damping and constant components. The results are shown in During the downstroke, the spring releases energy to the output,
Figs. 9–11.
From Fig. 9, adding a spring into the system reduces the re-
quired peak input power by 83% from 0.18 N m / s to 0.03 N m / s.
For the mechanism without a spring, the motor has to supply and
absorb the energy required by and released from the system’s
inertia. The required input power is, therefore, the same as that
stored or released by the system’s inertia. Adding a spring into this
type of system will help reduce the input power demand. The
spring will absorb the energy when the links decelerate and re-
lease kinetic energy, thus reducing the effort needed by the motor
to absorb the energy. On the other hand, when the links need to
accelerate, the spring releases the stored elastic energy, reducing
the power demand from the motor.
Case III. The mechanism subject to damping output force only
The values of parameters are shown in Table 5. In this case, all
links’ masses are zero. The output force only has a damping com- Fig. 11 For Case II „inertial force only…, the spring added to the
ponent. The results of power analysis are shown in Figs. 12–14. system helps absorb and supply energy as the links decelerate
and accelerate, thus reducing the power requirement from the

Table 4 Parameters for Case II „inertial force only… Table 5 Parameters for Case III „damping output force only…
mAB = 0.25 g co = 0 N s / m k = 0 or 226 N / m mAB = 0 g co = 0.207 N s / m k = 0 or 38 N / m
mBC = 0.50 g fo = 0 N l0 = 117 mm mBC = 0 g fo = 0 N l0 ⫽ 50 mm
mCF = 2.75 g mCD = 0 g

Fig. 9 For Case II „inertial force only…, adding a spring reduces Fig. 12 For Case III „damping output force only…, adding a
peak input power by 83% spring increases peak input power by 78%

Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1067

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
Fig. 13 For Case III „damping output force only… without a Fig. 16 For Case III „damping output force only…, adding a
spring, the motor has to supply energy equal to the amount spring of any free length, in this case with a stiffness of
extracted by the output force 38 N / m, results in undesired increase in peak input power. At
best it keeps the peak input power as low as that resulting from
not adding a spring into the system.
reducing input power requirement from the motor. The entire flap-
ping cycle, therefore, requires more input power if a spring is
added to this type of system. In fact, this should always be true for contrary, the existence of a damping force reduces the effective-
the following situation. If the only source that provides positive ness of elasticity to reduce the peak input power. From Table 6,
work into a massless system is the input, then the energy required the optimal values of spring’s properties for the case of a com-
to deform the spring will have to be from this source and, there- bined constant-inertial-damping force 共stiffness of 38 N / m and
fore, the source needs to provide this extra power to the system in free length of 50 mm兲 are between those of a pure constant output
addition to the power required from the dissipative output. Figures force 共Case I兲 and a pure inertial force 共Case II兲. It can be further
15 and 16 show that for the case of damping output force only, the noticed that these optimal values are closer to those for the case of
peak input power is minimal when there is no spring added to the a pure constant output force. This corresponds to the fact that the
system. The results of all case studies are summarized in Table 6. contribution of the input power due to a pure constant output force
If we consider each component of the load acting on the sys- 共0.45 N m / s兲 is more significant than that due to a pure inertial
tem, this study reveals that the existence of either a constant out- force 共0.18 N m / s兲. The presence of a damping component in the
put force 共Case I兲 or an inertial property of the system 共Case II兲 case of combined force brings the percentage of peak input power
allows the use of elasticity to reduce the peak input power. These reduction down to 42%. This study provides a visualization of
two types of load provide instantaneous positive work, which can how each load component affects the optimal values of the
be stored during one stage of a cycle and reused in another. On the spring’s properties.

Two Approaches to Exploit Elastic Energy

The study from the previous section illustrates the function of
an elastic component in a dynamic system as an energy storage
component. In Fig. 17共a兲, a given mechanical system with its
inertia is driven by an input actuator and performs a task at the
output port. The energy flow between these components affects
the required input power. Adding an elastic component to this
system, as shown in Fig. 17共b兲, will split the path of energy flow
in the original system. If the elastic component is added properly,
it will draw some energy from the output or inertia to itself when,
otherwise, the input has to absorb this energy. Similarly, this elas-
tic component will release some energy to the output or inertia
when, otherwise, the input has to supply this energy. If the elastic
Fig. 14 For Case III „damping output force only… with a spring,
component is used to split energy flow via path B, its role is
besides supplying energy to the output force, the motor also referred to as generative-load exploitation. This approach of ex-
has to supply energy to be stored in the spring ploiting elastic energy is useful to describe the function of the
elastic component when the output load provides positive work
into the system during a certain portion of the motion cycle. The
aerodynamic load used in this study is an example of generative
load, which can be beneficial for peak input power reduction. If
the elastic component is used to split the energy flow via path C,
its role is referred to as reactance cancellation. Even though KER
and PER are not exactly cancelled, we use this term in referring to
this approach because it involves power reduction between the
elastic component and inertia, which are reactive components
found in a general spring-mass-damper system.
The two approaches of generative-load exploitation and reac-
tance cancellation may be implemented on the design of elastic
components at the same time. In fact, they should be implemented
simultaneously when both generative load and system’s inertia are
Fig. 15 For Case III „damping output force only…, adding a significant. A design approach based on resonance frequency,
spring of any stiffness, in this case with a free length of 50 mm, which corresponds to the reactance cancellation approach as re-
results in undesired increase in peak input power ferred in this paper, can result in a suboptimal design when a

1068 / Vol. 129, OCTOBER 2007 Transactions of the ASME

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
Table 6 Summary of peak input power reduction using optimal springs in four cases. A spring
is useful for peak input power reduction when either constant output force or inertial property
exists, but not when only damping output force exists.

Optimal spring

k l0 without spring Pin,peak reduction
Load case 共N/m兲 共mm兲 共N m/s兲 共%兲

Constant output force only 共Case I兲 35 41 0.45 84

Combined constant-inertial-damping force 38 50 1.00 42
Inertial force only 共Case II兲 226 117 0.18 83
Damping output force only 共Case III兲 0 N/A 0.55 0

generative load is present. In this study, energy flow across the u2共t兲 = U2 sin共␻t兲 + Ū2 共4兲
system’s inertia is small compared to energy flow across the out-
put load 共Fig. 5兲. The optimal design of a spring leans toward the
design obtained by the generative-load exploitation. However, for
the operation at higher frequency, approaching toward resonance, f 1共t兲 = F1 sin共␻t兲 + F̄1 共5兲
energy flow across the system’s inertia may become larger than
energy flow across the output load. In this case, the optimal design
of a spring will lean toward the design obtained by the reactance f 2共t兲 = F2 sin共␻t兲 + F̄2 共6兲
Even though the distinction in the two approaches described in The equations of motion for the two masses can be written in a
this paper may seem unnecessary if a general optimization method matrix form. The static and harmonic components can be written
is applied, this distinction is helpful for energy flow visualization. separately, yielding the following systems of equations:
It will also help set up a design framework and may be useful for

冋册冋 册冋 册
guiding the design optimization process, leading to a more effi-
cient and robust design tool. This is necessary especially for the F̄1 k1 + k2 − k2 Ū1
design of compliant mechanisms, where nonlinear responses are = 共7兲
F̄2 − k2 k2 + k3 Ū2
computationally expensive.

冋册冉 冋 册冋 册冊冋 册
Power Analysis and Design of a 2DOF System
F1 m1 0 k1 + k2 − k2 U1
Two different approaches to describe the role of elasticity, = − ␻2 + 共8兲
namely, generative-load exploitation and reactance cancellation, F2 0 m2 − k2 k2 + k3 U2
have been defined based on the investigation of power flow in the
four-bar flapping mechanism with a spring. The concept of these Assuming that the system is subject to an input displacement 共U1,
two approaches cannot be immediately generalized to compliant Ū1兲 and the harmonic component of an output force 共F2兲 is zero,
mechanisms because compliant mechanisms are elastic systems
consisting of an infinite number of degrees of freedom 共DOF兲. we can solve for Ū2, U2, F̄1, and F1. The results are:
Unlike rigid-body mechanisms, the kinematics of compliant
mechanisms depends on system’s properties and operating condi-
tions, such as mass, stiffness, and operating frequency. We, there- F̄2 + k2Ū1
Ū2 = 共9兲
fore, use a 2DOF system, which represents the simplest form of a k2 + k3
compliant mechanism, to demonstrate the validity of the two ap-
proaches in exploiting elasticity. The system consists of two
masses: m1 and m2 and three linear springs: k1, k2, and k3, as k2
shown in Fig. 18. Forces f 1共t兲 and f 2共t兲 are applied to masses m1 U2 = U1 共10兲
− ␻ m2 + k2 + k3
and m2, resulting in displacements u1共t兲 and u2共t兲, respectively.
To solve for steady-state responses, we assume that the solu-
tions are:
共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2
F̄1 = 共11兲
u1共t兲 = U1 sin共␻t兲 + Ū1 共3兲 k2 + k3

Fig. 17 Concept of using elasticity to reduce peak input power requirement in dynamic

Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1069

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
␻4m1m2 − ␻2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3
F1 = U1 共12兲
− ␻ 2m 2 + k 2 + k 3
The requirement of this system is that it has to generate the output amplitude of U2,req. From Eqs. 共10兲–共12兲, to satisfy this requirement,
the amplitude of input displacement, average input force, and amplitude of input force, respectively, must be:
− ␻ 2m 2 + k 2 + k 3
U1 = U2,req 共13兲

共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2

F̄1 = 共14兲
k2 + k3

␻4m1m2 − ␻2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3

F1 = U2,req 共15兲
The instantaneous input power can be calculated from:
p1共t兲 = f 1共t兲u̇1共t兲 共16兲
Substituting Eqs. 共3兲 and 共5兲 into Eq. 共16兲, we have:
p1共t兲 = 关F1 sin共␻t兲 + F̄1兴关␻U1 cos共␻t兲兴 = ␻F1U1 sin共␻t兲cos共␻t兲 + ␻F̄1U1 cos共␻t兲 = ␻F1U1 sin共2␻t兲 + ␻F̄1U1 cos共␻t兲 共17兲
Equation 共17兲, along with Eqs. 共13兲–共15兲, shows that the instantaneous input power is a function of frequency, stiffness, mass, and
average input displacement. It is no longer a function of the input displacement amplitude due to the requirement on the output
displacement amplitude.
Since we are interested in designing a system for given operating frequency, output force, and required output displacement, the
design variables are stiffness, mass, and average input displacement. Equation 共17兲 can be written in the form:

p1共t兲 = G1共k1,k2,k3,m1,m2兲sin共2␻t兲 + G2共k1,k2,k3,Ū1兲cos共␻t兲 共18兲

1 关␻4m1m2 − ␻2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3兴关− ␻2m2 + k2 + k3兴 2
G1 = ␻ U2,req 共19兲
2 k22

关共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2兴关− ␻2m2 + k2 + k3兴 1 共k1k2 + k1k3 + k2k3兲共k2 + k3兲 2
G2 = ␻ U2,req G1 = ␻ U2,req 共23兲
k2共k2 + k3兲 2 k22
共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2
and G2 = ␻ U2,req 共24兲
k1 ⱖ 0, k2 ⬎ 0, k3 ⱖ 0, m1 ⱖ 0, m2 ⱖ 0, − ⬁ ⬍ Ū1 ⬍ ⬁ Since G2 has an extra variable 共Ū1兲, which can be any real
共21兲 number, G2 may be considered independent of G1. The peak of
We will determine optimal designs for two cases: one is when p1共t兲, or P1,peak, can be minimized when 兩G1兩 and 兩G2兩 are mini-
the system has no mass and the other is when the system has no mized. Since the minimum value that 兩G1兩 and 兩G2兩 may reach is
output force. The first case represents an attempt to use elasticity zero, we first attempt to find conditions that make G1 and G2 be
based on generative-load exploitation only, while the second case zero. From Eq. 共23兲, G1 is set to zero. We have:
represents an attempt to use elasticity based on reactance cancel-
lation. 共k1k2 + k1k3 + k2k3兲共k2 + k3兲
=0 共25兲
Case I. m1 = m2 = 0 and F̄2 ⫽ 0 共generative-load exploitation兲 k22
Equations 共18兲–共20兲, respectively, reduce to:
k 1k 2 + k 1k 3 + k 2k 3 = 0 共26兲
p1共t兲 = G1共k1,k2,k3兲sin共2␻t兲 + G2共k1,k2,k3,Ū1兲cos共␻t兲 共22兲
Substitute Eq. 共26兲 into Eq. 共24兲 to find conditions that make G2
be zero:

0 − k2F̄2
=0 共27兲

F̄2 = 0 共28兲

This is impossible since we have F̄2 ⫽ 0. Therefore, P1,peak cannot

be made zero.
Fig. 18 A 2DOF system used to validate the concept of We then attempt to find conditions that make G1 and G2 ap-
generative-load exploitation and reactance cancelation in com- proach zero. From Eq. 共23兲, when G1 approaches zero, we have:
pliant mechanisms
1070 / Vol. 129, OCTOBER 2007 Transactions of the ASME

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
共k1k2 + k1k3 + k2k3兲共k2 + k3兲
→0 共29兲

k 1k 3 k1k23 k23
k1 + 2 + k3 + 2 + → 0 共30兲
k2 k2 k2
Condition 共30兲 is true only when:
k1 → 0

k3 → 0
Fig. 19 A general plot of 円G1円 and 円G2円 as functions of k3 when
k1 is zero
k 1k 3
→0 共31兲
Substitute 共31兲 into Eq. 共24兲 and find conditions that make G2 From Eqs. 共36兲 and 共37兲, however, k2 appears only in G1 be-
approach zero: cause k1 has been assumed zero. The value of 兩G1兩 decreases when
k2 increases. The optimal k2 that minimizes the P1,peak, therefore,
共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2 approaches ⬁. This result implies, for the case when there is no
→0 共32兲 input stiffness, that the compliance between the input and the
output should be minimized and a spring should be attached to the
共k1k2 + k2k3兲Ū1 − k2F̄2 In reality, there is usually stiffness associated with the input,
→0 共33兲
k2 such as the stiffness of an actuator. In compliant mechanisms,
input stiffness exists due to the members connecting the input to
F̄2 the ground. In this case, k1 is not zero. G1 and G2 from Eqs. 共36兲
Ū1 → 共34兲 and 共37兲, respectively, are not valid. We have to use Eqs. 共23兲 and
k1 + k3 共24兲. From Eq. 共23兲, 兩G1兩 decreases when k2 increases. However,
this is not always true for 兩G2兩 in Eq. 共24兲. We rewrite Eq. 共24兲 as
Ū1 → ± ⬁ 共35兲 the following:
This analysis shows that P1,peak → 0 when k1 → 0, k2 ⬎ 0, k3 → 0,
and Ū1 → ± ⬁. By similar analysis, it can also be shown that k1 or
k3, but not both, can be zero to make P1,peak → 0.
G2 = ␻U2,req 再冉 k1 +
k 1k 3

+ k3 Ū1 − F̄2 冎 共38兲

However, Ū1 cannot physically be ±⬁. For simplicity, to inves- When k2 approaches ⬁, G2 will approach ␻U2,req兵共k1 + k3兲Ū1
tigate the case when there is a bound on Ū1, we assume that F̄2 − F̄2其, which will contribute to P1,peak. 兩G2兩 can be minimized
⬎ 0 and the upper bound Ū1,bound ⬎ 0. For the case when k1 = 0,
Eqs. 共23兲 and 共24兲 become:
1 k3共k2 + k3兲
冉 k1 +
k 1k 3

+ k3 Ū1 − F̄2 = 0 共39兲
G1 = ␻U2,req
2 k2 k 1k 3
k2 = 共40兲
G2 = ␻U2,req共k3Ū1 − F̄2兲 共37兲 F̄2/Ū1 − 共k1 + k3兲
From Eq. 共36兲, 兩G1兩 can be reduced when k3 is reduced. From Even though P1,peak depends on both G1 and G2 in this case,
Eq. 共40兲 shows that the optimal k2 does not necessarily approach
Eq. 共37兲, 兩G2兩 can be made zero when k3 ⱖ F̄2 / Ū1,bound. When k3 is
⬁ when there is input stiffness and a bound on average input
reduced further such that k3 ⬍ F̄2 / Ū1,bound, Ū1 that minimizes 兩G2兩 displacement. For this situation, compliance between the input
is Ū1,bound and 兩G2兩 increases to ␻U2,reqF̄2 as k3 approaches zero. and output will help reduce the peak input power.
A general plot of function 兩G1兩 and 兩G2兩 is shown in Fig. 19. Case II. m1 ⬎ 0, m2 ⬎ 0, and F̄2 = 0 共reactance cancellation兲
Figure 19 implies that an optimal k3 that minimizes P1,peak does Equations 共18兲–共20兲 become:
not necessarily approach zero when there is a bound on Ū1. The
analysis for the case when k3 = 0 yields a similar result, in which p1共t兲 = G1共k1,k2,k3,m1,m2兲sin共2␻t兲 + G2共k1,k2,k3,Ū1兲cos共␻t兲
the optimal k1 does not necessarily approach zero. 共41兲

1 关␻4m1m2 − ␻2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3兴关− ␻2m2 + k2 + k3兴 2
G1 = ␻ U2,req 共42兲
2 k22

共k1k2 + k1k3 + k2k3兲Ū1

G2 = ␻ U2,req 共43兲

To minimize P1,peak, Eq. 共43兲 shows that one of the solutions is when Ū1 = 0 and the other one is when k1 = k3 = 0, which will make G2
be zero.
For the case when Ū1 = 0, Eqs. 共41兲 and 共42兲 become:

p1共t兲 = G1共k1,k2,k3,m1,m2兲sin共2␻t兲 共44兲

Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1071

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
1 关␻4m1m2 − ␻2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3兴关− ␻2m2 + k2 + k3兴 2
G1 = ␻ U2,req 共45兲
2 k22
We determine conditions that make G1 be zero. By setting Eq. 共45兲 to be zero, we have:
关␻4m1m2 − ␻2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3兴关− ␻2m2 + k2 + k3兴
=0 共46兲

The solutions to Eq. 共46兲 are: show that the generative-load exploitation is applicable to a com-
pliant mechanism when an input stiffness and a bound of average
k2 = ␻ m2 − k3
input displacement exist. Since k2 in Case II 共G兲 may approach
or infinity, this confirms the concept of reactance cancellation in a
rigid-body mechanism. Both Cases II 共G兲 and II 共H兲 show that the
␻4m1m2 − ␻2共m1k3 + m2k1兲 + k1k3
k2 = 共48兲 concept of reactance cancellation is applicable to a compliant
␻2共m1 + m2兲 − k1 − k3 mechanism, regardless of the existence of an input stiffness or a
Depending on the values of m1, m2, k1, and k3, it is possible that bound of average input displacement. In addition, Eqs. 共47兲 and
there is k2 that reduces P1,peak to zero. In addition, it can be proved 共48兲 indicate that the reactance cancellation corresponds to two
that Eqs. 共47兲 and 共48兲 correspond to the natural modes of the types of resonance, one is associated with a free-input response
system with m1 fixed and m1 free, respectively. and the other is associated with a fixed-input response.
For the case when k1 = k3 = 0, Eqs. 共41兲 and 共42兲 reduce to: Even though the analysis of a 2DOF system validates the con-
cept of generative-load exploitation and reactance cancellation in
p1共t兲 = G1共k2,m1,m2兲sin共2␻t兲 共49兲 compliant mechanisms, the analysis procedures are not applicable
to analyze and design compliant mechanisms. Most compliant
1 关␻4m1m2 − ␻2共m1k2 + m2k2兲兴关− ␻2m2 + k2兴 2 mechanisms undergo large displacements and require nonlinear
G1 = ␻ U2,req
2 k22 analysis to evaluate their performance. In addition, stiffness and
mass in a compliant mechanism are coupled through geometry
共50兲 and cannot be designed independently. The analysis of a 2DOF
We determine conditions that make G1 be zero. By setting Eq. system in this study, therefore, only provides mathematical under-
共50兲 to be zero, we have: standing in the use of elasticity. The analysis and design of com-
pliant mechanisms with large displacements require more sophis-
关␻4m1m2 − ␻2共m1k2 + m2k2兲兴关− ␻2m2 + k2兴 ticated methods.
=0 共51兲
The solutions to Eq. 共51兲 are: Optimization Strategies
k 2 = ␻ 2m 2 共52兲 Unlike a linear system, most compliant mechanisms undergo
or large displacement, and geometric nonlinearity must be taken into
account in the analysis. The amplitude of input displacement re-
␻ 2m 1m 2 quired to provide the desired amplitude of output displacement
k2 = 共53兲
m1 + m2 cannot be determined directly. In addition, a harmonic analysis
and the principle of superposition are no longer applicable. In this
Equations 共52兲 and 共53兲 show that there is always k2 that reduces
study, we use nonlinear transient analysis based on an implicit
P1,peak to zero. The summary of the conditions that minimize the dynamic method to analyze and design a compliant mechanism.
peak input power is shown in Table 7. Since the amplitude and mean values of the input displacement
From Table 7, k2 approaching infinity in Case I 共D兲 represents a can be controlled during the operation, we do not consider them as
rigid-body connection between the input and output. This case true design variables in the context of a “mechanism design.”
confirms the concept of generative-load exploitation in a rigid- Instead, we use the term “performance parameters” to refer to the
body mechanism. Cases I 共E兲 and I 共F兲, with stiffnesses not ap- amplitude and mean values of the input displacement. Beam
proaching infinity, represent a compliant system. These cases heights and keypoint locations, on the other hand, cannot be con-
trolled once the system is built and are still considered “design
Table 7 Summary of the conditions that minimize peak input variables.” Based on the distinction between performance param-
power eters and design variables, there are at least three strategies for
optimizing compliant mechanisms.
F̄2 m1 m2 k1 k2 k3 Ū1 P1,peak 共1兲 Performance parameters and design variables are optimized
together in one optimization loop.
Case I ⫽0 =0 =0 →0 any k2⬎0 →0
→±⬁ +
→0 +
共B兲 共2兲 Performance parameters are optimized as an inner loop for
⫽0 =0 =0 =0 any k2 ⬎ 0 →0+ → ± ⬁ →0+
共C兲 each evaluation of a set of design variables being optimized
⫽0 =0 =0 →0+ any k2 ⬎ 0 =0 →±⬁ →0+
共D兲 ⫽0 =0 =0 =0 →⬁ →
” 0 Bounded ⬎0
in an outer loop.

共3兲 Performance parameters are fixed and only design variables
共E兲 ⫽0 =0 =0 →” 0 any k2 ⬎ 0 =0 Bounded ⬎0 are optimized.

共F兲 ⫽0 =0 =0 → ⬎0 ⬎0
Strategy 共1兲 is based on a modern design automation approach,
”0 →
” 0 Bounded
which does not have to distinguish between performance param-

”⬁ →
”⬁ →
eters and design variables. Strategy 共2兲 is based on a traditional
Case II 共G兲 =0 ⬎0 ⬎0 ⬎0 共47兲, 共48兲 ⬎0 =0 ⱖ0 design approach, in which a designer evaluates each design prop-
共H兲 =0 ⬎0 ⬎0 =0 共47兲, 共53兲 =0 ⫽0 =0 erly before moving and comparing to the next design. Strategy 共3兲
is based on simplicity, which ignores the effects of input on the

1072 / Vol. 129, OCTOBER 2007 Transactions of the ASME

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
Fig. 22 An example of input displacement with gradually in-
creasing amplitude to facilitate the convergence of nonlinear
transient analysis
Fig. 20 Front view of the compliant flapping mechanism
showing the initial design before optimization

applied at the center of the wing, is a simplified aerodynamic

design. The primary motivation in distinguishing these strategies force previously used in the analysis of the four-bar flapping
is to find the most efficient optimization scheme to search design mechanism, as described by Eq. 共2兲. The values of co and f o are
space. For example, it is possible that a continuously changing 0.207 N s / m and 0.27 N, respectively. The required flapping
amplitude or mean value without restarting the analysis from zero angle ␪flap, measured at the location where the output force is
input may reduce computational time by shortening the time dur- applied, is 44.8 deg, the same as in the four-bar flapping mecha-
ing transient state. In this case, Strategy 共2兲 may be preferred. nism. This amount of flapping angle is assumed to provide suffi-
However, there is no straightforward manner to predict the most cient lift at the specified frequency to sustain the vehicle in the air
efficient strategy to use. and is the primary functional requirement of the mechanism. The
gravity is ignored in the analysis.
Design of a Compliant Flapping Mechanism The mechanism is subject to a sinusoidal input displacement
In this study, we use Strategy 共1兲, optimizing performance pa- uin共t兲, with amplitude A and average C. However, to facilitate the
rameters and deign variables together in one loop, to design a convergence in finding solutions of the analysis, we impose the
compliant flapping mechanism. The initial design, shown in Fig. input displacement with gradually increasing amplitude, ex-
20, is an application of a patented compliant stroke amplifier pressed in the form:
关27–29兴, in which the wing is directly connected to the output uin共t兲 = 共1 − e−t/␶H兲A关cos共2␲ ft兲 − 1兴 + 共1 − e−t/␶S兲共C + A兲
link. It is modeled and analyzed in ABAQUS, which is commercial
finite element analysis software. Instead of having a uniform beam 共54兲
width, the wing section of the mechanism has been modified to For this study, ␶H = 1 / 4, ␶S = 1 / 4, and f = 4 Hz. An example of
redistribute the mass and stiffness, as shown in Fig. 21. The total input displacement is shown in Fig. 22.
mass of the wing section is 2.50 g, the same as that in the four-bar There are a total of 37 design variables for this optimization
flapping mechanism, but a significant amount of mass is distrib- problem: two are performance parameters including A and C from
uted to the end of the wing section. This will promote the use of Eq. 共54兲, ten are keypoint locations, and 25 are beam heights, as
reactance cancellation, in addition to the use of generative-load shown in Fig. 23. The objective of the optimization problem is to
exploitation. The wing’s dimensions are shown in Table 8. minimize the peak input power requirement, subject to the flap-
The mechanism is modeled using structural beam elements. Its ping angle of 44.8 deg.
overall dimensions are approximately the same as those of the Each analysis is run until the mechanism reaches a steady state,
four-bar flapping mechanism studied in the previous section. We which is predicted by comparing responses in one cycle to others.
assume that the entire mechanism is made of carbon fiber com- In doing this, criteria for analysis termination have to be specified.
posites whose Young’s modulus is 520 GPa, Poisson’s ratio is When the differences in responses 共i.e., flapping angle, input
0.28, and density is 1570 kg/ m3. The output force Fo共t兲, which is power, and output power兲 between cycles are within the specified
values, the analysis is terminated and the mechanism is considered

Fig. 21 Top view of the compliant flapping mechanism show-

ing a wing section with nonuniform mass and stiffness

Table 8 Dimensions of the wing

Wing segment Height, h 共mm兲 Width, w 共mm兲 Length, l 共mm兲

A 1.00 0.50 82.50

B 1.00 1.50 82.50
C 1.00 3.50 82.50 Fig. 23 Design variables include „a… ten variables of x and y
D 1.00 10.44 45.83 locations for five keypoints, „b… 25 variables of beam heights
E 1.00 18.00 36.67 for nine beam segments, and two variables for amplitude and
average values of input displacement „see Fig. 20…

Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1073

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
Fig. 24 Optimal design of a compliant flapping mechanism obtained from
the use of NOMADm and SQP algorithms

to reach a steady state. The flapping angle and peak input power of a Mesh Adaptive Direct Search 共MADS兲 algorithm, was first
can be determined after the steady-state criteria have been met. used to search the design space. One of the features of this algo-
The optimization process is accomplished through the use of rithm that is suitable for this problem is that the algorithm is
Matlab®. NOMADm 关30兴 algorithm, which is the implementation intended for a problem whose objective and constraint functions
are computationally expensive to evaluate. The solution was ob-
tained after 3814 analyses, which took 71 h on a computer with
Table 9 Keypoints of the optimal mechanism 2.8 GHz Pentium® IV processor and 512 MB of RAM. This so-
lution was then used as an initial design for local search using
Keypoint x-coordinate 共mm兲 y-coordinate 共mm兲 Sequential Quadratic Programming 共SQP兲. The solution was ob-
tained after 4958 analyses, which took 96 h on the same com-
1 0.00 0.00 puter. The optimal design is shown in Fig. 24, along with num-
2 8.69 12.00
3 21.52 6.01
4 29.98 26.62
5 5.20 23.65
6 31.00 39.97
7 3.00 50.00
8 −2.00 50.00
9 3.00 14.00

Table 10 Segments of the optimal mechanism „width

= 2.00 mm…

Segment Height 共mm兲

1 0.23
2 0.24
3 0.27
4 0.24
5 0.25 Fig. 25 The mechanism produces a flapping angle of 44.8 deg,
6 0.19
7 0.26 generating a sufficient lift for the vehicle at 4 Hz
8 0.27
9 0.25
10 0.24
11 0.24
12 0.32
13 0.25
14 2.55
15 0.31
16 0.24
17 0.30
18 0.25
19 0.24
20 0.25
21 0.24
22 0.25
23 0.27
24 0.26
25 4.90 Fig. 26 Wing rotation over a flapping cycle at steady state „af-
ter seven cycles…

1074 / Vol. 129, OCTOBER 2007 Transactions of the ASME

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm
ping Mechanism: A Minimum Torque Approach,” Mech. Mach. Theory,
41共1兲, pp. 3–16.
关2兴 Madangopal, R., Khan, Z. A., and Agrawal, S. K., 2005, “Biologically In-
spired Design of Small Flapping Wing Air Vehicles Using Four-Bar Mecha-
nisms and Quasi-Steady Aerodynamics,” ASME J. Mech. Des., 127共4兲, pp.
关3兴 Krouse, J. K., 1980, “Compliant Mechanisms—A New Class of Mechanical
Devices,” Mach. Des., 52共2兲, pp. 86–90.
关4兴 Thompson, L. H., 1983, “Compliant Positioning Mechanism,” IBM Tech.
Discl. Bull., 25共11A兲, pp. 5444–5445.
关5兴 Parkin, R. E., and Hutchinson, W. K., 1985, “Compliant Mechanical Gripper,”
Robotics Age, 7共5兲, pp. 11–13.
关6兴 Andeen, G. B., and Kombluh, R., 1988, “Design of Compliance in Robotics,”
Proceedings of the 1988 IEEE International Conference on Robotics and Au-
tomation, IEEE, New York, pp. 276–281.
关7兴 Holl, J., Koffler, W., and Midha, A., 1983, “Compliant Mechanism Synthesis
Fig. 27 Variation of different power components over a flap- in Prosthetic Design,” AMD Symposia Series, ASME, New York, Vol. 56, pp.
ping cycle at steady state 63–66.
关8兴 Trease, B. P., Moon, Y., and Kota, S., 2005, “Design of Large-Displacement
Compliant Joints,” ASME J. Mech. Des., 127共4兲, pp. 788–798.
bering for keypoints and segments, the values of which can be 关9兴 Howell, L. L., Rao, S. S., and Midha, A., 1994, “Reliability-Based Optimal
Design of a Bistable Compliant Mechanism,” ASME J. Mech. Des., 116共4兲,
found in Tables 9 and 10. The optimal amplitude A is 0.612 mm
pp. 1115–1121.
and average C is 0.770 mm. The extreme positions of the wing are 关10兴 Larsen, U. D., Sigmund, O., and Bouwstra, S., 1997, “Design and Fabrication
shown in Fig. 25, confirming that the wing is deformed and ro- of Compliant Micromechanisms and Structures With Negative Poisson’s Ra-
tated as expected. The plot of wing rotation over time is shown in tio,” J. Microelectromech. Syst., 6共2兲, pp. 99–106.
Fig. 26. The plot of various power components is shown in 关11兴 Kikuchi, N., Nishiwaki, S., Ono Fonseca, J. S., and Nelli Silva, E. C., 1998,
“Design Optimization Method for Compliant Mechanisms and Material Mi-
Fig. 27. crostructure,” Comput. Methods Appl. Mech. Eng., 151共3-4兲, pp. 401–417.
This design example of a compliant flapping mechanism re- 关12兴 Calin, M., Chaillet, N., Bourjault, A., Bertsch, A., and Zissy, S., 1996, “Design
quires only 0.49 N m / s of input power, compared to 1.0 N m / s and Control of Compliant Microrobots,” IEEE Symposium on Emerging Tech-
for the four-bar flapping mechanism without a spring, and nologies and Factory Automation, ETFA, IEEE, Piscataway, NJ, Vol. 2, pp.
0.58 N m / s for the one with a spring 共50% and 15% peak input 489–492.
关13兴 Kota, S., Ananthasuresh, G. K., Crary, S. B., and Wise, K. D., 1994, “Design
power reduction, respectively兲, in performing the same task. Note and Fabrication of Microelectromechanical Systems,” ASME J. Mech. Des.,
that the two peaks of input power are evened out. This is the same 116共4兲, pp. 1081–1088.
behavior found in the four-bar flapping mechanism when an ap- 关14兴 Su, X. S., and Yang, H. S., 2001, “Design of Compliant Microleverage Mecha-
propriate spring is added. The peak output power is only nisms,” Sens. Actuators, A, 87共3兲, pp. 146–156.
关15兴 Lu, K., and Kota, S., 2003, “Design of Compliant Mechanisms for Morphing
0.75 N m / s, compared to 1 N m / s in the four-bar flapping mecha- Structural Shapes,” J. Intell. Mater. Syst. Struct., 14共6兲, pp. 379–391.
nism. The reduction of peak output power can be explained by the 关16兴 Kota, S., Lu, K., Kreiner, Z., Trease, B., Areanas, J., and Geiger, J., 2005,
use of sinusoidal input displacement, which produces more effi- “Design and Application of Compliant Mechanisms for Surgical Tools,” J.
cient flapping velocity profile, generating the same amount of lift Biomech. Eng., 127共6兲, pp. 981–989.
with lower output power. The 15% reduction of peak input power 关17兴 Osborn, R. F., Kota, S., Hetrick, J. A., Geister, D. E., Tilmann, C. P., and Joo,
J., 2004, “Active Flow Control Using High-Frequency Compliant Structures,”
in the compliant flapping mechanism over the four-bar flapping J. Aircr., 41共3兲, pp. 603–609.
mechanism with a spring can, therefore, be attributed to the use of 关18兴 Tantanawat, T., Li, Z., and Kota, S., 2004, “Application of Compliant Mecha-
elasticity and a more efficient input velocity profile. nisms to Active Vibration Isolation Systems,” Proceedings of the ASME De-
sign Engineering Technical Conference, ASME, New York, Vol. 2B, pp. 1165–
Conclusions 1172.
关19兴 Du, H., Lau, G. K., Lim, M. K., and Qui, J., 2000, “Topological Optimization
In this work, we have investigated the concept of peak input of Mechanical Amplifiers for Piezoelectric Actuators Under Dynamic Motion,”
power reduction through the use of elasticity. From the power Smart Mater. Struct., 9共6兲, pp. 788–800.
关20兴 Nishiwaki, S., Saitou, K., Min, S., and Kikuchi, N., 2000, “Topological Design
analysis of a four-bar flapping mechanism, we make a distinction Considering Flexibility Under Periodic Loads,” Struct. Multidiscip. Optim.,
between two different approaches that can be used to describe the 19共1兲, pp. 4–16.
role of an elastic component: 共1兲 generative-load exploitation and 关21兴 Maddisetty, H., and Frecker, M., 2004, “Dynamic Topology Optimization of
共2兲 reactance cancellation. Only one or both of them may be ap- Compliant Mechanisms and Piezoceramic Actuators,” ASME J. Mech. Des.,
plied at the same time, depending on the nature of the load. These 126共6兲, pp. 975–983.
关22兴 Li, Z., and Kota, S., 2002, “Dynamic Analysis of Compliant Mechanisms,”
two approaches have been shown to be valid to describe the role Proceedings of the ASME Design Engineering Technical Conference, ASME,
of elasticity in compliant mechanisms as well. We then propose a New York, Vol. 5A, pp. 43–50.
compliant mechanism as an alternative to a rigid-body mechanism 关23兴 Lai, E., and Anathasuresh, G. K., 2002, “On the Design of Bars and Beams for
with attached springs. Through the use of NOMADm and SQP Desired Mode Shapes,” J. Sound Vib., 254共2兲, pp. 393–406.
关24兴 Boyle, C., Howell, L. L., Magleby, S. P., and Evans, M. S., 2003, “Dynamic
algorithms, we obtain a compliant flapping mechanism that can
Modeling of Compliant Constant-Force Compression Mechanisms,” Mech.
reduce the peak input power requirement by an additional 15% Mach. Theory, 38共12兲, pp. 1469–1487.
over the four-bar flapping mechanism with a spring, while gener- 关25兴 Handley, D. C., Lu, T., Yong, Y. K., and Zhang, W. J., 2004, “A Simple and
ating the same amount of lift. Efficient Dynamic Modeling Method for Compliant Micropositioning Mecha-
The computational requirement seems to be too high. There- nisms Using Flexure Hinges,” Proc. SPIE, Vol. 5276, pp. 67–76.
关26兴 Yu, Y., Howell, L. L., Lusk, C., Yue, Y., and He, M., 2005, “Dynamic Mod-
fore, in an attempt to develop an efficient and robust design tool, eling of Compliant Mechanisms Based on the Pseudo-Rigid-Body Model,”
we are currently investigating various problem formulations and ASME J. Mech. Des., 127共4兲, pp. 760–765.
optimization methods to reduce computational time. Other design 关27兴 Kota, S., Hetrick, J., Li, Z., and Saggere, L., 1999, “Tailoring Unconventional
issues, such as average power consumption, sensitivity to operat- Actuators Using Compliant Transmissions: Design Methods and Applica-
ing condition and dimensional error, the coupling between system tions,” IEEE/ASME Trans. Mechatron., 4共4兲, pp. 396–408.
关28兴 Kota, S., Hetrick, J., Rodgers, S., and Li, Z., 2001, “Compliant Displacement
design and aerodynamic load, and stress constraint will be ad- Amplification Apparatus for Micro Electro Mechanical Systems,” U.S. Patent
dressed in the future. No. 6,175,170.
关29兴 Hetrick, J., and Kota, S., 2003, “Displacement Amplification Structure and
Device, U.S. Patent No. 6,557,436.
References 关30兴 Abramson, M. A., 2006, “NOMADm version 4.0 User’s Guide,” http://
关1兴 Khatait, J. P., Mukherjee, S., and Seth, B., 2006, “Compliant Design for Flap- www.afit.edu/en/ENC/Faculty/MAbramson/NOMADm.html

Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1075

Downloaded 10 Nov 2007 to Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm