Design of Compliant Mechanisms for Minimizing Input Power in Dynamic Applications Tanakorn Tantanawat e-mail: tanakorn@umich.edu Sridhar Kota e-mail: kota@umich.edu Department of Mechanical Engineering, The University of Michigan, Ann Arbor, MI 48109 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 storage and transfer between the input, external load, and strain energy stored within the compliant transmission. The goal is to design compliant mechanisms for dynamic applications by exploiting the inherent energy storage capability of compliant mechanisms in the most effective manner. We present a detailed case study on a flapping mechanism, in 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 present two approaches: (1) generative-load exploitation and (2) reactance cancellation, to describe the role of stored elastic energy in reducing the peak input power requirement. 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, input power reduction Introduction ant mechanisms may be efficient when they are operated under a quasi-static condition, but not a dynamic one. Analysis and synthesis tools based on a dynamic condition are, therefore, necessary to extend the applications of compliant mechanisms. While the purpose of this ongoing research is to develop a systematic method of designing compliant mechanisms based on dynamic performance, this paper presents the results, as a preliminary step, of an attempt to understand the benefits of elasticity in dynamic systems. This understanding will be fundamental for the development of an efficient and robust design tool during the course of the research. Due to its multifunctional structure that combines the function of a mechanism and an energy storage component together, a compliant mechanism provides a lightweight and compact system. Its monolithic structure leads to scalability, no wear, and low manufacturing cost. In addition, its ability to incorporate unconventional actuators makes it very attractive to several applications. Nonetheless, if a distributed compliant mechanism is used, further benefit can be gained. Since stress and strain are more uniformly distributed than those in lumped compliant mechanisms, there is less stress concentration but higher energy storage capacity in distributed compliant mechanisms. Because of all these benefits, a compliant mechanism is expected to be an excellent alternative to a rigid-body mechanism, especially for autonomous robots, whose size and weight are critical. In this study, a compliant flapping mechanism, which can be applied to flapping-wing micro air vehicles developed by other researchers, will be used as a case study during the development of the design method. When a compliant mechanism is deformed to transmit force and motion, some of the input energy is stored in the mechanism in the form of strain energy, while the rest is transferred to the output load. This stored energy is commonly perceived as energy loss, preventing full energy transfer from the input port to the output port. One of the approaches to design efficient compliant mechanisms for a quasi-static condition, therefore, attempts to minimize this stored strain energy. The scenario is completely different if a compliant mechanism is operated under a dynamic condition, in which strain energy is stored and released during a cycle of operation. Most of the stored strain energy, if not all, can be recycled. An approach used to design a compliant mechanism for a dynamic condition should, therefore, be different from the one used for a quasi-static condition. Khatait et al. 关1兴 suggest the use of compliant mechanisms to reduce input torque requirement in a flapping mechanism. The study is, however, limited to the case of unloaded lumped compliant mechanisms. In addition, Madangopal et al. 关2兴 optimize a four-bar flapping mechanism with attached springs in an attempt to minimize the input torque requirement of a flapping-wing micro air vehicle. However, how the springs enable input torque reduction has not been clearly explained. These two examples have demonstrated the benefits of elasticity in reducing input torque requirement, and lead us to believe that a distributed compliant mechanism holds a potential to be an alternative for dynamic systems. Since most of the early research and development in analysis and synthesis tools for compliant mechanisms were based on a quasi-static assumption, the performance of the designed compli- Related Work Contributed by the Mechanisms and Robotics Committee of ASME for publication in the JOURNAL OF MECHANICAL DESIGN. Manuscript received July 13, 2006; final manuscript received October 25, 2006. Review conducted by Larry L. Howell. Paper presented at the ASME 2006 Design Engineering Technical Conferences and Computers and Information in Engineering Conference 共DETC2006兲, Philadelphia, PA, September 10–13, 2006. The use of compliant mechanisms to transmit force and motion has been studied for several decades. This type of mechanism evolved from work in robotics, where manipulator arms must perform various tasks on workpieces whose locations and dimensions are not precisely known 关3兴. In this application, the use of compliance 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 141.213.247.101. 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 robotics 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 reduction 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 wellbehaved compliant joints were proposed. Since compliant mechanisms 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 leverage 关13,14兴, structural shape morphing in smart structures 关15兴, surgical tools 关16兴, active flow control 关17兴, and vibration isolation 关18兴. Various methods of analysis and synthesis of compliant mechanisms can be found in the literature including Pseudo-Rigid-Body Models 共PRBMs兲 and continuum mechanics methods for topology, geometry, and size optimization. Much of the reported work on analysis and design assumes a quasi-static condition. It was not until recently that researchers began to incorporate dynamic behaviors of compliant mechanisms, such as Du et al. 关19兴, Nishiwaki et al. 关20兴, and Maddisetty and Frecker 关21兴. Li and Kota 关22兴 proposed a systematic method for dynamic analysis of compliant mechanisms. Dimensional synthesis based on desired mode shapes is presented in 关23兴 by Lai and Anathasuresh. Boyle et al. 关24兴, Handley 关25兴 et al., and Yu et al. 关26兴 presented the development in dynamic modeling of compliant mechanisms based on PRBMs. Khatait et al. 关1兴 used dynamic PRBMs to investigate the reduction of torque requirement in a flapping mechanism. None of the previous work, however, directly addressed the issue of input power requirement. Since it is the power requirement that governs various component designs and affects the overall system’s performance, we use the power requirement as one of the criteria in the design method. Even though a design approach based on torque requirement is equivalent to a design approach based on power requirement when the operating frequency is constant, it is not the case when the frequency changes during the operation, or when the frequency is one of the design variables during a system design. A design approach based on the power requirement will still be applicable to these situations. Power Analysis of a Four-Bar Flapping Mechanism To understand the behavior of input power reduction, we first perform power analysis on a four-bar flapping mechanism optimized by Madangopal et al. 关2兴. In their work, the aerodynamic model is based on quasi-steady blade element analysis. The model includes unsteady wake effects, camber and partial leading edge suction effects, and post-stall behavior. The grounds for the mechanism are assumed inertially fixed. For a given set of parameters, the optimal values of spring stiffness k, free length l0, and location of attachment a were obtained. We model this mechanism in ADAMS, which is dynamic mechanical system simulation software. 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. The input displacement in共t兲 at the operating frequency f can be written as in共t兲 = 2 ft 共2兲 The values of the links’ dimensions and operating frequency are shown in Table 1. Journal of Mechanical Design Note that the links’ cross sections are adjusted so that their masses match those in Madangopal’s model. The operating frequency is adjusted to match the peak values of inertial torque. Finally, a damping constant and a constant force component are adjusted to match the peak value of the lift. Since the weights of the links are small compared to the aerodynamic force on the wing, the gravity effect is ignored in this study to simplify the analysis. Once the basic understanding is gained, more accurate results can be obtained by including the gravity. The input power can be either positive or negative during a flapping cycle. A positive input power indicates that the motor is supplying energy into the system, while a negative input power indicates that the motor is absorbing energy from the system. Since the focus of this study is to capture the instantaneous input power experienced by the motor, which affects the design of motors and electrical components, we ignore the energy loss in the motor and mechanism. In other words, we assume that all energy absorbed by the motor can be fully recovered. In addition, we assume that instantaneous power experienced by the motor is equivalent in both directions 共supplying and absorbing energy兲. The quantity being investigated as an objective function to be minimized in this study is, therefore, the maximum of the absolute value of instantaneous input power, or a peak input power. The values of the spring stiffness and free length are optimized 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 torque and lift obtained from this simplified model are shown in Figs. 2 and 3, respectively. Plots of input powers for the mechanism with and without a spring are shown in Fig. 4. Various power components for the mechanism with a spring are shown in Fig. 5. Figure 4 shows that adding a spring reduces peak input power by 42% from 1.00 N m / s to 0.58 N m / s. Figure 5 indicates that energy flow in and out of the link’s mass 共kinetic energy rate 共KER兲, spring potential energy rate 共PER兲, and aerodynamic force 共Pout兲兲 all significantly contribute to the required input power 共Pin兲. The results themselves, however, do not provide deeper insight into the role of a spring and various components of energy Table 1 Links’ dimensions and operating frequency l1 = 6 mm x = 15 mm l2 = 20 mm y = 21 mm l3 = 330 mm a = 50 mm f = 4 Hz b = 165 mm 共1兲 The simplified aerodynamic force, parametrized by a damping constant co, constant force component f o, and normal velocity at the center of the wing vo can be written as: Fo共t兲 = − covo共t兲 + f o Fig. 1 Four-bar flapping mechanism adapted from Madangopal et al. †2‡ Table 2 System parameters and optimal values of spring properties used in the analysis of the simplified four-bar flapping mechanism mAB = 0.25 g mBC = 0.50 g mCD = 2.75 g co = 0.207 N s / m f o = 0.27 N k = 38 N / m l0 = 50 mm OCTOBER 2007, Vol. 129 / 1065 Downloaded 10 Nov 2007 to 141.213.247.101. 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 motion exchanged within the system components. We, therefore, investigate the effect of a spring when the system is subject to three different cases: 共I兲 with constant output force only 共II兲 with inertial force only 共III兲 with damping output force only. In each case, the values of spring stiffness and free length are optimized for minimum peak input power. The resulting power variations in the mechanism without a spring and with a spring are then compared. Based on these results, we attempt to visualize the path of the energy flow within the system. 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 zero to investigate the interaction of the spring and the constant component of output force. The results are shown in Figs. 6–8. From Fig. 6, adding a spring reduces the peak input power by 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 flapping cycle for each case. For the mechanism without a spring Fig. 5 Various power components of the mechanism with a spring, including kinetic energy rate „KER…, potential energy rate „PER…, input power „Pin…, and output power „Pout… 共Fig. 7兲, the constant output force, as being upward, produces positive work into the mechanism during the upstroke. To maintain the steady-state motion, the motor supplies negative work into the mechanism. In other words, it has to absorb energy flowing from the output. During the downstroke, the output force is still upward but the motion of the wing is downward. The output force is producing negative work into the system, or extracting the energy from the system. When a spring is added into the system 共Fig. 8兲, the output force still produces positive work into the mechanism during the upstroke. However, instead of having the motor absorb this work from the output, the spring stores this Table 3 Parameters for Case I „constant output force only… mAB = 0 g mBC = 0 g mCD = 0 g co = 0 N s / m f o = 0.27 N k = 0 or 35 N / m l0 = 41 mm 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. 4 Input powers of the mechanism with and without a spring. Adding a spring reduces peak input power by 42%. 1066 / Vol. 129, OCTOBER 2007 Fig. 7 For Case I „constant output force only… without a spring, the motor has to supply and absorb the full amount of energy to balance the energy extracted and injected by the output force Transactions of the ASME Downloaded 10 Nov 2007 to 141.213.247.101. 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, the presence of a spring helps the motor supply and absorb energy extracted and injected by the output force, thus reducing the effort needed by the motor to balance the system positive work as elastic energy. The motor, then, needs less effort in balancing the steady-state motion. During the downstroke, the output force extracts energy out of the system. During this period, the spring releases the stored elastic energy to the output. In this situation, adding a spring to the system reduces the peak input power requirement. Case II. The mechanism subject to inertial force only The values of parameters are shown in Table 4. In this case, the links’ masses are nonzero. The output force is set to be zero for both damping and constant components. The results are shown in Figs. 9–11. From Fig. 9, adding a spring into the system reduces the required 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 release 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 component. The results of power analysis are shown in Figs. 12–14. Table 4 Parameters for Case II „inertial force only… mAB = 0.25 g mBC = 0.50 g mCF = 2.75 g co = 0 N s / m fo = 0 N k = 0 or 226 N / m l0 = 117 mm Fig. 9 For Case II „inertial force only…, adding a spring reduces peak input power by 83% Journal of Mechanical Design Fig. 10 For Case II „inertial force only… without a spring, the motor has to supply and absorb the full amount of energy stored and released by links’ masses From Fig. 12, adding a spring into the system increases the required peak input power by 78% from 0.55 N m / s to 0.98 N m / s. For the mechanism without a spring 共Fig. 13兲, the damping output force extracts energy from the system both during upstroke and downstroke. The motor always has to supply this amount of energy to maintain the steady-state motion. The peak input power is, therefore, equal to the peak output power. When a spring is added into the system 共Fig. 14兲, the motor has to supply energy both to the output and to the spring during the upstroke and, therefore, requires more power than it does without a spring. During the downstroke, the spring releases energy to the output, Fig. 11 For Case II „inertial force only…, the spring added to the system helps absorb and supply energy as the links decelerate and accelerate, thus reducing the power requirement from the motor Table 5 Parameters for Case III „damping output force only… mAB = 0 g mBC = 0 g mCD = 0 g co = 0.207 N s / m fo = 0 N k = 0 or 38 N / m l0 ⫽ 50 mm Fig. 12 For Case III „damping output force only…, adding a spring increases peak input power by 78% OCTOBER 2007, Vol. 129 / 1067 Downloaded 10 Nov 2007 to 141.213.247.101. 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 spring, the motor has to supply energy equal to the amount extracted by the output force reducing input power requirement from the motor. The entire flapping cycle, therefore, requires more input power if a spring is added to this type of system. In fact, this should always be true for the following situation. If the only source that provides positive work into a massless system is the input, then the energy required to deform the spring will have to be from this source and, therefore, the source needs to provide this extra power to the system in addition to the power required from the dissipative output. Figures 15 and 16 show that for the case of damping output force only, the peak input power is minimal when there is no spring added to the system. The results of all case studies are summarized in Table 6. If we consider each component of the load acting on the system, this study reveals that the existence of either a constant output force 共Case I兲 or an inertial property of the system 共Case II兲 allows the use of elasticity to reduce the peak input power. These two types of load provide instantaneous positive work, which can be stored during one stage of a cycle and reused in another. On the Fig. 16 For Case III „damping output force only…, adding a spring of any free length, in this case with a stiffness of 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. contrary, the existence of a damping force reduces the effectiveness of elasticity to reduce the peak input power. From Table 6, the optimal values of spring’s properties for the case of a combined constant-inertial-damping force 共stiffness of 38 N / m and free length of 50 mm兲 are between those of a pure constant output force 共Case I兲 and a pure inertial force 共Case II兲. It can be further noticed that these optimal values are closer to those for the case of a pure constant output force. This corresponds to the fact that the contribution of the input power due to a pure constant output force 共0.45 N m / s兲 is more significant than that due to a pure inertial force 共0.18 N m / s兲. The presence of a damping component in the case of combined force brings the percentage of peak input power reduction down to 42%. This study provides a visualization of how each load component affects the optimal values of the spring’s properties. Two Approaches to Exploit Elastic Energy Fig. 14 For Case III „damping output force only… with a spring, besides supplying energy to the output force, the motor also has to supply energy to be stored in the spring Fig. 15 For Case III „damping output force only…, adding a spring of any stiffness, in this case with a free length of 50 mm, results in undesired increase in peak input power 1068 / Vol. 129, OCTOBER 2007 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 elastic component will release some energy to the output or inertia when, otherwise, the input has to supply this energy. If the elastic component is used to split energy flow via path B, its role is referred to as generative-load exploitation. This approach of exploiting 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 reactance 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 significant. A design approach based on resonance frequency, which corresponds to the reactance cancellation approach as referred in this paper, can result in a suboptimal design when a Transactions of the ASME Downloaded 10 Nov 2007 to 141.213.247.101. 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 properties Load case Constant output force only 共Case I兲 Combined constant-inertial-damping force Inertial force only 共Case II兲 Damping output force only 共Case III兲 k 共N/m兲 l0 共mm兲 Pin,peak without spring 共N m/s兲 Pin,peak reduction 共%兲 35 38 226 0 41 50 117 N/A 0.45 1.00 0.18 0.55 84 42 83 0 generative load is present. In this study, energy flow across the system’s inertia is small compared to energy flow across the output 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, 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 cancellation. Even though the distinction in the two approaches described in this paper may seem unnecessary if a general optimization method is applied, this distinction is helpful for energy flow visualization. It will also help set up a design framework and may be useful for guiding the design optimization process, leading to a more efficient and robust design tool. This is necessary especially for the design of compliant mechanisms, where nonlinear responses are computationally expensive. Power Analysis and Design of a 2DOF System Two different approaches to describe the role of elasticity, namely, generative-load exploitation and reactance cancellation, have been defined based on the investigation of power flow in the four-bar flapping mechanism with a spring. The concept of these two approaches cannot be immediately generalized to compliant mechanisms because compliant mechanisms are elastic systems consisting of an infinite number of degrees of freedom 共DOF兲. Unlike rigid-body mechanisms, the kinematics of compliant mechanisms depends on system’s properties and operating conditions, such as mass, stiffness, and operating frequency. We, therefore, use a 2DOF system, which represents the simplest form of a compliant mechanism, to demonstrate the validity of the two approaches in exploiting elasticity. The system consists of two masses: m1 and m2 and three linear springs: k1, k2, and k3, as shown in Fig. 18. Forces f 1共t兲 and f 2共t兲 are applied to masses m1 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 solutions are: u1共t兲 = U1 sin共t兲 + Ū1 共3兲 u2共t兲 = U2 sin共t兲 + Ū2 共4兲 f 1共t兲 = F1 sin共t兲 + F̄1 共5兲 f 2共t兲 = F2 sin共t兲 + F̄2 共6兲 The equations of motion for the two masses can be written in a matrix form. The static and harmonic components can be written separately, yielding the following systems of equations: 冋册冋 F̄1 = F̄2 冋册冉 冋 F1 F2 = − 2 k1 + k2 − k2 − k2 k2 + k3 m1 0 0 m2 册冋 + 册冋 Ū1 Ū2 册 k1 + k2 − k2 − k2 k2 + k3 共7兲 册冊冋 册 U1 U2 共8兲 Assuming that the system is subject to an input displacement 共U1, Ū1兲 and the harmonic component of an output force 共F2兲 is zero, we can solve for Ū2, U2, F̄1, and F1. The results are: F̄2 + k2Ū1 k2 + k3 共9兲 k2 U1 − m2 + k2 + k3 共10兲 共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2 k2 + k3 共11兲 Ū2 = U2 = F̄1 = 2 Fig. 17 Concept of using elasticity to reduce peak input power requirement in dynamic systems Journal of Mechanical Design OCTOBER 2007, Vol. 129 / 1069 Downloaded 10 Nov 2007 to 141.213.247.101. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm F1 = 4m1m2 − 2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3 U1 − 2m 2 + k 2 + k 3 共12兲 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 U2,req k2 共13兲 共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2 k2 + k3 共14兲 4m1m2 − 2共m1k2 + m1k3 + m2k1 + m2k2兲 + k1k2 + k1k3 + k2k3 U2,req k2 共15兲 U1 = F̄1 = F1 = The instantaneous input power can be calculated from: p1共t兲 = f 1共t兲u̇1共t兲 共16兲 1 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2t兲 + F̄1U1 cos共t兲 2 共17兲 Substituting Eqs. 共3兲 and 共5兲 into Eq. 共16兲, we have: 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2t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 2 k22 共19兲 where G2 = 关共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2兴关− 2m2 + k2 + k3兴 U2,req k2共k2 + k3兲 1 共k1k2 + k1k3 + k2k3兲共k2 + k3兲 2 G1 = U2,req 2 k22 共23兲 共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2 U2,req k2 共24兲 共20兲 G2 = and k1 ⱖ 0, k2 ⬎ 0, k3 ⱖ 0, m1 ⱖ 0, m2 ⱖ 0, − ⬁ ⬍ Ū1 ⬍ ⬁ 共21兲 We will determine optimal designs for two cases: one is when the system has no mass and the other is when the system has no output force. The first case represents an attempt to use elasticity based on generative-load exploitation only, while the second case represents an attempt to use elasticity based on reactance cancellation. Case I. m1 = m2 = 0 and F̄2 ⫽ 0 共generative-load exploitation兲 Equations 共18兲–共20兲, respectively, reduce to: p1共t兲 = G1共k1,k2,k3兲sin共2t兲 + G2共k1,k2,k3,Ū1兲cos共t兲 共22兲 Fig. 18 A 2DOF system used to validate the concept of generative-load exploitation and reactance cancelation in compliant mechanisms 1070 / Vol. 129, OCTOBER 2007 Since G2 has an extra variable 共Ū1兲, which can be any real number, G2 may be considered independent of G1. The peak of p1共t兲, or P1,peak, can be minimized when 兩G1兩 and 兩G2兩 are minimized. Since the minimum value that 兩G1兩 and 兩G2兩 may reach is zero, we first attempt to find conditions that make G1 and G2 be zero. From Eq. 共23兲, G1 is set to zero. We have: 共k1k2 + k1k3 + k2k3兲共k2 + k3兲 k22 k 1k 2 + k 1k 3 + k 2k 3 = 0 =0 共25兲 共26兲 Substitute Eq. 共26兲 into Eq. 共24兲 to find conditions that make G2 be zero: 0 − k2F̄2 =0 k2 共27兲 F̄2 = 0 共28兲 This is impossible since we have F̄2 ⫽ 0. Therefore, P1,peak cannot be made zero. We then attempt to find conditions that make G1 and G2 approach zero. From Eq. 共23兲, when G1 approaches zero, we have: Transactions of the ASME Downloaded 10 Nov 2007 to 141.213.247.101. 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 k1k23 k23 k 1k 3 + k3 + 2 + → 0 k2 k2 k2 共30兲 k22 k1 + 2 Condition 共30兲 is true only when: k1 → 0 k3 → 0 k 1k 3 →0 k2 共31兲 Substitute 共31兲 into Eq. 共24兲 and find conditions that make G2 approach zero: 共k1k2 + k1k3 + k2k3兲Ū1 − k2F̄2 →0 k2 共32兲 共k1k2 + k2k3兲Ū1 − k2F̄2 →0 k2 共33兲 Ū1 → F̄2 k1 + k3 共34兲 Ū1 → ± ⬁ 共35兲 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. However, Ū1 cannot physically be ±⬁. For simplicity, to investigate the case when there is a bound on Ū1, we assume that F̄2 ⬎ 0 and the upper bound Ū1,bound ⬎ 0. For the case when k1 = 0, Eqs. 共23兲 and 共24兲 become: k3共k2 + k3兲 1 2 G1 = U2,req 2 k2 Fig. 19 A general plot of 円G1円 and 円G2円 as functions of k3 when k1 is zero From Eqs. 共36兲 and 共37兲, however, k2 appears only in G1 because k1 has been assumed zero. The value of 兩G1兩 decreases when k2 increases. The optimal k2 that minimizes the P1,peak, therefore, approaches ⬁. This result implies, for the case when there is no input stiffness, that the compliance between the input and the output should be minimized and a spring should be attached to the output. In reality, there is usually stiffness associated with the input, such as the stiffness of an actuator. In compliant mechanisms, input stiffness exists due to the members connecting the input to the ground. In this case, k1 is not zero. G1 and G2 from Eqs. 共36兲 and 共37兲, respectively, are not valid. We have to use Eqs. 共23兲 and 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 the following: G2 = U2,req 冊 k 1k 3 + k3 Ū1 − F̄2 k2 冉 k1 + 冊 k 1k 3 + k3 Ū1 − F̄2 = 0 k2 k2 = k 1k 3 F̄2/Ū1 − 共k1 + k3兲 共38兲 共39兲 共40兲 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 ⬁ when there is input stiffness and a bound on average input displacement. For this situation, compliance between the input and output will help reduce the peak input power. Case II. m1 ⬎ 0, m2 ⬎ 0, and F̄2 = 0 共reactance cancellation兲 Equations 共18兲–共20兲 become: p1共t兲 = G1共k1,k2,k3,m1,m2兲sin共2t兲 + G2共k1,k2,k3,Ū1兲cos共t兲 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 2 k22 G2 = 冎 − F̄2其, which will contribute to P1,peak. 兩G2兩 can be minimized when: 共37兲 From Eq. 共36兲, 兩G1兩 can be reduced when k3 is reduced. From Eq. 共37兲, 兩G2兩 can be made zero when k3 ⱖ F̄2 / Ū1,bound. When k3 is reduced further such that k3 ⬍ F̄2 / Ū1,bound, Ū1 that minimizes 兩G2兩 is Ū1,bound and 兩G2兩 increases to U2,reqF̄2 as k3 approaches zero. A general plot of function 兩G1兩 and 兩G2兩 is shown in Fig. 19. Figure 19 implies that an optimal k3 that minimizes P1,peak does 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 the optimal k1 does not necessarily approach zero. k1 + When k2 approaches ⬁, G2 will approach U2,req兵共k1 + k3兲Ū1 共36兲 G2 = U2,req共k3Ū1 − F̄2兲 再冉 共k1k2 + k1k3 + k2k3兲Ū1 U2,req k2 共42兲 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2t兲 Journal of Mechanical Design 共44兲 OCTOBER 2007, Vol. 129 / 1071 Downloaded 10 Nov 2007 to 141.213.247.101. 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 2 k22 共45兲 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 k22 The solutions to Eq. 共46兲 are: k2 = m2 − k3 共47兲 2 or k2 = 4m1m2 − 2共m1k3 + m2k1兲 + k1k3 2共m1 + m2兲 − k1 − k3 共48兲 Depending on the values of m1, m2, k1, and k3, it is possible that there is k2 that reduces P1,peak to zero. In addition, it can be proved that Eqs. 共47兲 and 共48兲 correspond to the natural modes of the system with m1 fixed and m1 free, respectively. For the case when k1 = k3 = 0, Eqs. 共41兲 and 共42兲 reduce to: p1共t兲 = G1共k2,m1,m2兲sin共2t兲 共49兲 1 关4m1m2 − 2共m1k2 + m2k2兲兴关− 2m2 + k2兴 2 G1 = U2,req 2 k22 共50兲 We determine conditions that make G1 be zero. By setting Eq. 共50兲 to be zero, we have: 关4m1m2 − 2共m1k2 + m2k2兲兴关− 2m2 + k2兴 k22 =0 共51兲 The solutions to Eq. 共51兲 are: 共52兲 or 2m 1m 2 m1 + m2 共53兲 Equations 共52兲 and 共53兲 show that there is always k2 that reduces P1,peak to zero. The summary of the conditions that minimize the peak input power is shown in Table 7. From Table 7, k2 approaching infinity in Case I 共D兲 represents a rigid-body connection between the input and output. This case confirms the concept of generative-load exploitation in a rigidbody mechanism. Cases I 共E兲 and I 共F兲, with stiffnesses not approaching infinity, represent a compliant system. These cases Table 7 Summary of the conditions that minimize peak input power F̄2 Case I 共A兲 共B兲 共C兲 共D兲 ⫽0 ⫽0 ⫽0 ⫽0 m1 m2 =0 =0 =0 =0 →±⬁ =0 →0 any k2⬎0 →0 =0 =0 any k2 ⬎ 0 →0+ → ± ⬁ =0 →0+ any k2 ⬎ 0 =0 →±⬁ =0 =0 →⬁ → ” 0 Bounded → ”⬁ =0 → ” 0 any k2 ⬎ 0 =0 Bounded → ”⬁ =0 → ⬎0 ”0 → ” 0 Bounded → ”⬁ → ”⬁ → ”⬁ 共E兲 ⫽0 =0 共F兲 ⫽0 =0 Case II 共G兲 =0 ⬎0 ⬎0 共H兲 =0 ⬎0 ⬎0 共46兲 show that the generative-load exploitation is applicable to a compliant mechanism when an input stiffness and a bound of average input displacement exist. Since k2 in Case II 共G兲 may approach 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 concept of reactance cancellation is applicable to a compliant mechanism, regardless of the existence of an input stiffness or a bound of average input displacement. In addition, Eqs. 共47兲 and 共48兲 indicate that the reactance cancellation corresponds to two types of resonance, one is associated with a free-input response and the other is associated with a fixed-input response. Even though the analysis of a 2DOF system validates the concept of generative-load exploitation and reactance cancellation in compliant mechanisms, the analysis procedures are not applicable to analyze and design compliant mechanisms. Most compliant mechanisms undergo large displacements and require nonlinear analysis to evaluate their performance. In addition, stiffness and mass in a compliant mechanism are coupled through geometry and cannot be designed independently. The analysis of a 2DOF system in this study, therefore, only provides mathematical understanding in the use of elasticity. The analysis and design of compliant mechanisms with large displacements require more sophisticated methods. Optimization Strategies k 2 = 2m 2 k2 = =0 k1 k2 + ⬎0 =0 k3 Ū1 + 共47兲, 共48兲 共47兲, 共53兲 1072 / Vol. 129, OCTOBER 2007 ⬎0 =0 =0 ⫽0 P1,peak →0 →0+ →0+ ⬎0 + ⬎0 ⬎0 ⱖ0 =0 Unlike a linear system, most compliant mechanisms undergo large displacement, and geometric nonlinearity must be taken into account in the analysis. The amplitude of input displacement required to provide the desired amplitude of output displacement cannot be determined directly. In addition, a harmonic analysis and the principle of superposition are no longer applicable. In this study, we use nonlinear transient analysis based on an implicit dynamic method to analyze and design a compliant mechanism. Since the amplitude and mean values of the input displacement can be controlled during the operation, we do not consider them as true design variables in the context of a “mechanism design.” Instead, we use the term “performance parameters” to refer to the amplitude and mean values of the input displacement. Beam heights and keypoint locations, on the other hand, cannot be controlled once the system is built and are still considered “design variables.” Based on the distinction between performance parameters and design variables, there are at least three strategies for optimizing compliant mechanisms. 共1兲 Performance parameters and design variables are optimized together in one optimization loop. 共2兲 Performance parameters are optimized as an inner loop for each evaluation of a set of design variables being optimized in an outer loop. 共3兲 Performance parameters are fixed and only design variables are optimized. Strategy 共1兲 is based on a modern design automation approach, which does not have to distinguish between performance parameters and design variables. Strategy 共2兲 is based on a traditional design approach, in which a designer evaluates each design properly before moving and comparing to the next design. Strategy 共3兲 is based on simplicity, which ignores the effects of input on the Transactions of the ASME Downloaded 10 Nov 2007 to 141.213.247.101. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm Fig. 20 Front view of the compliant flapping mechanism showing the initial design before optimization design. The primary motivation in distinguishing these strategies is to find the most efficient optimization scheme to search design space. For example, it is possible that a continuously changing amplitude or mean value without restarting the analysis from zero input may reduce computational time by shortening the time during transient state. In this case, Strategy 共2兲 may be preferred. However, there is no straightforward manner to predict the most efficient strategy to use. Design of a Compliant Flapping Mechanism In this study, we use Strategy 共1兲, optimizing performance parameters and deign variables together in one loop, to design a compliant flapping mechanism. The initial design, shown in Fig. 20, is an application of a patented compliant stroke amplifier 关27–29兴, in which the wing is directly connected to the output link. It is modeled and analyzed in ABAQUS, which is commercial finite element analysis software. Instead of having a uniform beam width, the wing section of the mechanism has been modified to redistribute the mass and stiffness, as shown in Fig. 21. The total mass of the wing section is 2.50 g, the same as that in the four-bar flapping mechanism, but a significant amount of mass is distributed to the end of the wing section. This will promote the use of reactance cancellation, in addition to the use of generative-load exploitation. The wing’s dimensions are shown in Table 8. The mechanism is modeled using structural beam elements. Its overall dimensions are approximately the same as those of the four-bar flapping mechanism studied in the previous section. We assume that the entire mechanism is made of carbon fiber composites whose Young’s modulus is 520 GPa, Poisson’s ratio is 0.28, and density is 1570 kg/ m3. The output force Fo共t兲, which is Fig. 22 An example of input displacement with gradually increasing amplitude to facilitate the convergence of nonlinear transient analysis applied at the center of the wing, is a simplified aerodynamic force previously used in the analysis of the four-bar flapping mechanism, as described by Eq. 共2兲. The values of co and f o are 0.207 N s / m and 0.27 N, respectively. The required flapping angle flap, measured at the location where the output force is applied, is 44.8 deg, the same as in the four-bar flapping mechanism. This amount of flapping angle is assumed to provide sufficient lift at the specified frequency to sustain the vehicle in the air and is the primary functional requirement of the mechanism. The gravity is ignored in the analysis. The mechanism is subject to a sinusoidal input displacement uin共t兲, with amplitude A and average C. However, to facilitate the convergence in finding solutions of the analysis, we impose the input displacement with gradually increasing amplitude, expressed in the form: 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 共54兲 For this study, H = 1 / 4, S = 1 / 4, and f = 4 Hz. An example of input displacement is shown in Fig. 22. There are a total of 37 design variables for this optimization problem: two are performance parameters including A and C from Eq. 共54兲, ten are keypoint locations, and 25 are beam heights, as shown in Fig. 23. The objective of the optimization problem is to minimize the peak input power requirement, subject to the flapping angle of 44.8 deg. Each analysis is run until the mechanism reaches a steady state, which is predicted by comparing responses in one cycle to others. In doing this, criteria for analysis termination have to be specified. When the differences in responses 共i.e., flapping angle, input 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 showing a wing section with nonuniform mass and stiffness distribution Table 8 Dimensions of the wing Wing segment Height, h 共mm兲 Width, w 共mm兲 Length, l 共mm兲 A B C D E 1.00 1.00 1.00 1.00 1.00 0.50 1.50 3.50 10.44 18.00 82.50 82.50 82.50 45.83 36.67 Journal of Mechanical Design Fig. 23 Design variables include „a… ten variables of x and y locations for five keypoints, „b… 25 variables of beam heights for nine beam segments, and two variables for amplitude and average values of input displacement „see Fig. 20… OCTOBER 2007, Vol. 129 / 1073 Downloaded 10 Nov 2007 to 141.213.247.101. 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 can be determined after the steady-state criteria have been met. The optimization process is accomplished through the use of Matlab®. NOMADm 关30兴 algorithm, which is the implementation Table 9 Keypoints of the optimal mechanism Keypoint x-coordinate 共mm兲 y-coordinate 共mm兲 1 2 3 4 5 6 7 8 9 0.00 8.69 21.52 29.98 5.20 31.00 3.00 −2.00 3.00 0.00 12.00 6.01 26.62 23.65 39.97 50.00 50.00 14.00 Table 10 Segments = 2.00 mm… of the optimal mechanism Segment Height 共mm兲 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 0.23 0.24 0.27 0.24 0.25 0.19 0.26 0.27 0.25 0.24 0.24 0.32 0.25 2.55 0.31 0.24 0.30 0.25 0.24 0.25 0.24 0.25 0.27 0.26 4.90 1074 / Vol. 129, OCTOBER 2007 of a Mesh Adaptive Direct Search 共MADS兲 algorithm, was first used to search the design space. One of the features of this algorithm that is suitable for this problem is that the algorithm is intended for a problem whose objective and constraint functions are computationally expensive to evaluate. The solution was obtained after 3814 analyses, which took 71 h on a computer with 2.8 GHz Pentium® IV processor and 512 MB of RAM. This solution was then used as an initial design for local search using Sequential Quadratic Programming 共SQP兲. The solution was obtained after 4958 analyses, which took 96 h on the same computer. The optimal design is shown in Fig. 24, along with num- „width Fig. 25 The mechanism produces a flapping angle of 44.8 deg, generating a sufficient lift for the vehicle at 4 Hz Fig. 26 Wing rotation over a flapping cycle at steady state „after seven cycles… Transactions of the ASME Downloaded 10 Nov 2007 to 141.213.247.101. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm 关2兴 关3兴 关4兴 关5兴 关6兴 Fig. 27 Variation of different power components over a flapping cycle at steady state 关7兴 关8兴 bering for keypoints and segments, the values of which can be found in Tables 9 and 10. The optimal amplitude A is 0.612 mm and average C is 0.770 mm. The extreme positions of the wing are shown in Fig. 25, confirming that the wing is deformed and rotated as expected. The plot of wing rotation over time is shown in Fig. 26. The plot of various power components is shown in Fig. 27. This design example of a compliant flapping mechanism requires only 0.49 N m / s of input power, compared to 1.0 N m / s for the four-bar flapping mechanism without a spring, and 0.58 N m / s for the one with a spring 共50% and 15% peak input power reduction, respectively兲, in performing the same task. Note that the two peaks of input power are evened out. This is the same behavior found in the four-bar flapping mechanism when an appropriate spring is added. The peak output power is only 0.75 N m / s, compared to 1 N m / s in the four-bar flapping mechanism. The reduction of peak output power can be explained by the use of sinusoidal input displacement, which produces more efficient flapping velocity profile, generating the same amount of lift with lower output power. The 15% reduction of peak input power in the compliant flapping mechanism over the four-bar flapping mechanism with a spring can, therefore, be attributed to the use of elasticity and a more efficient input velocity profile. Conclusions In this work, we have investigated the concept of peak input power reduction through the use of elasticity. From the power analysis of a four-bar flapping mechanism, we make a distinction between two different approaches that can be used to describe the role of an elastic component: 共1兲 generative-load exploitation and 共2兲 reactance cancellation. Only one or both of them may be applied at the same time, depending on the nature of the load. These two approaches have been shown to be valid to describe the role of elasticity in compliant mechanisms as well. We then propose a compliant mechanism as an alternative to a rigid-body mechanism with attached springs. Through the use of NOMADm and SQP algorithms, we obtain a compliant flapping mechanism that can reduce the peak input power requirement by an additional 15% over the four-bar flapping mechanism with a spring, while generating the same amount of lift. The computational requirement seems to be too high. Therefore, in an attempt to develop an efficient and robust design tool, we are currently investigating various problem formulations and optimization methods to reduce computational time. Other design issues, such as average power consumption, sensitivity to operating condition and dimensional error, the coupling between system design and aerodynamic load, and stress constraint will be addressed in the future. 关9兴 关10兴 关11兴 关12兴 关13兴 关14兴 关15兴 关16兴 关17兴 关18兴 关19兴 关20兴 关21兴 关22兴 关23兴 关24兴 关25兴 关26兴 关27兴 关28兴 关29兴 References 关1兴 Khatait, J. P., Mukherjee, S., and Seth, B., 2006, “Compliant Design for Flap- Journal of Mechanical Design 关30兴 ping Mechanism: A Minimum Torque Approach,” Mech. Mach. Theory, 41共1兲, pp. 3–16. Madangopal, R., Khan, Z. A., and Agrawal, S. K., 2005, “Biologically Inspired Design of Small Flapping Wing Air Vehicles Using Four-Bar Mechanisms and Quasi-Steady Aerodynamics,” ASME J. Mech. Des., 127共4兲, pp. 809–816. Krouse, J. K., 1980, “Compliant Mechanisms—A New Class of Mechanical Devices,” Mach. Des., 52共2兲, pp. 86–90. Thompson, L. H., 1983, “Compliant Positioning Mechanism,” IBM Tech. Discl. Bull., 25共11A兲, pp. 5444–5445. Parkin, R. E., and Hutchinson, W. K., 1985, “Compliant Mechanical Gripper,” Robotics Age, 7共5兲, pp. 11–13. Andeen, G. B., and Kombluh, R., 1988, “Design of Compliance in Robotics,” Proceedings of the 1988 IEEE International Conference on Robotics and Automation, IEEE, New York, pp. 276–281. Holl, J., Koffler, W., and Midha, A., 1983, “Compliant Mechanism Synthesis in Prosthetic Design,” AMD Symposia Series, ASME, New York, Vol. 56, pp. 63–66. 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. 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, pp. 1115–1121. Larsen, U. D., Sigmund, O., and Bouwstra, S., 1997, “Design and Fabrication of Compliant Micromechanisms and Structures With Negative Poisson’s Ratio,” J. Microelectromech. Syst., 6共2兲, pp. 99–106. Kikuchi, N., Nishiwaki, S., Ono Fonseca, J. S., and Nelli Silva, E. C., 1998, “Design Optimization Method for Compliant Mechanisms and Material Microstructure,” Comput. Methods Appl. Mech. Eng., 151共3-4兲, pp. 401–417. Calin, M., Chaillet, N., Bourjault, A., Bertsch, A., and Zissy, S., 1996, “Design and Control of Compliant Microrobots,” IEEE Symposium on Emerging Technologies and Factory Automation, ETFA, IEEE, Piscataway, NJ, Vol. 2, pp. 489–492. Kota, S., Ananthasuresh, G. K., Crary, S. B., and Wise, K. D., 1994, “Design and Fabrication of Microelectromechanical Systems,” ASME J. Mech. Des., 116共4兲, pp. 1081–1088. Su, X. S., and Yang, H. S., 2001, “Design of Compliant Microleverage Mechanisms,” Sens. Actuators, A, 87共3兲, pp. 146–156. Lu, K., and Kota, S., 2003, “Design of Compliant Mechanisms for Morphing Structural Shapes,” J. Intell. Mater. Syst. Struct., 14共6兲, pp. 379–391. Kota, S., Lu, K., Kreiner, Z., Trease, B., Areanas, J., and Geiger, J., 2005, “Design and Application of Compliant Mechanisms for Surgical Tools,” J. Biomech. Eng., 127共6兲, pp. 981–989. 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,” J. Aircr., 41共3兲, pp. 603–609. Tantanawat, T., Li, Z., and Kota, S., 2004, “Application of Compliant Mechanisms to Active Vibration Isolation Systems,” Proceedings of the ASME Design Engineering Technical Conference, ASME, New York, Vol. 2B, pp. 1165– 1172. Du, H., Lau, G. K., Lim, M. K., and Qui, J., 2000, “Topological Optimization of Mechanical Amplifiers for Piezoelectric Actuators Under Dynamic Motion,” Smart Mater. Struct., 9共6兲, pp. 788–800. Nishiwaki, S., Saitou, K., Min, S., and Kikuchi, N., 2000, “Topological Design Considering Flexibility Under Periodic Loads,” Struct. Multidiscip. Optim., 19共1兲, pp. 4–16. Maddisetty, H., and Frecker, M., 2004, “Dynamic Topology Optimization of Compliant Mechanisms and Piezoceramic Actuators,” ASME J. Mech. Des., 126共6兲, pp. 975–983. Li, Z., and Kota, S., 2002, “Dynamic Analysis of Compliant Mechanisms,” Proceedings of the ASME Design Engineering Technical Conference, ASME, New York, Vol. 5A, pp. 43–50. Lai, E., and Anathasuresh, G. K., 2002, “On the Design of Bars and Beams for Desired Mode Shapes,” J. Sound Vib., 254共2兲, pp. 393–406. Boyle, C., Howell, L. L., Magleby, S. P., and Evans, M. S., 2003, “Dynamic Modeling of Compliant Constant-Force Compression Mechanisms,” Mech. Mach. Theory, 38共12兲, pp. 1469–1487. Handley, D. C., Lu, T., Yong, Y. K., and Zhang, W. J., 2004, “A Simple and Efficient Dynamic Modeling Method for Compliant Micropositioning Mechanisms Using Flexure Hinges,” Proc. SPIE, Vol. 5276, pp. 67–76. Yu, Y., Howell, L. L., Lusk, C., Yue, Y., and He, M., 2005, “Dynamic Modeling of Compliant Mechanisms Based on the Pseudo-Rigid-Body Model,” ASME J. Mech. Des., 127共4兲, pp. 760–765. Kota, S., Hetrick, J., Li, Z., and Saggere, L., 1999, “Tailoring Unconventional Actuators Using Compliant Transmissions: Design Methods and Applications,” IEEE/ASME Trans. Mechatron., 4共4兲, pp. 396–408. Kota, S., Hetrick, J., Rodgers, S., and Li, Z., 2001, “Compliant Displacement Amplification Apparatus for Micro Electro Mechanical Systems,” U.S. Patent No. 6,175,170. Hetrick, J., and Kota, S., 2003, “Displacement Amplification Structure and Device, U.S. Patent No. 6,557,436. Abramson, M. A., 2006, “NOMADm version 4.0 User’s Guide,” http:// www.afit.edu/en/ENC/Faculty/MAbramson/NOMADm.html OCTOBER 2007, Vol. 129 / 1075 Downloaded 10 Nov 2007 to 141.213.247.101. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm