^{1}

^{2}

^{3}

^{4}

^{4}

^{1}

^{2}

^{1}

^{2}

^{1}

^{2}

^{3}Department of Mechanical Engineering,

^{4}

The dielectric elastomer actuator (DEA) is one type of emerging soft actuator that has the attractive features of large actuation strains, high energy density, and inherent compliance, which is desirable for novel bio-inspired and soft robotic applications. Due to their inherent elasticity, when stimulated by an alternating current voltage with a frequency matching the natural frequency of the DEA system, the DEAs can exhibit resonant responses which maximize the oscillation amplitude. Silicone elastomers are widely utilized for resonant actuation applications for their reduced viscous damping hence better dynamic performance compared to VHB elastomers. However, the low pre-stretch ratios adopted by silicone elastomers could induce loss-of-tension of the mem-branes in high amplitude oscillations, yet its effects on the dynamic responses of a DEA are not fully understood. By using a numerical dynamic model, this work studies the effects of the loss-of-tension on the frequency response of the antagonistic pure-shear DEAs. A subharmonic frequency response curve isolated from the main response branch is uncovered for the first time in a parametrically forced DEA system, which causes a sudden jump in the oscillation amplitude and serves as a severe threat to the dynamic stability and controllability of the DEA system. By using a global analysis method, the evolution of the isolated response curve against the excitation components and system physical parameters is also investigated numerically.

Dielectric elastomer actuators (DEAs) are an emerging type of soft actuator that possess several advantages over conventional actuators in terms of large actuation strain, high theoretical efficiency, inherent compliance, and low cost^{[1,2]}. An idealized DEA consists of two pieces of compliant electrodes with one layer of dielectric elastomer in between. When a voltage is applied across the electrodes, opposing charges are accumulated on the two sides of the membrane, which cause it to contract in thickness and expand in area. Many configurations of DEAs have been proposed; for instance, stacked DEAs^{[3]}, conical DEAs^{[4,5]}, rolled DEAs^{[6,7]}, and the dielectric elastomer minimum energy structures^{[8,9]}. Many robotic applications based on DEAs have been developed in soft robotic locomotion^{[10-16]}, fluidic pumps^{[17,18]}, and grippers^{[19,20]}.

Pure-shear DEAs represent one class of planar and linear DE actuation system that utilizes the active area expansion of the membrane in one planar direction as the output. Such a design has a compact configuration which is desirable for many applications in robotics such as a flexible morphing wing^{[21]}, a vibration shaker^{[22]}, and a fishlike airship^{[23]}. In order to maintain the tension on the membrane, one end of it is usually fixed while the opposing end is typically attached to a biasing mass^{[24-26]} or an equivalent force^{[27]}, a spring^{[28-30]}, an antagonistic pure-shear DEA^{[31,32]}, or a bistable mechanism^{[33,34]}.

Large actuation strains of the DEAs can reduce the system complexity and robot size (i.e., no strain amplification mechanism required) and also increase the range of motion of the robot designs. Hence, several works have focused on optimizing the quasi-static linear actuation strains of the pure-shear DEAs. Lu ^{[35]} utilized the anisotropic pre-stretch and stiff fiber constraints to achieve an output strain up to 142%. By using the same anisotropic pre-stretch principle and adding an optimal mass to the membrane, a maximum actuation strain over 500% was documented by Goh ^{[26]}.

Due to the inherent elasticity, when stimulated by an alternating current (AC) voltage signal, the DEAs can generate oscillatory motions. As the excitation frequency approaches the natural frequency of the DEA system, it can exhibit a dramatic increase in its oscillation amplitude. This soft oscillation technology can have advantages over other conventional rigid mechanical oscillators in terms of inherent compliance, which can be readily integrated with soft robotic systems to allow active vibration control and dynamic robotic locomotion^{[36]}. As a result, more works have been concerned with the nonlinear dynamics of the pure-shear DEAs.

For example, Li ^{[37]} have extensively investigated the nonlinear dynamics of pure-shear DEAs with specific focus on the effects of biasing elements, membrane strain-stiffing^{[29]}, and the viscosity-induced drifting^{[38]}. Zhang ^{[27,39]} studied the effects of dissipating factors of a pure-shear DEA (viscous damping and leakage current) on its dynamic responses. By using an antagonistic pure-shear DEA configuration, Zhang ^{[40]} also investigated the loss-of-tension due to the large deformation of the membranes during vibrations and demonstrated that the loss-of-tension can be eliminated by increasing the pre-stretch ratio.

However, although loss-of-tension in pure-shear DEAs has been demonstrated, no work has looked at the effects of loss-of-tension on the frequency responses of such actuators. For silicone elastomers, which usually adopt low pre-stretch ratios (typically between 1-2; e.g., 1.0 in^{[14]}, 1.02 in^{[15]}, or 1.2 in^{[19]}) and exhibit low viscous damping^{[18,41,42]}, loss-of-tension could be expected in high-amplitude vibrations, particularly near the resonance of the system where amplitude can be dramatically increased. Hence in this work, by adopting the antagonistic pure-shear DEA configuration, we use a numerical model to analyze the effects of the loss-of-tension mechanism on the nonlinear dynamic response of the DEA under low pre-stretch conditions. From this, we uncover the resonant responses that are detached from the main response branch - these isolated resonant responses cannot be detected by conventional continuation methods but, once triggered, they will cause a sudden change in the oscillation amplitude and response phase, which serves as a real threat to the dynamic stability and controllability of such DEA systems; hence they demand careful and dedicated investigation.

The rest of this paper is structured as follows. In section

The basic structure of the antagonistic pure-shear DEA is shown in _{1}, _{2}, and _{3} in direction 1, 2, and 3, respectively. Two identical DE membranes, I and II, are pre-stretched along direction 1 by _{p} and then bonded to the constraint frames. One mass, _{I} and _{II}. The actuation principle of the antagonistic pure-shear DEA relies on the force balance of the two membranes on the mass. When a voltage, _{I}. The force imbalance leads to a translation of the mass along direction 1 until another force balance is achieved between the two membranes, as is shown in _{I} and _{II}, are defined as the ratios of the deformed lengths in direction 1 and the initial length, _{1}.

(A) Schematic diagram of the antagonistic pure-shear DEA and its actuation principle. (B) Demonstration of loss-of-tension in the membrane induced by large deformation (

The equation of motion that describes the dynamic responses of this system has been described in detail in previous studies^{[24,26,36,40,43]}. Hence, in this work, the development of the equation of motion is introduced briefly and the final equation used for the rest of the study is given in the end of this subsection.

The two DE membranes and mass are considered separately at this stage with the subsystem membrane I being investigated first. A thermodynamic framework is adopted, which indicates that the variation of the free energy of the DE during actuation is equal to the work done jointly by the voltage, tensile force, the viscous damping, and the inertial force of the membrane.

To derive this model, it is assumed that a voltage, _{I}, is applied across the compliant electrodes of the membrane and a charge, _{I}, is built on the electrodes. A tensile force, _{I}, by conservation of energy

where W is the Helmholtz free energy density, and can be expressed using the Gent model^{[44]} as

where _{I} is the electric displacement of membrane I and is given as

Note that the Helmholtz free energy density equation given here is simplified due to there being no stretch in direction 2 on the DE membrane.

By partially differentiating _{I} and substituting to

It is worth noting that a membrane is a type of element that can only withstand tension in direction 1 by neglecting the effects of shear stress in the membrane due to the sufficiently high area-to-thickness ratio. Hence if the tensional force on the membrane in the deformation direction becomes zero, the membrane loses tension in this direction, as illustrated in _{I} < 1, the tensile force, _{I} remains > 1. However, in this study, where a low pre-stretch value is used, the deformation, particularly at high-amplitude vibrations near the resonance of the system, can be large enough to cause loss-of-tension on the DE membrane which creates non-smoothness in the force function thus discontinuity in the stiffness. To take this effect in to account, a piecewise force function is used in this work, and is written as

where _{I} is the true tensile force in membrane I.

Due to the constraint applied to the two DE membranes, the stretch of membrane II can be written as

Then the true tensile force, FII, can be obtained by substituting

where _{II}, is the voltage applied to membrane II.

The equation of motion for the mass can be obtained as

By performing nondimensionalization and neglecting the inertia force of the membranes by assuming a sufficiently large _{1}_{2}_{3}, the equation of motion can be rewritten as

where

_{t} _{I} + _{II}, against the stretch offset from its passive equilibrium (with no actuation voltage), _{I} = _{I} - _{p}. Note from _{I} or _{II} becomes zero, which corresponds to _{I} and _{II} = 1, respectively. It is noteworthy that, in the region where -(_{p} - 1) < _{I} < (_{p} - 1), the system appears to be “stiffer” (the slope is steeper) than other regions as the two membranes act on the mass antagonistically in this region. As a voltage is applied to membrane I, the discontinuity points do not necessarily correspond to _{I} and _{II} = 1, as illustrated in

Force-displacement relationship of the antagonistic pure-shear DEA. (A) Total force vs. stretch offset with pre-stretch _{p}_{I}_{II}_{I}_{p}_{II}

In this work, the Gent model parameters (^{[42,45,46]} and is given as _{1} = 20 mm, _{2} = 200 mm, and _{3} = 0.1 mm. The mass ^{-11} F/m.

By solving the nondimensionalized equation of motion,

In this section, the nonlinear dynamic responses of the pure-shear DEA under AC voltage excitation are investigated. In this study, only membrane I is subject to a time-varying voltage excitation, while no actuation voltage is applied to membrane II. The forcing term of this system, _{I}, hence this system is excited parametrically^{[47]}. It is noteworthy that the forcing term is also a function of the square of the voltage input. Therefore, for an non-dimensional AC voltage signal given as _{I} = _{DC} + _{AC} cos(_{e}_{DC} and _{AC} are the non-dimensional direct current (DC) and AC voltage amplitudes respectively and ^{[46]})

where

From ^{[42,48]}, which are termed _{a}. As a result, varying either _{AC} or _{DC} could result in a change in the biasing excitation component, which then leads to a shift in the equilibrium point, and hence the resonant behavior. To analyze the dynamic responses of the DEA system consistently, the biasing forcing term, _{a}, AC to DC amplitude ratio, _{AC}/_{DC} and the excitation frequency, _{e}, are considered as three independent variables. In this section, the frequency response of the DEA system is investigated by varying the excitation frequency while fixing the _{a} value at 0.2^{2}, and two _{b} are 0.041 and 0.053 respectively, while _{c} is changed from 0.006 to 0.013 as _{a} remains unchanged, the increase in _{P} = 1.02. It is noteworthy that, with the material parameters adopted in this work, the nominal electrical breakdown electric field is about 90 MV/m with no pre-stretch^{[49]}, while the applying nominal amplitude adopted in this study is equivalent to about 40 MV/m, which is significantly lower than the critical value. As a result, it is believed that no electric breakdown occurs within this study.

In the first study, forward and backward frequency sweeps are conducted for the

(A) Frequency sweep results of the

The frequency components and their amplitudes in the forward frequency sweeps for the _{r}, (the frequency component in the response with the highest amplitude, represented by the dark-red color) is equal to the excitation frequency, _{e}, with the exception near the first resonant peak, where _{r} = 2_{e}, demonstrating a super-harmonic response. The fundamental response frequency in the case _{r} = 0.5_{e} relationship, which corresponds to a

The frequency sweep results demonstrated that, apart from super-harmonic and harmonic resonance, which result from the square of the sinusoidal excitation function, there exists the subharmonic response that is not directly correlated to the excitation. The results also illustrated that multiple stable periodic solutions might exist at a single frequency for this dynamic system (e.g., the region between the jump-up and down points in

^{[50-54]}). Yet to the best of the authors’ knowledge, this is the first time that an isolated subharmonic response curve being demonstrated theoretically using a parametrically forced DEA system. The mechanism leading to this type of isolated curve has not been fully understood, but it is believed to a result of the multi-valuedness of the nonlinear system at some frequencies^{[50,51]}. The oscillation amplitude of the isola can be over 10-fold higher than the primary response at the same frequency, once triggered, it will cause a sudden jump in the oscillation amplitude, which serves as a real threat to the dynamic stability and controllability of such DEA systems. In practical applications, this could lead to a system failure and catastrophic outcomes.

Frequency response, time histories and phase paths at _{e} = 1.2, _{e} = 1.8, and _{e} = 3.7 for

Note from _{e} = 1.2, _{e} = 1.8, and _{e} = 3.7 are also plotted in _{e} = 1.2, for both high and low amplitude solutions, the mass oscillates twice as the excitation signal repeats itself once, demonstrating a clear 2:1 super-harmonic behavior (i.e., _{r}:_{e} = 2:1). At _{e} = 1.8 where the primary resonance occurs, the high amplitude solution lags behind the excitation signal by 90 degrees, while the low-amplitude one is approximately in-phase with the excitation. At _{e} = 3.7, two types of periodic solutions exist, where the high amplitude solution corresponds to a 1:2 subharmonic response (i.e., the mass oscillates once in every two excitation cycles). The low amplitude solution shows a 1:1 anti-phase response with the excitation signal.

The same frequency response analysis is conducted for the ^{[52]}, in other words, the unstable solution between the two stable solutions lowers and eventually merges with the 1:1 harmonic branch as

Frequency response, time histories and phase paths of the three stable periodic solutions at _{e}

It is noteworthy that, in the _{e} ≈ 3.65 to 3.75. Note that, at _{e} = 3.7, there exist three stable periodic solutions, and the time history and phase path of each solution are plotted in

In the previous study, multiple stable periodic solutions of the DEA system at some excitation frequencies have been demonstrated for both ^{[55]} is adopted by dividing the initial condition region of interest into discrete points and numerically computing the steady-state response that each initial condition leads to. By plotting the initial conditions and the corresponding steady-state response,

Two cases are considered in this basin of attraction study: (1) _{e} = 3.7; and (2) _{e} = 3.6. In the first case, there are two stable periodic responses (i.e., a 1:1 harmonic and 1:2 subharmonic) where the latter one is on the isola, as is demonstrated in

The basins of attraction for the two cases are plotted in

Basins of attraction for a parametrically forced DEA system with (A) _{e} = 3.7 and (B) _{e} = 3.6. The black, white, and red regions represent the stable periodic solutions with the 1:1 harmonic, 1:2 subharmonic, and 1:3 subharmonic responses, respectively.

In the last section, the frequency response of the DEA system with _{a}), the damping, and the pre-stretch ratio of the DE membrane on the frequency response of the system. Particular attention is given to the evolution of the isola and its interaction with the main branch.

In this subsection, the effects of the excitation parameters (_{a}) on the frequency response of the system are studied. First, the _{a} as 0.2^{2}, and the results are presented in _{p} = 1.02 and the non-dimensional damping coefficient is _{I} in a steady-state periodic oscillation cycle.

Effects of _{p}_{a}

As the _{b} and super-harmonic forcing amplitude _{c} in

In the next study, the _{a} and the results are plotted in _{a} value increases both the AC and DC components of the excitation simultaneously, which causes higher _{b} and _{c} forcing components; this leads to increases in the amplitudes of both the harmonic and super-harmonic responses. The 1:2 subharmonic response also emerges as _{a} increases, first as an isola, before attaching to the main branch as _{a} rises further. Note that, as _{a} becomes sufficiently high, the displacement from the equilibrium to one end could be greater than 1, which violates the physical constraints of this system (i.e., the mass will collide with the constraint frames). For the case with _{p} = 1.02, _{a} should be less than 0.25^{2} to avoid collision.

Effects of Ea value on the frequency response of the DEA system. _{p}

In this subsection, the effects of a physical parameter, i.e., damping coefficient _{a}, and _{p} values are set as 0.6, 0.2^{2}, and 1.02, respectively, while

The frequency response of the DEA systems with different damping coefficients are shown in ^{[15,16]}. It should be noted that silicone DEA systems require comprehensive dynamic analyses to ensure their dynamic behavior is fully understood due to their low viscous damping.

Effects of damping coefficients on the frequency response of the DEA system. _{p}_{a}

Apart from the damping coefficient, another important physical parameter for the DEA system is the pre-stretch ratio, _{p}. In this study, the pre-stretch ratio is varied from 1.01 to 1.06, while the other parameters are fixed at _{a} = 0.2^{2}, and _{p} is increased to 1.02. It is noteworthy that increasing _{p} leads to a reduction in the size of the subharmonic response curve, such that the peak amplitude reduces and the bottom of the curve separates further from the main branch. At _{p} = 1.06, the subharmonic curve disappears under the current forcing conditions.

Effects of pre-stretch ratio on the frequency response of the DEA system. _{a} = 0.22, and

The study of the effects of the damping coefficients and the pre-stretch ratios demonstrate that, under the same forcing conditions, by increasing either the damping coefficients or the pre-stretch ratios, the subharmonic response is suppressed and eventually vanishes. However, it cannot be concluded that subharmonic responses, either in isolated or attached form, do not exist in the parametrically forced DEA system when the physical parameters reach those values. For insistence, as is demonstrated in the last subplot in _{p} = 1.02) and (_{p} = 1.06) cases, no subharmonic response is observed with the forcing condition of _{a} = 0.2^{2}. However, it is found that, by increasing the forcing amplitude, a subharmonic response can be triggered again, as is demonstrated in

Examples showing that, by increasing the forcing amplitude, the vanished subharmonic response can be triggered again. (A) _{p}_{p}_{a} is increased from 0.22 to 0.252.

The structure of the subharmonic response exhibits a complex interaction between the two physical parameters (_{p}) and the forcing parameters (_{a}). Note that, by using the control variate technique, the effects of each parameter may be analyzed separately. However, the subharmonic response of this system is due to the interaction between different parameters. As shown in _{p} reaches 1.2, even though the forcing amplitude is higher (_{a} = 0.3^{2}), there is no subharmonic response. Hence a future work that fully investigates the five-dimensional space (_{e}, _{a}, _{p}) could enhance the understanding of the isolated resonance responses in the DEA system.

Frequency response of the DEA system with _{p}_{a} = 0.32 are adopted in this case.

This paper has investigated the nonlinear dynamic behavior of an antagonistic pure-shear DEA with a low pre-stretch ratio. Under high-amplitude oscillations, such a configuration can lead to the loss-of-tension of the DE membranes, which induces the discontinuity in the force-displacement function. By simulating the effect of initial conditions, the isolated subharmonic resonant responses have been uncovered; these behaviors cannot be captured using conventional continuation methods (e.g., frequency sweeps or steps). The main contributions of this work are: (1) reporting isolated subharmonic response curves in a DEA system for the first time; (2) developing a comprehensive method for the analysis of isolas in DEA systems, including global analysis and an electro-mechanically coupled DEA dynamics model; and (3) demonstrating the correlation between the system and forcing parameters and the evolution of isolas. The key findings of this paper are listed as follow:

The pure-shear DEA exhibits multiple types of frequency response, including 2:1 super-harmonics, 1:1 harmonics, 1:2 subharmonics, and 1:3 subharmonics.

The 2:1 super-harmonic and 1:1 harmonic responses are the result of the square of the excitation function.

Subharmonic resonant responses do not occur naturally from the forcing function, instead, it is a result of the complex interaction between the forcing and response.

Increasing the forcing amplitude first leads to the growth of the subharmonic response curve outside the main frequency response curve, before causing it to attach to the main branch.

Under the same forcing condition, increasing the damping coefficient and pre-stretch ratio could reduce the size of the subharmonic response curve until it vanishes.

Isolated response curves have received increasing attention in the nonlinear dynamics community as they often represent high amplitude responses which could lead to catastrophic outcomes^{[56]}. For DEA dynamic systems, isolas can be easily overlooked by using continuation methods and in controlled environments with no perturbations, such as static lab conditions. However, in practical applications where the environment is dynamically changing, adequate perturbations could trigger isolas hidden in DEA systems, causing the response to jump to a high amplitude, which then leads to loss of control of the system. As a result, a careful selection of the system parameters and a comprehensive study of the dynamic responses are essential to avoid such outcomes in resonating DEA applications where dynamic stability and controllability are crucial. In the meantime, as highlighted in^{[50,51]}, the isolas demonstrated in the antagonistic pure-shear DEA system could also be exploited, for instance, in highly versatile dynamic mechanical oscillation systems with multiple resonant modes and amplitudes. Such oscillating systems could have practical applications in highly programmable robotic locomotion where the multiple resonant modes and amplitudes can serve as means for steering and speed regulation. The initial condition sensitivity of the isolas within these systems can be utilized as a passive collision controller for robots. For example, when a robot collides with other objects, the sudden change in system states can trigger isolas which increases the oscillation amplitude of the actuator, thus allowing for rapid escape of the robot. These systems can also be used in smart energy harvesting where the isolas can be actively triggered to increase bandwidth and efficiency of energy conversion.

Made substantial contributions to conception: Cao C, Hill TL, Gao X

Design of the study: Cao C, Chen G, Li B, Gao X

Data analysis and interpretation: Cao C, Hill TL, Wang L, Gao X

Data acquisition: Cao C

Administrative, technical, and material support: Cao C, Hill TL, Wang L, Gao X

Manuscript writing: Cao C, Hill TL, Li B, Chen G, Wang L, Gao X

Not applicable.

This work was supported in part by the National Key Research and Development Program of China under Grant 2019YFB1311600; in part by the National Natural Science Foundation of China, under Grant 62003333; in part by the Guangdong Basic and Applied Basic Research Foundation 2020A1515110175; and in part by the Shenzhen Fundamental Research Project JCYJ20200109115639654.

All authors declared that there are no conflicts of interest.

Not applicable.

Not applicable.

© The Author(s) 2021.