You are on page 1of 8

Appl Phys A (2009) 94: 507–514

DOI 10.1007/s00339-008-4946-8

I N V I T E D PA P E R

Finite element modelling of dielectric elastomer minimum energy


structures
Benjamin O’Brien · Thomas McKay · Emilio Calius ·
Shane Xie · Iain Anderson

Received: 13 August 2008 / Accepted: 13 October 2008 / Published online: 6 November 2008
© Springer-Verlag 2008

Abstract This paper presents an experimentally validated PACS 46.32.+x · 77.65.-j · 83.80.Va
finite element model suitable for simulating the quasi-static
behaviour of Dielectric Elastomer Minimum Energy Struc-
ture(s) (DEMES). A DEMES consists of a pre-stretched Di- 1 Introduction
electric Elastomer Actuator (DEA) adhered to a thin, flex-
ible frame. The tension in the stretched membrane causes Introduced in their modern form by Pelrine et al., Dielectric
the frame to curl up, and when a voltage is applied, the Elastomer Actuator(s) (DEA) consist of compliant capaci-
frame returns to its initial planar state thus forming a use- tors formed by elastic dielectrics sandwiched between flex-
ful bending actuator. The simulation method presented here ible electrodes. When a large electric field is applied across
incorporates a novel strain energy function suitable for sim- the dielectric, in-plane charge repulsion and across dielec-
ulating general DEA actuator elements. When compared tric charge attraction give rise to expansion and contraction
against blocked force data from our previous work, the new of the dielectric, respectively (Fig. 1) [1–3]. The electrostatic
model provides a good fit with an order of magnitude reduc- energy density Uel of a DEA is given in (1), where ε0 is the
tion in computational time. Furthermore, the model accu- absolute dielectric permittivity, εr is the dielectric constant
rately matched experimental data on the free displacement or real part of the relative permittivity, and E is the elec-
of DEMES formed with non-equibiaxially pre-stretched tric field strength in the membrane. Due to the incompress-
VHB4905 membranes driven by 2500 V. Non-equibiaxially ibility of DEA membranes, dielectric isotropy, and compli-
pre-stretching the membranes allowed control of effective ance of the electrodes this is often written as an equivalent
frame stiffness and bending moment, this was exploited by Maxwell pressure P max acting in the thickness direction and
using the model to optimise stroke at 2500 V in a hypothet- given by (2) [4]. DEA have earned the moniker “Artificial
ical case study. Dielectric constant measurements for non- Muscle” due to their excellent general performance char-
equibiaxially stretched VHB4905 are also presented. acteristics, such as high energy density, large stresses and
strains, kHz bandwidth, light weight, silent operation, high
efficiency, and the ability to operate in sensor and generator
B. O’Brien () · T. McKay · I. Anderson
The Auckland Bioengineering Institute, University of Auckland,
modes [5–9].
Level 6, 70 Symonds Street, Auckland, New Zealand
ε0 εr E 2
e-mail: ben.obrien@auckland.ac.nz Uel = (1)
Fax: +64-9-3737599 2
E. Calius Pmax = ε0 εr E 2 (2)
Industrial Research Limited, Brooke House, 24 Balfour Road,
P.O. Box 2225, Auckland 1140, New Zealand Arrays of bending actuators, such as cilia, are ubiquitous in
nature due to their enormous success as systems for manip-
S. Xie ulation, pumping and locomotion. Until recently there has
Department of Mechanical Engineering, School of Engineering,
The University of Auckland, Private Bag 92019, Auckland Mail been no way to utilise DEA in the form of artificial cilia suit-
Centre, Auckland 1142, New Zealand able for these tasks. Introduced by Kofod et al., Dielectric
508 B. O’Brien et al.

present a Finite Element Analysis (FEA) based approach


which recreated the blocked force produced by a range of
DEMES designs [13]. However, it was unable to accurately
predict free displacement as the Maxwell pressure was sim-
Fig. 1 DEA Actuation Principle. From left to right: A DEA consists of ulated as a global average, rather than locally in the DEA.
a compliant capacitor. When a high voltage is applied, the capacitor is The current paper proposes a methodology to improve the
charged. Electrostatic repulsion and attraction give rise to elastomeric Maxwell pressure simulation in a FE model and thus more
expansion accurately capture both the blocked force and free displace-
ment of DEMES. The results are experimentally validated
and the method is computationally cheaper than the ap-
proach presented in [13].

Fig. 2 Stages in the DEMES erection process. From left to right ten- 2 Materials and methods
sion in the membrane causes the frame to curl up

The DEMES presented here for experimental validation


used VHB 4905 for the membrane as it is a high performing
and readily available DEA material. Nyogel 756G carbon
grease was used for the electrodes, and 100 micron thick
laser-cut PET for frames. Frame geometry is presented in
Fig. 3 A DEMES under 2500 V sinusoidal actuation Fig. 4. The bases were held flat with laser-cut brackets and
the electrodes were addressed with aluminium foil intercon-
Elastomer Minimum Energy Structures (DEMES) provide nects.
an elegant solution to this problem. DEMES consist of pre- The membranes were pre-stretched to two different non-
stretched dielectric elastomer membranes adhered to flexi- equibiaxial stretch states of 2:4 and 3:5, where X:Y refers to
ble frames [10–12]. The tension in the membrane causes the a stretch ratio of X in the x-direction and Y in the y-direction
frame to bend until the bending energy in the frame is equal (Fig. 4). The stretch ratio λ of a material is given by the ratio
to the strain energy in the DEA and the structure is at equi- of the current length of the material in one direction divided
librium (Fig. 2). When the DEA is actuated electrical energy by its original length (3), where λ is the stretch ratio, l is
is added to the system and thus the structure moves (Fig. 3). the current length, and l0 is the initial length. Six DEMES
DEMES have a number of advantages as bending actu- in each stretch state were built to help quantify experimental
ators: They provide lightweight, 3D movement and can be scatter.
fabricated from flat sheets of material. They operate with the
l
membrane in a pre-stretched state, synergistic with the ben- λ= (3)
l0
efits of pre-stretching which has been shown to improve the
dielectric breakdown strength of membrane materials [2]. DEMES were powered with an Ultravolt High Power “C”
They are, in general, stronger than bilayer bending actua- Series high voltage power pack to 2500 V for blocked force
tors, as the membrane acts to bend the frame from a greater
distance, resulting in a stronger bending moment.
Before DEMES can be used as artificial cilia it is desir-
able to create models for use in actuator design and con-
troller optimisation. This incorporates challenges associated
with nonlinear electrostatics, hyper-viscoelastic membrane
behaviour, high aspect ratios, limited axes of symmetry, geo-
metric nonlinearities, and high sensitivity to boundary con-
ditions and quality of fabrication. Due to these issues exper-
imentally intensive manual design of DEMES is problem-
atic, and modelling should thus provide an essential tool in
DEMES design.
Kofod et al. presented a simple analytical model of
DEMES in [12] which is useful for developing an under-
standing of DEMES operation, but of limited use for sim-
ulating real devices. In a previous paper [13], the authors Fig. 4 DEMES geometry, frame is 100 µm thick
Finite element modelling of dielectric elastomer minimum energy structures 509

density, AB is the strain energy density due to stretch in the


material, and ES is the electrostatic energy density. This ap-
proach is similar to that taken by Goulborne et al. where
the Cauchy stress was expressed as the sum of the mater-
ial stress and the Maxwell stress in the material [14]. It is
different in that we use an expression for strain energy den-
Fig. 5 Calculating the free displacement of the DEMES. A triangle is sity, which is therefore applicable to membrane elements in
formed by the three red markers; from this the actuator angle can be ABAQUS,
calculated
U = AB + ES (4)
and 3000 V for free displacement. Blocked force was de-
termined using a copper cantilever with a laser displace- Before discussing the SEF, the following definitions must be
made: In a membrane λ1 and λ2 are the planar stretch direc-
ment transducer tracking its tip position. When DEMES
tions, and λ3 is the thickness stretch. VHB can be treated as
push against the cantilever the small deflections were related
incompressible so that (5) holds. The first and second prin-
to force through the stiffness of the cantilever [13]. Free
ciple stretch invariants are given as I¯1 and I¯2 , respectively,
displacement was measured photographically by attaching
in (6) and (7).
a coloured marker to the tip of the DEMES and recording
displacement with National Instruments Labview 7.1. The 1
voltage was applied in a square wave between zero and an λ1 λ2 λ3 = 1, λ23 = (5)
λ21 λ22
incrementally increasing value with a period of ten seconds.
Photos were taken just before every voltage change, giv- I¯1 = λ21 + λ22 + λ23 (6)
ing the DEMES five seconds to settle towards a final po- 1 1 1
sition. Stroke was defined as change in opening angle θ of I¯2 = 2 + 2 + 2 (7)
λ1 λ2 λ3
the DEMES as depicted in Fig. 5.
To the best of the authors’ knowledge, there is no data in The SEF was implemented into ABAQUS using the sub-
the literature on the dielectric constant of non-equibiaxial routine UHYPER which describes hyperelastic materials
stretch states in VHB. To successfully simulate the pre- based on the first and second stretch invariants. Voltage was
sented DEMES, this dielectric constant was measured for controlled in the simulation using the subroutine UDFLD,
several stretch states. Capacitors of known geometry were which is time dependent. UDFLD also provides a mecha-
formed by pre-stretching VHB and adhering the resultant nism for accessing the thickness of the membrane elements.
membrane to perspex frames. Thin and uniform gold elec- Wissler and Mazza’s model of VHB based on the Arruda–
trodes were then sputter-coated onto the membrane surface. Boyce strain energy and Prony series relaxation functions
The capacitance was measured with a Fluke PM6306 pro- was used to describe the material stretch strain energy den-
grammable automatic RCL meter and the dielectric permit- sity, AB [15, 16]. A scaling factor k0 was applied to the
tivity calculated using the standard capacitor equation. Four Arruda–Boyce stiffness coefficient to allow for apparent
samples of each stretch state were used with 1:1 as a ref- over-stiffness after long term relaxation. A k0 value of 0.86
erence and three non-equibiaxial cases of 2:4, 2.5:5.5, and was found empirically to be a good match. Equation (1) was
3:5. used to describe the electrostatic energy density ES.
The FEA package ABAQUS was used to simulate In ABAQUS, the Prony series relaxation function oper-
DEMES erection and actuation. Membrane and shell ele- ates directly on the SEF. For this reason a pre-multiplier ginv
ments were used to describe the DEA and frame, respec- corresponding to the inverse of the Prony series relaxation
tively. Shell elements are used in cases where the thickness ratio G was applied to the electrostatic term. It was difficult
dimension is much smaller than the in-plane dimensions, to find the true value of G inside the UHYPER subroutine.
allowing for simplified element formulation. Membrane el- To allow for this a constant value corresponding to the to-
ements are a subset of shell elements with the additional tal long term value of G was used. This limited the model
properties of zero bending stiffness and zero stress in the to quasi-static rates of applied voltage and loads. The SEF
thickness direction. is given in (8), the electrostatic energy variable K is given
Typically, DEA actuation is simulated by applying a in (9), and the Prony series in (10).
mechanical Maxwell pressure on either face of the mem-
5
Ci  ¯i  K
brane (2); however, this approach does not work with mem- U = AB + ES = μ I − 3i + 2 (8)
2i−2 1 λ3
brane elements as they are designed to simulate a state of λ
i=1 m
plane stress. To overcome this limitation, the electrostatic
energy density was implemented directly into the Strain En- ε0 εr V 2
K = ginv (9)
ergy Function (SEF) (4), where U is the total strain energy 2t02
510 B. O’Brien et al.

n −t series which has up to n terms. For implementation into
G=1− gi e τi (10)
UHYPER, a variety of partial derivatives were required from
the model, namely: ∂U/∂ I¯1 , ∂U/∂ I¯2 , ∂ 2 U/∂ I¯12 , ∂ 2 U/∂ I¯22 ,
i=1

In these equations, V is the voltage applied across the mem- and ∂ 2 U/∂ I¯1 ∂ I¯2 . This created a challenge because U is de-
brane in volts, t0 is the material thickness in meters before pendent on λ3 , and not just on I¯1 and I¯2 . To address this
pre-stretching, and ε0 and εr are the absolute and relative di- we rewrote U in terms of I¯1 and I¯2 giving (11) which is
electric constants, respectively. C1−5 are constants found in the characteristic equation derived in Appendix. From (11)
the Arruda–Boyce strain energy function, and λm and μ are the required partial derivatives can be found (12)–(18).
material properties. gi and τi are parameters for the Prony

  3   2   
U − AB I¯1 − I¯2 K U − AB I¯1 + I¯1 K 2 U − AB I¯1 − K 3 = 0 (11)
∂U K 2 (AB − U ) ∂AB
= + (12)
¯ ¯ ¯
∂ I1 (3(U − AB) − 2I2 K(U − AB) + I1 K )
2 2 ∂ I¯1
∂U K(U − AB)2
= (13)
∂ I¯2 (3(U − AB)2 − 2I¯2 K(U − AB) + I¯1 K 2 )

∂ 2U ( ∂∂U
I¯1
− ∂AB
∂ I¯1
)2 (2I¯2 K − 6(U − AB)) + 2K 2 ( ∂AB
∂ I¯1
− ∂U
∂ I¯1
) ∂ 2 AB
= + (14)
∂ I¯12 (3(U − AB)2 − 2I¯2 K(U − AB) + I¯1 K 2 ) ∂ I¯12

∂ 2U 4K(U − AB) ∂∂U


I¯2
− (6(U − AB) − 2I¯2 K)( ∂∂U
I¯2
)2
= (15)
∂ I¯22 (3(U − AB)2 − 2I¯2 K(U − AB) + I¯1 K 2 )

∂ 2U 2K(U − AB)( ∂∂UI¯1


− ∂AB
∂ I¯1
) − K 2 ∂∂U
I¯2
− (6(U − AB) − 2I¯2 K) ∂∂U ( ∂U −
I¯2 ∂ I¯1
∂AB
∂ I¯1
)
= (16)
∂ I¯2 ∂ I¯1 (3(U − AB)2 − 2I¯2 K(U − AB) + I¯1 K 2 )
 
∂AB(I¯1 ) 2C2 I¯1 3C3 I¯12 4C4 I¯13 5C5 I¯14
= μ C1 + + + + (17)
∂ I¯1 λ2m λ4m λ6m λ8m
 
∂ 2 AB(I¯1 ) 2C2 6C3 I¯1 12C4 I¯12 20C5 I¯13
=μ + 4 + + (18)
∂ I¯12 λ2m λm λ6m λ8m

The FEA model was defined as shown in Fig. 4 with an try was fixed in the width direction degree of freedom, and
additional cut down its central line of symmetry (Fig. 6). The in all rotational degrees of freedom for the frame.
membrane was connected to the inside edge of the frame The initial stress in the membrane was found by instan-
and the model thus ignored the membrane material bonded taneously stretching to the required stretch state a square of
to the frame, which had a negligible influence on bending material in a separate simulation. The initial stress was then
stiffness, and assumed a perfect bond. imported into the DEMES model and the membrane was al-
The membrane was formed with eight node quadrilateral lowed to relax for 2000 seconds whilst being held flat. Bias
quadratic reduced integration membrane elements (M3D8R) forces were applied to the frame to ensure that it found the
and the frame with eight node doubly curved quadratic re- correct equilibrium state. This is important as DEMES are
duced integration thick shell elements (S8R). The membrane poly-stable systems and this step is also required in the fab-
used the UHYPER material law developed in this paper and rication of the real device. Once the shape was in the region
the frame used a linear elastic law. The base of the DEMES of the desired equilibrium state the bias forces were released
was fixed in all degrees of freedom, and the line of symme- and the structure was allowed to relax over 2000 seconds
into its final position. At this point a boundary condition at
Fig. 6 Model geometry. The the tip was applied when blocked force data was desired,
membrane is connected to the
or the tip was left to be free to move if free displacement
inner surface of the frame with a
perfect bond, and the model is data was desired. The UDFLD subroutine in ABAQUS was
cut down its line of symmetry used to ramp the voltage from 0 to 2500 or 3000 V over
Finite element modelling of dielectric elastomer minimum energy structures 511
Table 1 Dielectric constant data

Stretch state Average Standard deviation

1:1 4.13 0.11


2:4 4.18 0.15
2.5:5.5 4.1 0.12
3:5 3.76 0.11

Table 2 DEMES equilibrium position, model versus experiment

Stretch state Model angle Experimental


Mean angle Standard deviation
Fig. 7 Blocked force versus voltage for 2:4 DEMES
2:4 75.81 79.01 3.27
3:5 74.17 75.71 2.49

300 seconds and the response of the model was recorded.


Finally, the data was extracted using the standard ABAQUS
post processing tools and exported to Excel for comparison
with experiment.

3 Results

The measured VHB dielectric constants are presented in Ta-


ble 1. The 2:4 and 2.5:5.5 stretch states did not significantly Fig. 8 Blocked force versus voltage for 3:5 DEMES
reduce the dielectric constant of the membrane; however, a
stretch state of 3:5 did. Further work to study dielectric con-
stant behaviour in VHB for non-equibiaxial stretch states
should be carried out to improve the accuracy of DEMES
simulation for device design.
The equilibrium position (θ in Fig. 5) of the DEMES was
compared between the model and experimental data. The
results are presented in Table 2. The model agreed well, with
a predicted equilibrium position falling within one standard
deviation of the experimental mean for both stretch states.
The model was used to calculate the blocked force for
two of the three DEMES membrane stretch states presented
in [13]. The ‘new’ model had a comparable match with the
‘old’ model presented in [13] for the 2:4 membrane stretch Fig. 9 Free displacement versus nominal electric field strength for 2:4
and 3:5 DEMES
state and a better match than the 3:5 membrane stretch
state as can be seen in Figs. 7 and 8. Note that, to ensure
good comparison with the old model, dielectric constants case, and the 3:5 case up to around 2500 V, where it began
from [13] were used. The old model took 97 minutes to sim- to diverge.
ulate blocked force and the new model took approximately
6 minutes for the same simulation.
Free displacement data for two different DEMES stretch 4 Case study
states, 2:4 and 3:5, were gathered from twelve DEMES, six
in each stretch state. All experimental data is presented in Once the model had been validated by comparison with the
Fig. 9 to show experimental scatter. In Figs. 10 and 11, the test results, it was used to find how the free displacement de-
model is compared with the experimental data for the 2:4 pends on the longitudinal stretch ratio at a fixed voltage of
stretch state and 3:5 stretch state cases, respectively. The 2500 V. To investigate this we chose a width stretch of 5.5
model matched the experimental results well for the 2:4 and length stretch ratios ranging from 1.5 to 5 in 0.5 incre-
512 B. O’Brien et al.

Fig. 12 Maximum stroke for Y:5.5, Y ranges from 1.5 to 5


Fig. 10 Free displacement versus voltage for 2:4 DEMES, comparing
model with experiment

Fig. 13 Equilibrium angle for Y:5.5, Y ranges from 1.5 to 5

Fig. 11 Free displacement versus voltage for 3:5 DEMES, comparing


model with experiment

ments and used the model to solve for the maximum stroke
in each case. A width stretch of 5.5 provides a reduction in
thickness which increases the applied electric field and the
breakdown strength of the membrane. For simplicity the di-
electric constant was assumed to remain fixed at εr = 4.1.
Fig. 14 DEMES side views showing the effect of different membrane
As can be seen in Fig. 12, there is a clear optimum stretch
stretch ratios (1.5:5.5 and 4.0:5.5) on the initial equilibrium state. The
state of 3:5.5 at 2500 V. This optimum produces over 50% apparent sharpness of the tip is an artifact of the view angle
more stroke than the highest and almost equibiaxial stretch
state 5:5.5, and six times the stroke of the most anisotropic
stretch state 1.5:5.5. Whilst the prediction will not be perfect The connection involved twisting of the frame and its ef-
due to the use of a fixed dielectric constant, the model has fect on bending stiffness. As the longitudinal stretch ratio in-
identified the approximate design space to be pursued with creased from 1.5, the membrane shifted from a planar form
experimentation. to one resembling a saddle. In the saddle form, the angle
The equilibrium angle (θ in Fig. 5) showed large vari- of membrane attachment to the frame combined with the
ations as a function of stretch ratio (Fig. 13). It was non- tension in the stretched membrane to produce a torque that
intuitive that a low lengthwise stretch of 1.5 could lead to twisted the frame, increasing its bending stiffness as shown
such a low equilibrium angle, comparable to that exhibited in Fig. 15. An increase in bending stiffness led to an increase
by a lengthwise stretch of 4, as shown in Fig. 14. Equally in equilibrium angle (less curl) which explains the trend de-
unexpected was the trend of increasing equilibrium angle picted in Fig. 13. With a stretch state of around 3:5.5 a tip-
with increasing lengthwise stretch exhibited in the 1.5 to 3 ping point was observed where the increasing bending mo-
cases. Comparing Figs. 12 and 13, it was clear that this non- ment overwhelmed the increase in bending stiffness. The net
intuitive behaviour was connected to the existence of a peak result was that the equilibrium angle started to decrease with
in the free displacement. increasing length tension past this point.
Finite element modelling of dielectric elastomer minimum energy structures 513

and the ability to simulate Maxwell pressure in membrane


elements, which are better suited to DEMES modelling.

6 Conclusions
Fig. 15 A schematic of the cross-section depicted in Fig. 14. As the
length stretch ratio increases, it twists the frame leading to an increase A new finite element modelling approach suitable for sim-
in bending stiffness and a decrease in curl
ulating DEA membranes in DEMES has been presented.
The models are an order of magnitude faster than the pre-
5 Discussion vious attempt; more accurate; and can be used to simulate
the equilibrium state, free displacement, and blocked force
The equilibrium position predicted by the model (Table 2) of DEMES. These features render the model suitable for use
fell within one standard deviation of the mean for both the in a design environment.
2:4 and 3:5 stretch states. This was a significant achieve- The approach is limited to the quasi-static case by the
ment and important metric as the behaviour of DEMES is way viscoelasticity is incorporated into the material strain
highly sensitive to equilibrium position. The model provided energy function. This could be overcome with implementa-
a good fit for the blocked force data, falling inside one stan- tion in an ABAQUS UMAT subroutine. The mathematics re-
dard deviation of the mean for both the 2:4 and 3:5 stretch quired to fully implement the model in ABAQUS have been
states at the higher voltages in the experiment, and falling presented in the hope that other researchers may find them
just outside at lower voltages (Figs. 7 and 8). The overes- useful. The results presented incorporate the influence of the
timation at low voltages was possibly caused by the quasi- non-equibiaxial stretch state on the dielectric constant.
steady state assumption, as the relaxation function may not Use of non-equibiaxial pre-stretching for optimisation of
have sufficient time to decay, leading to an overestimation DEMES behaviour was demonstrated. The validated model
of electrostatic energy density. was shown to be a powerful tool in this process through
The free displacement data was evenly distributed with- its use in maximising free displacement by manipulating
out any obvious clumping (Fig. 9) and a larger spread in the DEMES stretch state for a 2500 V applied voltage.
2:4 stretch states. A significant component of the experimen-
Acknowledgements The authors thank: The New Zealand Tertiary
tal uncertainty was caused by the photographic and image
Education Commission Bright Futures Scheme, The Auckland Bio-
processing experimental setup, and this was larger in the 2:4 engineering Institute, Industrial Research Limited, The Vice Chan-
cases which undergo smaller strokes. cellors University Development Fund, Todd Gisby, Scott Walbran,
The model predicted the free displacement to within one Pete Blythe, Michael Byrne, Cormac Flynn, Michelle Farrell, Mary
O’Brien, and Maurice O’Brien.
standard deviation for up to approximately 2500 V (Figs. 10
and 11). Above this voltage the model prediction and ex-
perimental results diverged for the 3:5 stretch states. This
Appendix
error may be due to increased sensitivity of the angle to
changes in system energy at high strokes. A preliminary in- Derivation of the strain energy function characteristic equa-
vestigation using the model to test this theory was incon- tion:
clusive and further investigation is beyond the scope of this Sub (5) into (8) and rearrange
paper. However, the study did predict the greatest sensitivity
for a 3:5 DEMES at around 90° which is consistent with U − AB(I¯1 ) 1
= 2 = λ21 λ22 (19)
observed behaviour. A second option is that the laser-cut K λ3
DEMES frames have a thicker melt zone around the edge
of the frame. This zone will affect the bending stiffness of Sub (5) into (6)
the frame and is potentially the source of the error at higher 1
strokes, where the moment arm of the free membrane is I¯1 = λ21 + λ22 + (20)
λ21 λ22
lower. Despite this divergence the model should still pro-
vide a useful tool for optimising the quasi-static behaviour Sub (5) into (7)
of DEMES.
1 1
Use of membrane and shell elements with 8 nodes and I¯2 = 2 + 2 + λ21 λ22 (21)
built-in electromechanical coupling instead of the 20-node λ1 λ2
solid elements used in previous models [13] reduced the Sub (20) into (21)
computational effort required to run the simulation by 94%.
1
The new model is also more accurate due to simulation of I¯2 λ21 λ22 = I¯1 − + λ41 λ42 (22)
Maxwell pressure corresponding to local element thickness λ21 λ22
514 B. O’Brien et al.

Sub (19) into (22) Polymer Actuators and Devices (EAPAD), San Diego, California,
USA. Proc. SPIE (2006)
U − AB(I¯1 ) 7. L.A. Toth, A.A. Goldenberg, Control system design for a dielec-
I¯2 tric elastomer actuator: the sensory subsystem, in Smart Structures
K and Materials 2002: Electroactive Polymer Actuators and Devices
 
K U − AB(I¯1 ) 2 (EAPAD). Proc. SPIE (2002)
= I¯1 − + (23) 8. R. Pelrine, R.D. Kornbluh, J. Eckerle, P. Jeuck, S. Oh, Q. Pei, S.
U − AB(I¯1 ) K Stanford, Dielectric elastomers: generator mode fundamentals and
applications, in Smart Structures and Materials 2001: Electroac-
Rearrange (23) tive Polymer Actuators and Devices. Proc. SPIE (2001)
  3   2 9. R. Pelrine, R. Kornbluh, Q. Pei, S. Stanford, S. Oh, J. Eckerle,
U − AB I¯1 − I¯2 K U − AB I¯1 R. Full, M. Rosenthal, K. Meijer, Dielectric elastomer artificial
   muscle actuators: toward biomimetic motion, in Smart Structures
+ I¯1 K 2 U − AB I¯1 − K 3 = 0 (24) and Materials 2002: Electroactive Polymer Actuators and Devices
(EAPAD), San Diego, California, USA. Proc. SPIE (2002)
10. G. Kofod, M. Paajanen, S. Bauer, Self-organized minimum-
energy structures for dielectric elastomer actuators. Appl. Phys.
References A: Mater. Sci. Process. 85, 141–149 (2006)
11. G. Kofod, M. Paajanen, S. Bauer, New design concept for di-
1. R.E. Pelrine, R.D. Kornbluh, J.P. Joseph, Electrostriction of poly- electric elastomer actuators, in Smart Structures and Materials
mer dielectrics with compliant electrodes as a means of actuation. 2006: Electroactive Polymer Actuators and Devices (EAPAD), San
Sens. Actuators A 64, 77–85 (1998) Diego, California, USA. Proc. SPIE (2006)
2. R. Pelrine, R. Kornbluh, Q. Pei, J. Joseph, High-speed electrically 12. G. Kofod, W. Wirges, M. Paajanen, S. Bauer, Energy minimization
actuated elastomers with strain greater than 100%. Science 287, for self-organized structure formation and actuation. Appl. Phys.
836–839 (2000) Lett. 90 (2007)
3. R. Pelrine, R. Kornbluh, J. Joseph, R. Heydt, Q. Pei, S. Chiba, 13. B. O’Brien, E. Calius, S. Xie, I. Anderson, An experimentally val-
High-field deformation of elastomeric dielectrics for actuators. idated model of a dielectric elastomer bending actuator, in Smart
Mater. Sci. Eng. C 11, 89–100 (2000) Structures and Materials & Nondestructive Evaluation and Health
4. G. Kofod, Dielectric elastomer actuators. Department of Chem- Monitoring 2008, San Diego, California, USA. Proc. SPIE (2008)
istry, The Technical University of Denmark (2001) 14. N. Goulbourne, E. Mockensturm, M. Frecker, A nonlinear model
5. J.D.W. Madden, N.A. Vandesteeg, P.A. Anquetil, P.G.A. Mad- for dielectric elastomer membranes. J. Appl. Mech. 72, 899–906
den, A. Takshi, R.Z. Pytel, S.R. Lafontaine, P.A. Wieringa, I.W. (2005)
Hunter, Artificial muscle technology: physical principals and 15. M. Wissler, E. Mazza, Electromechanical coupling in dielectric
naval prospects. IEEE J. Ocean. Eng. 29(3), 706–728 (2004) elastomer actuators. Sens. Actuators A 138, 384–393 (2007)
6. R. Heydt, R. Kornbluh, J. Eckerle, R. Pelrine, Sound radia- 16. M. Wissler, E. Mazza, Mechanical behavior of an acrylic elas-
tion properties of dielectric elastomer electroactive polymer loud- tomer used in dielectric elastomer actuators. Sens. Actuators A
speakers, in Smart Structures and Materials 2006: Electroactive 134, 494–504 (2007)

You might also like