Professional Documents
Culture Documents
Mechanical Responses of Ecoflex Silicone PDF
Mechanical Responses of Ecoflex Silicone PDF
incompressible behaviors
D. Steck,1 J. Qu,1 S. B. Kordmahale,2 D. Tscharnuter,3 A. Muliana ,1 J. Kameoka2
1
Department of Mechanical Engineering, Texas A&M University, College Station, Texas 77853
2
Department of Electrical Engineering, Texas A&M University, College Station, Texas 77853
3
Polymer Competence Center Leoben GmbH, Roseggerstrasse 12, 8700, Leoben, Austria
Correspondence to: A. Muliana (E-mail: amuliana@tamu.edu)
ABSTRACT: Silicon rubbers are widely used in a variety of products ranging from cooking utensils and electronics to medical devices and
implants. Recently, they have sparked an interest among soft robotics researchers as they can be easily formed into various shapes and
actuated in a relatively fast and easy way. In this article, we examine the nonlinear elastic response of a silicon rubber, Ecoflex, under
both compressible and incompressible constraints. An experimental test on a uniaxial tension indicates a slight compressibility, and the
compressibility increases with stretching. Five different constitutive material models are considered to describe the nonlinear elastic
responses of Ecoflex under both compressible and incompressible conditions. In addition, finite element (FE) analysis is presented to
analyze multiaxial response of structures or devices made of Ecoflex under complex boundary conditions. This study highlights the vari-
ations in the multiaxial response of structures at large deformations from different constitutive models under different compressible and
incompressible constraints. For a high precision control in soft robotics applications, there is a need to understand the multiaxial
response of silicon rubbers, especially under large deformations. © 2018 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2018, 135, 47025.
CONSTITUTIVE MODELS
We first consider a recently developed constitutive model20,21 to
describe the mechanical responses of the Ecoflex sample. The
new constitutive model expresses the left Cauchy–Green defor-
mation tensor (B) in terms of the Cauchy stress (T), whereas
constitutive models for nonlinear elastic behaviors often express
Figure 2. Experimental results from uniaxial stretching with a stretch rate the stress in terms of the kinematical quantities, that is, strains or
of 1.2/min. [Color figure can be viewed at wileyonlinelibrary.com] stretch. With regard to the material parameter characterizations
from experiments, a straightforward characterization is shown for
the new model that expresses the stretch tensor in terms of stres-
ses, which will be shown later. We also compare the responses
condition, the Ecoflex specimens show full recovery upon
generated from the classical hyperelastic models wherein the
unloading, as depicted in Figure 3. The Cauchy stress measure is
stress is derivable from the stored energy. The new model is in
considered in presenting the experimental data. Good repeatabil-
general described by four material constants in the case of a com-
ity is also observed in the testing results with marginal variations.
pressible isotropic elastic body. When the motion is restricted to
The amount of volume changes are examined by calculating the
the isochoric case, the response can be completely characterized
determinant of the deformation gradient (det F) during the uni-
by just two material constants.
axial tests, shown in Figure 2(c). The Ecoflex sample exhibits
slight compressibility, and the compressibility increases with Let x denote the current position of a particle that is at Xin a
stretching. From these observed experimental results, we consider stress-free reference configuration.1 Let x ¼ χκR ðX, t Þ be the
the Ecoflex as a nonlinear elastic material. We will assess both motion of a particle at current time t in a configuration κR with
compressible and incompressible conditions and their implication the deformation gradient F defined as
1
In this study, we take the initial configuration as our reference configuration, and we assume that in the initial configuration, the
material is stress- and strain-free.
∂χκR The constants μ and δ are determined directly from the axial
F¼ ð4:1Þ response.
∂X
We also consider the classical hyperelastic models wherein the
The left and right Cauchy–Green deformation tensors B and stress is derivable from the stored energy to compare the
Care defined through responses obtained from the new model. We first consider a very
B≔FF T , C≔F T F ð4:2Þ popular model by Ogden,23,24 where, in general, the stored energy
The new nonlinear elastic model for isotropic materials is in terms of principal stretches for a compressible material is
given as N μi
pffiffiffi
WO ¼ Σ ½Λ1 αi + Λ2 αi + Λ3 αi − 3 + Wvol ð4:6Þ
i¼1 αi
pffiffiffi 1 −e − δ I2
B ¼ ga ðTÞ ¼ I −κ 1− e −α I2 I + μ pffiffiffiffi T, ð4:3Þ
X
I2 1 2 1 N
Wvol ¼ λ J − J + − μi lnJ ð4:7Þ
whereα, δ, κ and μ are constants, and I2 = tr(T2).. It is seen in 2 2 i¼1
eq. (4.3) that when T = 0, B = I. As discussed by Muliana
where J = det F, and μi, αi, λ are the material constants. The axial
et al. 20 and Muliana et al.,21 the above model reduces to the clas-
component of stress is given as:
sical linearized elastic model upon linearization and imposing
small displacement gradients. Mansouri and Darijani22 have also Λk ∂W ∂W
Tkk ¼ + ; k ¼ 1,2,3 ð4:8Þ
considered an exponential function for strain energy in terms of J ∂Λk ∂J
the first and second invariants of the Cauchy stretch tensor. Their For the uniaxial tensile test, the axial stresses in the compressible
model is shown to be capable of capturing a multiaxial response Ogden model are
of nonlinear elastic behavior of soft materials with a relatively XN
μi αi αi =2 − αi =2
small number of material parameters. When compared to various T11 ¼ Λ1 − J Λ1
well-established models, the Mansouri and Darijani22 model gave J
i¼1
ð4:9Þ
much better predictions of experimental data. XN
μi αi =2 −αi =2
T22 ¼ T33 ¼ J Λ1 − 1 + λðJ − 1Þ ¼ 0
J
The uniaxial tensile test is defined by the following deformation i¼1
2 3
Λ1 0 0 When an incompressible condition is imposed (J = 1), the axial
gradient tensor, F ¼ 4 0 Λ2 0 5, where Λ1 and Λ2 are the stress reduces to
0 0 Λ2 X
N
− α =2
stretch in the axial (x1 axis) and transverse (x2 axis) directions. TO ¼ TO11 ¼ μi Λα1 i − Λ1 i ð4:10Þ
Assuming isotropic behavior, the transverse response in the x3 i¼1
axis is assumed to be the same as the one in the x2 axis. Once Another model that describes well the nonlinear elastic response
the deformation gradient has been determined, different stress of materials with a limiting stretch behavior is a model by Gent.25
and strain measures can be constructed. In the uniaxial tensile The stored energy in terms of stretch for a compressible material
test, the nonzero component of the Cauchy stress is T11 ¼ ðΛP11Þ2 , is given as
2
μ Jm J
obtained. T11 ¼ Λ21 − ð4:14Þ
J ðJm −J1 + 3Þ Λ1
When we impose the isochoric motion (det F = 1) to the uniaxial
loading, we have Λ2 ¼ Λ3 ¼ p1ffiffiffiffi
Λ
, and the constitutive model in When an incompressible condition is considered J1 ¼ Λ21 + 2=Λ1 ,
1
eq. (4.4) reduces to the axial stress in eq. (4.14) reduces to:
1 ð1− e − δT Þ μJm 1
Λ21 − ¼μ T ¼ μ 1 −e − δT ð4:5Þ TGT ¼ TGT11 ¼ Λ21 − ð4:15Þ
Λ1 T ðJm − J1 + 3Þ Λ1
When an incompressible condition is considered, the material is not exactly zero (Figure 6) unlike in the new model; however,
parameters in the Ogden and Gent models, eqs.(4.10) and(4.15), the values are negligible compared to the magnitude of the axial
respectively, can be easily determined from fitting the uniaxial stress in the experiment.
data. However, when a compressible condition is considered,
Furthermore, the absolute percent errors of the above models,
determining the material parameters is quite tedious because we
when compared with experimental data, are presented in
need to concurrently fit the axial and lateral experimental data
Figure 7. The percent errors from both incompressible and
from the following relations:
compressible assumptions are presented. It is seen that the pre-
T11 ¼ f1 ðΛ1 , Λ2 Þ sent model gives relatively small errors, especially for the com-
ð4:16Þ
T22 ¼ f2 ðΛ1 , Λ2 Þ ¼ 0 pressible condition. As expected, percent errors from the
In the new model where we express the stretch in terms of stres- Mooney–Rivlin model are quite large. Other models show varia-
ses in eq. (4.4), determining the material parameters for com- tions in the percent errors with the higher values at early
pressible bodies is straightforward, as discussed above. loading.
In addition, two additional models, that is, Yeoh and Mooney– In summary, the uniaxial tension tests of Ecoflex silicone rubber,
Rivlin, that are commonly considered for describing nonlinear which indicate slightly compressible behavior, have been used to
elastic response of soft materials are also considered. The stored examine the nonlinear elastic response of materials under com-
energy functions for the Yeoh and Mooney–Rivlin models are pressible and incompressible conditions. Several hyperelastic
models are used to describe the response of Ecoflex. The hyper-
X
3 X
3
W¼ Ci ðI1 − 3Þi + Di ðJ − 3Þ2k ð4:17Þ elastic models considered in this study are capable of capturing
i¼1 k¼1 the nonlinear elastic response of Ecoflex under compressible
and incompressible conditions. However, when the compress-
W ¼ C1 I1 −3Þ + C2 I2 − 3Þ + DðJ − 1Þ2 ð4:18Þ ible condition is considered, calibrating material parameters in
the hyperelastic models is quite challenging not only because
where I1 ¼ J − 2=3 I1 , and Ci and Di are the material constants.
the models involve more material parameters but also because
Figure 4 illustrates the uniaxial response of Ecoflex when the the need to concurrently satisfy the axial and lateral conditions
body is assumed as incompressible. Different constitutive mate- during the curve fitting of experimental data. As a consequence,
rial models are considered for describing the mechanical the traction-free condition on the lateral surfaces cannot be
responses of Ecoflex. The material parameters calibrated from the exactly satisfied, as depicted in Figure 6. The new model, which
above models are listed in Table I. We now consider the Ecoflex is shown to be capable of describing the mechanical responses
as a compressible material as suggested by the volume changes in of Ecoflex, has several advantages with regard to calibrating the
Figure 2(c). The volumetric strain energy functions in eqs.(4.7) material parameters: (1) it has a relatively few material parame-
and (4.12) are considered in the Ogden and Gent constitutive ters, (2) it eliminates the need to impose an isochoric motion in
models, respectively. Figure 5 presents the axial and transverse calibrating the material parameters, and 3) it directly satisfies
responses of Ecoflex with five constitutive models being consid- the boundary conditions.
ered. It is seen that all models are capable of capturing both axial
Furthermore, we investigate the nonlinear elastic response of Eco-
and lateral responses of Ecoflex, except for the Mooney–Rivlin
flex under uniaxial stretching using ABAQUS FE analyses. 3D con-
model that shows relatively large deviations. The calibrated mate-
tinuum elements (C3D20RH) are used. The Ogden constitutive
rial parameters are given in Table II. For the hyperelastic models,
models for both compressible and incompressible cases are consid-
where the stress is expressed in terms of stretch, the lateral stress
ered. The stored energy in terms of the principal stretches is given
h α i P h i
as WO ¼ ΣN μi Λ ^1 i + Λ^2 αi + Λ
^3 αi −3 + M 1 ðJ − 1Þ2j , where
i¼1 αi j¼1 Dj
H 1 μJm 1 1
pGT ¼ 2 Λ 2
− ; I1 ¼ 2Λ21 + ð5:4Þ
R Λ31 Jm − I1 + 3 1
Λ41 Λ41
Figure 8 illustrates the response of an inflated sphere with R/
H = 10 when incompressible constraint is considered for the
Figure 5. Uniaxial response of Ecoflex treated as a compressible material. models. All models show a similar trend, in which, initially, the
[Color figure can be viewed at wileyonlinelibrary.com] pressure increases with the in-plane stretch, then a slight change in
pressure with continuously increasing stretch is seen, and finally
BOUNDARY VALUE PROBLEMS
the pressure starts increasing again with further stretching. The cor-
We now present boundary value problems to examine the multi- responding in-plane stress and stretch response indicates the limit-
axial response of the Ecoflex when compressible and incompress- ing stretch behavior, which should be expected from experimentally
ible conditions are considered. observing the uniaxial stretching of the Ecoflex specimens.
1X N
J
0¼ μi 2αi − 1 + λðJ − 1Þ ð5:7bÞ
J i¼1 Λ1
Finally, we also consider the Gent model with the following con-
stitutive relation:
H 1 μJm J2 J2
pGT ¼ 2 Λ2
− ; I1 ¼ 2Λ 2
+ ð5:8aÞ
R Λ31 Jm − I1 + 3 1
Λ41 1
Λ41
1 μJm J 2
0¼ + 2c1 −μ + 4c2 lnJ + 4c3 J 2 ð5:8bÞ
J Jm − I1 + 3 Λ41
In both Ogden and Gent models, two equations need to be solved
simultaneously to determine the response of the inflated sphere.
In obtaining the response of the inflated sphere for compressible
cases, eqs. (5.6)–(5.8) are solved numerically.
i μi (kPa) αi Di
1 16.9 1.3 1.156
2 0.08 5.0 0.0001
Figure 7. Percent errors of several models compared to experimental data.
3 1.0 −2.0 -
[Color figure can be viewed at wileyonlinelibrary.com]
1.49
Figure 12. Response of sphere inflation using Ogden models (analytical and Figure 14. Response of sphere inflation using Ogden models (analytical and
FE simulation with 3D continuum elements). [Color figure can be viewed at FE simulation with shell elements). [Color figure can be viewed at
wileyonlinelibrary.com] wileyonlinelibrary.com]
13. Suzumori K, Hama T, Kanda T (2006), In Proceedings of 19. Muliana AH, Rajagopal KR, and Tscharnuter, D, “A Non-
IEEE International Conference on Robotics and Automa- linear Integral Model for Describing Responses of Visco-
tion, p 1824. elastic Solids” Int. J. Solids and Structure, 2015 58, pp.
14. Wakimoto S, Ogura K, Suzumori K, Nishioka Y (2009). In 146–156.
Proceedings of IEEE International Conference on Robotics 20. Muliana, A.; Rajagopal, K. R.; Tscharnuter, D.; Pinter, G.
and Automation, p 556. Int. J. Solids Struct. 2016, 100, 95.
15. Wakimoto, S.; Suzumori, K.; Ogura, K. Adv. Robotics. 2011, 21. Muliana, A.; Rajagopal, K. R.; Tscharnuter, D.;
25, 1311. Schrittesser, B.; Saccomandi, G. Rubber Chem. Technol.
16. Zhang J, Wang H, Tang J, Guo H, and Hong J (2015). In 2018, 91, 375–389.
Proceedings of IEEE International Conference on Informa- 22. Mansouri, M. R.; Darijani, H. Int. J. Solids Struct. 2014, 51,
tion and Automation, p 2460. 4316.
17. Polygerinos, P.; Wang, Z.; Overvelde, J. T. B.; 23. Ogden, R. W. Proc. R. Soc. London A. 1972a,
Galloway, K. C.; Wood, R. J.; Bertoldi, K.; Walsh, C. J. IEEE 326, 565.
Trans. Robotics. 2015, 1, 1552–3098.
24. Ogden, R. W. Proc. R. Soc. London A. 1972b, 328, 567.
18. Ogden, R. W.; Saccomandi, G.; Sgura, I. Comput. Mech.
2004, 34, 484. 25. Gent, A. N. Rubber Chem. Technol. 1996, 69, 59.