Professional Documents
Culture Documents
ABSTRACT: A micro-mechanical hyper-elastic constitutive theory for open-cell foams at large strains is
developed. The solid phase in the foam is represented by a network of struts each connecting to two vertex
points. The strut deformation is assumed to depend directly on the macroscopic deformation and the forces
carried by a strut are linked to the change of its vertex-to-vertex vector. The constitutive model was compared
with a compression experiment on an isotropic polyether urethane foam and good agreement was obtained.
ii Tension
cf ci c
-1= ii
IV III II I
Compression
101
Copyright 2005 Taylor & Francis Group plc, London, UK
and Lagzdin (1992) used a similar approach. A pin 2 THEORY
joined straight linear elastic single strut is adopted as
the structural element and affine motion of strut end- The material occupying a region 0 in the reference
points is assumed. This model allows for truly large configuration is on the micro-level understood to con-
deformations, including strut reorientation, and also sist of two phases, one solid phase s0 and one pore
incorporates a more sophisticated treatment of the phase f 0 = 0 s0 . The solid phase s0 is assumed
response due to strut buckling. Warren and Kraynik to consist of a network of struts r i each connecting to
(1991), introduced the more realistic and less restric- two vertex points X i 0 and X i 0 .
tive assumption of affine motion of strut midpoints Under deformation the region 0 is mapped to the
rather than endpoints. This allows for the non-affine current configuration such that the vertex points
stretching of a strut which is due to bending of the con- X i and X i can be found at x i and x i . Our
necting struts. The model is more complicated than approach is to characterize each strut by the vectors
the single strut models, since it involves solving the
forces on four struts simultaneously. No strut buck-
ling is included in their analysis. Wang and Cuitio
(1999) use the same kinematic assumption as Warren and
and Kraynik, but the solution is obtained by minimiza-
tion of energy for the unit-cell consisting of four struts.
This allows the authors to examine more general con-
figurations, although with the inherent limitation of where N i and ni are the material and spatial directors
periodicity. Despite their proven utility, one of the of the strut respectively, such that
drawbacks of the cell-model approach is that most
foams tend to have a random distribution of nuclei
which results in random cell sizes in contrast to peri-
odically distributed nuclei which are represented by
and
cell-models.
In RVE-models, the balance equations on the RVE
are typically solved by the finite element method
(FEM). Relevant background here is Shulmeister
(1998) and Zhu andWindle (2001), where the foam is i.e. ||N i ||, ||ni || = 1, i = 1 . . . N , see figure 3.
modeled as a three-dimensional framework of slender The average Cauchy stress within the strut can be
struts, regular as well as random based on the Voronoi expressed as
technique. A benefit of such computations is that one
may distinguish different mechanisms and study the
effect of non-uniform distributions of strut parameters
and connectivity. Thus Shulmeister (1998) shows that
Youngs modulus and Poissons ratio increase strongly Further f i is the resultant force that act on the strut i at
with increasing disorder. Further, Van der Burg (1997) the vertex point x i and vi is the volume of the strut i.
conclude that the normal deformation of the struts The macroscopic stress in the continuum is obtained
in the foam is of increasing importance relative to
bending deformation of struts for foams with higher x(X,t)
density. The main drawback of RVE-models is the high Ni
computation cost. Another limitation is the difficulty
of choosing the appropriate boundary conditions for ni
the RVE.
We propose here to develop the single strut model, ri0
because it will be considerably easier to generalize to X'i
inelastic responses, it is more adaptable with respect x'i
to random micro-structures than multi-strut models ri
and is numerically inexpensive. The model developed Xi
may be viewed as a generalization of the Zilauts and t=0 xi
Lagzdin (1992), by (i), inclusion of transverse forces
and (ii), a simple correction of the generally over-
stiff affine assumption. Like Zilauts and Lagzdin we t>0
assume that the struts are free to deflect between their
ends; so the model is not applicable to the densification
regime i.e. region IV in figure 2. Figure 3. Affine motion.
102
Copyright 2005 Taylor & Francis Group plc, London, UK
as the volume average of the local stress within the where is the macroscopic stretch along the direction
struts, of the strut, defined by
103
Copyright 2005 Taylor & Francis Group plc, London, UK
ft 0
m
2
4
kN/m2
fn fn
6
8
m 10
ft
12
Figure 4. Forces and couples on strut. 0.2 0.4 0.6 0.8 0
h/h0
buckling regime is roughly linear in the strain 1.
We thus propose a bi-linear response for the axial Figure 5. Three hysteresis loops on top of each other.
response,
3 EXPERIMENTAL
104
Copyright 2005 Taylor & Francis Group plc, London, UK
0 model and, therefore influences the values of the
parameters. The average strut length r0 that is imple-
mented in the model does not take into account the
2 geometry of the vertex points, mentioned as dead vol-
ume in Gent and Thomas (1959). The dead volumes
kN/m2
vectors: N i R3 , i {1 . . . N }. Then the parameter k1 This work was financed by the vehicle research pro-
was determined with respect to the initial linear elastic gram (ffp) and the following participating companies:
region I,II, and c was given by the onset of buck- Volvo Car Corporation, Volvo AB, SAAB Automobile
ling. Further the parameters k2 and k3 were adjusted AB, Finnveden AB, Scania AB, Outokumpu Stainless
to fit the plateau, region III. In figure 6, the contribu- AB. The authors also wish to thank Gran Werner at
tion from the axial force to the macroscopic Cauchy Carpenter Sweden AB for the supplying the foams and
stress, and the contribution of the transverse to the Thomas Schuman for taking the SEM pictures.
macroscopic Cauchy stress, together with the total
macroscopic Cauchy stress are plotted against an uni- REFERENCES
axial compression test.The theory made a good fit with
the experimental results when the parameters were set Alkhagen, M. and S. Toll (2001). A triaxial rheometer for soft
to: k1 = 280.5N , k2 = 1.9N , k3 = 9.5N and c = 0.976. compressible solids. Journal of Rheology 46(1), 3147.
The influence factor c was set to 1.0. Dario Aristizbal-Ochoa, J. (2004). Large deflection stabil-
ity of slender beam-columns with semirigid connections:
Elasrica approach. Journal of Engineering Mechanics
5 CONCLUDING REMARKS 130(1), 274282.
Gent, A. N. and A. G. Thomas (1959). The deformation
of foamed elastic materials. Journal of Applied Polymer
The model is based on a representation of the foam Plastics 1(1), 107113.
by a network of struts. The forces at the vertex points Mattiasson, K. (1979). Numerical results from elliptic inte-
are assumed to be a function of the strut deforma- gral solutions of some elastica problems of beams and
tion, which is assumed to depend on the macroscopic frames. Publication 79:10 Department of Structural
deformation gradient. The model is so far under Mechanics Chalmers University of Technology 79:10.
the restriction of hyper-elasticity; thus no strain rate Shulmeister, V., Van der Burg, M.W. D. V. d. G. E. M.
effects are considered. R. (1998). A numerical study of large deformations of
The reason for choosing the single-strut model is low-density elastomeric open-cell foams. Mechanics of
that it will be considerably easier to generalize to Materials 30, 125140.
Van der Burg, M.W. D., Shulmeister, V. V. d. G. E.M. R.
inelastic responses and that it is more adaptable with (1997). On the linear elastic properties of regular and ran-
random micro-structures than multi-strut models. It is dom open-cell foam models. Journal of Cellular Plastics
also numerically less expensive than multi-strut and 33, 3154.
RVE models. Wang, Y. and A. M. Cuitio (1999). Three-dimensional non-
The overall stressstrain response of the foam is linear open-cell foams with large deformations. Journal
governed by primarily axial deformations of the struts of the Mechanics and Physics of Solids 48, 961988.
including buckling. The contribution of transverse dis- Warren, W. E. and A. M. Kraynik (1991). The nonlinear
placement is small initially but becomes significant at elastic behaviour of open-cell foams. Journal of Applied
larger strains. This is mainly due to strut reorientation Mechanics 58, 376381.
Zhu, H. X. and A. H. Windle (2001). Effects on cell irregu-
and strut buckling but also because of the increasing larity on the high strain compression of opencell foams.
fraction of struts in under compression, see figure 6. Acta Materialia 50, 10411052.
The hysteresis loops, see figure 5, show that energy Zilauts, A. F. and A. Z. Lagzdin (1992). Single-bar model of
is dissipated during the experiments. This is not taken cellular materials subjected to large elastic deformations.
into consideration in the hyper-elastic constitutive Mekhanika Kompozitnykh Materialov (1), 310.
105
Copyright 2005 Taylor & Francis Group plc, London, UK