Professional Documents
Culture Documents
A Dissertation
Presented to
In Partial Fulfillment
Doctor of Philosophy
Linling Chen
December, 2012
DEVELOPING CONSTITUTIVE EQUATIONS FOR POLYMER FOAMS
Linling Chen
Dissertation
Approved: Accepted:
____________________________ ____________________________
Advisor Department Chair
Dr. Michelle S. Hoo Fatt Dr. Celal Batur
_____________________________ ____________________________
Committee Member Dean of the College
Dr. Atef Saleeb Dr. George K. Haritos
_____________________________ ____________________________
Committee Member Dean of the Graduate School
Dr. Gregory Morscher Dr. George R. Newkome
_____________________________ ____________________________
Committee Member Date
Dr. Xiaosheng Gao
_____________________________
Committee Member
Dr. Mukerrem Cakmak
ii
ABSTRACT
by offering high bending stiffness and strength per unit weight. They can also dissipate
energy when subjected to shock and impact loadings. The current foam models are based
on metallic foams, which can describe polymer foam behavior under monotonic loading
but not under cyclic or vibration loadings. Cyclic compression and pure shear tests were
done on Divinycell PVC H100 foam to produce hysteresis data in compression and shear
in the in-plane and out-of-plane directions of the foam with strain rates ranging from
0.0005 s-1 to 5 s-1. The compressive strain amplitudes were limited to less than 10%,
while the engineering shear strain amplitudes were less than 20%.
Two one-dimensional constitutive models were developed from the test data. The
element and in series a Prandtl element. Material properties were introduced for
(Mullins effect). This constitutive model was able to capture yielding, viscoelastic and
viscoplastic response, as well as hysteresis of the foam, but it failed to describe the strain-
rate dependency that was found in the experiments. The second constitutive model
consisted of an equilibrium spring only in parallel with a Maxwell element. This model
iii
was based on the concept of damage initiation and evolution of the foam. The damaged
material properties were found to depend on the magnitude and history of the
deformation. Good agreement with out-of-plane compression test results was reported
iv
ACKNOWLEDGEMENTS
This study was funded in part by ONR Grant No. N00014-11-1-0485. I would
like to acknowledge the kind support of Dr. Yapa D. S. Rajapakse, Program Manager of
The greatest thanks to my advisor, Dr. Hoo Fatt, from whom, I learned how to
conduct research that is needed to meet the high standards in the field. Her continuous
dedication to her work have always inspired me. I would like to thank her for the pleasant
and stimulating discussions, and the time she spent in guiding me throughout my
graduate studies.
members: Dr. Saleeb for his great and most helpful advice on material modeling and
about composites and his generosity of allowing me using his lab equipment; Dr. Gao for
continuum mechanics advice; Dr. Cakmak for his suggestions of the importance in
exploring the micro-structure of polymers. I want to thank you all for taking the time in
v
reading this dissertation. Your useful suggestions and constructive comments have
support throughout my doctoral program. I would like to thank Mr. Dale Ertley, together
with who we designed test apparatus and achieved an invention disclosure. I would like
also thank to Mr. Stephen Gerbetz, Mr. Clifford Bailey, Mrs. Christina L. Christian and
Mrs. Stacy Meier for their technical supports and personal care.
Linling CHEN
vi
TABLE OF CONTENTS
Page
LIST OF TABLES……………………………………………………………………...…x
LIST OF FIGURES………………………………………………………………………xi
CHAPTER
I. INTRODUCTION……………………………………………………………………....1
2.3.2 Viscoelasticity………………………………………………………...…….25
III. EXPERIMEMTS…………………………………………………………………….32
3.1 Specimens…………………………………………….………………………..…33
xv
3.3 Compressive Tests……………………………………………...…………..…...37
4.1.1 Viscoelasticity…………………………………………………….…....….73
4.1.2 Viscoplasticity…………………………….……………………………….81
REFERENCES……………………………………………………….……………...…108
APPENDICES …..……………………………………………….….…………..….….114
viii
APPENDIX F INELASTIC STRAIN VALUES……………………………………….132
ix
LIST OF TABLES
Table Page
x
LIST OF FIGURES
Figure Page
3.2 Strain history at a strain rate 5 s-1 and strain amplitude 4%.........................................36
xi
3.3 PVC H100 glued on Aluminum gripes………………………………………………39
3.10 Out-of-plane compressive stress-strain curves for glued and unglued specimens
at strain rate of 0.1s-1………………………………………………………………..45
3.16 Compressive stress-strain curve at a strain rate of 0.5 s-1 with various
strain amplitudes: (a) Out-of-plane and (b) In-plane………………………......…..50
3.17 Out-of-plane first cyclic of strain rates of 0.5 s-1 with various
strain amplitudes over lamping on the primary curve……………………………...51
3.18 Out-of-plane fourth cycle hysteresis at 10% strain at various strain rates………….51
3.19 PVC H100 crushed with 10% strain amplitude at strain rate of 0.0005 s-1
in the in-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at
A2 and (c) Damaged micro-structures at various locations……….………………...54
3.20 PVC H100 crushed with 10% strain amplitude at strain rate of 0.0005 s-1
in the out-of-plane direction: (a) Crushed specimen (b) Undamaged micro-structure
at A2 and (c) Damaged micro-structures at various locations.…………………….55
xii
3.21 PVC H100 crushed with 10% strain amplitude at strain rate of 5 s-1 in the in-plane
direction: (a) Crushed specimen (b) Undamaged micro-structure at A2 and
(c) Damaged micro-structures at various location…………….…………………56
3.22 PVC H100 crushed with 10% strain amplitude at strain rate of 5 s-1
in the out-of-plane direction: (a) Crushed specimen (b) Undamaged micro-structure
at A2 and (c) Damaged micro-structures at various locations…………………...…57
3.23 Crushed specimens loaded to 10% strain amplitude in the in-plane direction with
five increasing strain rates………………………………………………………….58
3.24 Crushed specimens loaded to 10% strain amplitude in the out-of-plane direction
with five increasing strain rates…………………………………………………….58
3.25 PVC H100 crushed with 30% strain amplitude at strain rate of 0.0005 s-1
in the in-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at
A2 and (c) Damaged micro-structures at various locations……..……….…………59
3.26 PVC H100 crushed with 30% strain amplitude at strain rate of 0.0005 s-1
in the out-of-plane direction: (a) Crushed specimen (b) Undamaged micro-structure
at A2 and (c) Damaged micro-structures at various locations……………….…….60
3.28 Comparison of out-of-plane shear stress-strain curves of PVC H100 foam core at
strain rate =0.001 s-1 with data from Steeves and Fleck (2004) and Danel (2009)…65
3.29 Out-of-plane shear fracture of PVC H100 foam core at strain rate =0.001 s-1……..66
3.31 Pure shear stress-strain curves of strain amplitude of 20% at various strain rates:
(a) Out-of-plane and (b) In-plane……………………………………………...……68
3.32 Pure shear stress-strain curves at a strain rate of 0.5 s-1 with various strain
amplitudes: (a) Out-of-plane and (b) In-plane……………………………...………69
xiii
4.4 Mechanical analogs for a viscoelastic………………………………..………………75
4.5 Relationship of strain verses time for loading, unloading, and reloading………..….77
4.8 Plastic hardening (a) Isotropic hardening and (b) Kinematic hardening………...….82
4.14 The experimental curve verses the simulation of Mullins damage at strain rate of
0.0005 s-1 in the out-of-plane direction…………………………………………….93
4.15 Comparison of the hysteresis at different strain rates in out-of-plane direction with
r=1.29, m=0.035, C1=2 and C2=220. (a) 0.0005 s-1, (b) 0.005 s-1, (c) 0.05 s-1,
(d) 0.5 s-1and (e) 5 s-1…………………….………………………………………….95
4.21 Comparison of experimental data with simulation curves at strain amplitude of 10%
in the out-off-plane direction: (a) 0.0005 s-1, (b) 0.005 s-1, (c) 0.05 s-1, (d) 0.5 s-1and
(d) 5 s-1…...…………………………………………………………………………104
xiv
4.22 The comparison at various strain rates of strain amplitude of 10% in the out-of-plane
direction………………………………………………………………………..….105
4.23 Comparison of simulation curve and test data at various strain amplitudes and fixed
strain rate of 0.0005 s-1 in the out-of-plane direction………………………….......105
xv
CHAPTER I
INTRODUCTION
Natural disasters and terrorists’ bomb blasts injure and kill people and destroy
property, causing billions of dollars in damages. How to prevent and limit the damage
due to them has become a big issue to both political leaders and scientists. As part of
these efforts, scientists are studying ways to improve the structures’ resistance to high
designs not only because it offers high bending stiffness and strength per unit weight, but
it may dissipate energy when subjected to shock and impact loadings. In 1820, a
Frenchman, Duleau [1] first discussed the advantages of using two co-operating faces
separated by some distance to create a sandwich structure. It was not until 110 years later
that the concept was first applied commercially. During World Wars I and II, the use of
sandwiched between two stiff facesheets. The sandwich material options were numerous,
such as metallic and composite facesheets with honeycomb, balsa wood and foam cores,
and a large range of adhesives. Reports [2] claims that the facesheets carry bending
1
stresses and give the structure a hardwearing surface, while the core material absorbs the
transverse shear stresses generated by loads and distribute them over a larger area.
For decades, sandwich constructions with foam cores have been utilized in
strengthen components ranging from bridges, buildings, ships, trains, and even
windmill blades. The sandwich structures have high specific stiffness and
have been used in high-speed truck trailers and high-speed trains for the safety of
building that will be subjected to wind loads [3] or even terrorist’s bomb blasts.
On a daily basis, external structures experience rain, sleet, snow and hail; when
people on the inside. Under bomb blast, not just the blast but fragments or debris
from walls, windows and fixtures, become high-velocity projectiles causing injury
and death. A sandwich panel can provide a nesting zone for these fragments.
Indeed, the study of sandwich structures that have both high static strength and
Foam-core panels are stronger than solid laminates, only if they are flat.
Curved foam-core panels are weaker than solid glass panels, especially when the
2
curved sandwich hulls in sailboats, found out that curved foam-core panels are poor in
bending [4]. Hence, the interest of this research is focused on foam-core flat panels with
various polymer foam core materials, such as polyvinylchloride (PVC). Nowadays, when
cost is the typical driving force in design, polymer foams are superior to other core
materials, such as metal honeycomb and balsa wood, because they are cheap and require
no maintenance. For example, PVC foam is half the price of aluminum honeycomb and
PVC foam has “closed cells,” which makes this material water-resistant and anti-
eco-friendly material [3]. For instance, Foamacell TM Green is recycled PVC foam made
with 70% recycled PVC and 30% virgin petrochemicals. Furthermore, PVC foam is
wall buckling, fracture, friction and viscous air/fluid flow. This ability of the polymer
foam is what makes it the ideal material for safety helmets, packaging and blast
protection. In these applications the foam is not bonded to two facesheets, as in typical
destroyed. The same cushioning and energy absorption of the foam core, however, also
become important when sandwich panels are under dynamic loading such as impact or
pressure pulse (impulsive or blast loading). The difference in the case of a sandwich
panel is that the foam is bonded to two facesheets, and energy can be absorbed in the
form of kinetic and strain energy when the entire sandwich construction vibrates. The
foam in a sandwich structure dissipates energy through hysteresis from subsequent stress
3
wave propagation and structural vibrations. It is not intended to be a disposal part
of the system.
low-velocity impact, including the analysis of the influence of the foam core.
challenging.
program, which is time consuming and expensive. Therefore the use of analytical
[7] and numerical [8] models to analyze the perforation of sandwich structures is
critical to reduce cost and time in the design processes. The main advantage of
the high-velocity impact behavior of sandwich structures. Since the foam core
plays a major role in absovbing energy, the characteristics of the foam need to be
4
studied well. Experimental and analytical investigations of the crushing of foam are done
in this study.
There are two well-established constitutive models being used on polymer foams
today: crushable foam model [9-13] and hyperfoam model [14-17]. Both of these
crushing response of the foam. Neither of them address viscoelasticity nor hysteresis,
although the hyperfoam model is often coupled with finite strain viscoelastic models to
produce hysteresis response. The hyperfoam model can also be combined with a Mullins
damage model to address stress softening during subsequent cyclic loads [18]. Several
has been validated with experimental results obtained from quasi-static and low-velocity
impact tests [19, 20]; however, the validity of this model to reproduce the high-velocity
The goal of this research is to develop a simple constitutive model that will
describe the behavior of the sandwich core material subjected to cyclic loads. This model
should address the micro-inertia and rate effects, as well as permanent crushing or
foams, especially the rate-dependent ones. Chapter III describes experiments and includes
the description of a test plan, apparatus design, test setup and experimental results. Two
types of tests are done: compression and shear. The tests are strain-controlled, cyclic tests
5
under uniaxial compression and pure shear loadings, with both in-plane and out-off-plane
directions tested under room temperature. Chapter IV discusses two rheological models:
one addresses viscoelasticity, viscoplasticity together with Mullins damage; the other is
based on the theories of damage mechanics. Conclusions are finally made in Chapter V.
6
CHAPTER II
LITERATURE REVIEW
structures subjected to blast loads [21-24]. Because of the difficulties of getting test
results under blast loads, computational simulations have been widely used. Current
models that utilize crushable elastic-plastic constitutive models to simulate the behavior
of the foam core in the sandwich panels or beams may address the material behavior
dissipation that take place in polymer foams under dynamic loading characterized by
In this chapter finite element analysis (FEA) packages for simulating crushable
foams in ABAQUS and LS-DYNA are reviewed. Both the validation and limitation of
these two software packages are discussed and compared. The application of crushable
foam plasticity models implemented in ABAQUS will be further analyzed. Before the
numerical review, this chapter starts with revealing some foam core material behaviors
derived from the experiments, which will be helpfully for later discussion.
7
2.1 Experimental Investigation
regions: a linear elastic region, a plateau region and a densification region. As shown in
Figure 2.1, the stress-strain curve indicates that the material behaves linearly elastic with
a slope equal to the Young’s modulus of the material when the strain is very small. As the
load increases, a stress plateau is generated. This roughly constant load is caused by the
collapse of foam cells. Depending on the mechanical properties of the cell walls, the
collapse may be caused by elastic buckling, plastic yielding or brittle crushing [25]. As
the cell walls keep on collapsing, the opposing cell walls meet and touch each other and
this procedure causes the stress to increase very rapidly with strain in the densification
region. Most current researchers are interested in studying the material behavior before
8
For applications, such as in the marine industry, it is necessary to study core
it should be made clear if strengths are different in tension and compression, and whether
the material is isotropic or anisotropic. A study of the foams under multiaxial stress
conditions was done by Gdoutos et al. [26]. They performed uniaxial tensile,
compressive and shear tests along the in-plane and the out-of-plane directions on
Divinycell PVC H100 and H250 foams. Their results indicated that the elastic modulus
under uniaxial tensile and compressive loadings were almost equal for both Divinycell
PVC H100 and H250. The authors also claimed that the low-density Divinycell PVC
H100 displayed nearly isotropic behavior, while the higher density Divinycell PVC H250
was strongly anisotropic with respect to stiffness and strength. The elastic modulus and
strength of Divinycell PVC H250 along the out-of-plane direction were much higher than
along the in-plane directions. The authors explained this was related to the microstructure
of the material where the cells were elongated in the thickness direction. The failure
surfaces (defined as yield) of Divinycell PVC H250 under combined normal and shear
stresses along the in-plane and out-of-plane directions were obtained from biaxial tests. It
was found that the failure envelopes in the principal plane were ellipses displaced along
the first quadrant of a graph with their long axis closer to the through-the-thickness than
the in-plane axis. This phenomenon was predicted well by the Tsai–Wu failure criterion
typically used in composite materials. Daniel et al. [27] studied the failure of PVC H250
as well, and drew the same conclusion that the Tsai-Wu theory could predict anisotropic
9
2.2 Finite Element Models for Foam
Based on the current experimental studies on foams, two FEA packages have
been widely utilized to simulate foam core behaviors: ABAQUS and LS-DYNA. In
ABAQUS, there are two well-established foam models: the crushable foam model and
the hyperfoam model [11]. The crushable foam model in ABAQUS is based on the
plasticity theory. It could be used to model foam deformation under compression due to
cell wall buckling process. On the other hand, the hyperfoam material model is based on
the hyperelastic theory, but it could not model the strain rate response of foam. The PVC
The foam model in LS-DYNA is similar to the one in ABAQUS. It allows the use
of a hysteretic unloading factor and a shape factor to represent the hysteretic behavior of
an elastomer foam material [28]. For crushable foam, hyperfoam, and low density foam,
the model only accepts test data of one stress-strain curve at a certain strain rate. The
modified crushable foam model in LS-DYNA and Fu-Chang foam model [29] accept test
data for stress-strain curves at differing strain rates. However, Pan [30] claimed the Fu-
Chang rate responsive material foam model could not be captured well under high strain
To evaluate and compare the validation of the two FEA packages, Ozturka and
Anlasb [31] applied both ABAQUS and LS-DYNA FEA simulations to study multiple
compressive loading and unloading with polystyrene (EPS) foam. They used crushable
foam material model in ABAQUS and low density foam material model (MAT 57) in
10
LS-DYNA. The research reported that both packages were accurate in predicting
maximum deceleration, force, and displacement during the first loading. However, the
authors claimed that the FEA packages need improvement for the case of unloading and
deceleration, force, and displacement for the same amount of absorbed energy at each
loading step. It was also found that LS-DYNA would provide accurate results if
parameters for controlling shape and hysteresis of unloading and reloading were
calibrated using test results. The force and displacement needed to be improved in the
For decades, researchers struggle to set up reliable FEA models that could
accurately simulate foam behavior under both quasi-static and impact loading conditions.
It is found that the crushable foam model implemented in ABAQUS could perform well
under quasi-static and low-velocity impact before unloading. Schubel et al. [32] studied
composite sandwich structures that subjected to low velocity impact damage. The
sheets and a PVC foam core, and instrumented panels impacted with a drop mass. The
comparison between low velocity impact and an equivalent static loading indicated that
regardless of localized damage, low velocity impact was generally quasi-static in nature.
Therefore, a quasi-static test, which is easier to perform and analyze, could be used to
predict related low velocity impact. In the Ref. [32], experimental results were also
compared with the model constructed in ABAQUS. The facesheets were modeled as
orthotropic laminate and the core was modeled as an isotropic elastic material until
11
yielding. After core yielding, the core was modeled as crushable foam. Finite element
analysis (FEA) using an elastic foundation compared well with experiments in the linear
region before the core yielded. However the researchers claimed that there was a need for
a FEA model to capture the response of the panel behavior after yielding.
Rizov [33] also applied the crushable foam and the crushable foam hardening
options of ABAQUS code to the plastic response of the foam core. He developed a 2D
finite element modeling procedure for analyzing the non-linear indentation behavior of
foam-cored sandwich composite beam. The indenter was modeled as a rigid cylinder, and
the facesheets were assumed linear-elastic and quasi-isotropic. The core was modeled as
obtained from uniaxial compression test in terms of true (Cauchy) stresses and volumetric
logarithmic plastic strains. The experimental data and the numerical analysis showed very
good agreement with the finite element modeling and could be used to predict the static
indentation response of foam cored sandwich beams. However, during the test it was
observed that after unloading the magnitude of the residual dent decreased with time. The
author claimed that it was due to the relaxation of the crushed foam core. For that reason,
it was concluded that the model was capable of predicting the residual dent at the instant
of impact.
High-velocity impact behavior is different from the low-velocity one, and the
comprehensive review by Abrate [6], high-velocity impact is defined as where the ratio
12
between impact velocity and the velocity of compressive waves propagating through the
thickness is larger than the maximum strain to failure in that direction. Abrate indicated
this implied that damage was generated during the first few travels of the compressive
wave through the thickness when overall plate motion was not yet established. Thus, it
impacts ware not applicable to high-velocity cases. Since the experiments under high-
velocity impacts are difficult, the theoretical models concerning them are immature and
not well-established.
Ivañez and Santiuste [34] studied the high-velocity impact response of sandwich
plates, with E-glass fiber/polyester face-sheets and foam core. They used 3D finite-
element model in ABAQUS Explicit to simulate the material behavior. For the
facesheets, they implemented Hou failure criteria [35] and a procedure to degrade
modeled it as a crushable foam material. The author concluded that under certain impact
velocities (280 in/s), the facesheets absorbed most of the impact energy and the foam
core’s influence was negligible. Fiber failure was found to be the main failure mechanism
in the composite facesheets, and it was responsible for the damaged area produced in the
facesheets.
Recent technological advances have been applied to increase the crushing strength
of polymer foams. If the facesheets and foam core do not fracture and are still bonded
13
together well, the foam-core will experience load-unload cycles and hysteresis due to
relative motion of facesheets. To better understand how hysteresis develops in the foam,
two separate finite element analyses of glass/vinylester-PVC H100 foam core sandwich
constructions are considered. They are described in Figure 2.2 and Figure 2.3, as a
projectile impact into a sandwich panel and blast loading of a sandwich shell,
respectively. The facesheets are orthotropic elastic and the PVC core is modeled as
2
Compressive Stress (MPa)
1.5
1
A● 0.5
B● 0 A
C● -0.5
B
C
-1
-1.5
-2
-0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Compressive Strain
(a) (b)
14
2
(a) (b)
shown for the impacted sandwich panel and the blast-loaded sandwich shell in Figures
2.2(b) and 2.3(b). Elastic-plastic waves propagate in the foam and reflect off the
interfaces within the sandwich structures. The elastic waves travel faster because of the
initial high modulus of the foam before it yields. There is hysteresis due to plasticity but
no viscoelasticity. The FEA simulations revealed some important facts about the foam.
The propagation of stress waves within the foam is controlled by the dynamics of the
facesheets and structure as a whole. In the case of projectile impact, the foam is
compressed beyond the crushing (yield) strength of the foam in the impact zone (Points A
or B in Figure 2.2 (a)) but this compressive stress is relieved when the projectile
15
rebounds. Tensile residual stresses and compressive permanent strain remain in the
impact region of the foam. Point C in Figure 2.2 (a) experiences only elastic vibrations.
Plastic deformation in a fully dense metal occurs at constant volume. Because of this, the
yield criterion that characterizes the plastic behavior in metals is independent of mean
stress. Deshpande and Fleck [37], who studied the yield criterion and the plastic response
of metal foams under various combinations of axial and radial compression, tension, and
shear loading, found that the yield surface in the stress space is adequately described by a
quadratic function of mean stress versus effective stress, with the hydrostatic yield
strength comparable to the uniaxial yield strength. Deshpande-Fleck claimed their foam
plasticity model could be applied for the open and closed-cell aluminum alloy foams.
They assumed the material is isotropic and the yield criterion is given as
̂ (2.1)
̂ [ ] (2.2)
[ ( ) ]
√ ⁄
| | (2.3)
16
and is a parameter that defines the aspect ratio of the ellipse in the - space. When
̂ reduced to . The value of is greater than zero and lies in a certain range.
For instance, in Alporas and Duocel foams, is in the range 1.35 to 2.08. The parameter
⁄
(2.4)
After defining the shape of the yield surface, the next step is to describe how the
yield surface evolves with strain. For an isotropic material, the applied associated flow
rule, which means that the change of stress with plastic strain, is the same in all
directions. Hence, the plastic strain rate is normal to the yield surface or
̇ ̇̂ (2.5)
where ̇̂ is the equivalent plastic strain rate, and it is the work conjugate of ̂, such that
̂ ̂̇ ̇ (2.6)
( ⁄ ) [ ̂ ̂
] (2.7)
̂̇ ( ⁄ ) [ ̇ ̇ ] (2.8)
17
A hardening curve is specified by the evolution of ̂ with ̂ . The uniaxial σ-ε curve is
often used to define ̂- ̂ .The true stress σ and logarithmic (true) plastic strain , may be
defined as
̇ ̇⁄ (2.9)
The crushable foam plasticity models in ABAQUS are intended for analyzing the
foams that are typically used as energy absorption structures. Two hardening models for
foam plastic behavior are available: the isotropic hardening model and the volumetric
hardening model. Both models use the von-Mises yield surface with an ellipse that
depends on deviatoric stress and meridional stress (p-q). The hardening curve describes
the uniaxial compression yield stress as a function of the corresponding plastic strain, and
the increment of plastic strain could be described as the plastic flow rule ( ̇ ̅̇ ).
The true (Cauchy) stress and logarithmic strain values should be given in defining the
As shown in Figure 2.4, the yield ellipse of the isotropic hardening model is
centered at the origin in the p–q stress plane. The yield surface evolves in a geometrically
self-similar manner. This phenomenological isotropic model was the one originally
developed for metallic foams by Deshpande and Fleck [9] and described earlier.
18
Figure 2.4 Crushable foam model with isotropic hardening:
yield surface and flow potential in the p–q stress plane [11].
The volumetric hardening model was proposed by Wilde [41] as shown in Figure
2.5, where the point on the yield ellipse in the meridional plane that represents
hydrostatic tension loading is fixed and the evolution of the yield surface is driven by the
However, under tension loading, the isotropic hardening model predicts the same
19
behavior in compression, while volumetric hardening model assumes a perfectly plastic
Current crushable foam plasticity models in ABAQUS could simulate the material
response under monotonic loading very well, but lack of the ability for simulating
hysteresis. Figures 2.6 (a) and (b) are illustrations of the stress-strain responses for
current crushable foam models. For both isotropic hardening and volumetric hardening,
yielding occurs with the initial yield stress of . For isotropic hardening, plastic flow
occurs after yielding up to ; this value affects the yield stress during unloading so that
it is a negative . For volumetric hardening, the yield stress in tension does not change
with plastic strain, but the yield stress for unloading is negative .
20
(a) (b)
Figure 2.7 illustrates the Bauschinger effect with isotropic hardening. The result indicates
that the unloading yield stress drops with Bauschinger effect to negative . For
example, an increase in tensile yield strength occurs at the expense of compressive yield
strength.
21
Figure 2.7 Bauschinger effect.
Figures 2.8 and 2.9 present the phenomena of modulus reduction and modulus
degradation, respectively. Figure 2.8 (a) is the modulus reduction shear stress- strain
( curve. It indicates that each time the material is loaded with an increasing shear
strain amplitude, the shear modulus ) decreases. This may be attributed to the changes
in the material's microstructure, for instance, the breakage and recovery of physical bonds
linking adjacent clusters. Figure 2.9 (a) is the shear stress- strain ( curve of one
specimen with a constraint shear strain amplitude. After each cycle, the material is
22
damaged. Therefore, the shear modulus ) decreases during the cyclic loading. The
material has its highest shear modulus after being loaded the first time. Both Figure
2.8 (b) and Figure 2.9 (b) indicate that the ratio of shear modulus ( will decrease
The unloading slope for both isotropic hardening model and volumetric hardening
are constant values, the Young’s modulus. Therefore, current crushable foam plasticity
(a)
(b)
23
(a)
(b)
The theories for metallic foams do not describe the polymer foam accurately.
There are different hystereses in the metallic and the polymer foams after compressive
yielding as shown in Figures 2.10 (a) and (b), respectively. The difference between
metallic foam and polymer foam is the unloading response, which in the metallic foam is
elastic but in the polymer foam is viscoelastic. In polymer foams, viscoelastic energy is
(permanent compressive strain) due to the collapse of cell walls. The PVC foam exhibits
not only viscoelasticity but also a Mullins effect. Such history effects have been well
24
Compressive Cauchy axial stress (Mpa)
True axial strain -
(a) (b)
2.3.2 Viscoelasticity
behavior of materials, such as metals, polymers, soils, concrete, rocks, etc. Viscoelasticity
consists of an elastic component (E) and a viscous component (η). As shown in Figs.
2.11 (a), (b) and (c), the Maxwell model, and the Voigt model are two phenomenological-
based models. The linear spring-Maxwell (Figure 2.11 (c)) is a three-parameter model
with a Maxwell element in parallel with an elastic spring. Unlike elastic materials, which
do not dissipate energy or have hysteresis after loads are removed, viscoelastic materials
do. Hysteresis is observed in the stress-strain curve, with the area of the loop being equal
to the energy lost during each loading-unloading cycle. The two main characteristics
associated with viscoelastic materials are stress relaxation and creep. Viscoelastic
25
behaviour may be modeled based on experimental results observed in a creep test or a
relaxation test.
deformation viscoelasticity which provided significant influence for the later work in this
field. The theory is compatible with the second law of thermodynamics in form of the
Clausius Duhem inequality. However, the nonlinear rate dependence and equilibrium
permanent strain. Ti et al. [40] reviewed various constitutive models that have been
Their study indicated that for simple analysis of modeling elasto-viscoplastic material, a
26
2.4 Energy Dissipation under Cyclic Loading
The stress-strain relation from preliminary experiments indicated that PVC foam
exhibit similar behavior as rubber under cyclic loadings, which is known as Mullins
effect. The Mullins effect was first investigated and studied by Mullins and his co-
workers back in 1947 [57]. Also known as Mullins softening, the Mullins effect is a
damage effect usually used to describe the mechanical response of rubbers-like material
under cyclic loadings. It results in a lower stress for the same strain after material
reloaded of each cycle. As shown in Figure 2.12, the closed loop of stress-strain curve
represents the hysteresis or the energy lost which is an irreversible phenomenon. The
stress t is plotted against λ, which is the eigenvalue of the Cauchy-Green tensor. The
curve is the primary loading plot under simple tension. Considered the loading
path which is the primary loading terminated at an arbitrary point , then unloading
followed path Ba. When the material is reloaded, the path is retraced as aB , and if
further loading is applied, it followed the path c, which is the continuation of the
primary loading path after . If further loading is applied to then the path Ca is
beyond is applied then the curve aC represents the subsequent material response,
which is then elastic. For loading beyond , the primary path is again followed and the
pattern described is repeated. Mullins effect, together with large deformation, non- linear
and hyperelastic behavior become the complexity of the mechanical behavior of rubbers.
27
Figure 2.12 Schematic loading-unloading curves in simple tension [45].
These may vary from a micro-mechanics level, such as the broken chain at the interface
between the rubber and the fillers, slipping of molecules, rupture of the clusters of fillers,
Most of the studies of Mullins effect are on rubbers under uniaxial tension as
opposed to multi-axial tension. Some remarkable articles in the literature are Ogden and
Roxburgh (1999) for Mullins effect [45] and the extension of this model by Dorfmann
and Ogden (2004) which includes the permanent set phenomenon. [58].
28
In Simo’s method[56], the strain energy function is augmented by a damage
function :
(4.31)
(4.32)
Both and are scalar functions that can be used to define microstructural damage of
Traditional Mullins effect theories do not consider the residual strain upon
energy function. Dorfmann and Ogden [58] proposed a stress softening model in rubber
together with associated residual strain effects. Their results showed how the stress
softening and residual strain change with the magnitude of the applied strain. They
modified the original Ogden-Roxburgh strain energy function by adding a new function
to describe Mullins effect with residual strain. The damage function includes two
variables in the energy function in order separately to capture the stress softening and
(4.33)
29
where N is the function introduced to characterize the residual strains and , are the
dissipation functions that are subjected to (1) = 0, (1) = 0. Note that when = 1,
which leads to the investigation of theories that could combine the existing theories for
different physical phenomena. Since individual models were developed within the pre-
models. Coupled elastoplasticity with continuum damage mechanics (CDM) has been
investigated back in 1987 by Simo and Ju [46] and later by Yazdani and Schreyer [47].
Hansen and Schreyer [48] developed a rigorous thermodynamic framework with coupled
elastoplastic and damage theories. It was the combination of anisotropic plasticity and
anisotropic damage formulations. The authors claimed that the concept of effective stress
was the critical mechanism for coupling these theories. The coupled formulations were
applied to an aluminum alloy, and it was found that the results were as expected. Borst et
al. [49] investigated two theories: one was the combination of gradient plasticity with
scalar damage and the other was combination of isotropic plasticity with gradient
mentioned the sensitivity of the results with respect to the discretization and analyzed
some model parameters. The two theories were valid for finite element simulations of
30
gradient-damage with hardening plasticity could be used to predict fracture mechanisms
31
CHAPTER III
EXPERIMENTS
material data are needed in various test modes. Special apparatus for determining the
and pure shear tests in both out-of-plane and in-plane directions are designed. This
chapter begins with an introduction of Divinycell PVC H100 foam and then describes the
3.1 Specimens
The foam used for the tests in this study is Divinycell PVC H100. Table 3.1 lists
the material properties provided by DIAB [2]. Divinycell PVC H100 is a solid closed-
32
Table 3.1 Characteristics Divinycell H100 [2].
The PVC H100 material was purchased from the Aircraft Spruce & Specialty
3-direction is defined as the out-of-plane direction, and the orientation of 1- and 2- are
defined as the in-plane directions. The specimens were cut by a slitting saw operated on
Bridgeport mill machine. Examine four specimens cut from the sheet as shown in Figure
3.1. If the load is applied along the 3-direction of the cubic specimen (left lower corner),
the test is an out-of-plane compressive test. When the load is applied along either 1- or 2-
direction of the cubic specimen (right lower corner), the test is defined as an in-plane
compressive test. If the load is applied along the solid arrow directions of the rectangular
specimen (right higher corner), the test is an out-of-plane shear test. When the load is
applied along the open arrow directions of the block specimen (left higher corner), the
33
Figure 3.1 Divinycell PVC H100 material sheet.
verify the manufacturer-published properties [2] and analyze the strain effect on PVC
H100 foam. In this research, the strain is limited to 20%, long before the densification
region, and strain rates from 10-4-0.1 s-1. At a strain rate of 0.1 s-1 , the monotonic
compressive test in the in-plane directions is compared with that of the out-of-plane
Cyclic compression tests and pure shear tests in both the out-of-plane and in-plane
directions were also done under different strain amplitudes and strain rates. The strain
rates were varied from 5×10-4 to 5 s-1, over five orders of magnitude. For each strain rate
of cyclic compression tests, the strain amplitudes ranged from 2-10% with 2%
increments. For the pure shear tests, the strain amplitudes ranged from 4-20% with 4%
34
increments on the foam specimen. A parametric study on the combination of strain
amplitude and strain rates that can be achieved by adjusting the PID (Proportional–
The cyclic compression strain history at strain rate equal to 5 s-1 with a strain
amplitude equal to 4% is shown as shown in Figure 3.2. The strain rate is given by
rate equal to 5 s-1 and strain amplitude equal to 4%, the MTS machine was required to
operate at a frequency of 62.5 Hz. For the same strain rate, the 2% strain amplitude test
would require the MTS machine to operate at a frequency of 125 Hz, etc. In addition to
the constant strain rate requirement, the waveform has to be a sawtooth shape and it is
very difficult to tune the MTS machine with this type of wave form at high frequency.
The PID tuning parameters of the MTS machine were adjusted of produce the correct
waveform over a range of strain amplitudes and frequencies. Finally, to collect enough
data points for the output files and capture the correct waveform at high frequency, the
data acquisition was also adjusted. Take again as an example, a strain rate equal to 5 s-1
with a strain amplitude equal to 4%, as demonstrated in Figure 3.2. If one requires 100
points during the first loading cycle, the data acquisition should be set at or ,
which means the time between two points is 0.00008 s or data per second.
35
T
ε
∆t
t
𝜀 5
-0.04
Figure 3.2 Strain history at a strain rate 5 s-1 and strain amplitude 4%.
Table 3.2 and Table 3.3 show the test frequencies for the cyclic compression test
and pure shear test, respectively, after tuning the MTS servo-hydraulic machine.
2 4 6 8 10
*Cyclic test under these combinations of strain amplitudes and frequencies could not
be achieved by the controller.
36
Table 3.3 Test frequency of MTS machine in pure shear tests.
4 8 12 16 20
The apparatus and setup for monotonic tests and compression cyclic tests are
basically the same. For monotonic tension tests and cyclic compression tests, the
specimen needs to be glued onto aluminum grips as shown in Figure 3.3 because of the
tensile loads. The adhesion agent used was LOCTITE® Quick SetTM. It can provide a
maximum of 3,211 psi shear strength on sand blasted, cold rolled steel after 24 hours of
curing [59]. It comes in two separate bottles. One is the resin, and the other is the
hardener. Every time the solvents should be mixed with a 1:1 ratio and blended well with
wood sticks. Experience indicated that the mixture should be used within 1 minute after it
is exposed to air. However, the foremost thing that should be seriously considered is that
the surfaces of foam and aluminum plates should be clean, dry and free of grease or oil.
First one side of the specimen was glued and cured for 30 minutes until the glue set. Then
37
to eliminate any MTS machine misalignment, the other side of the specimen was glued
and pre-set to cure in the MTS machine. After another 30 minutes, the specimen was
taken out of the MTS machine and cured for 24 hours in a room temperature
environment. After each test, the specimen is broken in two parts by MTS machine under
displacement control. The majority of the foam that left on the aluminum grips can be
removed by scraping with blades. Sandpapers are then used to further remove any
residual foam on the grips. Finally, the grip surface is cleaned with alcohol before the
There are two standard test methods for compression/tension tests of foams: one
is ASTM 1621 (Compressive Properties of Rigid Cellular Plastics) and the other is ISO
844 (Rigid Cellular Plastics Compression Properties). In this research, the ISO 844 (Rigid
shown in Figure 3.4. The specimen is a 1in x 1in x1in cube, and it was glued onto
aluminum grips, with diameter of 1.5in with 0.75in thickness. The directions of the
specimen are shown in Figure 3.1. A pair of “C” clamp was specially designed to connect
the actuator arm of MTS and the test piece, as shown in Figures 3.5. (a) and (b). The
hollow inside of the “C” clamps is a frisbee shape which can easily hold the actuator arm
of the MTS machine and the test piece. To eliminate the possibility of generating
additional stress along the vertical direction, the “C” clamps were connected by screws
orientated horizontally.
38
Figure 3.3 PVC H100 glued on Aluminum gripes.
Figure 3.4 Set up for compressive cyclic test for PVC H100.
39
(a)
(b)
The ISO 844 test standard gives a method to define the linear portion in the elastic
region on the stress-strain curve. As shown in Figure 3.6, the nonlinear portions at the
beginning and the end of elastic region are cut off, and the remaining portion is
40
defined as the linear elastic region. The slope of portion is used to define the Young’s
Figure 3.7 is the stress-strain curve obtained from the test at a strain rate of 0.1 s-1.
The test data was processed following the ISO 844 method mentioned above. As shown
in Figure 3.10, the original test data is the solid black curve, which has nonlinear portions
at the beginning and the end in the elastic region. After the cut off the nonlinear portions,
the Young’s modulus has been defined as 8,940.4 psi. The curve is then shifted to the
origin to get rid of the nonlinear portion at the beginning of the test. The yield strain
associated with the shifted curve is found where the linear portion ends, and this is
Tests were done at various strain rates to show that the material is viscoelastic and
viscoplastic, in Figure 3.8. Here, the viscoelasticity is signified by a slight change in the
41
Young’s modulus as the strain rate increases. The viscoplasticity is signified by the
change in the yield strength or flow stress as the strain rate increase.
Figure 3.7 Out-of-plane monotonic test under strain rate of 0.1 s-1.
42
Table 3.4 summarizes the strain rate effect for the material. The actual stress-
strain curves at different strain rates are given in Appendix A. As strain rates increase,
compressive modulus (slope of the elastic portion), compressive strength (stress at the
yield point) and plateau stress (stress in the plateau region) increase. For different strain
US Unit
Strain Rate Compressive Modulus Compressive Strength Plateau Stress Yield Strain
(s-1) (psi) (psi) (psi) (%)
0.0001 7629.9 237 240 3.06
0.0010 8506.1 259 256 3.00
0.0100 8780.9 268 272 3.02
0.1000 8940.0 289 295 3.08
With respect to the material directions defined in Figure 3.1, tests indicated the
material behaves the same in the 1- and 2- directions, but the stress-strain curve in the 1-
and 2-directions are different from that in the 3-direction. Hence, it could be concluded
that PVC H100 is a transversely isotropic material. Figure 3.9 shows a comparison of the
stress-strain curves in the out-of-plane and in-plane directions. It is obvious that the
modulus and strength are different. The strength in the in-plane direction is 40% lower
43
Figure 3.9 Out-of-plane vs. in-plane at strain rate of 0.1 s-1.
aluminum grips. Figure 3.10 shows the difference between these two situations at strain
rate of 0.1 s-1. The compressive modulus of the glued test (13,702 psi) is higher compared
with the unglued test (8,940 psi). The compressive strength (stress at yield), as shown by
Points A and B in Figure 3.10 occurred at the same stress value. Plateau stresses are
44
A
B
To verify the accuracy of test setup, the results has been compared to material
properties provided from DIAB in Table 3.1. Notice that DIAB’s tests were based on
ASTM standard D1621, which is different from the ISO 844 that was used in this
compressive modulus will be doubled following ASTM D 1621 when comparing to ISO
844. The ASTM D 1621 is the test standard under strain rate of 0.00167 s-1, and the
results obtained from the test standard in Table 3.1 are for unglued specimens under
different strain rates. An assumption was made to obtain the compressive modulus for
glued specimen at a strain rate of 0.001 s-1. Assume a scalar exists for compressive
modulus for glued and unglued conditions. The scalar is assumed to be the ratio of
compressive modulus for glued over compressive modulus for unglued. As displayed in
Figure 3.10, the scalar k=13,702/8,940. Apply this scalar to strain rate of 0.001 s-1, the
45
glued compressive modulus could be calculate out as 8,506 psi (Table 3.4), which is
roughly equals to 13,000 psi. Double this value and compare it with the data listed by in
Table 3.1. There will be a 30% over prediction for glued specimen at strain rate of 0.001
s-1. This difference is due to the size effect of the test specimen.
The tensile modulus has been compared with the compressive modulus after the
foam glued on the aluminum grips. As shown in Figure 3.11, tensile modulus is 15%
higher than compressive modulus. In the later study, this difference will be neglect.
As mentioned earlier, the PVC H100 was glued to aluminum grips so that the
specimen could be loaded in compression and tension. The cyclic compression tests in
both out-of-plane and in-plane directions were done on PVC H100 foams and the stress-
46
strain curves of different strain amplitudes at various strain rates were obtained. Details
of the results from these tests are provided in Appendix B. The stress-strain response at
strain amplitude of 10% at different strain rates in out-of-plane and in-plane are shown in
Figure 3.12 and Figure 3.13, respectively. The experiments indicated that the PVC H100
exhibits not only viscoelasticity but viscoplasticity with a permanent (plastic) strain upon
unloading to a zero stress state in both the out-of-plane and in-plane directions.
47
Figure 3.13 In-plane compressive stress-strain curves
at strain amplitude of 0.10 with various strain rates.
Figure 3.14 is the stress-strain curves of different cycles for strain rate of 0.5 s-1,
with strain amplitude of 0.1 in the out-of-plane direction. Figure 3.15 is the stress trend
for strain rate of 0.5 s-1 with various strain amplitudes in the out-of-plane direction. Both
figures indicated that PVC H100 has modulus degradation, as reviewed in Chapter II.
48
Figure 3.15 Out-of-plane compressive stress-strain curves
at strain rate of 0.5 s-1 with various strain amplitudes.
The cyclic behavior with strain rates and strain amplitudes in both out-of-plane
and in-plane directions are shown in Figure 3.16 (a) and (b). Ten loading-unloading
cycles were applied at strain amplitude of 2% with strain rate of 0.5 s-1, and this was
10%. In both out-of-plane and in-plane directions, it is noted that the first cycle of the
stresses at the same strain amplitude is higher than those following cycles, which means
Figure 3.17 shows the first cycle in out-of-plane direction at a strain rate of 0.5s-1
with various strain amplitudes, superimposed on the data with consecutively increasing
strain amplitudes. After the first reloading at strain amplitude of 2%, the stress-strain
response went back to the primary curve beyond 2%. This phenomenon repeats itself for
the different strain amplitudes. The phenomenon shown in Figure 3.17 is described as
Mullins damage [11]. The hysteresis curves of the fourth cycle in the out-of-plane
49
direction in these tests are given in Figure 3.18 to indicate rate-dependent behavior even
after Mullins damage. Similar responses were also found in the in-plane cyclic
compressive tests.
(a)
(b)
Figure 3.16 Compressive stress-strain curve at a strain rate of 0.5 s-1
with various strain amplitudes: (a) out-of-plane and (b) in-plane.
50
Figure 3.17 Out-of-plane first cyclic of strain rates of 0.5 s-1
with various strain amplitudes over lamping on the primary curve.
Figure 3.18 Out-of-plane fourth cycle hysteresis at 10% strain at various strain rates.
51
3.3.4 Material homogeneity
deformation and stress state. The ASTM and ISO standards mentioned earlier achieve
this requirement. However, these standards were written for monotonic, quasi-static
The PVC H100 cubic specimens were loaded in both the in-plane and out-of-
plane directions with strain amplitude of 10% at the lowest (0.0005 s-1) and highest (5 s-1)
strain rates, and the results are shown in Figures 3.19 - 3.22. With a magnification factor
8, specific regions at the top (A1), center (A2), bottom (A3) and the left/right midsection
magnified results from Figure 3.19 indicated that at strain rate of 0.0005 s-1, the damaged
cells are uniform in the in-plane direction with strain amplitude of 10%. A similar
conclusion is made for the out-of-plane direction shown in Figure 3.20. At strain rate 5 s-
1
and stain amplitude of 10%, the damaged microstructure is also uniform throughout the
specimen in both the in-plane and out-of-plane directions, shown in Figure 3.21 and
Figure 3.22, respectively. Therefore, the conclusion is that PVC H100 deformed
uniformly at strain amplitude of 10% with strain rates from 0.0005 s-1 and 5 s-1 in both
The permanently crushed specimens when loaded to 10% strain amplitude in the
in-plane direction at five increasing strain rates are shown in Figure 3.23. Here it can be
52
seen that more permanent strain results at higher loading rates. A similar trend appears
for the out-of-plane direction, as shown in Figure 3.23. The damage to the specimen
appears to be history-dependent.
For higher strain amplitudes, the specimen did not crush in a homogenous
manner. It was observed that when the strain amplitude increased to 30% at strain rate of
0.0005 s-1, the material was not crushed homogeneously. The crushed microstructure at
30% strain in the in-plane and out-of-plane directions are shown in Figures 3.25 and 3.26.
Both the top (A1) and bottom (A3) cells were crushed more than at the mid-center (A2)
and sides (A4/A5). This non-uniformity was also observed at larger strain rate of 5 s-1 ,
53
(a)
(b)
(c)
Figure 3.19 PVC H100 crushed with 10% strain amplitude at strain rate of 0.0005 s-1 in
the in-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at A2 and
(c) Damaged micro-structures at various locations.
54
(a)
(b)
(c)
Figure 3.20 PVC H100 crushed with 10% strain amplitude at strain rate of 0.0005 s-1 in
the out-of-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at A2
and (c) Damaged micro-structures at various locations.
55
(a)
(b)
(c)
Figure 3.21 PVC H100 crushed with 10% strain amplitude at strain rate of 5 s-1 in the in-
plane direction: (a) Crushed specimen (b) Undamaged micro-structure at A2 and (c)
Damaged micro-structures at various locations.
56
(a)
(b)
(c)
Figure 3.22 PVC H100 crushed with 10% strain amplitude at strain rate of 5 s-1 in the
out-of-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at A2 and
(c) Damaged micro-structures at various locations.
57
0.0005 s-1 0.005 s-1 0.05 s-1 0.5 s-1 5 s-1
Figure 3.23 Crushed specimens loaded to 10% strain amplitude in the in-plane direction
with five increasing strain rates.
Figure 3.24 Crushed specimens loaded to 10% strain amplitude in the out-of-plane
direction with five increasing strain rates.
58
(a)
(b)
(c)
Figure 3.25 PVC H100 crushed with 30% strain amplitude at strain rate of 0.0005 s-1 in
the in-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at A2 and
(c) Damaged micro-structures at various locations.
59
(a)
(b)
(c)
Figure 3.26 PVC H100 crushed with 30% strain amplitude at strain rate of 0.0005 s-1 in
the out-of-plane direction: (a) Crushed specimen (b) Undamaged micro-structure at A2
and (c) Damaged micro-structures at various locations.
60
3.4 Pure shear tests
The design of the pure shear test was based on the ASTM C273 [60] and ISO
1922 [61]. The specimen is a 6in x1in x0.5in block, which is half of the size of standard
specimen. The directions of pure shear tests are defined as shown in Figure 3.1. For each
direction, the 6in x1in surface is glued to the steel plates, which are 6in x 2in x 0.75in
steel block. Unlike the compressive cyclic tests, the specimens in pure shear tests do not
have to be cured on the MTS machine. Because the glued surface area is larger than the
compressive specimen, the curing time of the specimen for pure shear tests was extended
to 48 hours.
The above-mentioned standard test methods are limited to monotonic loading but
not cyclic loading. A specimen test fixture was designed to enable cyclic loading as
showed in Figure 3.27. The design of a 1 in-diameter brass ball encased in aluminum
seats and connected to a stainless steel 304 rod (3/8in-24 threaded) provides freedom of
rotation of the top and bottom edge of the specimen, while eliminating the undesired
displacement. With a safety factor of 2, ABAQUS/Explicit was used to check the failure
of the ball and housing at critical contact regions, such as Point A and Point B as shown
61
MTS Actuator
Brass Ball
Seats
Steel Rods
Steel Plates
Specimens
Load Cell
(a) (b)
Figure 3.27 Pure shear test apparatus: (a) Schematic of shear text fixture and
(b)Photograph showing experimental setup.
62
Monotonic and cyclic shear tests were performed in the out-of-plane and in-plane
directions as explained in Section 3.2. After each test, the specimen is broken into two
parts while still attached to the plates. Most of the foam will be removed by a cutting
saw. The remaining foam on the plates will then be baked at high temperature. After the
plates are cooled down to room temperature, the burnt foam could simply be scraped off.
Bead blasting was used to take any remaining foam off the steel surface. This is the
process of removing surface deposits by applying fine glass/sand beads at a high pressure
without damaging the surface. Acetone was then applied to do the final cleaning.
To verify the accuracy of the results from pure shear test apparatus, current results
were compared to those obtained from Steeves and Fleck [37],the DIAB data sheet [2],
Because of the similar dimensions of the specimen and test mode, the current test
results were compared with those of Steeves and Fleck (2004) at a strain rate of 0.001 s-1.
As shown in Figure 3.28, the current test result of the out-of-plane pure shear test has the
same plastic flow stress as that found by Steeves and Fleck. However, the two curves
have different shear modulus and different shear fracture strain at load drop. Steeves and
Fleck indicated the shear modulus of PVC H100 was 44 MPa, which is twice of current
out-of-plane shear modulus test of 22 MPa. The shear stress-strain curve in Steeves and
63
Data from DIAB was used to further validate the current out-of-plane shear test
results. The out-of-plane shear fracture strain was quoted as 40% in the DIAB data sheet,
which is the same fracture strain as found in the current test. However, the shear modulus
of PVC H100 in the out-of-plane direction from the DIAB data sheet was 35 MPa, which
does not match neither that of the current test nor that of Steeves and Fleck (2004).
The current test results were also compared with out-of-plane shear stress-strain
curve from Daniel (2009). Daniel used an Arcan loading fixture and specimen holder to
test PVC H100 under pure shear. As shown in Figure 3.28, the out-of-plane shear stress-
strain curve from Daniel indicated that the shear modulus of PVC H100 was 25 MPa,
which is very close to that in the current test. However, there is no discernible load drop
to indicate a shear fracture strain in the Arcan shear tress-strain curve. This is probably
due to the unique shear stress state imposed by the notched specimen in an Arcan test.
The current test results matched the flow stress of Steeves and Fleck (2004), the
shear fracture strain from DIAB and shear modulus by Daniel (2009). Therefore, the
conclusion that may be drawn here is that the present test results are comparable to those
obtained by independent studies on PVC H100. It was realized that the fracture strain
from Steeves and Fleck (2004) was half of that specified in the DIAB data sheet.
Assuming the DIAB data sheet is a more reliable source, the stress-strain curve from
Steeves and Fleck (2004) was modified by doubling the shear strain on the horizontal
scale. Now the modified stress-strain curve of Steeves and Fleck has the same shear
modulus, the same shear fracture strain and comparable flow stress as the current test
64
result. The difference of current test result versus Steeves and Fleck (2004) could be due
to the definition of the engineering shear strain by Steeves and Fleck. By the test standard
of ASTM C273, the engineering shear strain is the displacement between load plates
derived by the thickness of the core (γ=u/t). It is possible that Steeves and Fleck mistook
the engineering shear strain with tensoral shear strain, the later quantity being half of the
former.
shear strain)
Daniel (2009)
Figure 3.28 Comparison of out-of-plane shear stress-strain curves of PVC H100 foam
core at strain rate =0.001 s-1 with data from Steeves and Fleck (2004) and Danel (2009).
The through-thickness shear fracture of PVC H100 foam core at shear strain rate
of 0.001 s-1 is shown in Figure 3.29. The crack occurred at the center of the specimen and
this indicates that the specimen broke in a state of pure shear and that there was good
bonding between the foam and steel plates. Failure in these tests could be mistaken as
debonding between the foam and steel grips. Clearly this was not the case in these
experiments.
65
Shear Crack
Figure 3.29. Out-of-plane shear fracture of PVC H100 foam core at strain rate =0.001 s-1.
Figure 3.30 shows the out-of-plane and in-plane of monotonic shear stress-strain
response at strain rate=0.5 s-1 with strain amplitude=0.20. The material is again
transversely isotropic. The ratio of the yield stresses of the out-of-plane compare to the
66
Figure 3.30 Monotonic shear test at strain rate=0.5 s-1, strain amplitude=0.20.
This stress-strain curve can also be described by three phases: elastic, plastic and
hysteretic with Mullins damage. However, hardening occurs during the plastic flow phase
Cyclic shear stress-strain responses are given in Figures 3.31 and 3.32.
The results of cyclic pure shear tests are similar to cyclic compression test results. As
shown in Figures 3.31 (a) and (b), the cyclic pure shear test in both the out-of-plane
directions and in-plane direction indicate that the material is viscoelastic, viscoplastic and
experience Mullins damage. The pure shear stress-strain curves at one particular strain
rate of 0.5 s-1 with various strain amplitudes in the out-of-plane and in-plane directions
are shown as in Figures 3.32 (a) and (b). More results at different stain rates and various
67
(a)
(b)
Figure 3.31 Pure shear stress-strain curves of strain amplitude of 20% at various strain
rates: (a) out-of-plane and (b) in-plane.
68
(a)
(b)
Figure 3.32 Pure shear stress-strain curves at a strain rate of 0.5 s-1 with various strain
amplitudes: (a) out-of-plane and (b) in-plane.
69
3.5 Summary and Conclusions
directions show that PVC H100 is linear elastic-plastic. The modulus of elasticity
(2) The PVC H100 foam displays transversely isotropic behavior with the
higher than in the in-plane direction. The ratio of out-of-plane to in-plane yield
stress is 3:2.
(3) The modulus and yield strength in monotonic compression stress-strain curves
vary with strain rate and this indicates that the material is viscoelastic and
viscoplastic.
(4) Cyclic compression stress-strain curves derived at different strain rates reveal that
(5) Monotonic shear stress-strain curves also indicated that PVC H100 foam displays
(6) The cyclic shear test results have a similar stress-strain response as the cyclic
compressive tests. The material under cyclic shear is viscoelastic, viscoplastic and
hysteretic with Mullins damage. However, shear hardening occurs after yielding.
70
CHAPTER IV
CONSTITUTIVE MODELING
To simulate the behavior of polymer foam under cyclic compression, two one-
dimensional constitutive models are developed in this chapter. The first model is based
model is based on the theory of damage mechanics. Only the uniaxial compression mode
is considered in this chapter. Mullins and viscoelastic damage predictions will only be
made for the out-of-plane compressive mode since the in-plane compression mode is very
similar.
The stress-strain relation of PVC H100 under uniaxial loading indicates that the
as described in Figure 4.1. Based on this material behavior, the phenomenological one-
The mechanical analogs model is shown in Figure 4.2. An equilibrium spring with
modulus is in parallel to a Maxwell model (spring with modulus and dashpot with
71
η) to simulate viscoelastic response. These are then in series with a Prandtl element with
μ, for viscoplasticity.
Viscoplasticity
Viscoelasticity
Energy Dissipation
72
The total stress and strain at any time are given by
(4.1)
and
(4.2)
4.1.1 Viscoelasticity
Figures 4.3 (a) and (b) are the experimental results in both the out-of-plane and
in-plane directions. For each strain rate, the stress-strain relation is linear. The stress-
strain curves also show rate dependency. As strain rate increases, the modulus increases.
73
(a)
(b)
(fluid) and elastic behavior (solid). The simplest mechanical analogs for a viscoelastic
material are Maxwell model and Voigt model (also known as the Kelvin-Voigt model),
which were reviewed in Chapter II. The mechanical analog for viscoelastic in this
74
research is as shown in Figure 4.4. It could be described as a linear spring in parallel with
a Maxwell element.
dependent viscoelastic component. The elastic component consists of a linear spring with
spring constant of ; the rate dependent viscoelastic component consists another linear
spring with spring constant of , and a nonlinear viscous dashpot. As shown in Figure
4.4, it is clear from the geometry of the model that the total strain will be the strain of the
(4.3)
(4.4)
75
The equilibrium stress is described by a spring constant of . The spring with
spring constant of corresponds to the behavior of foam in the most relaxed state,
(4.5)
The overstress is the change in stress from the equilibrium stress associated with
̇ (4.6)
̇ (4.7)
where . Here, is a constant strain rate, as shown in Figure 4.5. Equation (4.7)
can be rewritten as
- (4.8)
76
ε
2εmax
-2εmax
Figure 4.5 Relationship of strain verses time for loading, unloading, and reloading.
Solving Equation (4.8) with the condition =0, when ε=0, one gets
[ ( ) ] (4.9)
77
[ ( )] (4.10)
[ ( )] (4.11)
(lowest strain rate in this research). In the out-of-plane experiments, was found to be
13940 psi, and in the in-plane experiments, was 7916 psi. The value of and η are
the two material properties which need to be defined from the monotonic compressive
OriginLab [51], was utilized for curve fitting. Table 4.1 is a summary of the values of
and η for different strain rates in the out-of-plane and the in-plane directions.
Table 4.1 Material properties of out-of-plane and in-plane after curve fitting.
Out-of-plane In-plane
Strain rate (s-1) (psi) η (psi·s) (psi) η (psi·s)
0.0005 1980 428 473 56
0.0050 2250 407 486 66
0.0500 2310 539 580 68
0.5000 2430 550 811 78
5.0000 2530 576 650 82
Average 2300 500 600 70
78
Note that there are variable sets of solutions for this nonlinear curve fit. In this
research, the solutions were set in the certain range and values of and η are assumed
to be constant in this research. After some preliminary studies, the average values are
Out-of-plane In-plane
(psi) 13940 7916
(psi) 2300 600
η (psi·s) 500 70
Figures 4.6 (a) and (b) are the results of fitted curves. Both the out-of-plane and
in-plane show the trend of viscoelsticity. More details are given in Appendix D. Table 4.3
provides the comparison results of test data to the fitted curves in out-of-plane and in-
plane directions. The comparison errors are within 2%, which is very good.
79
(a)
(b)
80
Table 4.3 Comparison of results for viscoelasticity in out-of-plane and in-plane
directions.
Out-of-plane In-plane
Strain rate Peak stress Predicted peak Diff. Predicted stress Predicted peak Diff.
(s-1) (psi) (psi) (%) (psi) (psi) (%)
0.0005 160.29 160.52 0.14 80.26 81.45 1.48
0.0050 162.53 162.2 -0.20 82.68 82.13 -0.66
0.0500 180.89 177.28 -2.00 84.25 84.15 -0.12
0.5000 186.73 186.48 -0.13 85.92 87.33 1.64
5.0000 192.04 188.76 -1.71 88.17 88.01 -0.19
4.1.2 Viscoplasticity
Experiments on metals show that under uniaxial loading, the strain at a given
stress has two parts: a small recoverable elastic strain ( ), and an irreversible plastic
Allan [52] reviewed five key concepts in modeling metal plasticity, as below:
81
2. Yield criteria, which predict whether the solid responds elastically or
plastically. For instance, for an isotropic metal, yield occurs when von Mises effective
3. Strain hardening rules, which control the way in which resistance to plastic
flow increases with plastic straining. There are two simple approaches, one is isotropic
hardening, and the other is kinematic hardening, as shown in Figure 4.8 (a) and (b).
Figure 4.7 (a) indicated that the isotropic hardening law is not useful where components
are subjected to cyclic loading. On the other hand, kinematic hardening law allows the
yield surface to translate, without changing its shape. This type of behavior better
(a) (b)
Figure 4.8 Plastic hardening: (a) Isotropic hardening and (b) Kinematic hardening.
4. The plastic flow rule, which determines the relationship between stress and
show that plastic flow is irreversible, and the material always dissipates energy.
82
The plastic response of foam differs fundamentally from metals. The models for
foams generally differ in three respects (i) the yield criterion; (ii) the strain hardening
The yield function defined the elastic domain, where the material behavior is assumed to
| | (4.12)
where is the yield stress. When | | , the material behavior is purely elastic. The
strain hardening law describe the changes in yield stress that result from plastic strain. It
could be describe as
̅ (4.13)
In this research, the yield stress was taken to be constant (perfectly plastic
behavior). This is because the compressive strains are limited to less than 10% and the
densification and hardening is insignificant at such small strain. The rate of term ̅ is
̇̅ ̇ (4.14)
̇ ̇ (4.15)
83
where sign is the signum function defined as
sgn(a)={ (4.16)
The plastic multiplier ̇ is a function describes how the rate of plastic strain varies
with the level of stress. There are many forms that are possible for ̇ . Based on the power
law form of the viscoplastic potential proposed by Perzyna [54] and Peric [55], it was
assumed that
| |
[( ) ]
̇ ̇̅ { (4.17)
dimensional rate-sensitivity parameter. Both μ and β are material constant and strictly
positive.
(4.18)
= (4.19)
(4.20)
( ) ̇ (4.21)
84
Hence, the total stress became
( ) (4.22)
Figures 4.9 (a) and (b) are the plastic responses of PVC H100 in the out-of-plane
and in-plane directions. Both graphs indicate that the material exhibits viscoplasticity.
(a)
(b)
Figure 4.9 Experimental stress-strain curves showing viscoplasticity:
(a) Out-of-plane and (b) In-plane.
85
Equation (4.20) is rewritten as
́ (4.23)
where .
( )
́ {[ ] } (4.24)
A MATLAB program, using ode45 solver, was written to curve fitting the plastic
Out-of-plane In-plane
β 0.0430 0.0606
μ (s) 27.5 E6 19.0 E6
Figure 4.10 (a) and (b) are the fitted curves for viscoplasticity. Table 4.5 provides
the comparison results of test data to the fitted curves in out-of-plane and in-plane
directions. More comparison details are shown in Appendix E. The difference between
test and predicted results for the out-of-plane direction is 5%. However, the difference for
86
(a)
(b)
87
Table 4.5 Comparison results for viscoplasticity in out-of-plane and in-plane directions.
Out-of-plane In-plane
Strain Rate Test Flow Predicted Flow Diff. Test Flow Predicted Flow Diff.
(s-1) stress (psi) stress (psi) (%) stress (psi) stress (psi) (%)
0.0005 243.47 241.053 -0.99 149.97 143.30 -4.44
0.0050 265.85 266.147 0.11 171.76 164.24 -4.38
0.0500 283.87 293.145 3.27 178.56 188.76 5.71
0.5000 310.94 324.256 4.28 196.07 217.09 10.72
5.0000 339.33 358.182 5.56 209.70 249.63 19.04
strain εmax. The coordinate position of Point A is (εmax, σys) and it becomes the beginning
point of the unloading. Before unloading, two values are defined: one is εv1, which is the
inelastic stain associated with εmax; the other is εp1, which is the plastic strain at εmax. The
values of εv1 and εp1, are given in Appendix F and were obtained from a MATLAB
program.
88
As reviewed in Chapter II, traditional methods to account for damage or so-called
Mullins softening are given by Simo [56] and Ogden-Roxburgh [45]. Both methods
derive the stress-strain relationship from strain energy density function. Because the
methods were initially intended to study the behavior of rubber-like materials under
cyclic loading, the Mullins effect was usually applied to materials under large
deformation, also called finite deformations. However, this research focuses on small
strain. Stress in finite deformations is usually correlated to the first and second Piola–
Kirchhoff stress tensors, the Biot stress tensor, or the Kirchhoff stress tensor. For small
strain, stress is related to the Cauchy stress tensor, which is the same as the second piola-
There are differences of the studies of rubber-like materials and PVC foam.
Theories for rubber-like materials are derived for finite strain, while the theory for PVC
foam is derived for small strain. Both rubbers and foams will exhibit residual strains. The
residual strain effect in rubbers during Mullins damage is introduced by including another
potential damage function [2004]. Residual strains are already incorporated in the foam
Test data is used to determine a modified damage model to be appropriate for the
(4.25)
89
-⁄
where , and and are material parameters. This function is
residual strains and stresses are already incorporated in the primary loading curve.
From experimental observations and test data analysis, the parameters , and
η which were defined from the initial compressive stress-strain curves now become
functions of d. The permanent strain is also a function of d for later unloading and
reloading. Therefore, the mechanical analog shown in Figure 4.2 is modified as shown in
Figure 4.12. The bars over , , η and denote that these are damaged functions of
d.
90
From the mechanical model showed in Figure 4.2, the stress-strain relation with
damage is
̅ ( ̅ ) ̅ ( ̅ ) (4.26)
̅ ̅ ( ̅ ) (4.27)
For unloading,
̅ ̅ ̅
( ̅ ̅
) [ ̅
] ̅
̅ (4.28)
̅ ̅ ̅
( ̅ ̅
) ̅ ̅̅̅̅̅
̅ (4.29)
For reloading,
̅ ̅ ̅
( ̅ ̅
) ( ̅
) ̅
̅ (4.30)
̅ (4.31)
91
Figure 4.13 demonstrates how to get the plastic strain damage function ̅ .
-
̅ - (4.32)
̅ { [ ] } (4.33)
̅ { [ ] } (4.34)
where and are two other material parameters, which could be evaluated by curve
fitting.
92
The OriginLab software program [34] was used for curve fitting. Figure 4.14
shows the experimental curve verses the simulation of Mullins damage at strain rate of
0.0005 s-1 in the out-of-plane direction. The priority of curve fitting is to match the shape
of hysteresis loop, catch the beginning and ending points of unloading and reloading, and
Figure 4.14 The experimental curve verses the simulation of Mullins damage at
strain rate of 0.0005 s-1 in the out-of-plane direction.
After curve fitting, the proper value of r, m, , and are found as, r=1.29,
m=0.035, C1=2 and C2=220. The values of r, m, and are material properties and
are therefore constants for one particular material. By keeping these 4 parameters the
same, but changing the strain rates, the stress-strain curves are shown as in Figure 4.15. A
comparison of the hysteresis at different strain rates in the out-of-plane direction indicates
that the method of applying Mullins effect to simulate the damage in polymer foam failed
93
to capture all the strain rate effect, as shown in Figure 4.15 (a)-(d). Furthermore, given a
set of values for r, m, , and cannot match the test data for different strain amplitudes
94
(a) (b)
(c) (d)
(e)
Figure 4.15 Comparison of the hysteresis at different strain rates in out-of-plane direction
with r=1.29, m=0.035, C1=2 and C2=220: (a) 0.0005 s-1, (b) 0.005 s-1, (c) 0.05 s-1, (d)
0.5 s-1and (e) 5 s-1.
95
(a) (b)
(c) (d)
96
4.2 Viscoelastic Damage Mechanics Model
criterion for the foam is associated with the onset of buckling or what was described as
(3.35)
The experiments indicate that for all the test strain rates, =0.023 for the out-of-plane
Figure 4.17 Damage initiations at , using actual test for out-of-plane compression.
The stress-strain response before damage at εcr is the same viscoelastic response
discussed in Section 4.1.1. In this new damage model, it is assumed that the damage
occurs right after strain exceeds εcr and the damage is controlled by the maximum strain
97
(εmax) and the history of loading or the flow stress ( ). Few things should be clarified,
such as how to determine the flow stress , and how to apply damage in the model. The
flow stress is determined at the end of viscoelastic response and just before damage
begins. As indicated in Figure 4.17, increases with increasing strain rate. This is also
shown in Figure 4.18 (a), where the stain rate effect is synonymous with flow stress or
history effect. The unloading curves for both strain rates are different because damage is
different. As the maximum strain increases at a fixed stain rate, the damage changes,
however. In general, there is more hysteresis with a larger maximum strain as explained
98
(a)
(b)
The critical strain at damage initiation from tests should not be used as the
predictive damage initiation criterion in the model because the previous predicted
response would not perfectly match test results. As shown in Figure 4.19, the dashed line
indicates that for different strain rates, the damage should initiate at a strain of 0.019 in
order to get similar flow stress as in the experiments. This is somewhat lower than the
critical strain from the test (see Fig. 4.17), but this is because the predicted viscoelastic
response yields higher stresses than the test. Once damage initiates, the flow stress is
constant with strain amplitude and only depends on the strain rate or previous history of
99
loading. For various strain rates, flow stresses are determined and these are given in
Table 4.1.
Figure 4.19 Determining flow stress at , using simulated viscoelastic response for
out-of-plane compression.
Strain rate
(s-1) 0.0005 0.005 0.05 0.5 5
Plastic stress
(psi) 245 264 287 305 325
Figures 4.20 (a) and (b). All the material parameters: , and η are modified with
the damage as indicated by the “bar”. From former discussion, it is known that all the
damaged parameters will be functions of the combination of and the history of the
100
(a)
(b)
From the mechanical model showed in Figure 4.20 (b), the stress-strain relation
with damage is
̅ ̅ ( -̅ ) (4.41)
The overstress is
̅ ̅ ̅ ̅ ̇ (4.42)
̅ ̅ ( - ) ̅
(̅ - -̅ ) [ ̅
] ̅
(4.43)
101
where, ̅ is the plastic strain at . This plastic strain depends on the damaged
, and becomes
̅ ̅ ̅
̅ (̅ - -̅ ) - ̅ ̅
(4.44)
The above equation is the initial condition for on reloading. For reloading,
̅ ̅ ̅
(̅ ̅
) (- ̅
) -̅ (4.45)
̅ ̅ ( -̅ ) (4.46)
-̅
̅ (4.47)
-̅
̅ -
(4.48)
̅ [ ( ) ]
(4.50)
102
Substituting Equations (4.48) and (4.49) into Equation (4.47) gives
-
-
̅ (4.50)
- - -
This new damage model can successfully describe both the strain rate and strain
amplitude effects found in the out-of-plane compression tests. The predicted curves are
comparable with test data at strain amplitude of 10% in the out-of-plane direction and
various strain rates. Results from Figures 4.21 (a)-(e) for strain rates 0.0005 s-1, 0.05 s-1
and 5 s-1 are superimposed in Figure 4.22 to show that the effect of increasing strain rate
The new damage model does not only capture strain rate effects but also the strain
amplitude effects. In Figure 4.23, predicted hysteresis at a fixed strain rate of 0.0005 s-1
and increasing strain amplitude or are compared with the out-of-plane test results.
Good agreement was found between the two, so that this new viscoelastic damage model
can simulate the correct response with increasing strain amplitude and fixed strain rate.
More comparisons of the simulation curves and experimental results are shown in
Appendix G.
103
(a) (b)
(c) (d)
(e)
Figure 4.21 Comparison of simulation curves with experimental data at strain amplitude
of 10% in the out-of-plane direction: (a) 0.0005 s-1, (b) 0.005 s-1, (c) 0.05 s-1, (d) 0.5 s-
1
and (d) 5 s-1.
104
Figure 4.22 Comparison of simulation curve and test data at various strain rates and fixed
strain amplitude of 10% in the out-of-plane direction.
Figure 4.23 Comparison of simulation curve and test data at various strain amplitudes and
fixed strain rate of 0.0005 s-1 in the out-of-plane direction.
105
CHAPTER V
constitutive model for polymer foams under cyclic loading. Monotonic and cyclic
compression and pure shear tests were done on Divnycell H100 foam to produce
hysteresis data in both in-plane and out-of-plane directions. The strain rates ranged from
0.0005 s-1 to 5 s-1. The compressive strain amplitudes were limited to less than 10%,
while the engineering shear strain amplitudes were less than 20%. Experimental data
indicates that PVC H100 foam displays transversely isotropic behavior under both
Two one-dimensional constitutive models were developed from the test data. The
first constitutive model is related to current crushable foam model which is based on
metallic foams. In this research, the model consisted of an equilibrium spring in parallel
with a Maxwell element and in series a Prandtl element. Damage functions were
introduced for permanent crushing of cells (plastic deformation) and progressive damage
of cells (Mullins effect). This constitutive model was able to capture yielding,
viscoelastic and viscoplastic response, as well as hysteresis of the foam, but it failed to
106
describe the strain-rate dependency that was found in the experiments. The second
constitutive model is based on damage initiation and evolution of the foam. It consisted
damaged spring and dashpot functions were found. The damaged functions were found to
depend on the magnitude and history of the deformation. Good agreement with out-of-
plane compression test results was reported. Similar modeling could be used for in-plane
constitutive modeling of polymer foams. The present study only addresses a one-
dimensional constitutive model for the foam. Future work must be done to examine the
the foam in multiaxial stress states and under cyclic loading. In addition to this, the
proposed model is limited to material behavior under a constant strain rate. In an actual
impact or dynamic problem the strain rates and strain at point may change in time. This
true history effect in the foam material has not been addressed, and would require further
107
REFERENCES
[1] D. Zenkert. (1997) The handbook of sandwich construction, The Chameleon Press
Ltd, London, United Kingdom.
[4] D.H. Pascoe. High tech materials in boat building, D. H. PASCOE & CO., Inc.
http://www.yachtsurvey.com/HiTech.htm
[5] L. Aktay, A.F. Johnson, and M. Holzapfel. (2005) Prediction of impact damage on
sandwich composite panels, Comput Mater Sci, 32: 252-60.
[6] S. Abrate. (1998) Impact on composite structures. Cambridge University Press, UK.
[7] U. Icardi, and L. Ferrero. (2009) Impact analysis of sandwich composites based on a
refined plate element with strain energy updating, Compos Struct, 89:35-51.
[9] V.S. Deshpande, and N.A. Fleck. (2000) Isotropic constitutive model for metallic
foams, Journal of the Mechanics and Physics of Solids, 48: 1253-1276.
[10] V.S. Deshpande, and N.A. Fleck. (2001) Multi-Axial Yield Behavior of Polymer
Foams, Acta Mater, 49: 1859-1866.
108
[11] ABAQUS, Inc. (1997) ABAQUS User Manual, Version 6.9.
[12] L.J. Gibson, and M.F. Ashby. (1982) The mechanics of three-dimensional cellular
materials, Proceedings of the Royal Society, London, 382: 43-59.
[13] L.J. Gibson, M.F. Ashby, G.S. Schajer, C.I. Robertson. (1982) The mechanics of
two-dimensional cellular materials, Proceedings of the Royal Society, London, 382: 25-
42.
[16] P.J. Blatz, and W.L. Ko. (1962) Application of finite elasticity to the deformation of
rubbery materials, Transactions of the Society of Rheology, 6: 223-251.
[17] R.W. Ogden. (1972) Large deformation isotropic elasticity – on the correlation of
theory and experiment for compressible rubberlike solids, Proceedings of the Royal
Society, London, 328: 567-583.
[18] R.W. Ogden, and D.G. Roxburgh. (1999) A pseudo-elastic model for the mullins
effect in filled rubber, Proceedings of the Royal Society, London, 455:2861-2877.
109
[21] M.S. Hoo Fatt, H. Surabhi, and Y. Gao. (2010) Blast response of sandwich shells
with crushable foam cores, in the Proceedings of the IMPLAST, Providence, RI, October
12-14, 2010.
[22] M.S. Hoo Fatt, and L. Palla. (2009) Analytical modeling of composite sandwich
panels under blast loads, Journal of Sandwich Structures and Materials, 11: 357-38.
[23] V.L.Tagarielli, V.S. Deshpande, and NA. Fleck. (2010) Prediction of the dynamic
response of composite sandwich beams under shock loading, International Journal of
Impact Engineering, 37: 854-864.
[24] D. Balkan, and Z. Mecitoglu. (2009) Dynamic response of sandwich plate with
viscoelastic core under blast load, in the proceedings of 7th EUROMECH Solid
Mechanics Conference, Lisbon, Portugal, September 7-11, 2009.
[25] Z. Xue, and J.W. Hutchinson. (2003) Preliminary assessment of sandwich plates
subjected to blast loads, International Journal of Mechanical Sciences, 45: 687-705.
[26] E. E. Gdoutos, I. M. Daniel, and K. A. Wang. (2002) Failure of cellular foams under
multiaxial loading, Applied Science and Manufacturing, 33(2):63-176.
[27] M. D. Isaac, J.J. Luo, and M.S. Patrick. (2007) Mechanical and failure
characterization of textile composites, in the Proceedings of 16th International
Conference on Composite Materials, Kyoto, Japan, July 8-13, 2007.
[28] F. Pan. (2008) Modeling of the Rate Responsive Behavior of Elastomer Foam
Materials, Journal of Engineering Materials and Technology, 130(1):1-6.
[30] F. Pan, S. Quander, and A. Ma. (2005) Use of material models for elastomer foams
in ABAQUS and LS-DYNA, in the Proceedings of the Fourth European Conference for
Constitutive Models for Rubber (ECCMR), 593-597, Stockholm, June 27-29, 2005.
110
[31] U. E. Ozturka, and G.Anlasb. (2011) Finite element analysis of expanded
polystyrene foam under multiple compressive loading and unloading, Materials &
Design, 32(2):773-780.
[32] P. M. Schubel, J. Luo, and I. M. Daniel. (2005) Low velocity impact behavior of
composite sandwich panels, Composites, 36:1389-1396.
[33] V.I. Rizov. (2006) Non-linear indentation behavior of foam core sandwich
composite materials-A 2D approach, Computational Materials Science, 35(2): 107-115.
[35] J.P. Hou, N. Petrinic, C. Ruiz, SR. Hallett. (2000) Prediction of impact damage in
composite Plates, Compos Sci Technol, 60(2):273–328.
[36] V.S. Deshpande, and N.A. Fleck. (2001) Multi-axial yield behaviour of polymer
foams, Acta Mater, 49:1859-66.
[37] V.S. Deshpande, and N.A. Fleck. (2000), Isotropic constitutive models for metallic
foams, Journal of the Mechanics and Physics of Solids, 48(6-7):1253-1283
[39] G. Holzapfel. and J. Simo. (1996) A new viscoelastic constitutive model for
continuous media at finite thermomechanical changes, Int. Solids Struct, 33: 3019-3034.
[40] K. Ti, B.B.K. Huat, J. Noorzaei, M. Jaafar and G.S. Sew (2009) A review of basic
soil constitutive models for geotechnical application, Electronic Journal of Geotechnical
Engineering, 14:1-17.
111
[42] K.K. Chawla (2008) Mechanical Behavior of Materials, second edition, Cambridge
University Press, United Kingdom.
[43] D.P. Zekkos, J.D. Bray, and M.F. Riemer. (2006) Shear modulous reduction and
material damping relations for municipal solid-waste, in the Proceedings of the 8th U.S.
National Conference on Earthquake Engineering, San Francisco, California, USA, Apirl
18-22, 2006.
[44] Y. Wang and T.D. O’Rourke. (2007) Interpretation of secant shear modulus
degradation characteristics from pressuremeter tests, Journal of Geotechnical &
Geoenvironmental Engineering, 133(12): 1556-1566.
[45] R.W. Ogden and D.G. Roxburgh. (1999) A pseudo-elastic model for the Mullins
effect in filled rubber, Mechanical Physics and Engineering Sciences, 455: 2861-2877.
[46] J.C. Simo, and J.W. Ju. (1987) Strain and stress based continuum damage models,
Part II: Computational Aspects”, Int. J. Solids & Struct., 23(7):841-869.
[47] S. Yazdani, and H.L. Schreyer. (2003) Nonlinear response of plain concrete shear
walls with damage, International Journal of IT in Architecture, 1(3): 251-258.
[49] R.D. Borst, P. Jerzy, and G.D. Marc. (1999) Geers On coupled gradient-dependent
plasticity and damage theories with a view to localization analysis, Eur. J. Mech.
A/Solids, 18: 939-962.
112
[52] A.F. Bower. (2008) Applied Mechanics of Solids, http://solidmechanics.org/.
[53] De Souza Neto, E.A., Peri´c, D. and Owen, D.R.J. (2008) Computational Methods
for Plasticity: Theory and Applications, John Wiley & Sons, Inc., Chichester, U.K.
[54] P. Perzyna. (1963) The constitutive equations for rate sensitive plastic materials,
Quart. Appl. Math., 20:321-332.
[55] De Souza Neto, E.A., Peri´c, D. and Owen, D.R.J. 1993. Some Aspects of
Formulation and Implementation of Ductile Damage at Finite Strains, ASME 93
Conference: Computational Mechanics in the U.K.
[57] Mullins, L., 1947. Effect of Stretching on the Properties of Rubber. Journal of
Rubber Research 16, 275-289.
[58] Dorfmann, A., Ogden, R.W., 2004. A constitutive model for the Mullins effect with
permanent set in particle-reinforced rubber. International Journal of Solids and
Structures 41, 1855–1878.
[60] ASTM C273 (2007): Standard test method for shear properties of sandwich core
materials.
[61] ISO 1922 (2001): Rigid cellular plastics –Determination of shear strength.
113
APPENDIX A
114
(a) (b)
(c) (d)
115
APPENDIX B
116
(a) (b)
(c) (d)
(e)
117
(a) (b)
(c) (d)
(e)
118
(a) (b)
(c) (d)
(e)
(c) (d)
(e)
121
(a) (b)
(c) (d)
(e)
(c) (d)
(e)
123
(a) (b)
(c) (d)
(e)
(c) (d)
(e)
126
(a) (b)
(c) (d)
(e)
(c) (d)
(e)
128
APPENDIX E
129
(a) (b)
(c) (d)
(e)
(c) (d)
(e)
Strain rate
0.0005 s-1 0.0050 s-1 0.0500 s-1 0.5000 s-1 5.0000 s-1
0.04 0.01729 0.02271 0.01909 0.02124 0.01849 0.01932 0.00429 0.01860 0.00052 0.01703
0.06 0.01729 0.04273 0.01909 0.04085 0.02054 0.03903 0.00711 0.03855 0.00094 0.03859
0.08 0.01729 0.06270 0.01909 0.06062 0.02097 0.05897 0.00961 0.05956 0.00132 0.05766
0.10 0.01729 0.08271 0.01909 0.08091 0.02106 0.07892 0.01153 0.07840 0.00170 0.07770
Strain rate
0.0005 s-1 0.0050 s-1 0.0500 s-1 0.5000 s-1 5.0000 s-1
0.04 0.01870 0.02093 0.02041 0.01928 0.00818 0.01728 0.00108 0.01429 0.00013 0.01090
0.06 0.01870 0.04134 0.02073 0.03928 0.01229 0.03608 0.00196 0.03548 0.00024 0.03157
0.08 0.01870 0.06122 0.02075 0.05938 0.01557 0.05683 0.00270 0.05388 0.00033 0.05047
0.10 0.01870 0.08130 0.02075 0.07925 0.01782 0.07658 0.00349 0.07426 0.00043 0.07066
132
APPENDIX G
133
(a) (b)
(c) (d)
(e)
(c) (d)
(e)
(c) (d)
136
(a) (b)
(c)
137