Beruflich Dokumente
Kultur Dokumente
To cite this article: M.T. Petre , A. Erdemir & P.R. Cavanagh (2006) Determination of elastomeric foam parameters for
simulations of complex loading, Computer Methods in Biomechanics and Biomedical Engineering, 9:4, 231-242, DOI:
10.1080/10255840600747620
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Computer Methods in Biomechanics and Biomedical Engineering,
Vol. 9, NO. 4, August 2006, pp. 231–242
†Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, 44106, USA
‡Department of Biomedical Engineering, Lerner Research Institute, Cleveland Clinic Foundation, Cleveland, OH, 44195, USA
¶Department of Orthopaedic Surgery and the Orthopaedic Research Center, Cleveland Clinic Foundation, Cleveland, OH, 44195, USA
Background: Finite element (FE) analysis has shown promise for the evaluation of elastomeric foam
Downloaded by [North West University] at 09:40 21 December 2014
Table 1. Examples of elastomeric foam modeling literature. Note that parameters are seldom determined from more than uni-axial compression test
data. The ABAQUSw hyperfoam model is the most common nonlinear model. Linear model parameters vary even for the same material. NAZNot
available.
Model description
Lyn and Mills Crash mat Remolded poly- ABAQUSw Hyperfoam 2nd order High strain uni-
(2002) urethane axial compression
Mills and Gilchrist Cushion Polyurethane foam ABAQUSw Hyperfoam 2nd order Compression and
(2000) shear
Sanders and Daly Prosthetic socket Pelitew ANSYSw Linear elastic EZ1.8 MPa, Uni-axial
(1993) nZ0.39 compression
Silver-Thorn and Prosthetic socket Pelitew MARCw Linear elastic EZ380 kPa, NA
Childress (1996) nZ0.49
Zhang et al. (1995) Prosthetic socket Pelitew ABAQUSw Linear elastic EZ380 kPa, nZ0.3 NA
and Zhang and
Roberts (2000)
Chen et al. (2003) Shoe insole PPTw, Plastazotew, MARCw Rubber foam 1st and 2nd order Uni-axial
Microcelw Puff, compression
Downloaded by [North West University] at 09:40 21 December 2014
Thermocork
Erdemir et al. Shoe insole Microcelw Puff, ABAQUSw Hyperfoam 2nd order Uni-axial
(2005) Microcelw Puff compression
Lite, Poronw
Lemmon et al. Shoe insole PPTw, Crepe ABAQUSw Hyperfoam 3rd order Uni-axial
(1997) (EVA/SBR blend) compression
Lewis (2003) Shoe PPTw, Plastazotew, Algorw Linear elastic EZ25 MPa, NA
Aliplast, XPE, nZ0.42
Spencow, Ucolin
Thomson et al. Shoe midsole Ethylene-vinyl- ABAQUSw Hyperfoam 1st order Uni-axial
(1999) acetate (EVA) compression
Verdejo and Mills Shoe midsole Ethylene-vinyl- ABAQUSw Hyperfoam 1st order High strain uni-
(2002) acetate(EVA) axial compression
Dionne et al. (1998) Wheelchair cushion Polyurethane foam ANSYSw Linear elastic NA Uni-axial
compression
Ragan et al. (2002) Wheelchair cushion Open-cell urethane ANSYSw Linear elastic EZ22 kPa, nZ0.1 Uni-axial
foam compression
any one of these inputs fails to represent its real world Past footwear modeling studies have used material
counterpart. definitions ranging from linear estimates to third order
Foam material representation is accomplished through hyperelastic to model foam materials that have highly
the use of constitutive models containing one or more non-linear large strain behavior (table 1). Most investi-
material specific parameters. When mixed loading is gators have obtained material parameters from com-
likely to occur, it is crucial that the model be able to pression tests only and then used the derived material
accurately predict all components of that loading. For models in mixed loading simulations. This trend is also
example, material definitions used in footwear modeling seen in studies modeling other personal protection devices
must be able to predict the combination of compression (table 1) (Zhang et al. 1995, Silver-Thorn and Childress
and shear created underneath the plantar surface of the 1996, Zhang et al. 1998a, Zhang and Roberts 2000, Lyn
foot (Perry et al. 2002). and Mills 2002, Ragan et al. 2002). This investigation will
Although the mechanical behavior of foams is known highlight the importance of multi-mode testing, briefly
(Gibson and Ashby 1988, Sanders et al. 1998), there is no investigate the effects of increasing material model order
source for well-defined material model parameters. Foam and provide model parameters and their associated fit
behavior is controlled by differing mechanisms under errors for some common footwear materials.
different modes of deformation and the technical literature
from FE analysis software developers strongly rec-
ommends using data from multiple tests to accurately 2. Methods
define foam parameters (Nonlinear Finite Element Ana-
2.1. Foam testing
lysis of Elastomers 2001, ABAQUS 6.4 Theory Manual
2004). Higher order equations with more parameters and Testing for this study concentrated on elastomeric foams
therefore degrees of freedom, may also be needed to used in therapeutic footwear. Twelve foams were selected
describe the non-linear elastic behavior and multiple modes based on common usage and availability (table 2).
of deformation. However, increases in model order may Multiple grades of each material were selected when
have an associated increase in computational cost. available. Cylindrical samples were punched from sheet
Determination of elastomeric foam parameters 233
Table 2. Manufacturer supplied information for foams used in this study along with their typical clinical use.
Trade name Grade Description Durometer (shore A) Common usage Thickness (mm)
w
Microcel Puff Regular Closed-cell (EVA) 35 Midsole, insole base 12.7
Microcel Puffw Lite Closed-cell (EVA) 25 Insole base, top layer 12.7
Poronw 4708: Cushioning Open-cell polyurethane 12 Shore 00 O 00 Top layer, cushioning 12.7
Poronw 4708: Firm Open-cell polyurethane 20 Shore 00 O 00 Top layer, cushioning 12.7
PPTw Regular Open-cell polyurethane 20 Asker Top layer, cushioning 9.53
Crepe Cloud Closed-cell EVA/SBR 35 Midsole, posting 12.7
blend
Crepe Regular Closed-cell EVA/SBR 55 Midsole, outsole 12.7
blend
Crepe Firm Closed-cell EVA/SBR 60 Midsole, outsole 12.7
blend
Cork Lightweight Natural cellulose 55 Midsole, posting 6.35
P-Cellw Regular Closed-cell x-linked 20 Top layer, conforming 12.7
polyethylene
Plastazotew Medium (pink) Closed-cell x-linked 20 Top layer, conforming 12.7
polyethylene
Downloaded by [North West University] at 09:40 21 December 2014
Plastazotew Firm (white) Closed-cell x-linked 35 Top, mid, or base layer 12.7
polyethylene
stock using a 25.4 mm circular punch. Five samples of strain of 0.25. Trials with failure at the glued interface were
each material were tested in each loading state. Data used repeated.
in parameter determination reflect the average of these Volumetric compression testing was performed in a
five trials. All tests were performed at room temperature custom-made hydrostatic chamber (figure 1c). Following
(w23 8C). sample submersion, the water filled compression chamber
Compressive and shear tests employed an actuator was purged of air. Volume changes were applied via
driven 8511 Instrone test machine (Instron Corp.) an actuator driven syringe (10 ml (G.1 ml)) and chamber
controlled by a custom Labview interface (National pressure was read using an analog gauge (690 kPa
Instruments, Inc.) with a 1 kN (G10 N) load cell. (G3.5 kPa)). The volumetric strain rate was approxi-
Compressive and shear samples were preconditioned by mately 100% sample volume/sec. This was the fastest
simulating 15 walking steps prior to final data collection. volumetric strain rate achievable with the experimental
Each simulated step consisted of deforming the sample at set-up and pilot studies showed that volumetric testing was
200% sample thickness/sec to the test strain, releasing and not as sensitive to strain rate as other tests.
allowing the sample to relax for 1 s (approx. time from Volumetric testing was straight-forward in the case of
heel-strike to heel-strike during walking). The final data closed-cell foams; however, volumetric compression of
were collected one second after preconditioning by open-cell foams using water was a more complex problem.
displacing to test strain and holding for 1.5 min. If the sample was simply submerged, some uncontrolled
Although stress-relaxation data were collected for use amount of air would be trapped within the open pore
in future studies, viscoelasticity is not addressed here. The structure. If the sample was first encapsulated in a flexible
parameters obtained are intended for use in quasi-static barrier such as latex (Zhang et al. 1998b), all air would be
modeling of footwear in walking or running conditions. retained and the foam would be effectively converted to a
Pilot tests conducted on Microcel Puffw showed that pre- closed-cell foam. Submerging the sample and removing all
loading the samples 10 times sufficiently attenuated air (if possible) would lead to water filled pores exerting a
transient effects. There was also little difference between uniform pressure on the incompressible polymeric struts.
tests conducted at 200% strain rate (simulating walking) Some air probably does escape from open-cell foams
and 300% strain rate (simulating running) (De Clercq within the shoe, although to our knowledge, this has not
et al. 1994), suggesting that the deformation rates present been quantified. In the absence of such information, it is
in walking are already faster than those to which the necessary to assume a testing condition, that is either sealed
viscous component of the foam can respond. with air, sealed with no air, or somewhere in between. The
For uni-axial compression tests, samples were pressed solution used here was to seal samples in a flexible plastic
between two parallel Teflonw lined platens to 50% of their bag and forcibly remove air until the bag was pulled as tight
original thickness (figure 1a). Lateral strain was not as possible against the sample without causing any visible
collected but information about compressibility was deformation. Air was removed using a 60 ml syringe
obtained from volumetric testing. To test in simple shear, capable of producing a vacuum of 600 mm Hg. Any air left
two pieces of foam were glued to either side of a center inside the barrier, such as the air in closed-cells or small
platen and then to equally spaced outside vertical platens amounts of residual air within open-cells, was considered
(figure 1b). The center platen was displaced vertically by intrinsic to the structure. The low-vacuum method
25% of the sample thickness to obtain a maximum shear described here was a compromise that allowed for uniform
234 M. T. Petre et al.
Figure 1. Experimental apparatus with arrows showing the direction of applied displacement. (a) Uni-axial compression test set-up with foam sample
in place. (b) Simple shear test designed to eliminate bending. (c) Volumetric compression chamber. A fast moving hydraulic actuator drove the syringe
plunger.
testing of all samples in a probable in-shoe condition— described by bi, which is related directly to Poisson’s
some air escapes the foam during compression and some ratio, vi, by equation (3). The most meaning is derived in
Downloaded by [North West University] at 09:40 21 December 2014
small amount of residual air is trapped inside. Pilot tests of the case where all bis are identical and an effective initial
closed-cell foams with and without the barrier provided Poisson’s ratio is prescribed.
similar results and inter-sample correlation was good for all bi
materials, implying that the method of encapsulation was vi Z (3)
1 C 2bi
consistent and did not artificially stiffen the samples. Both
open and closed-cell foams were encapsulated prior to Although some meaning can be assigned to the
testing. parameters at the beginning of the deformation, the
The loading portion of the force–displacement data for parameters are not related to physical quantities (density,
each sample was converted into nominal stress–strain data cell size, etc.).
by normalizing with respect to original sample dimen-
sions. Grouped samples were then averaged to obtain one
stress–strain data set for each material in each of the three 2.3. Parameter determination
testing modes.
Hyperfoam stress–strain relationships were derived for
each loading state by taking partial derivatives of the strain
2.2. Material model energy density function (Appendix A). These relationships
were then incorporated into custom Matlabw (Mathworks,
Test data were fit to a common compressible hyperelastic Inc.) scripts that read in the test data, used the test strain and
material model (Storakers 1986). The specific implemen- derived relationship to calculate model predicted stresses
tation of this model was the hyperfoam model embedded and then determined the error between experimental and
into ABAQUSw FE software (ABAQUS, Inc.) (ABAQUS model predicted stresses. In the case of uni-axial testing,
6.4 Theory Manual 2004). Similar models exist in other since lateral stretches were not measured during the test and
FE packages, such as ANSYSw (ANSYS, Inc.) and Marcw uni-axial stress is a function of the volume ratio, JZl1l2l3,
(MSC Software). The ABAQUSw hyperfoam strain the zero lateral stress constraint was used to compute lateral
energy potential is given by equation (1). strain as a function of the material parameters (see
X N
2mi ai ai ai 1 Kai bi Appendix A). Root mean square errors as a percentage of
UZ l1 C l2 C l3 K3 C J K1 maximum stress were produced for each test (equation (4)).
a2i bi el
iZ1 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(1) Pn
2
ðsmodKsexp Þ
1
where li are the principal stretches, N is the order, Jel is the RMS Z 100 !
n
(4)
volume ratio and m, a and b are material specific jsmax j
parameters. mis carry the units of stress, but ai and bi are
where n is the number of data points provided for the test,
both dimensionless. Comparison of this model to small-
smod is the model predicted stress, sexp is the experimental
strain elastic theory provides insight into each parameter’s
stress and smax is the maximum experimental stress. This
implications. The foam’s initial shear modulus is related to
error measure eliminates the effects of number of data
mi by equation (2).
points and data magnitude when comparing multiple tests.
X
N When multiple tests were fit simultaneously the objective
G0 Z mi (2) function was the sum of the RMS errors from individual
iZ1
tests.
The strain hardening/softening behavior is determined Matlabw was used to simultaneously solve for all
by the selection of ais. Material compressibility is parameters by minimizing the error between experimental
Determination of elastomeric foam parameters 235
data and model predictions for all test data provided. error less than 1% of max (see shaded area in table 3). The
Numerous local minima made the problem very sensitive to resulting parameters reproduced the test results when
initial values. To increase the likelihood of finding the applied to a simple, single element FE model of the test.
global minimum, a gradient-based optimization was However, in cases where parameters were determined by
completed with at least 50 different starting points. m and data from a single test mode, the predictive capability of
a parameter values were bound to [K10, 10] MPa the model for other deformation modes was often poor
and [K50, 50] (unitless), respectively such that the optimal (figure 2). Simultaneous fitting of two deformation modes
parameter values were unlikely to occur at the boundaries of was more difficult, but was possible in most cases with
the design space. a was not permitted to be equal to zero for RMS fit errors !10% of max. In cases where two modes
numerical stability. b parameter values were bound to were fit, there was often substantial error in the prediction
[K0.3, 1] (unitless) to reflect the highly compressible nature of the third mode. Fits which considered data from all
of the foams (Storakers 1986). b values were additionally three-deformation modes had the highest fit error, but
restricted by the constraint that all lateral stretches in the uni- generally the lowest total error, as all deformation modes
axial compression test should be greater than one. were reasonably represented over the strains encountered
Material stability was evaluated for the parameter set in the test data. The best fits for all 12 materials and their
with the lowest error using the ABAQUS Drucker stability associated errors are reported in Appendix B.
test, which ensures a positive definite material stiffness There was some computational cost to increasing the
Downloaded by [North West University] at 09:40 21 December 2014
matrix (ABAQUS 6.4 Theory Manual 2004). Any material order and the increased number of parameters failed to
stability warnings reported by ABAQUS were recorded. If decrease fit errors substantially. For the plane-strain
a parameter set was not stable over the strains supplied for indentation model with 3000 foam elements, use of first
each of the three tests, the parameter set with the next order foam properties resulted in a three trial average of
lowest error was selected. The final parameter sets 64 s of CPU time, second order 70 s of CPU time and third
selected for each material represent the lowest possible order 90 s of CPU time. Differences in peak pressure
error achieved while still preserving stability over the between the models were small and could be attributed to
range of supplied strains. differences in nodal alignment at the indenter/foam
interface (1st order: 1.83 MPa, 2nd order: 1.92 MPa, 3rd
2.4. Validation and application order: 1.83 MPa).
Using different material parameters can have large
Matlabw generated Poronw material parameters were effects on simulation results, particularly during complex
applied to a single element ABAQUSw model to evaluate loading scenarios (figure 3). During a simulated 50%
their predictive ability under compressive, shear and compression, both sets of material parameters predicted
volumetric deformations. The single element model also the compressive and shear stresses well. When a small
served as a validation of the derived Matlab stress–strain shear strain was added by displacing the indenter
formulations. A single element simulation was used rather horizontally by one-fifth of the foam block height, the
than a multi-element recreation of the test because the parameters determined without the benefit of shear data
constitutive model operates at the material (stress/strain) predicted a peak shear stress twice that of the parameters
level independent of sample geometry. determined with shear data. Additional indentation to 70%
Computational cost was investigated by solving a 3000 compression and a final horizontal displacement of 0.3
foam element plane-strain mat indentation simulation with times the block height (physically realizable in footwear)
1st, 2nd and 3rd order hyperfoam materials on a 2.8 GHz created large discrepancies in both compressive and shear
Pentium 4 processor with 512 MB of RAM dedicated to predicted stresses.
the solver. Although mesh convergence occurred around
1000 elements, more elements were used to exaggerate
time differences between simulations. The foam mat was 4. Discussion
75 mm wide by 10 mm thick. The parabolic indenter was
60 mm wide, 10 mm thick and considered rigid to isolate Our experience with the intrinsic ABAQUSw fitting
the effects of foam properties. The interaction between the algorithm suggested that it is less flexible than the one
indenter and the foam consisted of penalty friction with developed here. Mills and Gilchrist (2000) have also
coefficient of friction equal to 0.75 (Sanders et al. 1993). reported difficulty in obtaining parameters from ABA-
The plane-strain indentation simulation was also used to QUSw when supplying data from multiple tests. Deter-
evaluate the effect of multi-mode parameter determination mining parameters outside of the FE analysis software
on predicted shear and compressive stresses at the provides control over parameter limits and allows the
indenter/foam interface. compressibility parameter, b, to be determined rather than
specified beforehand via lateral strain (which is difficult to
3. Results measure for highly compressible materials) or direct input
of an effective Poisson’s ratio. The optimization approach
The hyperfoam model was capable of fitting any one of the used here is time consuming, but provides high confidence
individual Poronw deformation modes with an RMS fit that the best possible fit has been achieved. All parameter
236 M. T. Petre et al.
Table 3. Complete fit results for Poron Cushioning. The objective errors for each fit are shown in bold italics. The sum of these bolded numbers
represents the fit error. Truncated fit results for all materials tested are given in Appendix B.
Uni-axial error Shear error Volumetric error Fit error Total error
Data used for fit Order (RMS %max) (RMS %max) (RMS %max) (RMS %max) (RMS %max)
volumetric
2 5.30 2.58 2.83 10.71 10.71
3 4.24 1.30 2.04 7.58 7.58
sets reported were found to be stable over the range of relationship is purely analytic but increases in order still
strains in the provided test data although other limitations require some additional computations. Although simu-
may be present (Appendix B). It should be noted that lation time increased by 40% when using third order
stability in a deformation mode does not imply that the material properties rather than first, this increase was
model prediction is correct. For example, in many cases, negligible in practice, at less than 30 s of CPU time for a
the parameters determined from uni-axial compression model with 3000 foam elements. The real detraction from
data alone are stable in simple shear, but poorly predict increasing order to improve fit is the increased likelihood
shear behavior. of bad predictions of other deformation modes, i.e.
The hyperfoam model did converge as order was increasing total error. Although fit error consistently
increased, but gains in fit were generally small. decreased with increasing order, the only case where
Computational time was affected by order. Once para- total error regularly decreased with order was when all
meters have been determined, the stress–strain deformation modes were fit (table 3).
Figure 2. First order fits for Poronw Cushioning. The first row shows results when only compression data were fit and the second row shows results
when all tests were fit. When only compression data is used, the shear prediction of the model varies greatly from the experimental results. Fitting to all
deformation modes sacrifices some agreement in the compression test, especially at low and high strains. Some degree of improvement can be obtained
by increasing the model order (see table 3).
Determination of elastomeric foam parameters 237
Downloaded by [North West University] at 09:40 21 December 2014
Figure 3. A simple plane-strain indentation model used to visualize the effects of material parameters on model predictions. Three steps were
simulated: Step 1. A vertical displacement of one half the indenter height, h, is applied. Step 2. An additional horizontal displacement is applied to
the indenter to induce shear loading. Step 3. Additional vertical and horizontal displacements strain the foam block beyond the test data that was
provided during parameter determination. Deformed model geometries are displayed above the resulting stress distributions at the indenter/foam
interfaces (dark: parameters determined with compression only; light: parameters determined with compression, shear and volumetric). Results are
nearly identical when only compression is present (column 1), but when a small shear strain is applied (column 2), differences between the two
material definitions are apparent. Additional compression and shear beyond the strains of the test data (column 3) create even larger discrepancies
between the two simulations.
Foams do not meet the criteria for continuous media even When data from multiple tests are fit simultaneously,
at relatively large-scale (w0.01 mm) and are therefore the problem becomes one of multi-objective optimization
inherently difficult to model using continuum theory. The where the individual objectives (errors from each test) are
ABAQUSw hyperfoam model is quite capable of reprodu- competitive rather than cooperative. Multiple solutions
cing any single deformation mode, however, it is clear from with the same total error are possible because gains in fit
the present results that a good fit in one deformation mode for one mode of deformation are achieved only through
does not imply a good prediction of other deformation loss of fit in another mode. A Pareto optimality analysis
modes (figure 2). The shear prediction is particularly poor (Belegundu and Chandrupatta 1999) may be able to find
when shear data is not included during parameter the fit with the lowest overall error, but this fit may not be
determination. This is because the shear behavior at low- of practical interest. It is more important to know which
strains is nearly linear and difficult to represent with a deformation modes are the most active in the FE
model based on powers of stretches. Simultaneous fits of simulation and choose a material model that is weighted
compression and shear data are possible as are simul- in favor of these common modes. To this end, Appendix B
taneous fits of uni-axial and volumetric data, yet, when provides not only the overall fit error, but also fit errors for
multiple test modes are fit simultaneously, the errors for each testing mode.
each test are never as small as if that test were fit alone. This The prediction discrepancies seen in figure 3 have at
seems to be due to the common sensitivities of each mode to least three origins. In column two, shear stresses in the
m, a and b (Appendix A). It is possible that a more general single-test-fit simulation (dashed line) are likely over-
model may be able to better predict all deformation modes estimated because the lack of shear data results in a poor
simultaneously at the cost of an increased number of purely shear prediction (figure 2). Small differences in the
phenomenological parameters. compressive prediction are due to the dependence of the
238 M. T. Petre et al.
compressive stress on not only normal strains, but also This study focused on foam footwear materials but the
the poorly predicted lateral strains (Appendix A). When principles shown here also apply to other biomedical
the foam is strained beyond the limits of the provided modeling problems, which typically involve non-linear
test data, as in column three, extrapolation error can materials and geometry. For example, the realistic
occur. The compressive prediction of the multi-test-fit prediction of a shear stress distribution at the foot/shoe
simulation is nearly linear (figure 2). This creates a interface requires not only knowledge of foam shear
reasonable prediction within the bounds of the data behavior, but also that of the mating soft tissue. It is
provided, but at larger strains the actual foam behavior presently unknown how well current soft tissue models
increases exponentially and an extrapolation of the represent shear because the model parameters have
linear prediction greatly underestimates compressive generally been determined via compression testing alone
stress. Differences in final predicted geometry are also (Morag et al. 1997, Miller-Young et al. 2002).
present. It is clear that even “good” material definitions can
Although it is clear that large discrepancies exist result in poor simulation predictions, but it is difficult to
between the models used in figure 3, it is difficult to specify fixed requirements or criteria that will always
assign one or the other as “correct” because no multi-axial result in the best parameter determination. The needs of
experiments were performed here. Future work should each simulation are different and varying degrees of
include validation of the material model with a combined complexity are required.
Downloaded by [North West University] at 09:40 21 December 2014
2 3
ai
vt X
2 XN l
4 2g
j K1
Z 5 (13)
vmi jZ1 2 l2j K1 Kg2 iZ1 ai
vU 2X N
mi 1=3ai
PZ Z J KJKai ;bi (17) vP 2X N
vJ J iZ1 ai Z m JKai bi logðJÞ (20)
vbi J iZ1 i
vP 2X N
J 1=3ai KJKai bi
Z (18) Where all stretches are equal and, in the case of
vmi J iZ1 ai volumetric compression, less than one.
Downloaded by [North West University] at 09:40 21 December 2014
Appendix B: Material parameters
Table B.1. Hyperfoam material parameters for Microcelw Puff, Microcelw Puff Lite, Poronw Cushioning, Poronw Firm, PPTw and Crepe Cloud. UC, uni-axial compression; SS, simple shear; and VC, volumetric
compression. Lower or upper stability limits were generated by the ABAQUS Drucker stability test. If no limit is reported, no stability warning was returned by the Drucker stability test.
Total
Data error UC SS VC
used for (RMS (RMS % (RMS % (RMS m1 m2 m3 UC SS VC
Material fit Order %max) max) max) %max) (MPa) a1 b1 (Mpa) a2 b2 (MPa) a3 b3 (strain!) (strainO) (volume ratio!)
241
242
Table B.2. Hyperfoam material parameters for Crepe Regular, Crepe Firm, Cork, P-Cellw, Plastazotew Medium (pink plastazote) and Plastazotew Firm. UC, uni-axial compression; SS, simple shear; and VC, volumetric
compression. Lower or upper stability limits were generated by the ABAQUS Drucker stability test. If no limit is reported, no stability warning was returned by the Drucker stability test.
Total
error UC SS VC
Data used (RMS (RMS (RMS (RMS m1 m2 m3 UC SS VC
Material for fit Order %max) % max) % max) %max) (MPa) a1 b1 (MPa) a2 b2 (MPa) a3 b3 (strain!) (strainO) (volume ratio!)
M. T. Petre et al.
UCC 3 9.38 5.48 0.46 3.44 2.300 5.301 0.680 K1.63 1.792 1.00 0.738 K0.01 0.077 2.44 EK2
SSCVC
Cork UC 1 5933 1.34 5706 226 2.043 21.83 0.170 0.149
UCCSS 1 1000 2.52 6.92 991 0.936 8.428 0.680 0.104
UCC 3 9.56 4.32 2.72 2.52 1.158 0.232 0.080 K0.70 2.293 0.102 0.238 6.786 K0.089
SSCVC
P-Cellw UC 1 3.45 E4 0.46 34126 408 0.659 26.52 0.0498 0.111
UCCSS 1 235 3.52 5.31 226 0.253 7.886 0.120 2.74 EK3
UCC 3 8.40 3.55 2.86 1.99 0.070 0.045 K0.258 0.814 5.350 K0.162 K0.634 4.010 K0.291
SSCVC
Plastazotew UC 1 2.95 E4 1.07 28417 1081 0.488 26.09 0.070 0.133
Medium UCCSS 1 450 3.29 0.86 446 0.198 7.549 0.169 2.20 EK3
UCC 3 9.43 3.77 0.48 5.18 0.560 6.009 0.100 K0.424 5.042 0.134 0.073 0.942 0.039
SSCVC
Plastazotew UA 1 5.39 E4 0.18 53532 330 0.941 25.91 0.067 0.125
Firm UACSS 1 42.3 5.07 6.60 30.6 0.356 4.604 0.033
UAC 3 12.16 7.59 0.74 3.83 0.038 K0.383 0.380 0.333 3.364 0.058 K0.102 2.746 0.273
SSCVC