Using Abaqus/Explicit to model polymeric thin-film

mechanical property changes due to embossing

Colmar Wocke
Dow Chemical, Wolleraustrasse 15-17, CH-8807 Freienbach, Switzerland

Abstract: Embossing of polymeric films destined for usage in the personal care marketplace is an
industrial process that produces a very fine pattern, barely discernible to the naked eye, yet has a
significant influence on some market-driven properties; more bulk, a soft and smooth touch,
reduced crinkling noise and lower gloss. However this comes at a cost to the mechanical
properties such as stiffness and ultimate strength capability. Since the feature size of the
embossing pattern is so small, it is difficult to obtain all the information from an experiment alone,
making the Finite Element Method an ideal analytical tool to help gain additional insight into the
effects of some of the design and process parameters. The paper discusses the computational
approach, presents results of simulations performed in Abaqus/Explicit for an example 16 micron
film and compares to experimentally measured values.
Keywords: Constitutive Model, Damage, Experimental Verification, Forming, Fracture, Heat
Transfer, Plasticity, Polymer, Residual Stress, Springback, Thermal Stress, Tissue, Visualization.

1. Introduction
Off-line cold embossing is a stand-alone unit operation that consists of unwinding previously
blown and slit film, contact-heating it with a roll to a semi-molten state, passing it through a cold
nip formed between a hard engraving roll in contact with a soft backup roll and re-winding the
structured film as schematically shown in Figure 1. Embossing produces a very fine pattern on
the film, which is barely discernible to the naked eye, yet has significant influence on its
properties. Compared to the flat precursor film, the embossed film has more bulk, has a soft and
smooth touch, and reduced noise all much desired in hygiene tissue films used for diapers.
Furthermore the film has very low gloss, and features a reduced modulus, along with improved
dart impact strength.
In addition to the analysis reported in [Kamal, 1988], these authors have been involved in
additional studies along the same lines [Kamal, 1985; Kamal, 1992] that together form a good
basis for the analysis conducted here. Since the time of these investigations, the capability to
analyze these kinds of problems, using commercially-available Finite Element programs, has
advanced significantly [Nagarajan]; Dow has chosen to analyze the problem using the capabilities
available within Abaqus/Explicit from SIMULIA. Our analysis does not cover material structureproperty relationships or in-depth analysis of toughening effects; we perform an analysis using the
2008 Abaqus Users Conference

measured material test data, along different axes, as a basis for our input to the program. It is
unusual to have all the test data required (even for fairly basic material models) since the material
flows readily at the embossing temperature which makes testing very difficult; the complete
material data required to be input has thus been synthesized from measured data together with
some additional theoretical considerations and reasonable estimates.

Flattened tubular
and slit film
Blown film
Preheat roll
Windup
Radiation
Heaters
Extruder

Backup
roll

Embossing roll

Figure 1 Film embossing schematic, showing Abaqus analysis window [from

Kamal, 1988].

The primary mechanical property variable of interest is the alteration of the stiffness of the thin
film due to the embossed pattern. However, an explicit Finite Element analysis offers the
prospects of being able to conduct a failure analysis of the embossed film and to present this in
animated form to the requestor of the analysis.
Concisely stated, the problem statement is:
Is it possible to successfully model
1. The Process, to investigate various effects on the material.
2. The Geometric Effect.
and in so-doing, determine the final properties of the embossed film in sufficient detail to help
with some film embossing design decisions?
This paper covers some details associated with the parametric mesh generation of the embossing
pattern, the material tests conducted, selection of a suitable material model from the library
2

2008 Abaqus Users Conference

available in Abaqus and the synthesis of the required information in order to have sufficient
information available for the analysis.

2. Background information to the analysis.

The blown film, before it enters the heating and embossing phase, is already oriented and its
mechanical properties show a directional dependency. Subsequent operations may relax (heating
and cooling) some of this oriented structure, introduce new orientations (mechanical embossing)
and perhaps add internal damage or remove internal constraints to the material (heating and
cooling and mechanical embossing). One can, for example, see some fairly sharp edges after the
embossing process in the cross-sectional cut shown in Figure 2.
The modeling of any re-orientation, removal of internal constraints and damage effects within the
material structure as a consequence of thermal treatment is not within the scope of this analysis.
Its effects can perhaps be included, or rather accounted for in the data supplied to Abaqus.

16 micron true thickness

33 micron apparent thickness

after embossing

Figure 2 Cross-sectional dimensions of the embossed film.

2008 Abaqus Users Conference

3. Abaqus analyses
The major steps into which the analysis has been subdivided, is sketched in Figure 3. This
sequence follows the natural flowchart of the actual production process for such films, as can be
seen with reference to Figure 1.
Some factors which complicate the analysis are:

1. The exact state of internal stress in the thin film (and its distribution) both before and
after the embossing process is not known.
2. The material may experience constrained shrinkage on the heating rollers as it approaches
the embossing temperature.
3. Theoretical models able to consider alteration of microstructural details in the film are
not readily available and in any event would require quite some calibration to obtain
meaningful results from.
4. The finer details associated with the manufacturing process are not known; for example
what the modulus of the rubber on the backup roll actually is. The rubber modulus will
have an effect on the local distribution of pressure onto the thin film while it is being
formed into the bathtub depression on the engraving roll.

Thus, the approach followed here has been to perform the analysis using relatively-simple models
already available within Abaqus, but not so elementary so as to deliver useless results or give poor
insight into the mechanisms influencing the problem. The forming process is simply considered to
be due to a uniform pressure load applied to the top surface of the heat-softened polymer.
The material model chosen from those available in Abaqus is one that has not been specifically
developed to model polymer behavior, but rather that of metals typically showing elastic-plastic
behavior. The elastic part of the behavior is taken to be the linear part of the stress-strain curve,
while the plastic part is that portion extending past the traditional yield point. No
loading/unloading tests have been conducted to determine the true amount of recoverable strain.
With reference to the material test data curves, the elastic strain would be considered to extend to
ca. 10% strain this is in any case within the region of interest when predicting the stiffness
alteration of the film.

2008 Abaqus Users Conference

Step 1: Perform a heat

transfer analysis to make
some assessment of the
maximum temperatures seen
by the film, before embossing.

Step 3: Allow to cool

maybe even anneal
i.e. leave no internal
stress.
Step 4: Compare
tensile response
before and after the
process.

Step 2: Emboss the film

with some reasonable
pattern.

Figure 3 Step sequence of Abaqus analyses.

3.1

Step 1 Heat Transfer Analysis

The purpose of this analysis is to make some estimate of the maximum temperature experienced
by the thin film, before it enters the embossing rollers. Both the deformational properties of the
polymer and the total amount of shrinkage on cooling are dependent on this temperature. The film
is in contact with the roller for some fraction of the time it would take to cover the whole
circumference at the speed of winding. We consider the film and the roller to be flat having no
curvature and being infinitely wide along the roller with no gradient into the paper, thus a 2-D
analysis suffices. For constant values of the important heat-transfer coefficients, this problem has
a closed-form solution, of course. It was nevertheless solved by FEM - perhaps in anticipation of
handling more complicated cases in future.

Heat transfer effects are due to:

1.
2.

Heat flow into the sample due to its contacting the roller.
Heat flow out of the sample due to radiation and forced convection loss to the
surroundings.

There will be some forced convection cooling between the pre-heat roll and the embosser it is
assumed that in practice some radiator banks make up for this loss.

2008 Abaqus Users Conference

3.2

Step 2 Embossing Analysis

We examine the deformation of a representative volume element (RVE) of the thin film being
draped over and into a rigid mesh on the surface of the embossing roller in the nip area. The
embossing of the film is performed by applying a uniform pressure (in practice applied by a solid
surface from the rubber roll, its amplitude increasing and decreasing with time as the roller passes)
to the top surface of the film which eventually deforms the surface into a kind of catenary surface.
In the absence of more detailed information, this is considered a reasonable approach. The RVE
repeats many times over the surface, so we apply symmetrical boundary conditions on all its edges
as shown in Figure 4.
Y-symmetric
X-symmetric

X-symmetric
Y-symmetric

Figure 4 Outline of RVE with suitable boundary conditions in contact with

embossing bathtub mesh with uniform pressure applied.

3.3

Step 3 Cooling Analysis

The cooling analysis is included to account for any thermal shrinkage experienced on addition to
the elastic springback at the end of embossing. The cooling down also relaxes the internal forming
stresses experienced by the RVE; however the damage introduced into the material is not
healed. No annealing was performed, although this feature does exist in Abaqus. The amount of
embossing pressure that is required to be applied is dependent on the degree of orthotropy of the
material and how much shrinkage is experienced on cooling. Some iteration was required to bring
the final measured thickness of the RVE to 33 microns as shown in Figure 2.

3.4

Step 4 Comparative Tensile Testing Analysis

For these analyses, the force vs. displacement plots from the Abaqus/Explicit analyses are
compared for the two cases; a flat unembossed RVE and an embossed one. This analysis is run so
6

2008 Abaqus Users Conference

that it is representative of a quasi-static test; the ratio of kinetic energy to internal energy is kept
below 5% during the analysis. The material experiences internal damage as it is extended well into
the plastic zone and the failure can be monitored via the scalar damage variable (DUCTCRT).

3.5

Parameterized Abaqus input model

Since the shape is fairly regular, one can easily generate the mesh making up the rigid surfaces of
the bathtub using suitable *NGEN, *NFILL and *ELGEN commands. The required draft angle is
obtained by simply scaling the nodes defining the top rim of the bathtub via *NMAP and
shifting them by the required depth to the floor of the bathtub. The four triangularly-shaped
quadrants (NE, SE, SW and NW) around the central bathtub are generated through suitable
application of *NCOPY and *ELCOPY commands. The major variables required for
parameterized input as seen in Figure 5 for the rigid bathtub mesh generation, are listed in Table 1.
Base_Length

(1-Scale)*(Half_Base_Length - Crotch_Depth)

NW

Draft Angle

Radius
Floor

Crotch Width
Rim

Bathtub Wall

SE

Depth

Crotch_Depth

Figure 5 Major *PARAMETER references for generating the bathtub mesh.

Table 1 Parameters for generating the mesh

Parameter
Crotch_Depth
Crotch_Width
Base_Length
Scale
Radius
Depth
Draft_Angle = atan((1-Scale)*(Half_Base_Length-Crotch_Depth)/Depth)

2008 Abaqus Users Conference

4. Analysis material model

The material exhibits both elastic and plastic orthotropy along axes as shown in Figure 6; the
curves in Figure 7 clearly show the dependency of mechanical properties on orientation also in
the plastic region. In addition, the material stress vs. strain curves are, in general, functions of
temperature and strain rate. Test data, at various temperatures, for the thin film are taken at 0, 90
and at 45 to the dominant blowing/extrusion axis (machine direction axis). Together with the
theoretical material model considerations expounded here, these test data from the testing
laboratory are used to define the material to Abaqus, in the form Abaqus expects.
33

Through-thickness direction

22
Cross direction
11
Machine direction

Figure 6 Orthotropic axes definition.

4.1

Elastic and thermal expansion behavior

The in-plane compliance behavior of the film at an angle to the machine direction has the
following functional relationship [cf. Holmes], which may be used to estimate a realistic value
for G MC from a comparison of the test data at 0, 90 and at 45 to the machine direction axis.

2 MC
1
1
1
1
+
=
cos 4 +
sin 4 + (
) sin 2 cos 2
E E MD
ECD
E MD
GMC
with,

E MD = Measured elastic modulus along machine direction.

8

2008 Abaqus Users Conference

ECD = Measured elastic modulus along cross direction.

G MC = Shear modulus in the plane of film.

MC

= Poissons ratio in the plane of film defined as

CD

MD

when loaded along the machine

direction with the cross direction being traction free.

From thermodynamic considerations, the constraint MC

2

<

E MD
is also relevant.
ECD

We need to specify the through-thickness direction properties (cf. Figure 6) for Abaqus in order to
have a fully-specified material model. Most of the loading on the thin-film is in-plane and the
thickness direction has very little influence on the in-plane behavior in any event, so these
properties are estimated. We consider the material to be transversely-isotropic in the plane
containing the cross direction and thickness direction i.e. the 22-33 plane and thus

ETD = Estimated elastic modulus along the thickness direction = ECD .

MT

= Estimated Poissons ratio of the film defined as

TD

MD
direction with the thickness direction being traction free = MC

when loaded along the machine

G MT = Estimated transverse shear modulus of the film = G MC .

G CT = Estimated transverse shear modulus of the film is considered to be significantly less than
that of G MT , possibly by a factor of half.

The elastic behavior is entered using the Abaqus keyword *ELASTIC, TYPE =
ENGINEERING CONSTANTS and giving the associated Youngs modulus and Poisson ratio
data as a function of temperature, as required. After embossing, the material cools and will
generate internal stress owing to its being restrained in the nip area. The thermal expansion
behavior is entered using the Abaqus keyword *EXPANSION, TYPE = ORTHO and giving the
data as a function of temperature, as required.

2008 Abaqus Users Conference

4.2

Plastic behavior

The extension by Hill to the standard Von Mises plasticity model is able to model the directional
dependency of the plastic behavior. The theoretical treatment of this model and axes convention
follows those used in the Abaqus manuals. Each yield curve for loading along the other axes is
given as a ratio of some reference value, which for our analysis we have taken to be the 11direction (or machine direction). At first glance, the form of the test data curves in Figure 7 is
typical of those for a hyperelastic material or non-linear elastic material these data are taken at
high temperature. However a hyperelastic material would recover its initial position (zero
remaining strain) once the load is removed; the thermoplastic modeled here will not do this and
thus the elastic-plastic modeling approach is retained.
The yield function is:
2
f ( ) = F ( 22 33 ) + G ( 33 11 ) + H ( 11 22 ) + 2 L 23
+ +2M 312 + 2 N 122
2

The constants F,G,H,L,M and N are determined internally by Abaqus from the ratios, Rij, of yield
stress functions (they may be temperature dependent, as required) along the different test axes
given on the *POTENTIAL keyword.
In uniaxial tension

Rii =

ii
0

In shear

Rij =

ij 3
0

The plastic behavior defining the reference yield function along the 11-direction, or machine
direction, 0 , is entered using the Abaqus keyword *PLASTIC, RATE = <strain_rate>
and giving the associated data as a function of temperature and strain rate as required.

10

2008 Abaqus Users Conference

Figure 7 Typical test data - Stress vs. Strain (Hencky) curves illustrating directional
dependency.
4.3

Damaging behavior

As the plastic strain increases, internal damage accumulates in the material according to the
ductile damage model offered in Abaqus. This model assumes that the equivalent plastic strain,

Dpl

at the onset of damage in a material, is a function of stress triaxiality and strain rate. Thus:

pl
D

pl
= f ,

where

pl

p
, an index denoting the stress triaxiality
q
= Equivalent plastic strain rate.

2008 Abaqus Users Conference

11

The criterion for damage initiation is when an internal material state variable reaches the value
unity, as given by the following equation:
_

D =

pl
_ pl

f ,

=1

This is a monotonically increasing function with strain the material thus never heals upon
reversal of the strain state.
The model is described Abaqus keyword *DAMAGE INITIATION keyword. Once damaged,
the material follows a response given by a dependence on the total displacement at complete
failure (entered on the *DAMAGE EVOLUTION card in Abaqus).

4.4

Calibrating the Material Model

It is not possible to perfectly calibrate the material model in all directions, for all temperatures
thus a compromise has to be settled on. The machine direction at room temperature is the most
important to get right since it plays a major role in the stiffness of the film. The orthotropic Von
Mises model is unable to represent any behavior along the other axes which is different in
character to that along the machine direction because in essence these are all scaled versions of
the baseline behavior. For example, along the cross direction, the material shows the typical
kink (which is typical for these extruded films, due to internal re-arrangement of the polymer
ligaments) in the stress vs. strain curve as shown in Figure 8 and this detail cannot be captured
independently from that occurring along the machine direction. The tailoring of the failure
parameters is extremely difficult to perform, especially so for a material which is significantly
outside the range of normal metal plasticity there is, for example, some significant localization
behavior in the mesh used for material calibration. Since the exact prediction of this failure point
was not considered important (only as aid to show how the embossing pattern leaves a damaged
imprint in the film) this was not pursued to any degree more than necessary to get a reasonably
stable analysis.

12

2008 Abaqus Users Conference

Failure triggered by localization

Figure 8 Calibration curve showing excellent agreement along machine direction at

room temperature and compromise calibration curve along cross direction.

5. RESULTS
5.1

Step 1 Heat Transfer Analysis

As was intuitively to be expected, the thin film reaches steady state temperature conditions very
quickly and may be considered to be uniformly-heated throughout.

5.2

Step 2&3 Embossing and Cooling Analysis

During embossing, the material experiences three phases of deformation viz. a bending dominated
phase, a stretching dominated phase and finally some elastic springback as it leaves the forming
bathtub. These phases are sketched in the graphic in Figure 9 the apparent thickness is computed
through differencing the changes in displacement position between suitably-chosen mesh points
on opposite sides of the thin film. Initially there is no displacement difference (they are both zero)
and the curve starts at zero, but then eventually asymptotes to the apparent thickness after drape
forming. There is an additional change in dimensions caused by thermal expansion shrinkage in
this case, which is not shown on the graphic; it is similar to the elastic springback step.
2008 Abaqus Users Conference

13

End of Embossing

Springback

Stretching dominated

Bending dominated

Figure 9 Apparent thickness of film vs. time(s) during drape embossing.

Figure 10 Illustration of typical stages in drape forming process of the RVE.

14

2008 Abaqus Users Conference

5.3

Step 4 Comparative Tensile Testing Analysis

The comparative stiffness before and after embossing is illustrated in Figure 11 for the 2 different
directions and 2 different states of the film. It should be borne in mind that these results come
from different batches of the material the material data used for calibrating the analysis material
model is not from the same batch as that used to generate the experimental curves in Figure 11.
The analysis shows a surprisingly good ability to predict a change of stiffness in the machine
direction from the flat film state to when embossed, considering the complexity of the material. In
order to generate these points from the analysis output, the apparent thickness for each type of film
(33 microns and 16 microns) has been used to compute the cross-sectional stress from the forces
applied in the analysis. The simplified approach of simply forming the film via a pressure load on
the top surface has caused a more convoluted shape in the cross direction than in the machine
direction this illustration is not shown here. This has left the film to be relatively more flexible in
this cross direction and immediately shows up the relative importance of the geometric effect vs.
the material orthotropic effect for even these small strains.

MD and CD Unembossed Test

MD and CD Embossed Test

MD and CD Embossed Abaqus

Geometric
effect

Figure 11 Comparison between simulation and experimental stiffness.

A point to note is that up to the moderate levels of strain attained in the graphic in Figure 11, the
basic unembossed film is surprisingly isotropic in its behavior; only after ca. 7.5%-10% strain
does this start to change as the material starts to yield. This experimental behavior is consistent
with the test data from the batch used to generate the elastic data cards for the film at room
temperature.
2008 Abaqus Users Conference

15

6. CONCLUSIONS
It is possible using a fairly simple metal plasticity model in Abaqus to predict the stiffness change
due to embossing of the thin film. If predicting only the stiffness change for varying shapes of
embossing pattern is the goal, then one may consider this materials particular behaviour to be
isotropic rather than assuming it to be orthotropic as has been done here; this will reduce
dramatically the experimental support required.
The ultimate load carrying capacity is much more difficult to predict, even using fairly
sophisticated extensions to this material model. There is probably no sense in performing such an
analysis, using the most sophisticated extensions available to a Von Mises model, since the model
is fundamentally unable to account for many of the complicating factors in such materials at
higher strains. If this were to be a requirement in future, significant effort would need to be
invested to develop such a material model. This effort would constitute both significant theoretical
and experimental work, possibly resulting in the writing of a FORTRAN VUMAT subroutine
extension to Abaqus.

7. References

1. Holmes M. and Just D.J., GRP in Structural Engineering, Applied Science Publishers, ISBN
0-85334-232-6
2. Kamal M.R., Haber A. and Ryan M.E., An Engineering Study of the Film Embossing
Process, Polymer Engineering and Science, Mid-August, 1985, Vol. 25 Nr. 11.
3. Kamal M.R., Haber A. and Song Li, The Effect of Processing Conditions on the
Microstructure of Embossed Polyethylene Films, Polymer Engineering and Science, MidFebruary, 1988, Vol. 28 Nr. 3.
4. Kamal M.R., Haber A., Experimental Analysis of the Thermoplastic Film Embossing
Process, Advances in Polymer Technology, Vol. 11 Nr. 3, 159-180 (1992), John Wiley and
Sons.
5. Nagarajan P. et al., Rubber-Assisted Embossing Process, School of Polymer, Textile and
Fibre Engineering, Georgia Institute of Technology, Atlanta, Georgia (30332)

16

2008 Abaqus Users Conference