Anda di halaman 1dari 11

Proceedings of the 5th World Conference on 3D Fabrics and Their Applications

Delhi, India, 16-17 December 2013

Application of ABAQUS beam model to modelling mechanical properties of


woven fabrics.
[Authors List]

ABSTRACT
After describing current approaches to modelling fabric mechanical properties, the paper
will outline the way in which the ABAQUS beam model can be used for this purpose. The
methodology will propose applying energy minimisation to models of the various
combinations of crossing and floating at crossover elements in 2D woven fabrics. As a first
approximation, an affine deformation of the crossover elements will be adopted. The paper
will conclude with a discussion of how the methodology could be applied to 3D woven
fabrics.
Keywords: fabric modeling, ABAQUS beam model, 2D fabrics, 3D fabrics
1. Introduction
A barrier to the wider acceptance of 3D fabrics in demanding functions, such as composites
for aerospace applications, is the lack of computer aided design (CAD) procedures that are
suitable for use by textile manufacturers to predict the performance of the materials and
satisfy the needs of the engineering market. Software, such as Weave Engineer from TexEng
Software Ltd, enables weave structures to be designed and linked to weaving operations.
However, only schematic illustrations of the fabric geometry can be produced. More
advanced methodology is needed to show realistic geometries, which take account of yarn
shapes and curvatures, and predict force-deformation behaviour. Although 3D fabrics are
more complicated, the problem also exists for the simpler 2D fabrics despite a century of
academic work on the structural mechanics of textiles. Partly, it is a cultural problem. The
technical textiles industry has great practical skills and can meet the needs of most customers
through its accumulated knowledge and intuition backed by trial and error. The industry does
not see a need to use modern CAD procedures. Nevertheless to interact technically and
commercially with engineers in aerospace, automobile and construction industries, it is
essential that the culture changes. Efficient and easy-to-use software must tempt companies in
the textile industry to adopt 21st century CAD.
Early research into fabric mechanics used force and moment equilibrium methodologies to
predict deformation responses. More recently, it has been recognized that a better way is
minimization of the sum of the energies of yarn extension, bending, and cross-sectional
change, plus frictional energy when slip occurs. An initial problem is that the weave
specification does not represent the stress-free state of the fabric as made or after treatments
prior to use. An initial minimization routine is required to be followed by deformation steps.
Several software packages are available to carry out the necessary operations. TechText CAD
from the University of Manchester, which uses arbitrarily defined curvatures and shapes for
crossing and free-length sections of yarns, is independently programmed; TexGen from the
1

University of Nottingham uses ABAQUS in its more commonly known form of solid
elements; Wisetex from University of Leuven divides lengths of yarn into short finite
elements but is independently programmed. The aim of this paper is to introduce the use of
the ABAQUS beam model. In this way, the use of finite elements along a cylindrical beam,
which is a good way to represent fibres and yarns, can be combined with the many other
useful features of ABAQUS, which is widely used by engineers.
2. Application of ABAQUS beam model to composites
As an example of the use of the ABAQUS beam model, we describe its application to a
composite of parallel tows [1]. Continuous fibre reinforced composites undergo different
manufacturing processes. Most of these processes guide the composite part properties and
quality. Being a two-phase component for the embedding matrix and reinforcing fibre,
interactions are inevitable both at fibre and tow level. Hence, the authors attempted to study
those interactions at intra- and inter-tow level without the matrix the dry fibre mechanics of
the assembly. The phenomena like fibre entanglement, fibre migration mostly observed in the
dry yarns, take place before resin infusion and understanding such dry fibre mechanics is a
priority for deformation and structural integrity of the manufactured part. The dry compaction
is also observed during weaving when dry tows come in contact. To carry out these
investigations through a detailed experimental programme will be expensive. One alternative
is to use a limited number of experiments to validate a numerical model that can be used for
further study in a parametric sense. Finite element method is a popular numerical tool which
has been used in this area since a long time. However, proper choice of elements affects the
computational power and efficiency to a large extent. The authors in the present research
carried out initial studies on 3D solid finite elements but found the analysis to be
computationally expensive. This led them to utilize the capabilities of a beam element in
order to reduce the computation time. Phenomena like crimp interchange, shear locking had
so far been dealt with additional spring and truss elements which made the geometrical
modeling complex and computation time longer. The present researchers have proposed a
simpler approach of handling such issues while exploring the functionalities of a beam
element. The present article considered the compaction analysis of a bundle of filaments
(fibre scale) where the filaments were meshed with beam elements to predict the realistic
deformation. Nonetheless, the stick-slip friction, a vital aspect in this regard had also been
modeled using a subroutine with Abaqus v6-12 (a commercial finite element software) [2].
The main objective of this study was set to develop numerical models at fibre scale using
beam elements so that it can account the intra-tow deformations with less computational
effort and realistic assumptions. However, this approach can be furthered to fabric level by
representing the warp and weft yarns with 3D beam elements.
3. Review of existing models
Many complex weaves have already made their journey to the industries but the
investigation of the underlying fabric mechanics was not fully explored. Some of the
important assumptions found in most of the models are anisotropic continuum in order to
homogenize the composite and reduce the cost of computation. Although, this assumption
limits the effect of intra-yarn fibre friction, it saves the computation time. Steigmann [3] and
Baseu [4] proposed for the first time a general theory on small displacements for finite
deformations of elastic networks. This method is the forerunner of modelling the structural
2

displacements in case of woven fabrics. Xue et. al. [5] and Shockey et. al. [6] proposed
continuum models for woven composites and similar theories were also accounted for knitted
ones but they suffered from the intra-yarn fibre mechanics. The modelling approach can,
hence, be two-fold i) representing the yarn as a continuum or ii) representing every filament
in a yarn as a continuum. The former lacks important information like intra-yarn interactions,
whereas, the latter may capture all such phenomena at the cost of high computational effort
and time. Studies on structural modelling of interactions within a fabric did not account interyarn interactions which are equally important: a) Crimp interchange is a mechanism by
which the waviness of warp yarns is switched to that of weft yarns and vice-versa upon
loading, b) Shear locking takes place when fabric resists deformation as the interwoven yarns
restrict movement against each other c) Resistance to relative yarn rotation is the fabric
response to in-plane shear.

Figure 1. Basic fabric model geometry using beam and truss elements [7]
In 2005, King et. al. [7] proposed geometry for the fabric kin to that of Kawabata. Here, the
yarns were represented as a truss network which was connected by pin joints at crossover
points. The trusses were not co-planar as they captured the crimp interchange. Interactions
between the yarns were captured by crossover springs connecting the pin joints. These
crossover springs have two modes of deformation extension and contraction to study the
effects of cross-sectional deformation i.e., to allow the change in crimp amplitude. The spring
elements also offered to simulate in-plane fabric shear by its elastic and dissipative resistance
to in-plane rotation of yarn families. The major setback of their proposed model was that the
yarns were modelled as straight segments with sharp bends/crimps whereas in reality these
corners are smoothly wrapped. Figure 1 shows the basic fabric model geometry and the
model modified to take care of shear locking. In total, Kings model included elastic yarn
stretching, bending, and cross-sectional compression at the crossover points, locking and both
elastic and dissipative relative yarn rotation. However, it can be seen from the figure that the
modifications in the simple fabric geometry had made it even more complex. Few researches
have also been carried out in fibre level using digital elements however; inter-fibre friction
and fibre migration were not explored therein [8]. Here comes the necessity and justification
of the present study.
The current research has utilised beam elements that can take care of the functionalities of
springs in Figure 1, for filaments in a fibre bundle to study the compaction behaviour.
3

However, in the present scope the article has dealt with modelling of fibre bundles using
beam elements and is detailed in the subsequent sections.
The authors have also developed a 2D approach of multi-scale modelling of fibre assemblies
as a computationally effective approach [9, 10].
4. Proposed methodology
A fibre scale analysis with 3D beam elements was developed where the user can define the
number of filaments constituting the assembly, meshing, frictional behavior and loading. The
current research has been restricted to: 1) Straight and untwisted fibres, 2) Circular
arrangement of the fibre assembly, 3) No fibre migration and ideal void distribution.
Here, Abaqus v6-12 [2] was used as the numerical modeling tool. However, the preprocessing steps were implemented through a MATLAB program that automatically
generated an Abaqus input file. The program considers the following inputs a. Number of
filaments in a bundle, b. Filament length and diameter, c. Filament material properties, d. Top
and bottom platen dimensions for compaction, e. Platen cross-section is flexible, f. Friction
between filaments and between filament and platen, g. Compaction velocity of top platen, h.
No of finite elements per filament and fine meshing (if required), i. Contact detection
algorithm (developed for the sake of computational efficiency), j. Boundary conditions and
the load by which the filaments are pre-stressed. The output of the Matlab code is an input
file database that can directly integrate with the Abaqus interface.

(a)

(b)
Figure 2. (a) Finite element model of 37 filament assembly, and (b) Compacted model of
the assembly after rendering the beam profiles
4

Figure 2 shows the compacted model of an assembly of 37 fibres meshed with B31 Abaqus
Beam element and interactions defined by Abaqus ITT31 beam contact elements. The contact
detection and contact behavior are discussed later in this section. The top and bottom platens
in this case are of rectangular cross-section. The filament length is same as of 345 mm and all
the filaments are pre-stressed with 0.0004 cN load (calculated based on 50cN on a 12k tow as
used in experiments). In Figure 2 the 3D model represents the rendered cross-sectional
dimension of individual beam elements; however, the cross-sectional deformation was not
taken into account in the beam analysis. The analysis showed that for the meshing scheme
there is an optimal element length for the finely meshed zone i.e. the area where the platen
comes into contact with filaments. This element length should be less than the gap between
the filaments; otherwise numerical instabilities come into play. The top and bottom platens
were meshed with rigid beam elements RB3D2. The 37-fibre assembly FE model contained
about 900 elements per filament with approx 200 fine (at the zone of contact with the platens)
and 700 coarse elements. The beam-beam contact was modeled using a UFRIC friction
subroutine that took into account the stick-slip frictional behavior. The analysis was carried
out in two steps first step the filaments were pre-stressed with a pre-defined load (here,
0.0004 cN per filament) and in the second step the compaction is done with simply-supported
ends. In this case, material properties were anisotropic and the computational time for 37
fibre assembly takes about 3 hours for a computational power of 4-cores and 48GB RAM. At
this point, the authors decided to include the loading stiffness information rather than the
material stiffness in order to reduce the computation time. Thus, using the standard
generalized beam cross-section in Abaqus, axial, bending and torsional stiffness were
incorporated that reduced by 40% of the time taken in previous method. Another issue was
with the total number of interactions that had to be incorporated. In an assembly of 37
filaments the probable number of interactions is 37 filaments taken two at a time i.e. 666
interactions. However, in reality the number of interactions will be fewer than 666. For this a
contact detection algorithm was proposed and used for the current analysis that again reduced
the computation time.
5. Contact detection algorithm
This contact detection algorithm is a modification on the popular box search method. As the
initial fibre arrangement is a circular one so a contact circle was assumed that was drawn with
the filament centre as the circle centre and a radius equivalent to 2-3 times the centre-tocentre distance of the filaments in the bundle based on an ideal void distribution. This
assumed that the filament cannot bypass 2-3 concentric layers of filaments and come in
contact with a filament in a higher layer. This is quite a safe assumption as migration has not
been considered in the analysis. Figure 3 shows the contact detection algorithm with the
bounding circle round a filament under consideration.

Figure 3. Contact detection scheme for a 127 fibre assembly


The advantage of this contact detection scheme is to reduce the computation time without
affecting the accuracy of the problem. It has been seen that this contact detection algorithm is
challenging and useful when the fibre count in a bundle increases. That is, for an assembly of
37 filaments, the number of interactions reduces to 60% whereas for a 127 fibre assembly it
is 20%. This detection method improves the computation time by about 50%. However, the
time is still expensive for a 127 filament assembly model.
6. Contact behavior
In the present study, the frictional behavior between the fibres was attempted to model
using Abaqus user friction subroutine employing the exponential decay rule between the
coefficient of friction and slip rate. Eq. (1) portrays the decay rule with the following
parameters below.
k(s k)exp()

(1)

where, is the instantaneous coefficient of friction at a particular slip rate, k kinetic


coefficient of friction, s static coefficient, decay constant, was the slip rate. The values
of decay constant used in this study were obtained from a preliminary sensitivity study on
fibre friction that was validated with the experiments on fibre friction.
The parameters for the exponential rule were chosen from initial experimental studies as
follows. The static ( s ) and kinetic ( k ) coefficients of friction were taken from the
experimental results of fibre friction as the average of all the peaks and valleys above the
mean line that represents the average friction coefficient. The algorithm for the calculation of
tangential stresses due to friction is discussed below: The exponential decay rule guided the
calculation of instantaneous friction coefficient from the slip rate and the above mentioned
equation-specific constants. The finite element tool calculated the slip rate by dividing the
differential slip of an element in every time increment by the increment and plugged the value
in Eq. (1). Then the instantaneous friction coefficient is multiplied by normal pressure to get
the tangential stresses at a particular direction. A slip tolerance of 0.005 was also used default
so that if the slip has exceeded 0.005 times the characteristic element length then slip would
occur. Automatic stabilization was used to take care of numerical singularities during the
simulation.
7. Results
6

A comparison is made between three different finite element models 3D solid (brick)
elements, 2D solid linear quadrilateral elements [8-9] and 3D linear beam elements for a fibre
assembly of 37 filaments. As the beam elements are axially constrained and their crosssection cannot deform so the computation stops after certain steps as shown in Figure 2(b).
The following Figure 4(a) reflects the comparison of load-displacement response for three
different mesh types for 37 fibre assembly. In Figure 4(b) a column chart is provided which
reflects the improvement in computation time using beam elements. However, the analysis
with 2D elements took minimum time it suffered from certain limitations.

(a)
(b)
Figure 4. (a) Comparison of beam and solid elements for same fibre assembly, (b)
Computation time comparison chart
8. Value of the beam model
The present numerical model using 3D beam elements shows an approach of predicting the
deformation of yarns (here called as tows) that has significantly improved the computational
power. This work can be furthered to fabric level by developing an equivalent beam model
for warp and weft yarn that will include the information of friction between the filaments as
the material model. Further studies can also be conducted on the developed 3D model by
applying twist and migration to the straight tows into the Matlab code. Thus the predictability
of tow deformation at various stages of compaction in the composites manufacturing and
processing techniques were addressed in this research.
9. Application to 2D woven fabric modeling
As described in the introduction, woven fabrics have been used for thousands of years and a
great body of practical knowledge and skill exists in the textile industry. Digital technology
has been adopted for weave design and the link to control of electronic Jacquards on weaving
machines. Schematic 3D models of the weave structure can also be produced. However,
despite a century of academic studies, the two problems of predicting realistic geometry,
taking account of the bending and flattening as yarns cross over one another, and the
deformation under applied forces remain open for research. The need is to produce software
that can be used in industry. This is particularly important when woven fabrics are being
supplied for use in composites and other technical products. Engineers in these industries
expect to interact with suppliers in a 21st century computer-aided design (CAD) mode and not
the craft approach of the traditional textile industry.
This part of the pa per outlines a procedure by which the ABAQUS beam model could offer
7

a new way forward for fabric modelling. This is not surprising since fibres and yarns, which
make up the fabric, are slender flexible beams. The methodology is to minimise the sum of
the energies of yarn extension, bending and change of shape. In a plain weave, Figure 5(a),
yarns cross under and over each other at each intersection. In other weaves, yarns float past
at some intersections without crossing to the other side, as in a twill weave, Figure 5(b).

(a)
(b)
Fig. 5 (a) Plain 1/1 weave. (b) 2/2 twill weave. Repeat units are outlined.

As a first approximation to the modelling, it can be assumed that the repeat units shown in
Figure 5 deform affinely in a similar way to the overall fabric deformation. For a plain
weave, the procedures outlined above would generate a geometrical model that could then be
modelled mechanically by the ABAQUS beam model, taking account of yarn extension,
bending and change of shape, plus frictional slip, which is of particular importance in fabric
shear as rotation at crossovers occurs.
The problem is more complex for other weaves, which is perhaps why most of the literature
deals with plain weaves. In a forthcoming publication Hearle et. al. [11] propose a way of
describing weaves that identifies the forms of crossovers digitally. For simplicity, Figure 6
shows the application of this notation for a plain weave in which all the entries have the form
for a double crossover (0 or 1,11,00 or 11). A double float in both directions is given by (0
or1,00,00); a mixed float and cross would be given by such as (0,11,10) for a single float or
(0,10,10) for a double float. Figure 7 illustrates all the possible combinations of crossing and
floating. The principle of affine deformation could be carried one stage further by assuming
that it applies to each crossover in the repeat unit. A collection of ABAQUS beam models
would then provide a resource of sub-units to be incorporated in any weave.

Figure 6. Digital representation of a plain weave.


It is likely that the assumptions of affine deformation would give a reasonably accurate
prediction of real fabric geometry and deformation behaviour. However, it would be possible
to have additional stages of computation in which the dimensions of the repeat units or
internal sub-units were allowed to change to minimise the energy.
The first step in the energy minimisation would be to determine the initial stress-free
geometry. Then a small deformation of the fabric would be imposed and the new state of the
fabric computed. Another small step would then be made, so that the difference in energy
between the two would give the force or moment. The procedure would be repeated in a
succession of larger steps to generate the full force-deformation response.

Figure 7. All the possible combinations of crossing between under and over and floating.
With thanks to Haseeb Arshad for the drawing.
10. Application to 3D fabrics.
The same principles could be applied to 3D woven, or indeed braided or knitted, fabrics,
although there are many more forms in which yarns cross one another. The Weave Engineer
program would generate the topology from which the crossing sub-units could be extracted
for computation by by the ABAQUS beam model. For a simple 3D woven fabric, such as the
angle-interlock or multilayer in Figure 7 (a, b), the procedure would be similar to that for 2D
weaves but more complicated. Where the overall shape of the fabric is more complex, as in
the T-shape of the orthogonal weave in Figure 7(c) or the hollow fabric of Figure 7(d),
additional software would be needed. This would probably involve dividing up the structure
into separate, but linked, pieces.
11. Conclusion
If 3D fabrics are to achieve their full potential in the advanced engineering markets, such as
aerospace, automobile and construction, it is essential that the fabric manufactures adopt a
21st century approach to CAD that will provide the prediction of performamce that engineers
in the user industries require. CAD facilities will also provide the language for interaction
between producers and users as they jointly address the problems of a particular application.
To take just one example, Alan Prichard from Boeing in his presentations to the 3D fabric
conference has stressed the importance of finding the best way of joining a composite wing to
10

the fuselage of a plane.


Composites are inherently more complicated than homogeneous materials and 3D fabrics
introduce complications than are not found in simpler laminated composite structures. The
mission of the present paper is to encourage people to adopt the ABAQUS beam model, or
similar software, to provide efficient, validated and easy-to-use software to provide the tools
that the industry needs.
References
1. Mandal P., Chakladar, N. D., Potluri, P., Finite element modelling of fibre assemblies using beam
elements, First International Conference on Digital Technologies for the Textile Industries, Manchester,
2013.
2. Abaqus 6.12 [Computer Software], Providence, RI, Dassault Systemes Simulia.
3. Steigmann, D.J., Equilibrium of prestressed networks. IMA Journal of Applied Mathematics, 1992. 48:
p. 195-215
4. Baesu, E., Finite deformations of elaso-plastic filamentary networks. International journal of Nonlinear mechanics, 2003. 38: p. 1473-1479
5. Xue, P., X. Peng, and J. Cao, A non-orthogonal constitutive model for characterising woven composites.
Composites Part A: Applied science and manufacturing, 2003. 34: p. 183-193.
6. Shockey, D.A., D.C. Elrich, and J.W. Simons, Improved barriers to turbine engine fragments: Final
Report. 2002.
7. King, M.J., P. Jearanaisilawong, and S. Socrate, A continuum constitutive model for the mechanical
behaviour of woven fabrics. International journal of solids and structures, 2005. 42: p. 3867-3896.
8. Zhou, G., X. Sun, and Y. Wang, Multi-chain digital element analysis in textile mechanics. Composites
Science and Technology, 2004. 64: p. 239-244.
9. Chakladar, N.D., P. Mandal, and P. Potluri. Multi-scale modelling of compaction of fibre assemblies.
Proceedings of the International conference on designing against deformation and fracture of composite
materials: Engineering for integrity large composite structures(DF-6). 2013. Cambridge, England.
10. Chakladar, N.D., P. Mandal, and P. Potluri. Multi-scale modeling of fibre bundles. Proceedings of 19th
International conference on composite materials. 2013. Montreal, Canada.
11. Hearle, J.W.S., Ramgulam, R.B, Jiang, Y., Potluri, P., Chen, X., to be submitted to J. Textile Inst.

11