Sie sind auf Seite 1von 133

Nonlinear mechanical behavior of automotive airbag fabrics: an

experimental and numerical investigation

by

Steven Edward Zacharski


B.Sc., Villanova University, 2008

A thesis submitted in partial fulfillment of the requirements for the degree of

MASTER OF APPLIED SCIENCE

in

The Faculty of Graduate Studies

(Materials Engineering)

THE UNIVERSITY OF BRITISH COLUMBIA


(Vancouver)

December 2010
Steven Edward Zacharski, 2010
Abstract

Abstract
Over the past two decades, the airbag has become an essential safety device in automobiles. The
airbag cushion is composed of a woven fabric which is rapidly inflated during a car crash. The
airbag dissipates the passengers kinetic energy thereby reducing injury through biaxial
stretching of the fabric bag and escaping gas through vents. Therefore, the performance of the
airbag is greatly influenced by the mechanical properties of the fabric. Unlike traditional
engineering materials, airbag fabrics are composed of discrete constituents and have highly
nonlinear mechanical behavior that arises from both geometric deformations and material
nonlinearity. Henceforth, airbag designers are forced to make simplified assumptions regarding
the mechanical behavior of the fabric cushion. This incontrovertibly limits designers in taking
advantage of the full potential of the fabric system. In order to optimize the airbag design,
improve deployment simulations and overall dependability, a more sophisticated approach is
needed.

In this study, a simple unit cell model representing a single crossover of two orthogonal woven
yarns is developed to simulate the in-plane mechanical behavior of both coated and uncoated
plain weave airbag fabrics under multiple states of stress. Since the structural analysis of the
deployment of the airbag is performed using the finite element method, the proposed mechanistic
model is implemented as a User-Material-Model in the commercial code LS-DYNA. Here, the
unit cell model represents the constitutive behavior of a continuum membrane. The approach
results in capturing, in detail, the discrete nature of the fabric while retaining the computational
efficiency of simple membrane formulation compared to explicitly modeling each yarn within
the fabric.

The procedure to calibrate the model inputs, namely the yarn geometric and mechanical
properties for a given fabric is detailed. The sensitivity of the unit cell model and verification of
the finite element implementation is discussed. A series of experiments were performed to
validate the in-plane behavior of the model. The proposed model can be adopted by designers to
better represent the nonlinear mechanical behavior of the fabric. It can also be used as a tool to
design novel fabrics that are optimized for a particular application.

ii
Table of Contents

Table of Contents
Abstract ........................................................................................................................................... ii
Table of Contents........................................................................................................................... iii
List of Tables .................................................................................................................................. v
List of Figures ................................................................................................................................ vi
Nomenclature............................................................................................................................... viii
Acknowledgements......................................................................................................................... x
Chapter 1 Introduction ................................................................................................................. 1
1.1 Motivation............................................................................................................................. 1
1.2 Goals and objectives ............................................................................................................. 3
1.3 Outline .................................................................................................................................. 4
Chapter 2 Background ................................................................................................................. 6
2.1 Overview of airbag fabric technology .................................................................................. 6
2.2 Airbag fabric research........................................................................................................... 9
2.2.1 Summary ...................................................................................................................... 14
2.3 Modeling of the mechanical behavior of fabrics ................................................................ 15
2.3.1 Representative mechanistic models ............................................................................. 15
2.3.2 Continuum approaches................................................................................................. 19
2.3.3 Current state-of-the-art................................................................................................. 20
2.3.4 Summary ...................................................................................................................... 21
Chapter 3 - Development of a representative unit cell: theory and calibration ........................... 22
3.1 Background and approach .................................................................................................. 22
3.2 Unit cell definition .............................................................................................................. 24
3.3 Deformational mechanisms and constitutive relationship .................................................. 29
3.3.1 Yarn axial extension .................................................................................................... 29
3.3.2 Yarn bending................................................................................................................ 30
3.3.3 Coating extension......................................................................................................... 33
3.3.4 Unit cell in-plane shear behavior ................................................................................. 34
3.4 Numerical procedure........................................................................................................... 37
3.5 Characterization of constituent properties .......................................................................... 39
3.5.1 Microscopy .................................................................................................................. 39
3.5.2 Yarn extension test....................................................................................................... 42
3.5.3 Coating characterization .............................................................................................. 47
3.5.4 In-plane shear calibration............................................................................................. 48
3.6 Results................................................................................................................................. 50
3.7 Sensitivity analysis ............................................................................................................. 51
3.8 Summary............................................................................................................................. 54
Chapter 4 Implementation and verification ............................................................................... 56
4.1 Background and approach .................................................................................................. 56
4.2 Explicit dynamic finite element analysis ............................................................................ 60
4.3 Implementation of unit cell based UMAT for FE shells .................................................... 61
4.3.1 Continuum formulation................................................................................................ 61
4.3.2 Special considerations yarn separation, failure and element erosion ....................... 65
4.4 Verification ......................................................................................................................... 66
4.5 Summary............................................................................................................................. 68
Chapter 5 Validation .................................................................................................................. 70

iii
Table of Contents

5.1 Background and approach .................................................................................................. 70


5.2 Experimental evaluation ..................................................................................................... 70
5.2.1 Uniaxial extension ....................................................................................................... 70
5.2.2 Biaxial extension.......................................................................................................... 72
5.2.3 Bias extension .............................................................................................................. 78
5.3 Comparison of UC, FE-UC and experiments ..................................................................... 85
5.3.1 Uniaxial results ............................................................................................................ 85
5.3.2 Biaxial results............................................................................................................... 88
5.3.3 Bias results ................................................................................................................... 94
5.4 Summary............................................................................................................................. 99
Chapter 6 - Conclusion and recommendations ........................................................................... 100
6.1 Conclusions....................................................................................................................... 100
6.2 Recommendations and future work .................................................................................. 101
6.3 Summary........................................................................................................................... 102
References................................................................................................................................... 103
Appendix A UMAT *MAT card documentation..................................................................... 109
Appendix B UMAT pseudo code ............................................................................................ 117
Appendix C Justification of the selection of a linear unit cell over sinusoidal geometry ....... 120
Appendix D Conversion of membrane stress into specific stress............................................ 123

iv
List of Tables

List of Tables
Table 3-1: Existing center-line unit cell models vs. proposed model........................................... 23
Table 3-2: Unit cell configuration conditions............................................................................... 28
Table 3-3: Geometric properties of coated and uncoated airbag fabric as determined from
microscopy.................................................................................................................................... 42
Table 5-1: Test methods in the literature for evaluating the biaxial behavior of fabrics............. 73
Table 5-2: Biaxial extension of coated fabric experiment vs. simulation.................................. 91
Table 5-3: Biaxial extension of uncoated fabric experiment vs. simulation.............................. 92
Table 5-4: Comparison of bias deformation of coated sample: experiment vs. simulation ......... 97
Table 5-5: Comparison of bias deformation of uncoated sample: experiment vs. simulation ..... 98

v
List of Figures

List of Figures
Figure 1-1: Biaxial extension of airbag fabric vs. current linear elastic assumption...................... 3
Figure 1-2: Overview of the scope of this study............................................................................. 4
Figure 2-1: Pierce's unit cell (1937).............................................................................................. 16
Figure 2-2: Kawabata's unit cell (1973)....................................................................................... 18
Figure 3-1: Simple unit cell geometry in initial and deformed states.......................................... 24
Figure 3-2: Procedure for generating stress-strain behavior of fabric under multiple states of
stress using the unit cell approach ................................................................................................ 27
Figure 3-3: Bending geometry of an elastica................................................................................ 30
Figure 3-4: Typical shear stress-strain curve for single-coated and uncoated fabric .................. 35
Figure 3-5: Shear model behavior a) secant shear modulus as a function of strain b) regions in
the shear stress-strain curve .......................................................................................................... 36
Figure 3-6: Plain weave of a) single-coated fabric and b) uncoated fabric ................................. 40
Figure 3-7: Cross section of 350dtex coated airbag fabric a) warp direction and b) fill direction
....................................................................................................................................................... 41
Figure 3-8: Cross section of 350dtex uncoated airbag fabric a) warp direction and b) fill
direction ........................................................................................................................................ 41
Figure 3-9: Yarn specimen preparation procedure ...................................................................... 43
Figure 3-10: KES-G1 microtensile tester with loaded yarn sample ............................................. 44
Figure 3-11: Procedure of obtaining "pure" yarn load-elongation response ................................ 45
Figure 3-12: Average force-elongation curve for 350dtex nylon 6,6 airbag yarn ........................ 46
Figure 3-13: KES-FB1 textile shear tester shown in a) and b) shear deformation adopted in the
KES-FB1 testing system ............................................................................................................... 48
Figure 3-14: Shear stress-strain behavior for 350dtex nylon airbag fabric ................................. 49
Figure 3-15: Membrane stress-strain curves for 350dtex fabric produced by the unit cell model
....................................................................................................................................................... 51
Figure 3-16: Crimp parameter sensitivity for a) 1:1 biaxial extension and b) uniaxial extension52
Figure 3-17: Yarn bending rigidity parameter sensitivity for a) 1:1 biaxial extension and b)
uniaxial extension ......................................................................................................................... 53
Figure 3-18: Coating thickness parameter sensitivity for a) 1:1 biaxial extension and b) uniaxial
extension ....................................................................................................................................... 54
Figure 4-1: Outline of the basic concept of the unit-cell-membrane ........................................... 58
Figure 4-2: Effect of curvature on the internal structure of woven fabric ................................... 59
Figure 4-3: Numerical procedure of LS-DYNA with user material model option...................... 60
Figure 4-4: Verifying modes of shear: a) pure shear and b) rail shear ........................................ 67
vi
List of Figures

Figure 4-5: Verifying shear deformation behaviors of a single element with varying shear
stiffnesses...................................................................................................................................... 68
Figure 5-1: Average uniaxial stress-strain of 350dtex fabric........................................................ 71
Figure 5-2: Biaxial tester at Drexel University............................................................................. 74
Figure 5-3: Biaxial specimen dimensions..................................................................................... 75
Figure 5-4: Biaxial testing specimen configuration...................................................................... 76
Figure 5-5: Average biaxial stress-strain curve for coated and uncoated samples ...................... 77
Figure 5-6: Photographs of biaxial extension at approximately 12% strain for a) uncoated and b)
coated fabrics ................................................................................................................................ 78
Figure 5-7: Heterogeneous deformation of fabric bias extension................................................. 79
Figure 5-8: Bias extension-apparent shear angle recording set-up ............................................... 80
Figure 5-9: Load-elongation of bias coated and uncoated airbag fabric ..................................... 81
Figure 5-10: Bias shear -- applied specimen end stress vs. center specimen shear angle ........... 82
Figure 5-11: Typical bias shear deformation sequence - 50mm x 100mm coated sample.......... 83
Figure 5-12: Typical bias shear deformation sequence - 50mm x 100mm uncoated sample...... 83
Figure 5-13: Detailed low shear angle plot of applied stress vs. center shear of bias sample..... 84
Figure 5-14: Uniaxial extension of coated airbag fabric - model vs. experiments ....................... 86
Figure 5-15: Uniaxial extension of uncoated airbag fabric - model vs. experiment..................... 86
Figure 5-16: Macroscopic strains of simulated uniaxial loaded coated airbag fabric at 30%
extension a) warp direction and b) fill direction........................................................................ 87
Figure 5-17: Yarn strains of simulated uniaxial loaded coated airbag fabric at 30% extension
a) warp direction and b) fill direction ........................................................................................... 88
Figure 5-18: Biaxial extension of coated airbag fabric - model vs. ex periments ........................ 89
Figure 5-19: Biaxial extension of uncoated airbag fabric - model vs. ex periments .................... 89
Figure 5-20: Simulated yarn strains at approximately 12% equal biaxial extension in the- a) warp
direction and b) fill direction ........................................................................................................ 93
Figure 5-21: Bias load-elongation for coated and uncoated fabric - simulation vs. experiment .. 95
Figure 5-22: Applied shear stress vs. measured shear angle - simulation and experiments......... 95
Figure 5-23: Detailed low level applied shear stress vs. measured shear angle - simulation and
experiments ................................................................................................................................... 96
Figure A1: Shear model behavior regions in the shear stress-strain curve ............................... 113
Figure A2: Secant shear modulus as a function of strain........................................................... 114
Figure C1: Approximation of yarn height by linear and sinusoidal unit cell ............................. 120
Figure C2: Approximation of sin theta by linear and sinusoidal unit cell .................................. 121
Figure C3: Approximation of cos theta by linear and sinusoidal unit cell ................................. 121
Figure C4: Approximation of warp fabric membrane stress by linear and sinusoidal unit cell . 122
vii
Nomenclature

Nomenclature
( )i Index referencing warp yarn (i=1) and fill yarn (i=2)
Ai Specific area of yarn in the i material direction
c Speed of sound in the fiber material
c* Speed of sound across a crimped yarn
cri Yarn crimp in the i material direction
dfiber Diameter of a single fiber
Ec Elastic modulus of coating
(EI)yarn Yarn bending rigidity the i material direction
Fb,i Vertical force component contributed to yarn bending
Fc,i Vertical force component contributed to yarn extension
Fct,i In-plane force generated at the end of the yarn in the i direction by the coating
Fend,i Total in-plane force in-plane force generated at the end of the unit cell in the i
material direction
Ff,i In-plane force generated at the end of the yarn in the i direction by the yarn
Fy,i Yarn axial tension force
G Shear modulus
G1 Unit cell shear modulus that is contributed by coating shear
G2 Unit cell shear modulus that is contributed by yarn rotation
G3 Unit cell shear modulus that is contributed to locking of yarns
H0i Initial yarn centerline height at the center of the unit cell
Hi Current yarn centerline height at the center of the unit cell
K Bulk modulus
L0i Initial yarn length
Ld,i Linear density of the yarn
Li Current yarn length
Nfiber Number of fibers per yarn
ni Yarns per inch
Ni Normal in-plane membrane stress
p Packing factor of the yarn
rc Radius of curvature
S12 In-plane unit cell shear membrane stress
tc Coating thickness
wc Areal density of coated fabric
wu Areal density of uncoated fabric
y0i Initial horizontal spacing of yarn
yi Current horizontal spacing of yarn
Strain increment tensor
i Strain applied to the unit cell in the i material direction
yarn,i Yarn strain in the i material direction

yarn,
ult Ultimate yarn strain in the i material direction
i

i 0 Original angle between material vector i and the local material axes
it Angle between material vector i and the local material axes at time step t

viii
Nomenclature

1 Shear strain corresponding to the end of coating denominated shear stiffness


2 Shear strain corresponding to the beginning of yarn rotation dominated shear
stiffness
3 Shear strain corresponding to the onset of contact between parallel yarns
4 Shear strain at which shear locking occurs
i Total stretch in the i material direction
c Poissons ratio of the coating
Density of unit cell-shell

ix
Acknowledgements

Acknowledgements

I wish first to thank my supervisors Dr. Frank Ko and Dr. Reza Vaziri for all their mentoring,
guidance, ideas and assistance through the duration of this work. I am grateful for their
willingness to share with me the worlds of textile and computational mechanics. I would like to
thank the financial support from TRW Automotive on this project and the technical support from
Dr. Chuan Lee from TRW.

I also wish to thank Dr. Joseph Wartman and Mr. David Harmanos of the Civil, Architectural
and Environmental Engineering Department at Drexel University for sharing and their aid in
helping me operate the biaxial tester which was vital to validating the model developed in this
work. Additionally, I would like to thank Dr. John Gosline and Dr. Ken Savage from UBC
Zoology for sharing their tensile tester which was used to generate the uniaxial and bias data
used for validation.

I also offer much thanks to the past and present members of the UBC Composites Group and the
Advanced Fibrous Materials Laboratory for many fruitful discussions and advice.

Finally, I would like to thank my friends and family back east for their support and
understanding while I pursue to advance my education and career. I especially want to thank my
parents for their love and encouragement throughout the years.

x
Chapter 1 Introduction

Chapter 1 Introduction

1.1 Motivation
The first recorded automobile fatality occurred in Birr, Ireland in 1869 (Fallon & ONeill, 2005)
-- an event that arguably marked the dawn of the study of automobile safety. Over the past two
decades, the emergence of the airbag has established itself as an integral part in vehicle safety for
passengers. The United States National Highway Traffic Safety Administration (NHTSA)
estimates that as of 2009 more than 28,000 lives have been saved in the U.S. because of frontal
airbags (National Highway Traffic Safety Administration, 2009). While the airbag has seen
great success, the sobering statistic of 24,474 vehicle occupant fatalities in the U.S. during 2009
documented by the NHTSA illustrates that the airbag as well as other safety technologies need
continuing improvement (National Highway Traffic Safety Administration, 2010). The concept
of the airbag is fairly simple: upon collision, a charge is sent to ignite a gas explosion that rapidly
inflates the airbag. The airbag provides a cushion in which the impact energy is dissipated,
forces that act upon the passenger are distributed over a large area and excessive rotations of the
passenger are limited. When the forces and rotation that act on the passenger are kept to a
minimum, the likelihood of injury dramatically decreases.

Considerable efforts of the airbag designers and manufacturers have been focused on producing
airbag systems that are reliable and have predictable performance. Of fundamental importance
to the airbag performance is the mechanical properties of the airbag fabric. Airbag fabrics, which
are typically constructed of a simple plain weave of nylon yarns, exhibit unique characteristics
that differ from the traditional engineering materials. More specifically, airbag fabrics are
heterogeneous, anisotropic, have the ability to undergo large deformations and exhibit nonlinear
mechanical behavior. Heterogeneity and anisotropy lends itself to the geometric assembly of
discrete constituents while nonlinearity is due to both geometric deformations and material
nonlinearity of the constituents.

1
Chapter 1 Introduction

The applied pressure in the airbag is a follower-type loading, meaning the temporal and spatial
distribution of the load depends on the structural response of the fabric. Structural response is
governed by material behavior of the fabric as well as the operative boundary conditions. A great
amount of effort has been devoted by the computational mechanics community to develop
sophisticated gas models and fluid-structure interaction algorithms which have improved the
accuracy of simulating the operative boundary conditions. Still, simplified assumptions are
made regarding the mechanical behavior of the fabric during the design, analysis and simulation
of the airbag structure.

Of particular interest to the airbag industry is improving the structural analysis and design
through a better understanding of the mechanical behavior as well as the failure behavior of the
fabrics under deployment conditions. The most common simplified assumption in the industrys
structural analysis is to approximate the fabric as an orthotropic continuum with linear elastic
mechanical behavior (Drnhoff et al., 2008; Hirth, Haufe, & Olovsson, 2007; Wawa, Chandra, &
Verma, 1993). The assumed elastic modulus is taken from the initial modulus measured by
experimental data. This type of approach obviously neglects the internal changes of the fabric
structure and the material nonlinearity of the yarn (Wawa et al., 1993). Figure 1-1 illustrates the
shortcomings of the current linear elastic analysis capabilities to capture what is physically
observed in biaxially strained fabrics from this study. From the figure, it can be observed that
after 4% strain there are large portions of the fabric stress-strain curve where the assumed
stiffness is first over estimated and then at strains exceed approximately 15%, the assumed
mechanical behavior under predicts the stiffness of the physical fabric. The erroneous
estimations can therefore lead to false predictions regarding energy absorption of the system and
incorrect simulations of the airbag deployment and impact process. Additionally, there are no
clear criteria to describe the failure of the airbag fabric which is a concern for the industry.
Refinements in the formulation of the mechanical behavior that take into account the nonlinear
stress-strain behavior up to failure can result in better predictions regarding both the dynamics
and failure of the airbag structure.

2
Chapter 1 Introduction

40.00
Y Assumption
35.00
X Test
30.00
membrane stress (N/mm)

Test
25.00

20.00
Assumption
15.00

10.00

5.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25
strain

Figure 1-1: Biaxial extension of airbag fabric vs. current linear elastic assumption

1.2 Goals and objectives


The objective of this thesis is to model the nonlinear stress-strain behavior of airbag fabrics up to
failure under deployment-like conditions. The aim of this work is to develop a constitutive
model that accurately represents the anisotropy and nonlinearity of the airbag fabric material
under states of stress seen in the deployment of the airbag using simple inputs based on the
fabrics constituents. Since the industry utilizes airbag simulations that are performed using
dynamic explicit finite element codes, the material model is intended to be implemented into
such codes keeping in mind computational efficiency. Experimental evaluation of the fabric is
performed to generate input data for the material model as well as reference for validation.
Figure 1-2 illustrates the general strategy in this study. Successful modeling of the stress-strain
behavior can aid engineers in improving the safety and dependability of the airbag system.

3
Chapter 1 Introduction

Figure 1-2: Overview of the scope of this study

It should be made clear that the purpose of this work is to improve predictions of the elastic
behavior of the airbag fabric up to failure under multi-axial states of stress. While a more
sophisticated representation of the constitutive behavior of the airbag fabric may improve the
structural analysis and simulated deployment kinematics, any conclusions drawn about
improvements regarding post-impact kinematics or passenger energy dissipation should be made
cautiously as the work does not consider visco-plasticity (i.e. permanent and rate-dependant
deformations of the fabric).

1.3 Outline
This section presents a general outline of the thesis. Chapter 2 begins with a brief overview of
the airbag system along with the historical evolution of the fabric system. A literature review
presents the work that has been performed in either evaluating the mechanical behavior of airbag

4
Chapter 1 Introduction

fabrics experimentally or by modeling. Then a historical review of the classical literature in


textile mechanics pertinent to modeling the elastic behavior of woven fabrics and the current
state-of-art is presented.

Chapter 3 presents a representative unit cell model based on a simple linear approximation of the
yarn crossover geometry for woven fabrics. The structural mechanisms and constitutive
relations that empower the mechanical response of the unit cell are discussed in detail. The
procedures for characterization of constituent properties for a current generation airbag fabric are
presented. A sensitivity analysis of the unit cell inputs is performed using properties of the
airbag fabric as the nominal baseline.

Chapter 4 describes the continuum formulation for a finite element shell using the unit cell
model as the basis for the constitutive relationship. A brief overview of the importance of finite
element analysis in airbag and crashworthiness simulations is provided. The unit cell based
continuum formulation is implemented as a user material subroutine (UMAT) into the
commercial dynamic finite element code, LS-DYNA. The derivation of the continuum
formulation that transforms element strains to the unit cell and transforms the unit cell stresses
back to the element under large deformations is presented. A few verification case studies are
performed to confirm the successful implementation of the theory.

Chapter 5 discuses the validation of the modeling techniques derived in Chapters 3 and 4. The
fabric was tested under uniaxial, biaxial and bias extension to validate the model under different
states of stress. The experimental procedure is carefully laid out and the correlation between the
simple unit cell model predictions and the experimental results is discussed in detail.

Chapter 6 presents the conclusions drawn from the current study and recommends possible
future work in the area of modeling woven fabrics and their failure predictions under multiple
stress states leading to airbag simulations.

5
Chapter 2 Background

Chapter 2 Background

2.1 Overview of airbag fabric technology


During an automobile collision, the vehicle can undergo a rapid change in velocity while the
passengers continue to move until an opposing force occurs. Highly concentrated opposing
forces arising from the vehicle interior or seatbelt can result in serious injury to passengers. An
airbag is designed to minimize these concentrated forces and reduce excessive motion (mostly
upper body rotation) of a belted passenger. Upon collision, a charge is initiated which inflates
the airbag that forms a cushion between the passenger and vehicle interior. The deployment
sequence occurs rapidly between 25 to 30 milliseconds from the time of sensing the crash to full
inflation of the airbag where the speed of the deploying airbag can reach up to 160 km/hr (100
mph) (Crouch, 1994; Mukhopadhyay, 2008). At which time gas pressures can reach 70 kPa and
temperatures close to 600 C for a few milliseconds (Gon, 2010; Mukhopadhyay, 2008). The
bags internal pressure is uniformly applied to the occupant and the bag gently dissipates the
passengers kinetic energy while distributing the contact force over a large area and minimizing
rotations of the passengers articulated segments. The mechanisms that contribute to airbags
absorption and dissipation of energy are the mechanical biaxial stretching of the fabric and
escaping gases.

While historical references of the airbag can be traced back to the 1920s, the first patent in the
United States for an automotive safety cushion assembly was issued in 1953 (Crouch, 1994;
Mukhopadhyay, 2008). The concept was not executed in a commercial vehicle until the early
1970s when General Motors started to produce automobiles with the safety devices.
Unfortunately, the feature garnered little public support due to expensive and complicated
technology. With additional research and development of an effective and more economical
airbag system, the technology started to be implemented into vehicles extensively in the late
1980s. Starting in 1998, all new vehicles in the United States were required to have driver and

6
Chapter 2 Background

passenger airbags. Today, vehicles are outfitted with side curtain and thorax airbags for side and
rollover impacts (Mukhopadhyay, 2008).

The technical requirements for airbags are defined as: good foldability, high softness, resilient,
heat resistant, low air permeability, high dynamic stress resistance, low fabric weight, high fabric
strength, good abrasion resistance and stability upon aging (Gon, 2010; Mukhopadhyay, 2008;
Schwark & Muller, 1996; Crouch, 1994). To meet these requirements, the airbag is constructed
of woven textile membranes that are mostly constructed from multi-filament polyamid 6,6
(nylon) yarns. The woven fabric structure used in the airbag application is the plain weave;
where the warp and fill yarns are interlaced in a regular sequence of one under and one over.
The term warp refers to the direction of yarns that run continuously through the weaving
machine and are continuous for the entire length of the fabric roll. The fill yarns (sometimes
referred to as weft or woof) are the yarns that run transverse to the warp yarns.

The mechanical behavior of woven fabrics is anisotropic and highly nonlinear. The nonlinearity
is a result of internal structural changes during deformations where the deformations are finite
but can also be attributed to material nonlinearity of the yarn constituents. Woven fabrics
possess high flexibility which allows them to fold and conform to a variety of shapes. The low
bending stiffness is attributed to the interlacing of yarn without rigid bonding of the overlapping
points as well as thin parallel system of fibers which are only restricted by friction.

The yarn material system, polyamide (or better known by its trade name nylon) is the generic
term for any long-chain synthetic polymeric amide which has recurring amide groups as an
integral part of the main polymer chain (Harris, 1954). While this definition can cover a wide
range of structures, the most important group from which commercial fibers are constructed
include condensation polymers of an - straight-chained aliphatic diamine with an - straight-
chain aliphatic dicarboxylic acid. Specifically, nylon 6,6 which is most commonly used in
airbag fabrics is the condensation polymer of hexamethylene diamine with adipic acid (Harris,
1954).

It should also be mentioned that polyester (PET) and nylon 4,6 yarns are used in small quantity
for airbag fabrics and sewing threads in some parts of world. Nonetheless, nylon 6,6 is the fiber

7
Chapter 2 Background

of choice compared to other synthetic or natural fibers because it has the highest strength-to-
weight ratio at an economical price and other properties that are desirable in airbag applications.
Nylon 4,6 has similar mechanical and thermal behavior to nylon 6,6 but is more expensive (Gon,
2010; Mukhopadhyay, 2008). Polyester is a cheaper fiber that is used extensively in seatbelts
and other automotive textiles. However, properties that make polyester a suitable material
candidate for seatbelts does not translate into airbag applications. Polyester undergoes less
dimensional changes compared to nylon under moisture and temperature fluctuations which
allows for smooth uptake and pull out of the seatbelt (Crouch, 1993). Additionally, polyester is
more rigid than nylon which prevents excessive stretching of the belt during high loading that
occurs as a result of an impact. The high elongation of nylon is advantageous in airbags as it
promotes wider allocation of forces throughout the airbag as well as uniform stress distribution
along perimeter seams. Additionally, nylon has a relatively higher melting point, high heat of
fusion and absorbs 2-4% water by weight that provides quenching properties which aids in the
prevention of burn through from hot particulates that could possibly break free from the inflator
and travel into the bag (Crouch, 1993).

The airbag fabric can also be coated with an elastomer which lowers the permeability of the
fabric and provides an ablative shield to the fabric from the hot gases (Crouch, 1993; Crouch,
1994; Gon, 2010; Schwark & Muller, 1996). The first generation of airbags used a neoprene
coating while todays airbags are almost exclusively coated with silicone. Neoprene coatings
can release HCl over time which can degrade the nylon yarns. Additionally, neoprene has a
tendency to self-adhere which requires the airbag to be coated with talc while silicone does not
exhibit adhering so no talc is used. Silicone rubbers are more stable and therefore do not degrade
the fabric even after high temperature aging. This means the amount of coating required is
significantly less if silicone is used instead of neoprene, resulting in a thinner, lighter and more
compliant air bag (Gon, 2010). Silicone exhibits better wear and abrasion resistance compared
to neoprene coatings. From a manufacturing point of view, the silicone used in the airbag
application is a Pt-cured, low viscosity liquid rubber applied using knife coating equipment to
apply a thin coat upon one side of the fabric (Mukhopadhyay, 2008; Schwark & Muller, 1996).
After the coating is applied to the fabric, the fabric is pulled through an oven to induce
polymerization and adhesion between the fabric and coating.

8
Chapter 2 Background

There has been much debate by the airbag industry regarding the value of coated and uncoated
fabrics (Crouch, 1993; Gon, 2010; Mukhopadhyay, 2008; Schwark & Muller, 1996). The coated
fabrics offer better resistance to heat conductivity, tear performance, and low permeability. On
the other hand, the process to coat the fabrics is environmentally unfriendly. The fabric becomes
hard to handle due to the permeation of the solvent of the coating liquid. Uncoated fabrics are
more environmental friendly and are easier to fold and pack into small spaces.

Aside from the evolution from neoprene to silicone coatings, the fabric construction has changed
throughout the twenty years of airbag use (Crouch, 1993; Mukhopadhyay, 2008). The first
generation of fabric design consisted of 940 decitex (abbreviated as dtex) yarns. Decitex is a
measure of the linear density of the yarn and is defined as the weight in grams per 10,000 meters
of yarn. Another popular unit of linear density used in textile is denier, which is the weight in
grams of 9000 meters of yarn. The higher the decitex or denier, the thicker the yarn and coarser
the fabric weave. The first generations of fabrics were coarse, heavy and difficult to pack.

The second generation of fabrics used during the mid-1990s were made of high tenacity 470 dtex
nylon yarns. At this time, the transition from neoprene to silicone coatings was also seen. The
new fabrics were gentler on the passengers skin during impact than its predecessors and had
better packability. The fabrics were also lighter and had more controlled permeability.

Manufactures are now evaluating fabrics that are constructed of high to super high tenacity yarn
with linear densities of 235 to 350 dtex. The fabrics have improvements in weight reduction,
packability and softness. In the future, the trend of low density, high strength yarns will continue
that will result in bags becoming lighter, more robust and more compact (Gon, 2010;
Mukhopadhyay, 2008). As the fabric system continues to evolve, the analysis and design
procedures must as well.

2.2 Airbag fabric research


This section details the research related to the evaluation of airbag fabric and modeling of the
mechanical properties of airbag fabric chronologically. While there has been great effort in
modeling the deployment and impact kinematics of airbags, the literature related to experimental

9
Chapter 2 Background

evaluation and modeling of the mechanical properties of airbag fabrics is very limited.
Keshavaraj et al (Keshavaraj, Tock, & Nusholtz, 1995; Keshavaraj, Tock, & Nusholtz, 1996)
studied the biaxial properties of nylon 6, nylon 6,6 and polyester fabrics using a blister-inflation
device. The fabrics were of a balanced construction with the same amount of yarns in the warp
and fill direction although there has been no mention of the value of crimp in the yarns. The
blister-inflation technique used in the study is a quasi-steady-state measurement in which a flat
sheet of fabric is deformed into a semi-spherical blister via pressure drop across the fabric using
compressed air. The biaxial stretching of the fabric and changing permeability as the fabric
structure is inflated is recorded. A pressure gauge measures the internal pressure of the blister
while the blister height is recorded manually. The temperature of the inflating gas was collected
using a temperature sensor while volumetric flow rate is measured with an anemometer. The
fabric thickness is measured before the sample is loaded into the rig. The biaxial stress-strain is
then determined by a relationship previously derived for solid plastic films under blister inflation
and is dependent on the internal pressure, diameter of pressurized sphere, blister height and
fabric thickness. The biaxial stress-strain blister relationship is based on the assumption that a
constant volume of fabric sheet deforms (uniformly) from a flat configuration into a spherical
segment during the experiment.

Permeability behavior of 630 denier and 420 denier fabrics made of nylon 6,6 and nylon 6 were
compared under both low pressure drop and high pressure drop over a variety of temperatures.
The nylon 6 fibers exhibited higher permabilities than the nylon 6,6 fibers. The biaxial stress-
strain behavior was also evaluated for the same specimen matrix. The specimens were
pressurized at set intervals and held so that the blister height could be recorded before increasing
the pressure, hence the quasi-steady-state definition. During this test, the fabric was not
extended to rupture. A ball-burst rupture test was also performed to test the fabrics to failure
under both static and dynamic loading. The authors claim that the ball-burst experiments
provided a more realistic view of the performance of the fabric under biaxial condition. The
crosshead rates tested were 0.5 and 50 inches per minute.

As a continuation of the previous work, Keshavaraj et al (Keshavaraj, Tock, & Haycook, 1996)
developed an airbag fabric material model for nylon and polyester using a simple neural network
architecture. The neural nets formulated were intended to be used as a design tool in

10
Chapter 2 Background

determining permeability and biaxial stress-strain relationships for airbag fabrics. The authors
claimed that the advantage that neural networks have is the computation efficiency in handling
complex and nonlinear problems compared to the nonlinear finite element technology at the time
the paper was written. The inputs of the model included the experimental data obtained from the
blister-inflation experiments performed in a previous study (Keshavaraj et al., 1996) and the
geometric properties of the fabric (i.e. yarn linear density, yarns per inch). The authors reported
predictions that fitted well with the experimental data with extremely fast computation time. The
ability of the artificial neural network to obtain the mechanical behavior at different state of
stresses was not discussed.

Hong (Hong, 2003) used a mix of dynamic experiments with finite element simulation-
optimization techniques to back calculate the elastic properties of 315 denier (60 yarns x 60
yarns) silicone coated airbag fabric under deployment conditions. Hong discussed the anisotropy
seen in uniaxial tests taken in the warp, fill and bias (yarns oriented 45 degrees) directions in
which Hong argued that this can be problematic in deciding the proper values to use for the
mechanical properties in simulations. To find the elastic constants that govern the airbag fabric
response under deployment, Hong carried out an optimization simulation. In the optimization
procedure, the material properties of the airbag fabric are optimized to minimize the difference in
a drop tower test on a deploying airbag and finite element simulation results. The variables that
are minimized between the test and simulation are the acceleration, velocity, displacement and
force. Hong was able to find values for modulus of elasticity, shear modulus and Poissons ratio,
however, the results were influenced by the type of element formulation used. The mechanical
properties obtained were not compared to other mechanical tests.

Rohr et al (Rohr, Harwick, & Nahme, 2004) performed a series of experiments to determine the
strength and failure behavior of fabric under different loading rates and exposed temperatures. A
series of uniaxial tests on nylon 6,6 airbag fabric in the warp and fill directions was performed at
loading rates 0.08 mm/s and 500 mm/s using an Instron Type 8033 mechanical tester at
temperatures between -35 C and 85 C. Additionally, the fabric was tested at a loading rate of
9000 mm/s using a drop weight tower at temperatures of -35 C and 20 C. The fabric geometry
such as yarn linear density, areal density, yarns per inch in both the warp and fill directions was
not reported in the study. The force was normalized by assuming the fabric as a continuum with

11
Chapter 2 Background

constant cross-sectional area. The results of the uniaxial tests at room temperature, quasi-static
conditions found no differences in mechanical behavior between the warp and fill directions
however, the sample size was not mentioned nor any indication of the repeatability of the test.
The investigation of the strain-rate parameter found that as strain rate increased, the ultimate
strain decreased while failure stress increased but no empirical equations are derived based on
their experimental findings. When varying the exposed temperature, as the testing temperature
increased, the trend of the data showed that the failure stress decreased and failure strain
increased for all three strain rates. However, the authors only tested the fabrics at three
temperatures over a relatively small temperature range with respect to the possible range of
temperatures that the airbag structure can undergo during deployment. Additionally, the authors
carried out burst pressure tests on the fabrics. Little detail is given regarding the experimental
set up and procedure of the test, regardless, the failure pressure was reported to be between 4-5
bar.

In a subsequent study, Rohr et al (Rohr, Harwick, & Nahme, 2005) carried out biaxial cruciform
tests to determine the material behavior under 1:1 biaxial loading. The experimental rig used
consisted of four lever arms where one end of a lever is connected to a traditional mechanical
tester cross head while the other ends traveled on a track equipped with piezoelectric sensors to
record the pulling load from the fabric specimen. Different lever lengths can introduce different
ratios of biaxiality, however this study only used a 1:1 loading ratio. Like the preceding study
(Rohr et al., 2005), there is no reference to the fabric geometry or the size of the sample
population. It is also unclear how stresses and strains were normalized but it is assumed the
same procedures of the preceding paper were used. The results found little difference between
the warp and fill directions for the particular fabric system tested. The authors indicated that the
sample failed near the grips and at the corners of the specimen.

A group of German automotive researchers noted the importance of fabric modeling, particularly
in folding and deployment simulations (Drnhoff et al., 2008). The fabric models used in their
studies are orthotropic elasticity definitions which are implemented into the major dynamic finite
element codes (LS-DYNA, PAM-CRASH, MADYMO FE). Biaxial extension and picture frame
tests were performed to generate the stress-strain response of the fabric under extension and pure
shear. For the biaxial test, the fabrics were not loaded to failure, rather loaded to a given point

12
Chapter 2 Background

and the recovery of the fabric was monitored. The procedure outlining the picture frame test for
evaluation of shear properties of the fabric was discussed, although no results are given. The
authors claimed that using the generated stress-strain curves rather than mechanical data from the
fabric supplier, it was possible to get a better representation of the real life airbag kinematics.
While not a main focus in the paper, it discussed the need and challenge for accurate geometrical
description of the airbag structural components such as patches, straps and vents to improve
virtual airbag designs.

Behera and Goyal (Behera & Goyal, 2009) used an artificial neural network system to predict
performance parameter for five different airbag fabric systems coated with silicone and
polyurethane. The inputs of the model consisted mostly of construction parameters such as
fabric areal density, yarns per inch, warp/fill yarn linear density, yarn strength, yarn ultimate
elongation, fabric cover, warp/fill crimp, thickness and yarn flexural rigidity. To validate the
model output experimental testing followed ASTM standards for breaking load, tear strength,
bursting strength, air permeability and specific packability. After the network was calibrated, the
model was able to determine breaking loads, final elongations, permeability, tear strength, burst
strength and packability. The model was able to predict failure strength and elongation with
minimal error while air permeability and tearing strength outputs were found to have high
percentages of error. The authors state that the prediction performance of the model is purely
based on the amount of training or the amount of data available.

Brueggert and Tanov (Brueggert & Tanov, 2002; Tanov & Brueggert, 2003) proposed a user
defined material model for loosely woven airbag fabric in the finite element code, LS-DYNA,
which takes into account the non-orthogonal orientation as the fabric shears. The model was
based on a representative unit cell constructed of four bars to from a trellis that is diagonally
braced by two springs. The linear elastic bars represent the extensional behavior of the fabric
while the diagonal springs model captures the effect of the yarns locking upon excessive shear
deformation. The authors demonstrated that the internal pressure and kinematics of a side-
cylindrical shaped airbag recorded in experimental tests were in good agreement with a
simulation using the generated material model.

13
Chapter 2 Background

2.2.1 Summary

The literature regarding evaluation of airbag fabric and modeling the constitutive behavior is
quite limited. On the experimental evaluation side, airbag fabrics have been tested in uniaxial
tension under a variety of strain rates and temperatures, and under biaxial tension using the
inflation and cruciform techniques. The blister inflation technique assumes the stresses and
strains are equal for the warp and fill direction which could be invalid if the fabric is unbalanced
either with regard to fabric geometry or yarn mechanical behavior. The cruciform technique is
susceptible to stress concentrations but can be used to mimic different states of stress on
balanced and unbalanced fabrics. In the studies discussed, stresses are normalized by cross
sectional area under the assumption that the fabric is a continuum.

While the procedure of picture frame shear tests has been discussed, shear stress-strain results
have not been published for airbag fabrics. Additionally, other modes of shear deformations like
rail shear or shear under biaxially pre-stressed fabrics have not been investigated for airbag
fabrics to date. Also missing from the current literature is an investigation of the mechanical
properties of the airbag yarns as well as a direct comparison between coated and uncoated
fabrics.

The current avant-garde in terms of modeling elastic behavior of airbag fabrics has been
dominated by finite element orthotropic elastic definitions and to a lesser degree artificial neural
networks. The orthotropic elastic material models common in the commercial codes are as
useful as their inputs making them only accurate for the stress state at which the inputs are given.
The artificial neural networks have been more of an academic exercise that requires a wide range
of experiments with large sample sizes to be of use. Additionally, the ability of artificial network
to predict mechanical behavior and failure under multiple states of stress has not been studied.
The current cutting edge methods, a mix of unit cell constitutive relationships with finite element
continuum formulations have been proposed and demonstrated for use in airbags although
further study and validation of the material behavior is needed.

14
Chapter 2 Background

2.3 Modeling of the mechanical behavior of fabrics

2.3.1 Representative mechanistic models

Perhaps the grandfather of the study of mechanics of woven fabrics is Dietzius Haas (Haas,
1918). Haas was interested in studying the deformations of biaxially stressed textile fabrics for
use in airship design. Haas analysis considered the effects of crimp interchange, thread
straightening and thread shear on the forces seen in the principal axes of a fabric system.
Mathematical relationships to describe the equilibrium of forces were derived based on the basic
geometry of the fabrics. The consideration of fabric shear was based on the assumption that the
fabric would shear in a frictionless trellis like manner under a biaxial stress field until the
resultant force at the intersection of the warp and fill yarns aligned with the physical orientation
of the warp and fill yarn. The analysis was practical and significant as it improved the stability
of airship, particularly when attempting to make turns (Hearle, 1969).

A 1937 paper by F.T. Pierce is a classical paper in woven fabric mechanics that is highly cited in
the literature (Peirce, 1937). In the work, Pierce derived the geometry for a representative unit
cell model of woven fabric based on yarns with solid circular cross sections as seen in Figure
2-1. The unit cell model is able to capture the finite internal structural changes that occur within
the fabric under the condition that the force exerted by the warp yarn on the filling yarn equals
the force exerted by the filling on the warp. Thereby under the conditions of equilibrium and
continuity, the stress-strain curve of the fabric at a macro scale can be determined. However,
solving the geometric and kinematic equations has to be performed using numerical methods as
the relations are complex and nonlinear. Later, a study by Freeston developed a theoretical
analysis of the Pierce unit cell under biaxial loading to provide simpler analytical expressions
(Freeston, Platt, & Schoppee, 1967). The circular cross section and the assumptions of perfectly
flexible, incompressible yarns in the Pierce model limits the analysis to a small range of fabric
systems. Future improvements of this technique were made by Kemp (Kemp, 1958) for race-
track shaped yarns and Shanahan and Hearle (Shanahan & Hearle, 1978) for lenticular shaped
yarns that were compressible.

15
Chapter 2 Background

w
df hf

d
H
D/2

pf

Actual Fabric Idealized Unit Cell


Structure Structure
Figure 2-1: Pierce's unit cell (1937)

Grosberg was among the first to present a mechanistic model of shear for plain weave fabrics
(Grosberg & Park, 1966; Grosberg, Leaf, & Park, 1968). Grosberg recognized the factors that
influence shear rigidity, namely the resistance against change of the interlacing angle caused by
friction and elastic restriction. The model was based on a unit cell of intersecting warp and fill
yarns and considered elastic lateral bending of the yarns, slippage of the yarns at the intersection
and subsequent elastic rotation of the yarn crossovers. The frictional contact between the yarns
was assumed to be a line contact however the length of the contact had to be back calculated
from experiments for the slippage portion of the model due to simplifications made in the
geometry. In the elastic portion of the model, the line contact is computed with geometrical
relationships. Additionally, Grosberg recognized that the geometry of the fabric structure and
contact force changes when the fabric specimen is put under tension and provided a correction
within the model to account for the presence of a pre-tension. However, the correction cannot
account for changing tensions as the structural configuration of the unit cell is updated
throughout the analysis. While the model presented in the papers had good agreement with the
experimental results, greater deviations were seen in extremely tight and extremely loose fabrics.
The model has not been able to replicate the shear behavior of different fabric systems observed
by other researchers in the recent literature (Sun & Pan, 2005).

16
Chapter 2 Background

A simplified model of fabric shear and a discussion of the geometrical limits of fabric were
developed by Skelton (Skelton, 1976). In many ways, the simplified model is similar to the
Grosberg analysis in that it considered the frictional force required to rotate the yarns at the
intersection for the same unit cell geometry but neglecting the consideration of lateral bending of
the yarns. The theoretical maximum shear angle with respect to fabric tightness was determined
by taking into account the geometric construction of the fabrics under several assumptions: yarns
are thin interwoven strips, yarns are interwoven cylinders and yarns are in side-by-side contact.
Additionally, Skelton concluded that for fabrics that are tightly woven, the occurrence of shear
would not be contributed to yarn rotation but by yarn distortion, in which case, the fabric shear
behavior is more like a laminar sheet material.

The next significant contribution to a mechanistic approach of woven fabrics was made by
Kawabata (Kawabata, Niwa, & Kawai, 1973a; Kawabata, Niwa, & Kawai, 1973b; Kawabata,
1989) who considered both extensional and shear responses. The fabric is represented by simple
a unit cell constructed of straight bars connected by pins at the yarn crossover as seen in Figure
2-2. The structural parameters required to construct the unit cell are the warp and fill yarn
density and crimp. Similar to the equilibrium condition in the Pierce model, under biaxial
deformation, the contact force between the warp and fill yarn is balanced and the axial force in
the yarn is broken down into its components to compute the force seen at the fabric ends.
Adequate predictions were found comparing the experimental performance on cotton, wool and
polyester fabric systems under biaxial loads. However, the equilibrium condition cannot be
applied to solve uniaxial extension behavior because no tension is applied to the yarn in the non-
loaded transverse direction (Kawabata, Niwa, & Kawai, 1973b). Therefore, no resistance force
is preventing the straightening of the load yarn (geometric stiffening). It was found that in order
to capture the geometric stiffening a resistance force from the transverse yarn acts upon the
loaded yarn direction. By considering the bending rigidity and intra-fiber shear in the resistance
force of the transverse yarn, the straight line unit cell model could capture the uniaxial behavior
seen in experiments with reasonable accuracy.

17
Chapter 2 Background

X2 2y01
X3

X1 X2
H01
2y02
H02
y01

y02

X1
Actual Fabric Idealized Unit Cell
Structure Structure
Figure 2-2: Kawabata's unit cell (1973)

The same unit cell model was also used to describe the shear deformation of fabrics using some
empirical relationships of the yarns in rotation (Kawabata, Niwa, & Kawai, 1973c). The shear
model is capable of capturing the coupled behavior extension has on the shear resistance of the
fabric. In the Kawabata model, as the unit cell structure is deformed by extension and rotation,
the shear force is found by balancing the torque force required to rotate the intersecting yarns,
which is a function of contact force and friction. To calibrate the shear portion of the model,
specialized equipment was developed to determine the torque required to rotate a crimped yarn
under several magnitude of contact forces. This method was able to obtain reasonable results of
fabric under simple shear (rail shear) as well as shear combined with constant extension.

Overall, the Kawabata unit cell approach used to determine the extensional behavior is very
popular in present day research due to its simplicity and accuracy to describe the complex
nonlinear behavior of fabrics. The method has been modified to include coating extension and
inelastic effects (Stubbs & Thomas, 1984); sinusoidal profile of yarns instead of linear elements
(A. Shahkarami & Vaziri, 2006; A. Shahkarami, 2006; A. Shahkarami, 2006); and incorporated
as the constitutive relationship in some finite element analysis (Ivanov & Tabiei, 2004; King,
Jearanaisilawong, & Socrate, 2005; A. Shahkarami & Vaziri, 2007).

18
Chapter 2 Background

2.3.2 Continuum approaches

There have been considerable efforts by textile engineers to assume the fabric system as a
continuous sheet in order to use plate theory in the analysis and design of fabric structures.
Kilby (Kilby, 1963) developed planar stress-strain relationships of a simple trellis in which linear
elements pivoted together at the yarn intersection, although the passing over and under of the
yarns is not considered. The analysis found that the fabric can be treated as an elastic lamina and
yielded a single expression for the general modulus in any direction using the modulus of
elasticity in the warp and fill directions, shear modulus and Poissons ratio and angle between the
direction of extension and the warp yarn as input. Due to the nonlinearity of fabrics, the analysis
is limited to small strains.

Alley and Fasion (Alley & Faison, 1972) attempted to analyze the fabric response using the
generalized form of Hookes Law for continuum analysis of membrane structures, more
specifically parawings. The authors described that nine coefficients are needed to construct the
plane stress anisotropic constitutive relations, four of which are coefficients of interaction of first
and second kinds. The coefficients of interaction are analogous to the shear coupling parameters
seen lamina constitutive relationships. Unlike lamina, the resulting Hookian relationship for
fabric is asymmetric and a function of both axial and shear loads. A series of experimental tests
were performed on polyurethane coated nylon fabric to obtain the nonlinear coefficients for that
particular fabric system.

Shanahan, Lloyd and Hearle (Shanahan & Hearle, 1978) examined the uses of plate and shell
continuum formulations to describe the complex deformations of fabrics under the assumption
that fabrics behave as a sheet. Due to the orthogonal nature of the woven fabric geometry, it is
convenient to make use of structural axes. Therefore, Shanahan and co-workers were able to
construct the linear elastic stiffness matrix that had 13 independent stiffnesses that accounts for
extension, bending and coupling. Again, due to the complexity of fabric mechanical behavior,
the analysis is limited to small strain, linear problems though the authors state the analysis
provides the framework for solving for nonlinear situations. Aside from the computational
limitations of the analysis are the difficulties in evaluating the fabrics experimentally to obtain
the elastic constants. The authors argue that the large displacements and finite areas to obtain

19
Chapter 2 Background

measurable effects in addition to lack of rigidity cause impediments using traditional


measurement techniques for anisotropic engineering materials.

2.3.3 Current state-of-the-art

The current methods in fabric analysis rely heavily on the finite element method. Shockey et. al
(Shockey, Erlich, & Simons, 2000) were one of the earliest researchers to discretely model the
individual yarns as solid continuum orientated in a plain weave structure. A small patch of
fabric was modeled to capture the dynamic response of the fabric subjected to aircraft engine
fragmentations. A similar approach was preformed by Duan et. al. (Duan, Keefe, Bogetti, &
Cheeseman, 2005) for ballistic impact of fabrics under different boundary conditions.
Shahkarami (A. Shahkarami, 2006) proposed a less intricate model, where a unit cell of two
crossover yarns is analyzed to capture the detailed behavior of fabric under biaxial loading.
Lomov and Verpoest have developed sophisticated algorithms to generate finite element models
for a wide range of fabric structures ranging from plain weaves, twill weaves and 3D woven
structures (Lomov, Gusakov, Huysmans, Prodromou, & Verpoest, 2000; Lomov et al., 2001).
While these fully 3D analyses provide a wealth of detail regarding the mechanical and dynamical
behavior of the yarns and fabric, a significant amount of computational power and time are
required. Therefore, only small portions of the structures are modeled.

To remedy the computational requirements but retaining important details of the fabric
deformation, shell elements whose constitutive relationships are based on a mechanistic
representative unit cell have been developed by a number of researchers for a variety of
applications. Brueggert and Tanov (Tanov & Brueggert, 2003) developed a user material model
(UMAT) for shell element in the commercial code LS-DYNA where the constitutive relationship
was based on a square arrangement of linear springs capturing the extensional response of the
fabric while a set of diagonal springs represent the shear behavior of the fabric. Ivanov and
Tabiei (Ivanov & Tabiei, 2004) developed a micromechanical model for Kevlar fabric based on
the Kawabata unit cell geometry with a viscoelastic stress-strain yarn model to capture strain rate
dependency. The model was implemented as a UMAT for the LS-DYNA to simulate ballistic
impact of Kevlar fabric. King (King et al., 2005) developed a material model for the commercial
code ABAQUS/Standard to predict the behavior of Kevlar fabric under quasi-static uniaxial

20
Chapter 2 Background

extension, bias extension and picture frame shear. Shahkarami (A. Shahkarami & Vaziri, 2006;
A. Shahkarami, 2006) developed a UMAT for LS-DYNA intended for shell formulations based
on a unit cell with a sinusoidal yarn profile for ballistic impact simulations of Kevlar fabric
panels. Overall, this approach of using a respective micromechanical model as the constitutive
relationship of a shell finite formulation increases the computational efficiency while still
accurately capturing the internal structural changes that affect the macroscopic mechanical
behavior of the fabric.

2.3.4 Summary

The study of the mechanics of woven fabrics is nearly a century old and has been applied to the
study of everything from airships to bullet proof armor. The unit cell method has proven to be
effective in generating the extensional stress-strain behavior of fabrics while predicting shear
behavior has been less successful. The continuum approaches are easy to implement for the
analysis of structures but require a great deal of testing to capture the anisotropic and nonlinear
nature of the fabric. Fully 3D finite element models provide the most detail about the fabrics
mechanical behavior, but are too computationally demanding to model large structures. The
current state-of-the-art techniques discussed offer a favorable compromise of mechanistic unit
cell methods with finite element analysis that provides accuracy and detail without extensive
calibration of elastic constants.

21
Chapter 3- Development of a representative unit cell: theory and calibration

Chapter 3 - Development of a representative unit cell:


theory and calibration

3.1 Background and approach

The first step in developing a modeling approach is to establish a geometry that best represents
the fabric structure that is simple yet does not sacrifice accuracy or neglect reality. A unit cell
based on the center-line positioning of the yarn which is approximated by straight lines originally
proposed by Kawabata (Kawabata, Niwa, & Kawai, 1973a; Kawabata, Niwa, & Kawai, 1973b;
Kawabata, 1989) was chosen for this study. This simple center-line approach has been adopted
by several researchers over the years for a variety of applications as shown in Table 3-1. While
the center-line unit cell geometry has been repeated throughout the literature, the structural
mechanisms that empower the unit cell vary greatly depending on the fabric system and
application. For instance, some textile composite preforms have lower interlacing density and
lower yarn crimp which permits more fiber mobility in the yarn that in turn allows the yarn to
exhibit a more compressible behavior. (Lomov & Verpoest, 2000) However, tighter woven
fabric structures like those used in airbags tend to inhibit fiber mobility which results in an
incompressible yarn (Lomov, 2000). Likewise, fabric systems of fiber-glass and Kevlar can be
approximated to have linear elastic behavior while other fabrics are composed of highly
nonlinear materials such as nylon or polyester. Therefore, when implementing this type of unit
cell approach, it is important for the user to consider the application and material system in order
to incorporate the proper structural mechanisms to reflect the physical nuisances of the fabric.
Thus as novel applications and material systems are considered for this particular unit cell
theory, new mechanisms will be developed that will add to the library of knowledge that the
fabric designer or analyst can use.

22
Chapter 3- Development of a representative unit cell: theory and calibration

A goal in the proposed model is to implement mechanisms whose inputs are simple constitutive
properties of the fabric, more specifically fabric geometry and yarn stiffness. In that way, the
model can be used to analyze the in-plane behavior of a fabric without excessive or specialized
testing. Additionally, it can be utilized as a design tool to predict certain mechanical properties
of a virtual fabric.

The way the proposed unit cell model adds to the current body of knowledge is by including:
nonlinear yarn extension behavior for nylon, yarn bending rigidity, and coating extension. All of
which are based on structural mechanics using simple geometry and material properties
(Youngs Modulus, Poissons ratio). In Table 3-1, the proposed models features are compared
to the current models in the literature.

Table 3-1: Existing center-line unit cell models vs. proposed model

Researcher Fabric Application UC Shape Yarn Behavior Shear Coating Experimental


System Included Included Validation Methods
Kawabata Cotton; Apparel Linear Nonlinear Extension, Yes(EC) n/a Biaxial, uniaxial, rail
(1973) polyester Bending Rigidity, shear
Incompressible &
Compressible(EC)
Stubbs Teflon- Architectural Linear Linear Ext, Perfectly No Yes Biaxial
(1980) coated Fabrics Flexible,
fiberglass Compressible(EC)
Kato Teflon- Architectural Linear Linear Ext, Perfectly Yes(EC) Yes(EC) Biaxial, Picture Frame
(1999) coated Fabrics Flexible, Shear
fiberglass Compressible(EC)
Boisse Fiberglass Composite Linear Linear Ext, Perfectly No n/a Biaxial, Uniaxial
(2001) Preform Flexible,
Forming Compressible(EC)
Tabiei & Kevlar Ballistic Linear Linear Ext w/ Yes(EC) n/a None
Inanov Protection viscoelasticity,
(2004) Perfectly Flexible
King Kevlar Ballistic Linear Linear Ext, Bending Yes(EC) n/a Uniaxial, Bias
(2004) Protection Rigidity(EC),
Compressible(EC)
Shahkarami Kevlar Ballistic Sinusoidal Linear Ext, Perfectly Yes(NC) n/a Biaxial* , Ballistic
(2005) Protection Flexible, Impact
Compressible(EC)
Bridgens Teflon- Architectural Linear & Linear Ext, Perfectly No Yes Biaxial
(2008) coated Fabrics Sinusoidal Flexible,
fiberglass; Incompressible
PVC-
polyester
PROPOSED Nylon 6,6; Automotive Linear Nonlinear Extension, Yes(EC) Yes Biaxial, Uniaxial, Bias
MODEL Silicone- Airbag Bending Rigidity, Shear
coated Incompressible
Nylon 6,6
* Basic UC material behavior verified using data from Boisse
EC= Experimentally Calibrated
NC = Numerically Calibrated

23
Chapter 3- Development of a representative unit cell: theory and calibration

3.2 Unit cell definition

The simple unit cell geometry used in this study is shown in Figure 3-1. The coordinate system
origin is defined as the intersection of the warp and fill yarns such that the coordinate axis X1 is
along the neutral line of the warp direction, axis X2 in the fill direction and axis X3 is the through-
thickness direction of the fabric. The assumptions established in this studys formulation of the
unit cell are the following:
The yarns are elastic
Displacements are equal for each yarn end
No slippage occurs at the yarn crossovers
Environmental effects such as temperature or moisture changes are not
considered; i.e. the behavior of the fabric is the same as it is at room temperature
Strain rate effects are neglected for the time being due to lack of experimental
procedure and data
2y01
x3 Initial
State
x1 L01
x2
01 H01 02
2y02
tc
H02
y01
L02
y02

x3

FEnd,1 x2
Fc,1 FEnd,2
Fb,1

Fc,2
Fb,2 x1
FEnd,2 FEnd,1

Deformed
State
Figure 3-1: Simple unit cell geometry in initial and deformed states

24
Chapter 3- Development of a representative unit cell: theory and calibration

The construction of the straight line unit cell geometry requires two geometric inputs. The first
being the yarns per unit length in the warp and fill directions (in textiles, the common unit is
yarns/inch). From this parameter, the horizontal half spacing of the unit cell, y, can be calculated
as:

1 (3-1)
y01 =
2(n2 )
1 (3-2)
y02 =
2(n1 )
where n is the number of yarns per unit length. From here forward, the 1 and 2 indices will
indicate the warp and fill yarn, respectively. The 0 index represents the original undeformed
structure.

The second geometry input parameter is the percent crimp of the warp and fill yarns. The crimp
of the yarn is the quantification of the amount of undulation caused by weaving and can be found
by measuring a horizontal distance, Xi, between two points that are parallel to the yarn and the
actual length, l0i, of the yarn between the same two points. The crimp can be quantified as:

l0 i X i (3-3)
cri =
Xi
where i = 1,2.

Therefore the length of the yarn can be determined using the expression:
L0 i = y0 i (1 + cri ) (3-4)

The final geometric parameter to assemble the unit cell structure is the yarn height which can be
found simply by:
2 2 (3-5)
H 0 i = L0 i y0 i

To this point, the yarn geometry has been defined in the unit cell, now the coating geometric
parameters needs to be defined. The length and width of the coating are equal to the fabric unit
25
Chapter 3- Development of a representative unit cell: theory and calibration

cell spacing, y01 and y02, as shown in Figure 3-1. The thickness of the coating, tc, can be found
using the areal density fraction between the uncoated and coated fabric and the total height
(thickness) of the unit cell.

w (3-6)
t c = (2 H 01 + 2 H 02 ) c 1
wu
where wu is the areal density of the uncoated fabric and wc is the areal density of the coated
fabric. The equation (3-6) is only applicable to coated fabrics that have the same geometric
construction as the uncoated fabric. Additionally it is assumed that the coating maintains a
constant thickness throughout the unit cell.

When a displacement is applied to the unit cell, the coating and fabric are assumed to stretch by
the same amount. The tensile force developed at the end of the unit cell, denoted as Fend,i in
Figure 3-1, is the sum of forces generated by the fabric and coating and is expressed as:

Fend ,i = F f ,i + Fct ,i (3-7)

where Ff,i is the fabric end force and Fct,i is the end force from coating extension.

Finally, the unit cell end forces can be translated into the membrane stress (force per unit width
of fabric) by:

N i = Fend ,i ni (3-8)

The membrane stresses developed depend on the reconfiguration of the unit cell structure which
is governed by constitutive relations and mechanisms of the yarn and coating. To determine the
new structural configuration of the perturbed fabric structure requires solving a set of highly
nonlinear equations for which a numerical procedure is needed. A computer code was developed
to solve these equations for a particular strain or can be iterated to generate stress-strain curves
of the fabric system for uniaxial and multiple ratios of biaxial extension. Figure 3-2 shows a
flowchart of this code.

26
Chapter 3- Development of a representative unit cell: theory and calibration

Input n, c, yarn f-d,


EI, ult Specify Stress State

Biaxial Uniaxial

Specify Biaxial Ratio, Specify Warp or Fill


Strain increment Displacement,
Strain increment

1 =1i-1 + 1
N-R Procedure 1 =1i-1 +
2 =2i-1 + 2

Check > ult


End
< ult

< ult
Save
Ni

Figure 3-2: Procedure for generating stress-strain behavior of fabric under multiple states
of stress using the unit cell approach

Depending on the stress state, the solution procedure varies slightly due to conditional geometric
behavior of the unit cell. To be clear, under biaxial load the change of yarn spacing in both the
warp and fill direction can be determined from the strains while the change in the yarn length is a
function of the yarn height. Under uniaxial load the length of the yarn transverse to the applied
load is equal to its original length while the yarn spacing is a function of the yarn height. Under
these conditions, the yarn is not allowed to compress axially to support negative strains hence
compressive resistance is provided by the bending rigidity of the yarn. Table 3-2 summarizes
these conditional geometric properties of the fabric under different modes of extension. Finally,
a yarn failure criterion was established in which the code runs until either the warp and fill yarn
surpasses the specified ultimate yarn strain.

27
Chapter 3- Development of a representative unit cell: theory and calibration

Table 3-2: Unit cell configuration conditions

Variable Biaxial Uniaxial Warp Uniaxial - Fill


y1 = y01 + d1 = y01 + d1 f(H1, H2)
y2 = y02 + d2 f(H1, H2) = y02 + d2
L1 f(H1, H2) f(H1, H2) = L01
L2 f(H1, H2) = L02 f(H1, H2)
H1 Unknown Unknown Unknown
H2 Unknown Unknown Unknown

The center-line unit cell method has two main equations that need to be evaluated in order to
solve for the end forces that arise from the perturbed fabric structure: continuity and
equilibrium. The continuity equation for an incompressible yarn where the cross-section of the
yarns remain unchanged during stretching, regardless of the profile shape is given as:

H 01 + H 02 = H 1 + H 2 (3-9)

where H01 and H02 are the original height of the warp and fill yarn, respectively. H1 and H2 are
the height of the warp and fill yarns in the perturbed structure.

The equilibrium condition considers the sum of the vertical force components that arise between
the warp and fill yarns at the cross-over. The forces that are considered in this study are axial
extension whose vertical component exerts a contact force at the cross-over and the forces that
arise from bending the yarn. The equilibrium condition as shown in Figure 3-1 is expressed as:

Fc1 (H 1 ) Fb1 (H 1 ) = Fc 2 (H 2 ) Fb 2 (H 2 ) (3-10)

where the magnitude of the vertical cross-over forces are a function of the yarn height, Fc is the
contact force that arise from yarn extensions and Fb is the reaction force that arises from
bending.

28
Chapter 3- Development of a representative unit cell: theory and calibration

3.3 Deformational mechanisms and constitutive relationship

Before solving equations (3-9) and (3-10), the deformational mechanisms of the fabrics
constituents need to be derived and discussed in detail. Some of these mechanisms dictate the
equilibrium between the warp and fill yarns while others simply contribute to the end force
through super-position. Furthermore, the shear stress-strain behavior of the fabric needs to be
established.

3.3.1 Yarn axial extension

The yarn axial force is determined simply by the relation

Fy ,i = E ( ) Ai
(L i L0 i ) (3-11)
L0 i
where E ( ) i is the axial stiffness of the yarn, the magnitude of which is a function of strain. The

nonlinear relationship of E ( ) is determined experimentally and will be discussed in greater


detail in Section 3.5. The variable Ai is the specific area of the yarn which is determined through
the following relationship

Ld , i (3-12)
Ai =
9000
where Ld,i is the linear density of the yarn (denier) and is the density of the fiber material
(g/cm3) resulting in Ai being measured in (mm2).

The total contact force or reaction force at the yarn crossover that arises due to the extension can
be shown to be:

Hi (3-13)
Fc ,i = 2 Fy ,i
Li

29
Chapter 3- Development of a representative unit cell: theory and calibration

3.3.2 Yarn bending

While yarns are known for their high flexibility, the small bending rigidity is important to
consider capturing low level stress behavior, particularly under the uniaxial case where the only
force developed by the unloaded yarn to maintain equilibrium at the crossover is due to the
bending rigidity of the yarn. For determining the bending reaction force due to the changing
angle of the yarn, first consider the case of bending a straight linear elastica into a crimped form
as shown in Figure 3-3. With respect to the unit cell, the elastica shown in the figure is
composed of two unit cell halves from two crossovers. Hence, the bending force, Fb,i is divided
by a factor of two. The bending property of the yarn can be characterized by the linear equation:

(EI ) yarn (3-14)


M=
rc
where (EI)yarn is the bending rigidity of the yarn and rc is the radius of curvature.

dS


Fb,i/2

rc d

i

x+ dx

yi yi

Fb,i/2

Figure 3-3: Bending geometry of an elastica

30
Chapter 3- Development of a representative unit cell: theory and calibration

There has been a great deal of research performed concerning the mechanics of the bending
rigidity of yarn particularly on twisted yarns or cordage materials (Freeston & Schoppee, 1975;
Hearle, 1969; Platt, Klein, & Hamburger, 1959). Fortunately, the airbag fabric is constructed of
low-twist continuous filaments with a circular cross-section, therefore a simple treatment is
defensible. The theoretical bending rigidity of low-twist, non-blended yarns is bounded by two
values (Platt et al., 1959). The lower bound assumes each individual fiber has complete freedom
of motion in a frictionless manner and is defined as:

d 4fiber (3-15)
(EI ) LB
yarn = N fiber E fiber
64
where Nfiber is the number of fibers in the yarn, Efiber is the modulus of elasticity of the fiber and
dfiber is the diameter of the fiber.

The upper bound assumes the fibers have no freedom to move and act like a complete cluster
bonded by friction. Therefore the yarn bending rigidity can be computed using the number of
fibers in the yarn divided by the yarn packing factor times the rigidity of the lower bound. This
yields the expression:

(EI ) UB
yarn = E fiber I yarn (3-16)

where Efiber is the modulus of elasticity of the fiber and Iyarn is the second moment of inertia using
the cross-sectional geometry of the yarn.

To determine the force equilibrium between the reaction force and bending moment, from
Figure 3-3, it can be shown that

ds (3-17)
rc =
d
and
dx (3-18)
ds =
cos

31
Chapter 3- Development of a representative unit cell: theory and calibration

Substituting (3-17) into (3-14) and representing the moment by the reaction force and the
bending moment gives:
d F
M = (EI ) yarn = b x (3-19)
ds 2
Substituting (3-18) into (3-19) and integrating

Fb
(EI ) yarn cosd =
2
xdx (3-20)

Applying the boundary conditions seen in Figure 3-3 where at x = 0, = yields:

2(EI ) yarn
Fb = (sin sin ) (3-21)
x2
Therefore at the yarn intersection where x = y and = 0, and adding the second half of the
bending force, the total bending force at the crossover can be expressed as:

8(EI ) yarn i H i
Fb ,i = 2
(3-22)
yi Li
This expression is identical to the one proposed by Grosberg (Grosberg, 1966; Hearle, 1969) in
his analysis of the bending of fabrics. However, we are interested in the vertical force that arises
when the yarn is bent into a new configuration from its original woven structure. If we reference
the yarn in its woven state as datum (original configuration), the vertical force from bending out
of this state can be expressed as:

8(EI ) yarn i H 0i H i
Fb ,i = 2
(3-23)
yi L
0i Li

Once the vertical force components are known and balanced, the end horizontal force from the
yarns which are of interest can be found simply by:

yi
F f ,i = Fy ,i (3-24)
Li

32
Chapter 3- Development of a representative unit cell: theory and calibration

3.3.3 Coating extension

In the case of the coated fabric, the end force from the coating must be taken into account.
Assuming the coating is therein isotropic elastic continuum, the stresses under biaxial load using
Hookes law can be expressed as:
Ec
1 = (1 c 2 ) (3-25)
1 c2

Ec
2 = ( 2 c1 ) (3-26)
1 c2
where Ec is the modulus of elasticity for the coating and c is the Poissons ratio of the coating.

Resolving these stresses into an end force associated with the fabric unit cell geometry, the force-
strain relationship for the coating is:

Ec y2tc y1 y01 y2 y02


Fct ,1 = (3-27)
1 c2 y01 y02
c

Ec y1tc y2 y02 y y
Fct , 2 =
2
c 1 01 (3-28)
1 c y02 y01

The fabric and coating end force are superimposed to obtain the total end force per unit cell
Fend ,i = F f ,i + Fct ,i (3-29)

Finally, the unit cell end forces can be translated into the membrane stress (force per unit width
of fabric) by:
N i = Fend ,i ni (3-30)

33
Chapter 3- Development of a representative unit cell: theory and calibration

3.3.4 Unit cell in-plane shear behavior

Up to this point the constitutive behavior discussed pertains to the extensional behavior of the
fabric. The fabric can undergo shear deformations characterized by rotation of the yarns at the
crossover. At first glance the shear problem seems trivial -- the yarns bend laterally until
overcoming static friction at the crossover then rotating that mimicks the rotation of a trellis
structure. Several researchers have proposed simple mechanisms based on elasticity to describe
the shear forces that cause this behavior with varying degrees of success depending on the fabric
system and geometry. However, the culprits that contribute to the shear behavior are more
complicated than ones first instinctive analysis of the problem. Intra-fiber friction, yarn torque,
yarn sliding, intra-yarn shear of the fibers and the rate dependency of friction are issues that are
difficult to capture using simple unit cells.

The problem of shear becomes progressively more complicated when considering coated fabrics.
To the authors best knowledge, Farboodmanesh has probably conducted the most extensive
studies on the shear behavior of coated fabric (Farboodmanesh, 2003; Farboodmanesh et al.,
2005). The shear behavior of a typical single-coated fabric versus uncoated is illustrated in
Figure 3-4. Essentially, at low level shear stress the shear behavior is governed by a coating-
fabric interaction marked by a high shear modulus. As the shear strains grow, the shear stiffness
of the system decreases and transitions to a behavior that resembles the uncoated fabric.
Farboodmanesh (Farboodmanesh, 2003) attempted to model this behavior borrowing from
micromechanical models used in traditional fiber reinforced composites but had limited success.
Developing a robust analytical-constituent-based shear model for coated or uncoated fabric is an
enormous and difficult task which is the reason why many of the unit cell models that include
shear require experimental calibration.

34
Chapter 3- Development of a representative unit cell: theory and calibration

Single-Coated

Shear Stress
Fabric

Uncoated Fabric

Shear Strain
Figure 3-4: Typical shear stress-strain curve for single-coated and uncoated fabric

Due to the difficulties of evaluating fabric shear using the fabrics constituents, empirical data
from the fabric is warranted to empower the shear behavior of the proposed unit cell method.
This empirical data can be obtained either through picture frame test where the fabric undergoes
pure shear; rail shear test using the Kawabata Evaluation System (KES) (Kawabata, 1989) for
fabrics; or could potentially be numerically calibrated through a full 3D finite element model of
the fabric and its constituents. Each method has advantages and disadvantages. Picture frame
requires special fixture and the design of the fixture must be carefully done to ensure
reproducible results. Kawabata Evaluation System requires specialized equipment and some
post-processing to neglect the effects of tension on the specimen. The use of 3D finite element
models can be used to evaluate the pure shear behavior of virtually designed fabrics but
capturing the correct boundary conditions can be challenging.

Regardless of the approach used to obtain the shear data, the following method can be used to
represent the shear behavior of the unit cell. A similar approach was used by Shahkarami (A.
Shahkarami & Vaziri, 2006; A. Shahkarami, 2006) to describe the shear behavior of uncoated
fabrics and is modified for the current model to include coated fabrics. A spline fit of the shear
modulus as a function of shear strain illustrated in Figure 3-5a can be used to describe the shear
stress-strain behavior based on typical behavior seen in Figure 3-5b.

35
Chapter 3- Development of a representative unit cell: theory and calibration

G3

Shear Modulus
G1coated

G2
G1uncoated

1 2 3 4

Shear Strain
(a)

Single-Coated Fabric
Uncoated Fabric

Coated Transition Zone


Shear Stress

Locking Transition
Zone

12 3 4

Shear Strain
(b)
Figure 3-5: Shear model behavior a) secant shear modulus as a function of strain b)
regions in the shear stress-strain curve

For coated fabrics, the fabric-coating shear interaction modulus, G1 needs to be defined as well
as the extent of the transition zone which is bounded by 1 and 2. For uncoated fabrics, these
values can be set equal to zero. The pre-locking shear stiffness of the fabric, G2 is the same for
both uncoated and coated fabrics of the same construction. As the fabric continues to deform in
shear, the amount of rotation at the crossovers reach a geometric limit and begin to lock. This is

36
Chapter 3- Development of a representative unit cell: theory and calibration

a gradual process controlled by friction, the current packing state of the fibers within the yarn
and geometric features of the fabric. Within the spline model, this process is bounded by 3 and
4 which are the same for both the coated and uncoated fabric. Likewise the locking shear
modulus, G3, is the same for uncoated and coated fabrics. In the end, the spline fit can be
summarized as shown in equation (3-31).

G1 0 < 1
1 G G1
G1 + 2 ( 1 )
2
1 < 2
2 2 1
1
G + (G2 G1 )( 2 1 ) 2 < 3
S12 = 2
2 (3-31)
1 G G2 1
G2 + 3
2 4 3
( )
( 3 )2 + (G2 G1 )( 2 1 ) 3 < 4
2
1 1
G3 + (G3 G2 )( 4 3 ) + (G2 G1 )( 2 1 ) 4
2 2

It is important to note that this representation of the shear is purely elastic. In reality, the
overcoming of friction is a plastic behavior in the fabric which is exhibited by hysteresis in the
loading-unloading behavior of the fabric. Since the airbag is a one-time use structure subject to
high strain rates, the effect of friction and plasticity are assumed to be negligible at this time.

3.4 Numerical procedure


With the continuity equation and the deformational mechanisms that influence equilibrium
defined, a technique to solve the nonlinear equations to determine the fabric stress at an
associated strain can be discussed. While there are many methods in the literature to solve multi-
variable, nonlinear equations such as Brents method, the Newton-Raphson method was used for
its efficiency and speed at arriving at the roots of a system of nonlinear equations (Press, 1992).
First the functional equations are defined as

f1 (H 1 , H 2 ) = (H 1 + H 2 ) (H 01 + H 02 ) (3-32)

f 2 (H 1 , H 2 ) = (Fc1 Fb1 ) (Fc 2 Fb 2 ) (3-33)

37
Chapter 3- Development of a representative unit cell: theory and calibration

The essential part of the multivariable Newton-Raphson method is assembling the Jacobean
matrix for which partial derivatives of the functional equations are taken with respect to the yarn
height as shown in equation (3-34).

f1 f1
H 2
[J ] = Hf 1 f 2
(3-34)
2
H 1 H 2
where
f1 f1
= =1 (3-35)
H1 H 2

and
f 2 Fc1 Fb1
= (3-36)
H 1 H 1 H 1

and
f 2 F F
= c2 + b2 (3-37)
H 2 H 2 H 2

The derivative of the contact force with respect to the yarn height is given as:

Fci 2 E ( ) Ai L0 i H i2 L0 i
= 1 + (3-38)
H i L0i Li (H i2 + yi2 ) 2
3

The derivative of the yarn bending force with respect to the yarn height is given as:

Fbi 8(EI )i H 0 i H i2

= (3-39)
yi L0 i (H i + yi )

2 3
H i 2 2 2

With the elements of the Jacobean matrix known, the iterative procedure can be performed as
shown below.

38
Chapter 3- Development of a representative unit cell: theory and calibration

H 1new H 1guess f1
new = guess [J ] f
H 2 H 2 2 (3-40)

if k=0  H i
guess
= H 0 i else H iguess = H ik 1
The amount of iterations can be reduced between strain increments if the values of Hi from the
previous step are used as the guess for current step.

3.5 Characterization of constituent properties

To develop the stress-strain behavior of the fabric using the unit cell method described thus far, a
series of experiments were performed to characterize the constituent properties. The fabrics
investigated for this study are constructed of high tenacity nylon 6,6 multifilament yarns with a
linear density of 350dtex which was provided by TRW Automotive. One system has a single
side silicone coating which does not penetrate through the full thickness of the fabric while the
other system remains uncoated. Both fabric systems have identical geometrical features (i.e.
yarns/in, % crimp in warp and weft directions). The areal density of the coated fabric is 218.93
g/m2 and 171.56 g/m2 for the uncoated fabric as determined by ASTM D3776-95.

3.5.1 Microscopy
Due to the dense weave and small yarn width, it is difficult to measure features of the fabric
visually; therefore the use of optical microscopy was warranted. To obtain the number of yarns
per inch in the warp and fill directions, micrographs were taken in plan view with respect to the
fabric as shown in Figure 3-6a for a coated sample and Figure 3-6b for an uncoated sample.
Images were taken randomly throughout the fabric samples to collect a global set of data.

39
Chapter 3- Development of a representative unit cell: theory and calibration

(a)

(b)
Figure 3-6: Plain weave of a) single-coated fabric and b) uncoated fabric

The other geometric parameter of interest is the amount of crimp in each yarn. Using
microscopy, the fabric was set in epoxy so that the edge of the fabric was perpendicular to the
mold face. Upon hardening, the sample is sanded and polished until the fabric edge is exposed.
Images are taken of the sample and then using the software package ImageJ (Rasband, 1997-
2009) the yarn length and spacing are measured to calculate crimp according to equation (3-3).

Figure 3-7 shows images of the cross section of the fabric in the warp and fill directions for
coated and uncoated specimens, respectively. Additional parameters obtained from the sectioned
samples were the filament diameter and count for both warp and fill directions.

40
Chapter 3- Development of a representative unit cell: theory and calibration

(a)

(b)

Figure 3-7: Cross section of 350dtex coated airbag fabric a) warp direction and b) fill
direction

(a)

(b)
Figure 3-8: Cross section of 350dtex uncoated airbag fabric a) warp direction and b) fill
direction

41
Chapter 3- Development of a representative unit cell: theory and calibration

Table 3-3 summarizes the geometric parameters for both the coated and uncoated systems as
measured from the micrographs. From these parameters, the geometry of the unit cell and some
constitutive properties can be determined.

Table 3-3: Geometric properties of coated and uncoated airbag fabric as determined from
microscopy

Coated Uncoated
Warp Fill Warp Fill
Yarns/in 60.43 55.27 60.87 55.09
% Crimp 6.30% 8.90% 6.28% 8.79%
#Filaments/yarn 140 140 140 140
Filament Diameter 16.95 16.95 16.95 16.95
(m)
Coating Thickness 0.082 0.082 N/A N/A
(mm)

Overall, the Table 3-3 confirms that the coated and uncoated fabric samples used in the study
have identical geometric construction. From the figures, the plain weave geometry can be
observed and it can also be seen that the yarns have virtually no twist and are composed of many
continuous filaments. The silicone coating is very thin and does not penetrate through the entire
thickness of the fabric.

3.5.2 Yarn extension test

Besides geometric structure, the mechanical properties of the yarn are essential inputs to a
potential model of the fabric. To evaluate the mechanical properties of the yarns, namely failure
strain and force-elongation behavior, extension tests were performed. The magnitude of crimp
can also be quantified using this method and compared to the values obtained through
microscopy by recording the percent elongation at which the yarn becomes completely straight.

Warp and fill yarns were carefully extracted from the uncoated fabric sample by gently pulling
away neighboring yarns in the weave structure. Yarns were sampled from different locations of
42
Chapter 3- Development of a representative unit cell: theory and calibration

the fabric roll to generate a global population and to avoid any possible localized effects. The
sample population was 10 yarns in each of the warp and fill direction. A length of 150 mm (3 in)
was marked off using a felt tip marker on the fabric and then the yarns were withdrawn as shown
in Figure 3-9. This procedure of marking the length on the fabric as opposed to measuring on an
extracted yarn ensures that the gauge length is based off of the yarn in its crimped condition as
seen in the fabric structure. A condition that is important to keep true if one wants to back
calculate the percent crimp from the extension test.

Figure 3-9: Yarn specimen preparation procedure

It should be noted that yarns were not taken from the coated fabric due to the degree of difficulty
of peeling away the coating layer to get the yarns without causing significant damage to the
yarns. It was also found that extracting lengths greater than 25 mm was near impossible due to
the hindrance of the coating. Extension tests were attempted on a few samples that were suitable
for testing, however the reproducibility of the tests was not acceptable so they are not included in
the study. The other potential test considered was testing a coated yarn, but since the coating
does not penetrate the thickness of the fabric combined with the plain weave structure, the yarn
is not continuously coated for lengths required for testing.

The yarns were mounted onto paper frames as seen in Figure 3-9c using epoxy. The paper frame
ensures the proper gauge length distance between grips preventing additional slack or causing a

43
Chapter 3- Development of a representative unit cell: theory and calibration

pre-stress in the yarn. Additional slack or smaller gauge length than the yarn extracted can cause
erroneous results in the magnitude of crimp if one wanted to subtract the uncrimping stiffening
from the load-elongation curve, a procedure that will be discussed later on. A larger gauge length
can cause the yarn to stretch, causing a pre-stress in the yarn as well as providing incorrect
results of the magnitude of crimp. A Kato Tech KES-G1 microtensile tester (Figure 3-10) with
a 5kg load cell located at the Advanced Fibrous Materials Laboratory at The University of
British Columbia was used to obtain force-displacement information of the yarn up to failure.
Upon loading the sample into the grips, the paper frame is cut before extension is applied. The
elongation rate was 2mm/min at ambient conditions according to ASTM D 3883-04 (ASTM
D3883-04, 2008).

Controller

Load Cell

Sample

Data Acquisition

Figure 3-10: KES-G1 microtensile tester with loaded yarn sample

As the yarn is extended, the crimp in the yarn begins to straighten before the yarn undergoes
stretching, a process commonly referred to as uncrimping. Obviously, this uncrimping of the
yarn is a form of geometric stiffening that needs to be removed in order to obtain the pure
mechanical response of the yarn. Figure 3-11 illustrates the technique of determining the
44
Chapter 3- Development of a representative unit cell: theory and calibration

geometric stiffening region of the force-elongation curve as specified by Option C of ASTM D


3883-04 (ASTM D3883-04, 2008). The straight line portion of force elongation curve is
extrapolated by line AB. The point A represents the magnitude of the crimp in the yarn and the
elongation where the crimp is fully removed from the yarn. The Point C is obtained by
constructing a line parallel to the Force axis from Point A. Point C corresponds to the tensile
force required to remove crimp without stretching the yarn. The curve is graphically shifted so
that Point C becomes the origin, therefore obtaining the pure force-elongation behavior of the
yarn as seen in Figure 3-12.

Figure 3-11: Procedure of obtaining "pure" yarn load-elongation response

45
Chapter 3- Development of a representative unit cell: theory and calibration

25.00

Warp
20.00
Fill

15.00
Force (N)

10.00

5.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25
strain

Figure 3-12: Average force-elongation curve for 350dtex nylon 6,6 airbag yarn

Figure 3-12 shows the average force-elongation curves. Overall, the tests were reproducible
which suggests the yarn responses are uniform throughout the fabric and there are no regional
effects. All yarns failed within the middle of the specimen, away from the grips. The average
failure strains were 21.35% and 20.81% for the warp and fill yarn, respectively. The average
failure load was 21.12 N and 20.18 N. The average crimp for the warp and fill yarns obtained
from the procedure specified by ASTM D 3883 was found to be 6.33% and 8.74%, respectively.

It can be seen from Figure 3-12 that the yarn exhibits a hyperelastic-like behavior which
includes a high initial modulus, a transition period where the stiffness decreases and then as
undergoes higher stiffening at higher strains until failure. This behavior arises from the
semicrystalline structure of nylon polymer fibers. As a tensile load is applied to the yarns,
crystalline block segments separate from the lamellae in the polymer and both block and tie
chains become oriented in the direction of the tensile axis which results in stiffening of the
polymer.

46
Chapter 3- Development of a representative unit cell: theory and calibration

One may argue that it may be more beneficial to test non-crimp yarns (i.e. yarns that have not
been processed into fabric) to reduce the geometric stiffening effect from decrimping. While this
would reduce the amount of work associated with sample preparation and post-processing of
data, this approach would not capture any additional drawing that the polymeric yarns may go
under during the warping process. From Figure 3-12, it can be seen that the warp yarn is stiffer
and fails at a higher load. Considering the warp and fill yarns are identical in terms of linear
density and filament count, additional drawing of the warp yarn from manufacturing is a
plausible explanation for the difference in mechanical behavior of the two yarn directions. In the
end, the obtained response can be considered ex situ as we have obtained the response of the yarn
from a processed fabric. Of course there may be some behaviors, namely failure stress and
strain, which are affected by the in situ environment of the fabric. Pan (Pan, 1996) and
Shahpurwala (Shahpurwala & Schwartz, 1989) have discussed some of the in situ variables that
can impact yarn such as pressure-independent adhesion, a frictional component dependent on the
confinement pressure of the yarn and statistical distribution of fiber strength.

The values of crimp for the warp and fill yarns found using the extension tests are close to the
values obtained using the sectioning method. The extension test advantages in testing crimp that
it can be used to test a large sample population of yarns with relative ease regarding sample
preparation, testing and processing. However, the method is sensitive to keeping the proper
gauge length. The sectioning method can be time consuming for large sample populations
considering the effort of cutting samples, casting them in epoxy, polishing and imaging.
However, the images obtained can be used to measure crimp with a high level of assurance
assuming the fabric sample does not undergo dimensional changes from handling or the epoxy
environment.

3.5.3 Coating characterization

The silicone coating used on the fabric was not available for mechanical testing, although airbag
grade silicone mechanical properties are readily available in the literature (Crouch, 1994;
Schwark & Muller, 1996). The modulus of elasticity of the coating system is 2.75 MPa
(Schwark & Muller, 1996) and Poissons ratio is assumed to be 0.35. The thickness of the

47
Chapter 3- Development of a representative unit cell: theory and calibration

coating was found to be approximately 0.082 mm based on equation (3-6) and was also
confirmed using microscopy.

3.5.4 In-plane shear calibration

The properties of the fabric were evaluated in a previous study (Ko, December 2006) using a
KES-FB1 shear tester at Philadelphia College of Textiles. A fabric sample is held by two
parallel chucks on both edges of the fabric as shown in Figure 3-13a. The one edge moves
parallel to the fixed grip which applies a shear force to the fabric. A constant normal force is
applied to prevent buckling or wrinkling of the specimen. Figure 3-13b better illustrates the
boundary conditions of the test. The test was performed only on coated fabric but as previously
discussed, the low level shear behavior of the uncoated fabric was observed to be the same as the
coated fabric after the influence of the coating is overcame. Figure 3-14 is a reproduction of the
results in the study and it can be seen that the tests had a high reproducibility. For
characterization of the shear relationship described in equation (3-31), the loading portion of the
curve is used.

W2

Fs

2 2

1 ,1

(a) (b)

Figure 3-13: KES-FB1 textile shear tester shown in a) and b) shear deformation adopted in
the KES-FB1 testing system

48
Chapter 3- Development of a representative unit cell: theory and calibration

0.09

0.08
Membrane Shear Stress (N/mm)
0.07

0.06

0.05 ing
oad
L
0.04
g
0.03 a din
U nlo
0.02

0.01

0
0 0.05 0.1 0.15 0.2
Shear Strain (radians)

Figure 3-14: Shear stress-strain behavior for 350dtex nylon airbag fabric

Since the fabric undergoes a shear deformation under a constant load, the fabric sample is subject
to both shear and tensile forces. The model depends on the pure shear response of the fabric so
the tensile response measured in the fabric needs to be removed. The extension of the unit cell
can be used considering the boundary conditions that arises from KESF testing system as shown
in is as follows.

1 = 1 = 1 (3-41)

2
2 = (3-42)
cos

where i is the yarn stretch, i is the orthogonal stretch with respect to the initial fabric
configuration, and is the shear angle as shown in Figure 3-13b.

Now that the stretches applied to the unit are expressed as a function shear strain, Kawabata
demonstrated that the pure membrane shear stress-strain behavior that arises from yarn rotation,

49
Chapter 3- Development of a representative unit cell: theory and calibration

S12, can be obtained by removing the tensile forces that contribute to the recorded shearing force,
Fs as shown below (Kawabata, Niwa, & Kawai, 1973c; Kawabata, 1989):
n2
S12 = (FS W2 tan ) (3-43)
cos

where n2 is the fill yarns per inch, W2 is the constant normal load and Fs is the applied shear load.

The drawback to using the KES-FB1 for calibrating shear is that it can only give information
about low level shear behavior in the current shear model the tests can only help find G1 and
G2. At high shear stresses, the shear behavior is no longer governed by yarn rotation rather than
the shear stiffness of the yarn. Therefore, the jamming stiffness of the fabric needs to be
defined. Here it is proposed that the jamming shear stiffness of the yarn is related to the shear
stiffness of the fiber material knocked down by the fiber packing factor of the yarn:

E
G3 = p (3-44)
2(1 + )

where E is longitudinal tensile modulus of the yarn, is the Poissons ratio of the yarn and p is
the packing factor of the yarn.

3.6 Results

The simple unit cell model was run to evaluate the unit cell model for the cases of uniaxial stress
applied to the warp and fill directions and biaxial state of stress. Figure 3-15 shows the model
generated membrane stress-strain curves under equal biaxial and uniaxial state of stress for the
warp and fill directions.

50
Chapter 3- Development of a representative unit cell: theory and calibration

60.00 60.00
Model-Warp Model-Warp
Model-Fill
Model-Fill
50.00 50.00
membrane stress (N/mm)

membrane stress (N//mm)


40.00 40.00

30.00 30.00

20.00 20.00

1:1
10.00 10.00

0.00 0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
strain strain

(a) (b)
Figure 3-15: Membrane stress-strain curves for 350dtex fabric produced by the unit cell
model

3.7 Sensitivity analysis

A sensitivity analysis was performed to investigate the influence of the essential yarn input
parameters. For clarity of presentation, only the warp stress is shown but the same trends
witnessed in the warp direction are applicable to the fill direction. For the sensitivity study, the
nominal input values are those obtained through characterization and therefore are identical to
Figure 3-15.

Figure 3-16 shows the sensitivity of the unit cell model to yarn crimp for equal biaxial extension
and uniaxial load. Under uniaxial stress, increasing the yarn crimp increases the failure strain of
the fabric. The increase in crimp requires more elongation to uncrimp or geometrically deform
the yarn before material straining can occur. Decreasing crimp eliminates the amount of
geometric deformation and therefore requires less elongation to reach the fabrics failure point.
However, the change in crimp does not change the stiffness or failure load under uniaxial

51
Chapter 3- Development of a representative unit cell: theory and calibration

extension. For biaxial extension, the trend of increasing failure strain with crimp can be seen but
there is also a decrease in stiffness and the ultimate stress.

60.00 60.00

-Crimp -Crimp
+25%Cr +25%Cr
+10%Cr +10%Cr
50.00 Nominal 50.00 Nominal
-10%Cr -10%Cr
-25%Cr +25%Cr
+Crimp +Crimp
warp - membrane stress (N/mm)

warp - membrane stress (N/mm)


40.00 40.00

30.00 30.00

20.00 20.00

10.00 1:1 10.00

0.00 0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
strain strain

(a) (b)

Figure 3-16: Crimp parameter sensitivity for a) 1:1 biaxial extension and b) uniaxial
extension

Figure 3-17 shows the sensitivity to the bending rigidity of the yarn. If the yarn is considered to
be perfectly flexible, the unit cell is not able to capture the stiffening effect as the fabric
uncrimps. The yarn simply elongates until the crimp is removed and then forces are generated as
the straightened yarn elongates. Outside the low level stress behavior during the uniaxial
extension, variations up to two times the nominal value of the bending rigidity have little effect
on the mechanical behavior of the fabric regardless of the stress state.

52
Chapter 3- Development of a representative unit cell: theory and calibration

60.00 60.00
No Bending +EI No Bending
+EI
Nominal
Nominal
50.00 2x EI 50.00 2x EI
10X EI

warp - membrane stress (N/mm)


-EI
warp - membranestress (N/mm)

10X EI
-EI
40.00 40.00

30.00 30.00

20.00 20.00

10.00 1:1 10.00

EI=0
0.00 0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
strain strain

(a) (b)

Figure 3-17: Yarn bending rigidity parameter sensitivity for a) 1:1 biaxial extension and b)
uniaxial extension

Figure 3-18 shows the sensitivity to the thickness of the coating in the unit cell model. There is
virtually no effect on the extensional behavior between an uncoated fabric, the nominal values
and coating that penetrates fully through the thickness of the fabric. Intuitively, one would
expect this result since the stiffness of the coating is much less than that of the fabric.

53
Chapter 3- Development of a representative unit cell: theory and calibration

60.00 60.00

No Coating No Coating
Nominal Nominal
50.00 50.00
Full Coating
Full Coating
warp - membrane stress (N/mm)

warp - membrane stress (N/mm)


40.00 40.00

30.00 30.00

20.00 20.00

1:1
10.00 10.00

0.00 0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
strain strain

(a) (b)

Figure 3-18: Coating thickness parameter sensitivity for a) 1:1 biaxial extension and b)
uniaxial extension

3.8 Summary

A unit cell model based on a simple linear geometry was proposed and the governing
deformational mechanisms were derived. The characterization techniques to obtain inputs for
the model were carefully outlined and the geometric and mechanical properties of a 350dtex
airbag fabric are obtained. The results of the unit cell model represent the behavior of an ideal
fabric where deformation is uniform throughout.

The next step would logically be to perform a validation study comparing the results of the
simple unit cell model to experimental results. However, stress heterogeneity is seen in many of
the standard tests to evaluate the in-plane behavior of the fabric. Stress heterogeneity arises due
to specimen geometry, yarn orientation as well as the samples boundary conditions. This

54
Chapter 3- Development of a representative unit cell: theory and calibration

warrants further computation to accurately model the physical conditions of certain tests.
Additionally, besides simulating physical experimental specimens, we would like to use
mechanical response of the developed unit cell as the basis for structural analysis of the airbag in
the crash simulation. More specifically, the technique developed needs to be formulated in a
way that it can be of use with the current design tools. Therefore in the next chapter, a finite
element continuum approach using the unit cell as the constitutive behavior is presented that can
be utilized in modeling stress heterogeneity in test samples.

55
Chapter 4 Implementation and verification

Chapter 4 Implementation and verification

4.1 Background and approach

As established to this point, the airbag contributes significantly to the overall crashworthiness of
an automobile. The analysis and simulation of vehicles for crashworthiness is still fairly young
field and has evolved greatly the past 30 years (Khalil & Du Bois, 2004). Starting as simple
lumped parameter model, today models are based on 3D continuum mechanics with intricate and
complex geometry that often require the use of supercomputers for simulation. In the past
decade, the simulation of the airbag and occupant/airbag interaction has become a standard
option in many commercial simulation codes (Drnhoff et al., 2008; Hirth et al., 2007).

The high costs conducting and difficulty measuring the airbag deployment and impact makes
simulations desirable to evaluate a large number of parameters (Khan & Moatamedi, 2008). The
airbag deployment and impact, like most crashworthiness analysis of vehicles, involves highly
nonlinear structural mechanics due to large deformations, rotations and material nonlinearity.
Also considering the equations of state to model gas flow from the inflator, the governing
equations pose a great challenge in the analysis, warranting the use of a numerical method like
the finite element method.

The finite element method of structural dynamics solves the set of nonlinear partial differential
equations of motion for body or structure in a space-time domain along with the material stress
strain relationship within defined boundary conditions. The body or structure is descritized into
a finite number of components (elements) that are connected by points (most commonly referred
as nodes). The temporal and spatial discretization results in a solution of the equations in the
forms of the values of the field parameters at common nodes of the elements. Within the domain
of an element, the values of all the parameters are approximated through the use of the
56
Chapter 4 Implementation and verification

appropriate shape functions. The state of stress and strain in the material corresponds to the
minimum energy of the system which is an approximate representation of the overall equilibrium
of the system.

The finite element simulation of airbags is based on the assumption that the fabric can be
approximated using membrane shell elements with an orthotropic, linear stress-strain material
relationship (Bedewi, Marzougui, & Motevalli, 1996; Hirth et al., 2007; Khan & Moatamedi,
2008). The membrane assumption is valid since the out-of-plane mechanical properties of the
fabric are very low compared to the in-plane properties. However, the orthotropic material
assumption as stated numerous times before, is incorrect as the material is highly anisotropic as
the yarns can rotate to a non-orthotropic state. The current state-of-the-art fabric constitutive
model in LS-DYNA (*MAT_FABRIC) utilizes a nonlinear orthotropic stress-strain formulation
with non-orthogonal material axes (Hallquist, 2006). However, to calibrate the inputs for the
model requires cruciform biaxial test data for extensional responses and picture frame shear tests
for the shear response. This necessitates special equipment and multiple tests for different shaped
airbags which can be time consuming particularly when evaluating novel fabric systems.
Additionally, at stress ratios that deviate from the tested configuration, the constitutive behavior
may not be correct due to rearrangement of the fabrics internal structure.

The proposed alternative is to incorporate the unit cell model described in the previous chapter as
the basis for the constitutive behavior for shell-membrane elements. Figure 4-1 illustrates the
basic concept of the proposed implementation of the unit cell model as the basis for representing
the in-plane mechanical behavior of the fabric. The fabric structure is replaced by a continuum
of a homogeneous anisotropic material. Stretches and rotations are applied to the continuum
membrane which are identical to that occurring in the unit cell. One key aspect of the unit cell-
continuum formulation is to translate the continuum strains which are taken with to respect the
element coordinate system and compute the strains that occur in the fabric with respect to the
yarn configuration which may not coincide with the element coordinate system upon
deformation. Once the strains are transformed to the yarn orientation, the forces from the unit
cell can be computed. Then, the forces are normalized and translated back into the element local
coordinate system.

57
Chapter 4 Implementation and verification

Original Unloaded Structure LS-DYNA applies strain to element

n4 UMAT finds yarn strains from


0
element coordinate system n3
q2 n3

= y q1 0
n4
q2
q1
0
q2 q1 0
n1 n2 n2
x

UMAT applies yarn strain to y n1


Unit cell forces and
normalized stress are X3 unit cell algorithm x
calculated
X1 X2
Coating
H01

H02
y01 n3

y02
n4

n2
=
UMAT transforms unit cell
stress back to element n1
coordinate system y
x
Deformed Loaded Structure

Figure 4-1: Outline of the basic concept of the unit-cell-membrane

Therefore, the outlined approach can determine the fabric state of stress with consideration of the
internal changes the yarns undergo without explicit modeling of each yarn. This approach is
therefore very computationally efficient compared to other techniques that attempt to simulate
the mechanical response and contact behavior of every yarn within the fabric. The increase in
efficiency can lead to faster solution times or allow more resources to be dedicated to more
computationally intensive processes in the airbag simulation. Additionally, the history of the
evolution of certain fabric parameters such as the yarn tension, yarn failure, contact forces and
yarn orientation can be tracked for each element for the duration of the simulation.

While there are many advantages using this method, the assumptions made in the development
can impose limits on the use of the model. For instance, the continuum assumption is only valid
if the length scale of the element is larger than the yarns and empty spaces within the physical
fabric. More specifically, the element dimensions cannot be smaller than the dimensions of a
single crossover. This means more dense fabrics can use smaller elements in the simulation

58
Chapter 4 Implementation and verification

while still retaining the physical conditions. Since the airbag fabric is a very dense weave and
the evolution of new airbag fabrics are increasingly becoming denser, this approach has plenty
warrant in the structural analysis of the airbag deployment. If this method was applied to small
structures made of coarse woven fabrics, the limiting element size may be larger than the
required mesh size needed to obtain the correct results.

The other main assumption that can limit the capabilities of the model is the no-slip condition
between the yarns. During the failure of fabric structures, the yarns can slip and slide near the
location of the failure or tear to accommodate the equilibrium within the fabric. This obviously
changes the internal geometry of the fabric and the level of tension can vary throughout the yarn
length along a single crossover. Under the no-slip assumption, yarns crossover if affined to the
midpoint of both yarns and the yarn tension developed is constant through the length of the yarn
within the unit cell.

The final main assumption is that the unit cell remains in a flat planar configuration during out-
of-plane deformations. More clearly, the curvature of the structure is not reflected in the unit
cell geometry. This effect of curvature is significant as it can strain the yarn as well as change
the internal structure of the fabric as seen in Figure 4-2. This assumption is appropriate for
structures with large radius of curvatures with fabrics that have small yarn spacing shown where
the out-of-plane deformation is negligible. In the airbag application, the curvature of the
structure is typically fairly large and the density of the fabric is very high. However, if this
method were to be applied to small diameter tubes like those used in biomedical or petroleum
applications, the effects of the curvature on the unit cell geometry maybe more pronounced and
may affect the mechanical response of the fabric.

Figure 4-2: Effect of curvature on the internal structure of woven fabric

59
Chapter 4 Implementation and verification

4.2 Explicit dynamic finite element analysis

The numerical testbed for this study is the commercial nonlinear dynamic finite element code,
LS-DYNA. The code is used extensively in the automotive industry as it contains sophisticated
contact algorithms for modeling impact problems in addition to specialized capabilities for
modeling airbags, seatbelts and anthropomorphic test devices. Most importantly, LS-DYNA
allows users to code their own subroutines defining material models, more commonly referred to
as User Material Model (UMAT). Figure 4-2 shows the typical analysis process used by LS-
DYNA with the UMAT option invoked.

Update
Velocities

Update Update
Accelerations Displacements

Start

Apply Force
Process
Boundary
Contacts
Conditions

User Material Process


Model Elements

Figure 4-3: Numerical procedure of LS-DYNA with user material model option

The user-defined subroutine (UMAT) allows for formulation of constitutive relationships for
applications where the standard library of material definitions fail to accurately represent the
physical material behavior. The UMAT is coded in FORTRAN and contains a common block to
access the strain increments, material constants, element failure flags and load curves from the
main program. At each time step, LS-DYNA passes the element strain increments evaluated at
the shell element integration point to the subroutine. For this study, the integration points are
located at the center of the element for the constant stress membrane. The UMAT computes the

60
Chapter 4 Implementation and verification

stresses within the element and passes the values back to the main program to continue the
analysis.

4.3 Implementation of unit cell based UMAT for FE shells

4.3.1 Continuum formulation

The challenge in the continuum formulation of the constitutive behavior of fabric is that the
configuration of the fabric is continuously changing. The first issue to be addressed is the
measuring the kinematics of the changing configuration. From continuum mechanics, the
fundamental measure of a deformed body at a point in time can be given by the deformation
gradient which is defined for a two-dimensional body as

xi (t )
Fij (t ) = (4-1)
x 0j

The deformation gradient can be used to find the stretch and rotations that a body undergoes in
two points in time. To do so, the right Cauchy-Green deformation tensor needs to be constructed
as per:

C = FT F (4-2)

The Cauchy-Green deformation tensor is symmetric and positive-definite. The Green


deformation tensor can now be used to find the stretch and rotation in each yarn.

Within the constraints of the UMAT, the displacement gradient can be constructed at a time t
based on the increments of strain that are passed by the main program

x t = x t 1 + t (4-3)

where xt-1 is the displacement gradient of the previous time step and

t
xt xyt
= t (4-4)
xy yt

61
Chapter 4 Implementation and verification

t
where x y xy are the current strain increments in the local element coordinate system
t t

xyz.

The local axes system used in LS-DYNA is based on the shell node numbering where the x-axis
is defined parallel to the n1 - n2 edge as shown in Figure 4-1 (Hallquist, 2006). The z-axis is
therefore taken to be normal to the element mid-plane and the y-axis is determined by the cross
product:

y = zx (4-5)

The material deformation gradient can therefore be found using

Ft = xt + I (4-6)

where I is the identity matrix.

To track the configuration of the material axis, two vectors q1 and q2 are assumed to be parallel
to the original orientation of the warp and fill yarns respectively. Therefore, the vectors are
assigned based on the angle of the yarn

(4-7)
q0i = [cos i0 sin i0 ]
T

0
where i is the initial angle of the yarn taken with respect to the local x axis.

According to Bathe (Bathe, 1982), the stretch of each material yarn axis can be found using
equation (4-8)

ti = q0Ti Cq02 (4-8)

Therefore, with the stretch in each yarn vector calculated, the displacement along each yarn
within the unit cell at the current time step is determined below:

62
Chapter 4 Implementation and verification

d it = y0 i (ti 1) (4-9)

With the stretch of the warp and fill material yarn axis known along with the Cauchy-Green
deformation tensors, the angle between the yarn axes at the current time step can be found using
equation (4-10) (Bathe, 1982):

q01T Cq02
cos = (4-10)
12

With the updated angle between the two yarn vectors known, the vector coordinates need to be
updated. Assuming that a positive value of strain results in the yarn vectors becoming acute, the
change in the yarn angle is determined by equation below

( 0
f w0 )
= (4-11)
2

A positive change in yarn angle follows the rotation convention shown below for each yarn
vector:

1t = 10 (4-12)

2t = 20 (4-13)

Therefore the newly rotated yarn vectors for the warp and fill directions can be expressed
according to equation (4-14):

qi = [cos( it ) sin ( it )]
T
(4-14)

With the unit cell displacement and current angle between the yarns now known, the yarn strain
can be found and the resulting forces can be computed using the procedure outlined in the
previous chapter. The obtained stresses now must be transformed from the yarn coordinate
system back into the local element coordinate system. Keeping with the tensor notation, the
63
Chapter 4 Implementation and verification

local stress tensor can be computed back into the local element xyz system using the scalar
magnitude of the internal yarn forces and the dyadic product of the unit yarn vectors at the
current time step according to Peng and separately by King (King et al., 2005; Peng & Cao,
2005).

N1
= (q1 q1 ) + N 2 (q2 q2 ) + S12 (q1 q2 + q2 q1 ) (4-15)
h h h
where N1 is the magnitude of the force per unit length from the warp yarn end, N2 is the
magnitude of the force per unit length from the fill yarn end, S12 is the magnitude of the shear
force per unit length and h is the thickness of the shell element.

Specifically, the components of the stress tensor within the element coordinate system can be
shown to be:

N1 N S (4-16)
x = cos 2 1 + 2 cos 2 2 + 12 2 cos1 cos 2
h h h
N N S (4-17)
y = 1 sin 2 1 + 2 sin 2 2 + 12 2 sin 1 sin 2
h h h
N N S (4-18)
xy = 1 cos1 sin 1 + 2 cos 2 sin 2 + 12 (cos1 sin 2 + cos 2 sin 1 )
h h h
This form differs from the work of Shahkarami (A. Shahkarami & Vaziri, 2007) as the shear
component is included in the stress tensor rather than treating shear separately. Additionally, the
treatment of the material axes is consistent with the procedures for finite deformation, non-
orthogonal constitutive behavior in continuum mechanics (Bathe, 1982; Holzapfel, 2000).

Within the explicit finite element framework, the determination of the critical time step is crucial
for analysis. LS-DYNA requires the bulk and shear moduli of the material to be defined in order
to determine the critical time step as well as calculating transmitting boundary, contact stiffness
etc. Within the context of the proposed continuum model, the bulk and shear moduli are based
on published values of the speed of sound in bulk material the yarn is composed of and the
density of the fabric structure and can be computed using equations (4-19) and (4-20)
respectively.

64
Chapter 4 Implementation and verification

(4-19)
2c*
K=
3(1 2 )

2c* (4-20)
G=
2(1 + )
where is density of the fiber material , is the Poissons ratio of the yarn material. The values
of c* can be found using the expression:

c (4-21)
c* =
1 + cri
where c is the speed of sound in the fiber material and cr is the crimp of the yarn.

With respect to best practices, for unbalanced fabrics, the yarn crimp that results in the larger
values of K and G governs.

4.3.2 Special considerations yarn separation, failure and element erosion

Within the user material model framework, the displacements in the warp and fill direction will
always be given so that the yarn heights are unknown and the length of the yarns is a function of
the height. This is different from the previous chapter regarding uniaxial extension of the fabric
where the transverse displacement was unknown and was solved for assuming the transverse
yarn did not compress. In that case, it was necessary to establish certain unknown unit cell
conditions in order to generate a solution for uniaxial stress condition. In the user material
model framework, the axially incompressible yarn condition is still established however the yarn
spacing will always be determined from the continuum stretches. Therefore, under these
imposed conditions, during large negative displacements the algorithm will not find a solution
since continuity of the unit cell structure is broken. Here the yarns separate and are no longer in
contact. In this case, the yarns are decoupled so that the deformed geometry and forces of each
yarn can easily be computed.

65
Chapter 4 Implementation and verification

An instantaneous failure criterion based on the ultimate yarn strain is used to initiate failure of
the individual warp and fill yarns. Once the yarn surpasses the specified ultimate strain, the yarn
is flagged by the subroutine and the yarn is permanently eliminated from the unit cell. When a
yarn is removed from the unit cell, the shell element is no longer capable of carrying extensional
load in that direction or any shear loading. When both of the unit cell yarns have exceeded the
ultimate yarn strain, the shell element is then eroded from the finite element mesh. The erosion
results in the element being removed from the calculation while retaining its mass properties on
the nodes attached to the element. Nodes that become completely unconstrained due to the
erosion of elements are than also eliminated from the computation. Under certain conditions that
arise from single yarn failure, the elements can undergo an excessive and unrealistic deformation
which promotes numerical instabilities. To remedy possible instabilities, a user-defined criterion
was established in the user material model to erode unhealthy elements.

One last consideration of the UMAT formulation is the need to refer to the local x axis of the
element. This requires the 1-2 edge of each element to be parallel throughout the mesh to ensure
the yarn orientation is consistent throughout the simulated part. Therefore, the current UMAT
definition cannot be used in arbitrarily meshed structures. In the driver side airbag structure, the
bag is constructed of two circular pieces of fabric so the meshing within the constraints of the
UMAT is feasible for the majority of the bag. Possible problematic areas are near edges and
vent hole. Future work will need to address this limitation.

4.4 Verification

Code verification checks that the algorithms are working properly and that the solution of the
mathematical model is accurate. The addition of the unit cell algorithm with the continuum
formulation along with interactions between the subroutine and main program presents several
nuances that can cause problems. To verify the successful implementation of the code a series of
simple tests were performed. The first of these test simply check that the extensional behavior of
a single element to the simple stand-alone computer algorithm discussed in the previous chapter.
Prescribed displacements are applied to the element to confirm the biaxial and uniaxial response
of the implemented unit cell agree with the results from the simple model developed in the
previous chapter. The second set of simple tests consisted of applying displacements to a square

66
Chapter 4 Implementation and verification

patch of 5x5 elements to investigate the elements whose displacements are controlled by the
connectivity of the neighboring elements. This provides us with an opportunity to find arbitrary
displacement ratios where the algorithm has difficulties converging.

The final set of verification studies investigated the rotation-nonorthogonal behavior of the yarn
vectors as illustrated in Figure 4-4 and Figure 4-5. Three single element tests were performed:
1) pure shear, 2) rail shear and 3) a balanced trellis like shear varying the shear stiffness of the
yarn crossover. Under pure shear, one would expect that the only deformation that arises is
shear, therefore no stretch is generated in the yarns. From Figure 4-4a, the model correctly
exhibits the proper pure shear behavior. Under rail shear test there is positive stretch generated
in direction of the deformed edge in addition to shear stresses generated. In Figure 4-4b, the
element is sheared so that the deformed edge from shear is parallel to the fill yarns. Again, the
calculated response of the model mimics the expected behavior.

1.15 1.15

Warp Warp
Fill Fill
Yarn Stretch

Yarn Stretch

1.10 1.10

1.05 1.05

1.00 1.00
0.00 0.10 0.20 0.30 0.40 0.00 0.10 0.20 0.30 0.40
shear strain shear strain

(a) (b)
Figure 4-4: Verifying modes of shear: a) pure shear and b) rail shear

Finally, in Figure 4-5, an element with balanced yarn properties (meaning that the properties for
each yarn direction are identical) has two opposing corners with assigned displacements while
the remaining two corners are free. The purpose of this example is to examine the impact of
shear stiffness on the non-predefined rotation of the yarn. Assigning no shear stiffness or
resistance at the crossover, the crossover should behave like a pin. Thus, the crossover should

67
Chapter 4 Implementation and verification

act like a trellis where any deformation is attributed to rigid body rotation of the yarns. It can be
seen from Figure 4-5 that shear strain of the crossover results in no stretch of the yarns. If very
high or infinite shear stiffness is assigned, the crossovers should act like they are fixed and
remain in the orthogonal state. Therefore, any deformation of the element should be attributed to
stretch as seen in the calculations. Finally, an assigned realistic value of shear stiffness is applied
and it can be observed that the element experiences a mix of rotation and extension. Overall, this
confirms that the model matches deformational behaviors that one would expect under a
multitude of conditions.

1.30

1.25
Yarn Stretch

1.20

1.15

1.10 Infinite Shear Stiffness


No Shear Stiffness
1.05
Realistic Shear Stiffness
1.00
0.00 0.10 0.20 0.30 0.40
shear strain
Figure 4-5: Verifying shear deformation behaviors of a single element with varying shear
stiffnesses

4.5 Summary

The structural analysis of airbag and fabric structures requires a constitutive model that can
consider the complex and often heterogeneous deformations of the fabric structure, but at the
same time, the model needs to be simple enough to keep computational costs low. Here, the
mechanistic unit cell model is implemented as the material behavior for an efficient membrane
element in the finite element code LS-DYNA. Using the developed user material model, a wide
variety of fabric deformational behaviors can be obtained without explicitly modeling each yarn.

68
Chapter 4 Implementation and verification

The user material model has been carefully verified to ensure the in-plane extensional and shear
behaviors are computed correctly and are passed to LS-DYNA without any error. Now that the
unit cell model has been successfully implemented in terms of computational correctness, the
next step is to check the model for physical correctness. In the next chapter, a validation study is
performed demonstrating the models accuracy under a variety of deformations.

69
Chapter 5 Validation

Chapter 5 Validation

5.1 Background and approach

While the verification and sensitivity of the model has been performed, neither gauges the
competence of the model for representing the physical system. Validation of a computational
model is the process of determining the degree to which a model is accurate representation of the
real world from the perspective of the intended uses of the model (Schwer, 2002). The intent of
this chapter is to confirm correctness of the models quasi-static in-plane mechanical properties to
experimental data. Since published, publically available experimental data on the mechanical
properties of airbag fabric is very limited, especially with regards to the current state-of-art
fabrics, a series of experiments were performed to rigorously confirm the results of the models.
Of the experiments performed, the biaxial extension of the fabric is believed to be the closest to
mimicking the conditions that occur in the airbag. During the simulation of the quasi-static tests
using the user material model within the dynamic explicit finite element framework, the total
kinetic energy was kept to a minimum so that any influence from inertial effects is negligible.

5.2 Experimental evaluation

5.2.1 Uniaxial extension

The uniaxial extension of fabric is often performed due to its simplicity and can be used to
compare the stiffness of the warp and fill directions. The uniaxial tensile properties of the coated
and uncoated fabrics were measured according to the modified strip tensile test method as
specified by ASTM D5035-95. Five samples taken in each of the warp and fill directions for
both coated and uncoated fabrics were tested. The ends of the fabric were wrapped with layers
of masking tape to protect that fabric from crushing at the grips. The specimen width was 25

70
Chapter 5 Validation

mm with a gauge length of 100mm. An Instron 1122 testing machine located in the Department
of Zoology at The University of British Columbia was used for testing. The crosshead speed
was 100 mm/min.

The stress-strain curves in the warp and fill directions for the coated and uncoated samples are
shown Figure 5-1. The stress is normalized based on a membrane assumption, simply
corresponding to the force divided by unit length. The vertical error bar show the maximum and
minimum stress values at selected data points. The horizontal error bar at the failure point
illustrates the minimum and maximum failure strains recorded. Failure occurred within the
middle of the gauge length indicating uniform stress distribution. Overall the tests were very
reproducible. The sample population for the uncoated sample was reduced to four since one of
the specimens had broken near the grip.

70.00
Exp-Coated-warp
60.00
Exp-Coated-fill
membrane stress (N//mm)

50.00 Exp-uncoated-warp
Exp-uncoated-fill
40.00

30.00

20.00

10.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
strain

Figure 5-1: Average uniaxial stress-strain of 350dtex fabric

From Figure 5-1, it can be seen that there are no discernable difference between the
deformational behavior of coated and uncoated samples as observed in the biaxial extension

71
Chapter 5 Validation

tests. The stress-strain behavior is very different from the biaxial behavior of the fabric. The
uniaxial response is characterized by a sigmoidal shape which consists of a low initial modulus
followed by gradual stiffening into high linear modulus. Compared to the biaxial results, fabrics
under uniaxial loading have higher failure stresses but lower apparent modulus.

5.2.2 Biaxial extension

As discussed previously, the biaxial loading is the fundamental loading condition seen in an
airbag. To evaluate the biaxial behavior, there are three methods discussed in the literature: burst
test, inflated cylinder test, and cruciform plane biaxial extension test (Bassett, Postle, & Pan,
1999). These methods are better illustrated in Table 5-1 showing the original fabric
configuration and the deformed fabric during operation. In the burst test, stress and strains are
quantified based on the internal pressure within the fabric and the fabric shape and requires
impermeable fabrics. In the inflated cylindrical test, the results can be affected by the presence
of seams unless circularly woven samples are tested and also requires fabrics that are
impermeable. The cruciform biaxial extension test can evaluate both permeable and
impermeable fabrics at a wide range of stress ratios and can directly measure stresses and strains.
For this study, biaxial response of the fabric is evaluated using the cruciform biaxial extension
method.

72
Chapter 5 Validation

Table 5-1: Test methods in the literature for evaluating the biaxial behavior of fabrics

Test Method Original Deformed


Pressure Sample
Source Sample (Spherical)
(Flat)
Burst

Pressure
Sample
Source (Cylindrical)
Inflated
Cylinder

Sample
(Flat)

Cruciform

Biaxial extension experiments of coated and uncoated fabric samples were performed using a
custom-made large load capacity biaxial tester (see Figure 5-2) located at the Hess Research
Laboratories at Drexel University. The tester was designed with fibrous materials in mind, more
specifically geotextiles, but is suitable for a wide range of textiles and composites that require
large applied loads (Wartman, Harmanos, & Ibanez, 2005). The tester consists of a servo-
hydraulic test frame with two pairs of grip carriages oriented 180 degrees apart that move in
equal and opposite directions. The carriages are driven by a 67 kN hydraulic actuator powered
by a hydraulic manifold with high-pressure and low-pressure accumulators. Each actuator drives
the grip carriages via connection arms. Each grip carriage has a dynamic range of motion of
10.2 cm and peak velocity of 30 cm/s. A 2.0 cm threaded rod extends from each carriage to be
used as the grip attachment point. The grip widths are approximately 20.23 cm (8 inches) wide
and have pyramid-teethed attachment plates to provide sufficient grip. A computer system
consisting of a digital controller running ANCO Engineers ANIPC-400 software controls the

73
Chapter 5 Validation

tester through a 4 channel data acquisition system. A 32 channel analog-to-digital converter is


used to obtain the data.

Figure 5-2: Biaxial tester at Drexel University

Specimens were cut into a cruciform shape as shown in Figure 5-3 leaving a 20.32 cm x 20.32
cm (8 x 8) test area. Great care was taken while cutting the sample to ensure the fabric
remained orthogonal and true to the desired dimensions in order to avoid unwanted bias.
Masking tape placed in layers was used as a grip tab to protect the fabric from being damaged by
the grips. A sample size of 5 coated and 5 uncoated samples were prepared.

74
Chapter 5 Validation

Figure 5-3: Biaxial specimen dimensions

The alignment and calibration of the loading actuators was first performed. Using a straight edge
and a laser level, each extender arm from the actuator was measured and adjusted to ensure each
grip carriage and arm was equidistant. Upon confirmation of the actuator extensions calibration,
the grip assemblies were attached to each arm and were checked to ensure that the grip slots
were leveled. An overhead digital camera was placed over the sample to qualitatively measure
fabric displacements within the gauge length in addition to the recorded cross-head
displacements.

An extension rate of 5mm/sec (2.5mm/sec per actuator) and a sample rate of 50Hz were inputted
into the digital controller. On a separate digital controller, the sampling rate of the overhead
digital camera was defaulted at 2Hz. Samples were loaded into the grip assemblies with the
5.08cm grip tab fully inserted into the grip slot and secured. The orientation of the fabric (warp
or fill direction) with respect to the load cells (A,B,C,D) was recorded. To ensure there was no
bias or unbalance in the load cells, samples were rotated, as demonstrated in Figure 5-4 into
different configurations so that the load cells had the opportunity to measure warp and fill

75
Chapter 5 Validation

directions during the duration of testing. Results from using this procedure show that there was
no noticeable asymmetry between the load cells.

Figure 5-4: Biaxial testing specimen configuration

A pretension of 44.5 N (10 lbs) was introduced to each actuator which is equivalent to 2.2 N/cm
(1.25 lbs/in) of stress imposed on the sample. The pretension is an ad-hoc procedure to eliminate
excessive experimental noise obtained from the load cells early on in the testing sequence. The
samples were loaded until final fracture. Upon completion of the test, the failed sample was
removed and the alignments of the actuators were checked after the actuators had returned to
their original positions. A new sample was placed into the grip fixtures and the loading
procedure was repeated.

The load data recorded had some inherent noise which was eliminated using a moving average
filter. The average membrane stress-strain curve is shown in Figure 5-5. The stress is
normalized based on a membrane assumption, simply force divided by unit length. The vertical
error bars show the maximum and minimum stress values at selected data points. The horizontal
error bars at the failure point illustrates the minimum and maximum failure strains recorded.
Overall, the biaxial response for the coated and uncoated samples was shown to be reproducible.
The failure mode of the fabric was tearing which initiated at the corners of the samples.

76
Chapter 5 Validation

45.00
Coated-warp
40.00
Coated-fill
membrane stress (N/mm)

35.00 Uncoated Warp


Uncoated-warp
30.00
Uncoated-fill
25.00

20.00
Uncoated Fill
15.00 Coated Warp
Coated Fill
10.00

5.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25
strain
Figure 5-5: Average biaxial stress-strain curve for coated and uncoated samples

The fabric under biaxial extension exhibits a hyper-elastic like behavior very similar to the yarn
behavior exemplified by a high initial modulus, some softening before stiffening into a high
linear modulus until failure as seen in Figure 5-5. There is no drastic change in extensional
properties between the coated and uncoated samples. This is to be expected since the silicone
coating has a considerably lower stiffness compared to the yarn system. However, it was also
observed during the experiments that the coated fabric fractured at a lower strain than the
uncoated fabric. The probable cause of the discrepancy of the failure points between the coated
and uncoated sample is due to the difference in the stress concentrations seen in the corners of
the samples. The difference lies in the prohibitive nature of the coating to resist in-plane yarn
rotation and sliding. Qualitatively, yarn sliding at the edges of the sample apparent by fraying of
the edges occurs in the uncoated fabric which can be seen from the photographs taken and shown
in Figure 5-6a. This sliding and rotation allows the yarns to conform to the load path putting
less stress on the yarns. On the other hand, in Figure 5-6b, the coated fabrics edges have no
fraying. Additionally, the edges have a slight curvature caused by crimp interchange between
the warp and fill yarns due to no slip condition.

77
Chapter 5 Validation

(a)

(b)
Figure 5-6: Photographs of biaxial extension at approximately 12% strain for a) uncoated
and b) coated fabrics

5.2.3 Bias extension

The bias extension test refers to the uniaxial extension of a rectangular fabric specimen that has
been cut at 45 degrees to the yarn directions. The test is fairly simple to conduct compared to
other methods used for shearing a fabric such as rail shear or picture frame test and can provide
reasonably reproducible results. One drawback is that the strain is not homogeneous throughout

78
Chapter 5 Validation

the sample in which three distinctive shearing zones can be identified. Referring to Figure 5-7,
Zone A undergoes pure shear deformation until the yarns reach a jamming point. Zone B is mix
of yarn extension and shear while Zone C remains undeformed.

Zone C
Zone C

Zone B Zone B
Zone B Zone B

L Zone A Zone A

Zone B Zone B
Zone B Zone B

Zone C
Zone C
L
a) b)

Figure 5-7: Heterogeneous deformation of fabric bias extension

Five samples each of the uncoated and coated fabric specimens were tested. The specimens
dimensions were 50mm wide by 100mm long gauge length which has been used in other
research (Harrison, Clifford, & Long, 2004; Page & Wang, 2000; Spivak & Treloar, 1968). The
samples were cut vigilantly to ensure the 45 degree orientation and optical microscopy was used
to confirm the fabric angle. Masking tape was applied to the gripped portion of the fabric to
protect it from grip induced damaged. A 1cm x 1cm orthogonal grid was drawn on each sample
using a felt-tip permanent marker as a reference to measure rotation at the center of the specimen
as well as monitor the deformations in other regions in the fabric. Additionally, the grid square
in the center of the sample will be used as a Point of Reference to quantify shear strain verses the
applied stress to the specimen in the upcoming discussion.

The bias tests were performed on the Instron 1122 tester at room temperature with a digital
camera mounted onto a tripod to record the shear angle at the center of the specimen as shown in
79
Chapter 5 Validation

Figure 5-8. The camera recorded video at 30 frames per second at a resolution of 640x460
pixels. The video started to record before the extension was performed and a clicking noise
which acts as an audio marker was made simultaneously when the Instron loading was engaged.
Using the audio marker, the video images can be correlated to an exact data point recorded by
the Instron. Data acquisition from the Instron was performed at 100 Hz. A 500 kg load cell was
installed and an extension rate of 60 mm/min was used. After jamming, the samples started to
slip out of the grips, therefore they were not tested to failure. Since the low level shear behavior
is of interest, the range of data obtained when the specimen slips at higher shear strains is
ignored.

Instron Tester

Camera & Tripod

Bias Specimen

Figure 5-8: Bias extension-apparent shear angle recording set-up

Post-processing the video to obtain the shear strain can be a daunting task. The duration of each
bias test was about one minute and the video was recorded at 30 fps, which results in 1800
images available for measurement for one sample. Even using one frame per second, it would
require 60 images to manually measure for shear angles. Therefore, to reduce the amount of
images while retaining accuracy, images were measured at 1 fps for the first 10 seconds where
the majority of the shear deformation occurs and the one frame every five seconds is measured to
80
Chapter 5 Validation

until 50 seconds has past for the test. It was witnessed that after a minute slippage of the
specimens in the grips begins to occur in the sample, therefore any data points beyond that time
are not considered. The angle of the lower corner of the center square marked off on the
specimen is measured using the software ImageJ.

Figure 5-9 shows the average load-elongation behavior of the coated and uncoated bias fabric
samples. At low levels of extensions, the coated samples exhibit a stiffer behavior than the
uncoated fabric. After about 20mm, the stiffnesses of the coated and uncoated fabric are almost
identical.

8.00
Exp - Coated
7.00
Exp - Uncoated
Applied Stress (N/mm)

6.00

5.00

4.00

3.00

2.00

1.00

0.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00
Extension (mm)

Figure 5-9: Load-elongation of bias coated and uncoated airbag fabric

Figure 5-10 shows the average applied specimen stress and shear angle results for the coated and
uncoated specimens. To be clearer, the y-axis (applied shear stress) is the load recorded by the
Instron normalized by the specimen width and the x-axis (shear angle) is the measured change in
angle of the center square. The alphabetic markers signify the correlated video snapshots shown
in Figure 5-11 for the coated fabric and Figure 5-12 for the uncoated fabric so that the stress-
strain curve can be referenced with respect to the observed deformed configuration of the

81
Chapter 5 Validation

sample. The Point of Reference (P.O.R.) where the center shear angle was measured is indicated
in the first frame. Since the camera position moved during the testing, it is not possible to
superimpose semi-transparent images to illustrate the average deformed configuration of the
sample. Therefore in the construction of Figure 5-11 and Figure 5-12, the sample whose
applied stress-strain response was closest to the average is referenced.

16.00

Uncoated G
14.00
Applied shear stress (N/mm)

Coated
12.00

10.00
F
8.00
E
6.00
B C D
4.00

2.00

0.00
A
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Center Square Shear Angle (degrees)

Figure 5-10: Bias shear -- applied specimen end stress vs. center specimen shear angle

The horizontal error bars at each data point represent the deviations that occur mostly due to
human error in measuring shear angle and effects that result in yarn slippage. The error in the
shear angle for both specimens is acceptable, although the uncoated samples exhibit higher strain
error. Overall, Figure 5-10 attempts to quantify the local shear deformation of the pure shear
deformation zone as opposed to the total elongation of the sample which is mainly a global
response.

82
Chapter 5 Validation

Figure 5-11: Typical bias shear deformation sequence - 50mm x 100mm coated sample

Figure 5-12: Typical bias shear deformation sequence - 50mm x 100mm uncoated sample

83
Chapter 5 Validation

It can be observed that at lower levels of shear strain there is a significant difference in the
deformational behavior of the coated and uncoated samples. Figure 5-13 better illustrates this
low shear strain behavior for both of the fabric sample sizes tested. It can be observed that the
coated fabric has a much higher initial shear stiffness compared to the uncoated fabric. While
the coating has a low Youngs Modulus and is very thin, it still seems to inhibit yarn rotation
during low shear strains. At higher shear strains, the coated fabric starts to behave similarly to
the uncoated sample having a region of low shear stiffness before dramatically stiffening when
the yarns have rotated into a jammed (locked) condition.

2.00

1.80 Uncoated
1.60
Coated
Applied Stress (N/mm)

1.40

1.20

1.00

0.80

0.60

0.40

0.20

0.00
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0
Center Square Shear Angle (degrees)
Figure 5-13: Detailed low shear angle plot of applied stress vs. center shear of bias sample

84
Chapter 5 Validation

5.3 Comparison of UC, FE-UC and experiments

5.3.1 Uniaxial results

Figure 5-14 and Figure 5-15 show the stress-strain results for the coated and uncoated uniaxial
extension of the simple single unit cell model, simulated sample using the finite element material
model formulation and the experimental results. The simple single unit cell analytical model is
an adequate approximation of the uniaxial behavior of the fabric assuming that the deformation
is homogeneous. In reality, the deformation of the sample is not homogeneous due to the
transverse contraction along the free edges of the sample from crimp interchange of the yarns.
Figure 5-16a demonstrates the predicted strain heterogeneity in the coated fabric at 30% strain
in the warp direction. Figure 5-16b shows the fabric strain in the fill direction. For brevity,
only the warp-loaded coated simulation is shown but the following discussion is applicable to
fill-loaded direction and uncoated fabric. As it can be seen by comparing the warp and fill
strains, the simulation predicts an area of biaxial tension near the clamped portion of the sample
caused by constriction of the transverse yarns against yarn crimp interchange. Along the center
of the specimen, the strain is highest and uniform where the axial yarns have almost completely
decrimped. Stepping away from the macroscopic strains, Figure 5-17a and Figure 5-17b shows
the calculated warp and fill yarn strains. Since the failure criteria of the unit cell is based on the
ultimate yarn strain, studying the yarn strains developed can indicate the location where possible
fabric failure initiates. In the examples given in Figure 5-17, at approximately 30% fabric strain,
the warp yarns near the middle of the sample approach the ultimate warp yarn strain which was
the area in which failure had occurred in the experiments. The simulated sample using the finite
continuum formulation has the ability to replicate both the macroscopic and yarn deformations
that occur physically in the sample. Therefore, the predicted stiffness and failure from the finite
element simulation more closely resemble the behavior of to the experiments compared to the
calculated behavior from the single unit cell.

Outside the deformation mechanisms, the calculated stress-strain behavior of the coated and
uncoated fabric simulations are almost identical. This is to be expected as shown previously
from the sensitivity analysis of the unit cell, experimental observations and the lack of shear
deformation in the type of loading.

85
Chapter 5 Validation

70.00
Exp-Coated-warp
60.00 Exp-Coated-fill
Analytical UC-Coated-warp
membrane stress (N//mm)
50.00 Analytical UC-Coated-fill
FE UC-Coated-warp
40.00 FE UC-Coated-fill

30.00

20.00

10.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
strain

Figure 5-14: Uniaxial extension of coated airbag fabric - model vs. experiments

70.00
Exp-uncoated-warp
60.00 Exp-uncoated-fill
membrane stress (N//mm)

50.00 Analytical UC-uncoated-warp

Analytical UC-uncoated-fill
40.00
FE UC-Uncoated-warp
30.00 FE UC-Uncoated-fill

20.00

10.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
strain
Figure 5-15: Uniaxial extension of uncoated airbag fabric - model vs. experiment

86
Chapter 5 Validation

a)

b)

Figure 5-16: Macroscopic strains of simulated uniaxial loaded coated airbag fabric at 30%
extension a) warp direction and b) fill direction

87
Chapter 5 Validation

a)

b)

Figure 5-17: Yarn strains of simulated uniaxial loaded coated airbag fabric at 30%
extension a) warp direction and b) fill direction

5.3.2 Biaxial results

Figure 5-18 and Figure 5-19 show the stress-strain results of the simple unit cell model,
simulated biaxial cruciform test and the experimental cruciform test. Both the simple model and
finite element model were run until reaching their calculated ultimate load of the fabric. The
simulated cruciform sample is a quarter model of the actual test to keep computational time low.
Overall, the deformational behavior of the single unit cell and simulated model closely resemble
the global stress-strain characteristics seen in the experiments.

88
Chapter 5 Validation

60.00
Exp-Coated-warp
50.00 Exp-Coated-fill
membrane stress (N/mm)

UC Model-warp
40.00
UC Model-fill
30.00
FE Model-warp

20.00 FE Model-fill

10.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25
strain
Figure 5-18: Biaxial extension of coated airbag fabric - model vs. ex periments

60.00
Exp-uncoated-warp
50.00 Exp-uncoated-fill
membrane stress (N/mm)

UC Model-warp
40.00
UC Model-fill
30.00
FE Model-Warp

20.00 FE Model-Fill

10.00

0.00
0.00 0.05 0.10 0.15 0.20 0.25
strain
Figure 5-19: Biaxial extension of uncoated airbag fabric - model vs. ex periments

89
Chapter 5 Validation

Table 5-2 shows the simulated deformation of the sample compared to photographs taken during
the biaxial experiments. Likewise, Table 5-3 shows a similar comparison for the uncoated
fabric. Qualitatively, the simulations capture the deformational nuances of the corners of the
specimens, specifically the slight curvature that is developed. On the other hand, there are
visible differences between the uncoated simulation and the experiment. The uncoated sample in
the experiment exhibits sliding and pullout of the yarns; thereby giving the edges of the sample a
straight rather than curved appearance seen in the simulation. However, this discrepancy does
not greatly affect the global force-displacement relationship of the system as evident in the
calculated stress-strain curves.

While the single unit cell model and finite element material formulations have very good
agreement on the stress-strain behavior, the failure predictions by the two methods vary.
Notably, the finite element simulation of the coated fabric adequately predicts the point of failure
seen experimentally for coated fabric (around 12%). The single unit cell model greatly over-
predicts the failure point of the coated fabric, understandably as the edge effects and stress
concentrations are not considered. The failure of the simple unit cell model presents the failure
of a fabric under homogeneous biaxial displacement. Much like the uniaxial sample, the biaxial
sample has edge effects that cause heterogeneous displacement condition.

The initiation of failure in the simulation occurs at the corner of the specimen where yarns strains
are the highest. Upon the first onset of yarn failure, propagation of failure to neighboring yarns
occurs suddenly causing the simulation to become unstable. In the experiments, a similar
phenomena is exhibited -- failure typically initiating at the corner of the sample and rapidly
propagating across the cruciform arm. Figure 5-20 (a) and (b) shows the warp and fill yarn
strains for the coated failure at 12% biaxial strain respectively. From the figures, it can be seen
that the fill yarn near the corner of the specimen approaches its ultimate strain. It should be
noted that modeling the cruciform shape used in this study within the finite element framework
is quite difficult. While the global stress-strain behavior is believed to be correct, the local
behavior of the cruciform corner is a singularity point where calculating the correct magnitude of
stress becomes a very challenging task. Refining the mesh past dimensions that are smaller than
the dimensions of the unit cell results in earlier failures at the corner as one would expect. One
could bevel the corner to elevate the effect of the singularity, but this condition would also

90
Chapter 5 Validation

neglect the physical system. Therefore, the failure predictions of the biaxial simulation may
require further technique and investigation. Nonetheless, we will address possible weaknesses in
the user material model in predicting failure of biaxial samples assuming the singularity problem
is less evident in actual applications.

Table 5-2: Biaxial extension of coated fabric experiment vs. simulation

Strain Experiment Simulation

0%

~5%

~12%

91
Chapter 5 Validation

Table 5-3: Biaxial extension of uncoated fabric experiment vs. simulation

Strain Experiment Simulation

0%

~5%

~10%

92
Chapter 5 Validation

(a

(b
Figure 5-20: Simulated yarn strains at approximately 12% equal biaxial extension in the-
a) warp direction and b) fill direction

Interestingly, the failure predictions of the uncoated fabric by the simple unit cell and finite
element material model deviate greatly from what is seen in the experiments. The simple unit

93
Chapter 5 Validation

cell model over-predicts the failure point while the continuum material model under-predicts the
failure of the uncoated fabric. The shortcomings of predicting failure of the uncoated fabric
samples are apparent when considering the assumptions that are the basis for each model. The
simple unit cell model represents an ideal fabric. The failure stress predicted by the simple model
represents the failure of an infinitely long fabric sample under biaxial extension where the yarns
remain orthogonal with no edge effects or sliding and friction. On the other hand, the material
model is based on continuum assumptions that allow for non-orthogonal yarn configurations and
considers no sliding between the yarns. Henceforth, the actual fabric sample can be descritized
to consider the stress concentration from edge corners. This continuum assumption seems to be
valid for the coated fabric system where the coating reduces slip thus ensuring a more continuum
like structure. Additionally, as mentioned previously the yarn failure strain criteria is based on
the ex-situ response. Certainly the confining pressure of neighboring yarns and filament friction
are aspects not considered within the model that could potentially affect the failure criteria.

5.3.3 Bias results

The bias extension is a test to examine the shear behavior of the fabric. Due to the high
deformational heterogeneity, it is not possible to model the test using the simple unit cell model.
Figure 5-21 shows the simulated and average experimental results of the force-elongation
relation for the bias sample. A good correlation is seen between simulation and experiments for
both coated and uncoated samples. At higher extensions, around 35 mm, the response of the
model for both coated and uncoated fabrics becomes slightly stiffer than the experiments. While
the global force-displacement seems reasonable, examining shear deformation in detail suggests
weaknesses in the model. Figure 5-22 shows the applied stress versus the measured shear strain
at the center of the specimen. At low shear strains there is reasonable agreement between the
simulation and experiments. Figure 5-23 gives a more detailed view of the low level stress-
strain. However, at larger center shear strains the simulations of the coated and uncoated
samples deviate from the experiment, even though the global stiffness of the simulation and
experiments were found to be similar. Table 5-4 and Table 5-5 show the deformed sample
against the simulation at various extensions for the coated and uncoated fabric, respectively.
From the tables it can be seen that the simulation accurately captures the distinct deformation
zones seen in the bias experiments.

94
Chapter 5 Validation

35.00
Exp - Coated
30.00 Exp - Uncoated
FE - Coated
FE - Uncoated
25.00
Force (kgf)

20.00

15.00

10.00

5.00

0.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00
Extension (mm)
Figure 5-21: Bias load-elongation for coated and uncoated fabric - simulation vs.
experiment

16.00
Exp-Uncoated
14.00 FE-Uncoated
Applied shear stress (N/mm)

Exp-Coated
12.00
FE-Coated
10.00

8.00

6.00

4.00

2.00

0.00
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
shear strain (degrees)
Figure 5-22: Applied shear stress vs. measured shear angle - simulation and experiments

95
Chapter 5 Validation

1.60
Exp-Uncoated
1.40 FE-Uncoated
Exp-Coated
Applied shear stress (N/mm)

1.20 FE-Coated

1.00

0.80

0.60

0.40

0.20

0.00
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0
shear strain (degrees)

Figure 5-23: Detailed low level applied shear stress vs. measured shear angle - simulation
and experiments

The agreement between the simulation and experiments at low shear levels shown in Figure
5-23 is encouraging since the shear model at those regions is calibrated using experimental data.
At large shear deformations as shown in Figure 5-22, the approximation of the locking stress
seems to have an effect on the shear deformation. The locking shear does not consider possible
yarn deformations that have been documented in the literature such as inter-filament friction,
yarn torque, rapping of the yarns at the crossover and changes in fiber volume fraction in the
yarn (Grosberg et al., 1968; McBride & Chen, 1997; Skelton, 1976). Therefore, it can be argued
that the shortcomings in the simulations do not lie in the shear model but in the calibration
techniques. A possible and more suitable technique for calibrating the shear behavior of the unit
cell that would include both low level shear and the full locking shear effect is by testing the
fabric using a picture frame set-up as discussed in Chapter 3.

96
Chapter 5 Validation

Table 5-4: Comparison of bias deformation of coated sample: experiment vs. simulation

Extension Experiment Simulation


0 mm

10 mm

20 mm

30 mm

97
Chapter 5 Validation

Table 5-5: Comparison of bias deformation of uncoated sample: experiment vs. simulation

Extension Experiment Simulation


0 mm

10 mm

20 mm

30 mm

98
Chapter 5 Validation

5.4 Summary

Considering the lack of experimental data for the mechanical behavior of 350 dtex airbag fabric,
a series of three experiments were conducted to validate the model: uniaxial extension, biaxial
extension and bias extension. Of the three experiments, the biaxial extension most closely
resembles the condition seen in the inflated airbag. Overall, the deformational behavior of the
simulated fabric closely resembles that measured in the experiments. Additionally, the single
unit cell model incorporated in MATLAB can be particularly useful to quickly evaluate different
plain weave fabric parameters for their mechanical properties under simple deformations.

99
Chapter 6- Conclusion and recommendations

Chapter 6 - Conclusion and recommendations

6.1 Conclusions

The importance of accurately simulating the mechanical behavior of airbag fabric in the
structural analysis of the airbag and the shortcomings of the current methods has been
established in the previous literature. The proposed approach can closely predict the
macroscopic nonlinear mechanical behavior of airbag fabric using the constituent properties
under a variety of deformations. Additionally, not only does the approach successfully simulate
the macroscopic behavior of the fabric, it can also approximate deformations of the constituents
at the structural level such as crimp interchange and yarn strain all at vary small computational
cost. This is a significant improvement on the current constitutive definitions for airbag fabric.

The basis of the constituent based approach is the unit cell model the simplified approximation
of the plain weave fabric structure. The structural mechanisms that contribute to the
macroscopic load-elongation behavior that arises from deformation of the constituents are
carefully defined. The characterization of the constituent properties for the model is fairly
simple: geometric spacing and crimp; yarn load-elongation behavior; and in-plane shear
behavior of the fabric.

After implementation into the explicit finite element code, LS-DYNA, the unit cell model was
used to simulate three quasi-static tests. The first of the simulated test was a uniaxial extension
of a fabric sample a simple and popular test to quickly evaluate a fabric performance. Good
correlation of deformational behavior and failure was seen between the model and experiments.
To truly understand the models use in the airbag application, a biaxial extension test of the fabric
was conducted using a customized cruciform test set-up. Again, there was good agreement
between the simulated and recorded stress-strain behavior of the fabric. Finally, a bias extension
test, where a sample is loaded with the yarns oriented at 45 degrees from the loaded direction

100
Chapter 6- Conclusion and recommendations

was performed to evaluate the models ability to represent the shear behavior of the fabric. At
low levels of shear deformation, the model results closely resemble the experiments. However,
at higher shear strains while the global load-elongation of the model is consistent with
experiments, the measured shear at the center of the sample deviates from what is calculated in
the simulation.

6.2 Recommendations and future work

There are many parameters that can potentially affect the airbags performance which have not
been explored in the literature. Certain yarn structural mechanisms such as yarn bending are of
importance in quasi-static simulations but could be less influential on the airbag dynamics during
the rapid dynamic simulation. Outside of structural mechanisms, assumptions neglecting the
out-of-plane effects of the fabric should also be evaluated, particularly regarding any impact of
simulated unfolding of the airbag.

The environments that the airbag fabric experience can vary greatly during its lifetime which can
possibly affect its overall performance. Before deployment, the fabric and its constituents can
undergo the degradation effects from the environment such as temperature and humidity
changes. During deployment, the fabric undergoes high strain rates and can experience
temperatures up to 600 C, all of which are certainly different than the conditions measured in
the experiments. All these scenarios are not currently implemented in the current model. If
further experimentation shows sensitivity to these effects, additional unit cell mechanisms can be
added to the model to enhance its capabilities.

The other potential improvement of the proposed model is that it is purely elastic. Due to the
discrete and free movement nature of the constituents, fabrics exhibit hysteresis. Of course this
hysteresis acts as an energy dissipation agent. In the occupant crash simulations, documenting
the total energy dissipation between the impact of the occupant and bag are essential in
evaluating the likelihood of injury. The exact or estimated percentage of the fabric hysteresis
contribution to the total energy dissipation of the system is not documented in this study. Further
experiments and simulations should be conducted regarding this subject. Outside of energy

101
Chapter 6- Conclusion and recommendations

dissipation, quantifying yarn sliding seems to be important in predicting fabric failure under
complex modes of deformation.

6.3 Summary

Overall, considering the current methods used by airbag designers, the current approach can be a
powerful tool to better understand how the nonlinear behavior of the fabric can affect designs
and passenger safety. Not only is the nonlinear behavior of the fabric accurately represented, but
details into crimp interchange and yarn strain are implicitly computed providing a wealth of
information at low computational costs. While the studied application was intended for
automotive airbags, the developed approach and methodology can be applied to a variety of
textile structures such as architectural fabric roofs, ballistic armor, inflatable structures,
parachutes and the like.

102
References

References

Alley, V. L., & Faison, R. W. (1972). Decelerator fabric constants required by the generalized
form of Hookes Law. Journal of Aircraft, 9(3), 211-216.

ASTM D3883-04. (2008). Standard test method for yarn crimp and yarn take-up in woven
fabrics. West Conshohocken, PA: ASTM International.

Bassett, R. J., Postle, R., & Pan, N. (1999). Experimental methods for measuring fabric
mechanical properties: A review and analysis. Textile Research Journal, 69(11), 866-875.

Bathe, K. (1982). Finite element procedures in engineering analysis. Englewood Cliffs, N.J.:
Prentice-Hall.

Bedewi, N. E., Marzougui, D., & Motevalli, V. (1996). Evaluation of parameters affecting
simulation of airbag deployment and interaction with occupants. International Journal of
Crashworthiness, 1(4), 339.

Behera, B. K., & Goyal, Y. (2009). Artificial neural network system for the design of airbag
fabrics. Journal of Industrial Textiles, 39(1), 45-55.

Brueggert, M., & Tanov, R. R. (2002). An LS-DYNA user defined material model for loosely
woven fabric with non-orthogonal varying weft and warp angle. Proceedings of the 7th
International LS-DYNA Users Conference, 8-01-8-13.

Crouch, E. T. (1993). Evolution of airbag components and materials. Worldwide Passenger Car
Conference and Exposition, Dearborn, Michigan.

Crouch, E. T. (1994). Evolution of coated fabrics for automotive airbags. Journal of Industrial
Textiles, 23(3), 202-220.

Drnhoff, H., Baldauf, H., Beesten, B., Hirth, A., Luijkx, R., & Remensperger, R. (2008).
Relevance of fabric modeling for simulation of airbag deployment. Airbag 2008, Karlsruhe,
Germany.

Duan, Y., Keefe, M., Bogetti, T. A., & Cheeseman, B. A. (2005). Modeling friction effects on
the ballistic impact behavior of a single-ply high-strength fabric. International Journal of
Impact Engineering, 31(8), 996-1012.

Fallon, I., & ONeill, D. (2005). The world's first automobile fatality. Accident Analysis &
Prevention, 37(4), 601-603.

Farboodmanesh, S. (2003). Parametric studies of coated fabric shear behavior: Fabric


construction and coating penetration effects. (Master of Science, University of
Massachusetts - Lowell).

103
References

Farboodmanesh, S., Chen, J., Mead, J. L., White, K. D., Yesilalan, H. E., Laoulache, R., et al.
(2005). Effect of coating thickness and penetration on shear behavior of coated fabrics.
Journal of Elastomers and Plastics, 37(3), 197-227.

Freeston, W. D., Platt, M. M., & Schoppee, M. M. (1967). Mechanics of elastic performance of
textile materials. Textile Research Journal, 37(11), 948-975.

Freeston, W. D., & Schoppee, M. M. (1975). Geometry of bent continuous-filament Yarns1.


Textile Research Journal, 45(12), 835-852.

Gon, D. (2010). Air bags for automobiles - A textile challange. Textile Asia, (April), 25-29.

Grosberg, P. (1966). The mechanical properties of woven fabrics part II: The bending of woven
fabrics. Textile Research Journal, 36, 205-211.

Grosberg, P., Leaf, G. A. V., & Park, B. J. (1968). The mechanical properties of woven fabrics:
Part VI: The elastic shear modulus of plain-weave fabrics. Textile Research Journal, 38(11),
1085-1100.

Grosberg, P., & Park, B. J. (1966). The mechanical properties of woven fabrics: Part V: The
initial modulus and the frictional restraint in shearing of plain weave fabrics. Textile
Research Journal, 36(5), 420-431.

Haas, R. D., A. (1918). The stretching of the fabric and the deformation of the envelope in
nonrigid balloons No. 16)NACA - National Advisory Committee for Aeronautics.

Hallquist, J. O. (2006). LS-DYNA keyword user's manual : Version 971. Livermore, Calif.:
Livermore Software Technology Corp.

Harris, M. (1954). Handbook of textile fibers. Washington: Washington : Harris Research


Laboratories.

Harrison, P., Clifford, M. J., & Long, A. C. (2004). Shear characterisation of viscous woven
textile composites: A comparison between picture frame and bias extension experiments.
Composites Science and Technology, 64(10-11), 1453-1465.

Hearle, J. W. S. (1969). In Backer S. (Ed.), Structural mechanics of fibers, yarns, and fabrics.
New York: New York : Wiley-Interscience.

Hearle, J. W. S. (1989). Mechanics of yarns and nonwoven fabrics. In T. W. Chou, & F. K. Ko


(Eds.), Textile structural composites (pp. 27-64). Amsterdam: Elsevier Science Publishers.

Hirth, A., Haufe, A., & Olovsson, L. (2007). Airbag simulation with LS-DYNA: Past, present
and future. LS-DYNA Anwenderforum, Frankenthal.

Holzapfel, G. A. (2000). Nonlinear solid mechanics : A continuum approach for engineering.


Chichester; New York: Wiley.

104
References

Hong, S. (2003). A study on the modeling technique of airbag cushion fabric. New York, NY,
ETATS-UNIS: Society of Automotive Engineers.

Ivanov, I., & Tabiei, A. (2004). Loosely woven fabric model with viscoelastic crimped fibres for
ballistic impact simulations. International Journal for Numerical Methods in Engineering,
61(10), 1565-1583.

Kawabata, S. (1989). Nonlinear mechanics of woven and knitted materials. In T. W. Chou, & F.
K. Ko (Eds.), Textile structural composites (pp. 67-116). Amsterdam: Elsevier Science
Publishers.

Kawabata, S., Niwa, M., & Kawai, H. (1973a). The finite-deformation theory of plain-weave
fabrics. part I: The biaxial-deformation theory. Journal of the Textile Institute, 64(1), 21.

Kawabata, S., Niwa, M., & Kawai, H. (1973b). The finite-deformation theory of plain-weave
fabrics. part II: The uniaxial-deformation theory. Journal of the Textile Institute, 64(2), 47.

Kawabata, S., Niwa, M., & Kawai, H. (1973c). The finite-deformation theory of plain-weave
fabrics. part III: The shear-deformation theory. Journal of the Textile Institute, 64(2), 62.

Kemp, A. (1958). An extension of Peirce's cloth geometry to the treatment of non-circular


threads. Journal of the Textile Institute Transactions, 49(1), 44.

Keshavaraj, R., Tock, R. W., & Haycook, D. (1996). Airbag fabric material modeling of nylon
and polyester fabrics using a very simple neural network architecture. Journal of Applied
Polymer Science, 60(13), 2329-2338.

Keshavaraj, R., Tock, R. W., & Nusholtz, G. S. (1995). Comparison of contributions to energy
dissipation produced with safety airbags. New York, NY, ETATS-UNIS: Society of
Automotive Engineers.

Keshavaraj, R., Tock, R. W., & Nusholtz, G. S. (1996). A realistic comparison of biaxial
performance of nylon 6,6 and nylon 6 fabrics used in passive restraints - airbags. Journal of
Applied Polymer Science, 61(9), 1541-1552.

Khalil, T., & Du Bois, P. (2004). Finite element analytical techniques and applications to
structural design. In P. Prasad, & J. Belwafa (Eds.), VEHICLE CRASHWORTHINESS AND
OCCUPANT PROTECTION (pp. 111-158). Southfield, Michigan: American Iron and Steel
Institute.

Khan, M. U., & Moatamedi, M. (2008). A review of airbag test and analysis. International
Journal of Crashworthiness, 13(1), 67.

Kilby, W. F. (1963). 2Planar stress-strain relationships in woven fabric. Journal of the Textile
Institute Transactions, 54(1), 9.

105
References

King, M. J., Jearanaisilawong, P., & Socrate, S. (2005). A continuum constitutive model for the
mechanical behavior of woven fabrics. International Journal of Solids and Structures,
42(13), 3867-3896.

Ko, F. K. (December 2006). Characterization of the mechanical properties of airbag fabrics.


Drexel University: External Report for TRW (Unpublished)

Lomov, S. V., Gusakov, A. V., Huysmans, G., Prodromou, A., & Verpoest, I. (2000). Textile
geometry preprocessor for meso-mechanical models of woven composites. Composites
Science and Technology, 60(11), 2083-2095.

Lomov, S. V., Huysmans, G., Luo, Y., Parnas, R. S., Prodromou, A., Verpoest, I., et al. (2001).
Textile composites: Modelling strategies. Composites Part A: Applied Science and
Manufacturing, 32(10), 1379-1394.

Lomov, S. V., & Verpoest, I. (2000). Compression of woven reinforcements: A mathematical


model. Journal of Reinforced Plastics and Composites, 19(16), 1329-1350.

McBride, T. M., & Chen, J. (1997). Unit-cell geometry in plain-weave fabrics during shear
deformations. Composites Science and Technology, 57(3), 345-351.

Mukhopadhyay, S. (2008). Technical developments and market trends of automotive airbags.


Textile advances in the automotive industry CRC Press.

National Highway Traffic Safety Administration. (2009). Special crash investigations - counts of
frontal air bag related fatalities and seriously injured persons No. DOT HS 811 104).
Washington, DC: United State Department of Transportation.

National Highway Traffic Safety Administration. (2010). Fatality analysis reporting system
encyclopedia. Retrieved 10/10, 2010, from http://www
fars.nhtsa.dot.gov/People/PeopleAllVictims.aspx

Page, J., & Wang, J. (2000). Prediction of shear force and an analysis of yarn slippage for a
plain-weave carbon fabric in a bias extension state. Composites Science and Technology,
60(7), 977-986.

Pan, N. (1996). Analysis of woven fabric strengths: Prediction of fabric strength under uniaxial
and biaxial extensions. Composites Science and Technology, 56(3), 311-327.

Peirce, F. T. (1937). The geometry of cloth structures. Journal of the Textile Institute
Transactions, 28(3), 45.

Peng, X. Q., & Cao, J. (2005). A continuum mechanics-based non-orthogonal constitutive model
for woven composite fabrics. Composites Part A: Applied Science and Manufacturing,
36(6), 859-874.

Platt, M. M., Klein, W. G., & Hamburger, W. J. (1959). Mechanics of elastic performance of
textile materials. Textile Research Journal, 29(8), 611-627.

106
References

Press, W. H. (1992). Numerical recipes in C : The art of scientific computing. Cambridge; New
York: Cambridge University Press.

Rasband, W. S. (1997-2009). ImageJ. Bethesda, Maryland, USA: U. S. National Institutes of


Health.

Rohr, I., Harwick, W., & Nahme, H. (2004). Ermittlung des festigkeits- und
schdigungsverhaltens von airbaggewebe bei verschiedenen belastungszustnden und
dehnraten. Materialwissenschaft Und Werkstofftechnik, 35(9), 574-577.

Rohr, I., Harwick, W., & Nahme, H. (2005). Der biaxiale kreuzzugversuch zur ermittlung von
werkstoffkennwerten von airbaggeweben am beispiel von polyamid 6.6.
Materialwissenschaft Und Werkstofftechnik, 36(5), 195-197.

Schwark, J., & Muller, J. (1996). High performance silicone-coated textiles: Developments and
applications. Journal of Industrial Textiles, 26(1), 65-77.

Schwer, L. E. (2002). ASME standards committee on verification and validation in


computational solid mechanics, #293. Proceedings- The International Society for Optical
Engineering, 1(4753), 670.

Shahkarami, A. (2006). An efficient unit cell based numerical model for continuum
representation of fabric systems. Unpublished PhD, The University of British Columbia,

Shahkarami, A., & Vaziri, R. (2006). An efficient shell element based approach to modelling the
impact response of fabrics. 9th Int. LS-DYNA Users Conference, Dearborn, Michigan. 12.

Shahkarami, A., & Vaziri, R. (2007). A continuum shell finite element model for impact
simulation of woven fabrics. International Journal of Impact Engineering, 34(1), 104-119.

Shahpurwala, A. A., & Schwartz, P. (1989). Modeling woven fabric tensile strength using
statistical bundle theory. Textile Research Journal, 59(1), 26-32.

Shanahan, W. J., & Hearle, J. W. S. (1978). An energy method for calculations in fabric
mechanics part II: Examples of application of the method to woven fabrics. Journal of the
Textile Institute, 69(4), 92.

Shockey, D. A., Erlich, D. C., & Simons, J. W. (2000). Improved barriers to turbine engine
fragments No. U.S. Department of Transportation). Menlo Park, California: SRI
International.

Skelton, J. (1976). Fundamentals of fabric shear. Textile Research Journal, 46, 862-869.

Spivak, S. M., & Treloar, L. R. G. (1968). The behavior of fabrics in shear: Part III: The relation
between bias extension and simple shear. Textile Research Journal, 38(9), 963-971.

Stubbs, N., & Thomas, S. (1984). A nonlinear elastic constitutive model for coated fabrics.
Mechanics of Materials, 3(2), 157-168.

107
References

Sun, H., & Pan, N. (2005). Shear deformation analysis for woven fabrics. Composite Structures,
67(3), 317-322.

Tanov, R. R., & Brueggert, M. (2003). Finite element modeling of non-orthogonal loosely
woven fabrics in advanced occupant restraint systems. Finite Elements in Analysis and
Design, 39(5-6), 357-367.

Wartman, J., Harmanos, D., & Ibanez, P. (2005). Development of a versatile device for
measuring the tensile properties of geosynthetics. , 161(40782) 28.

Wawa, C. J., Chandra, J. S., & Verma, M. K. (1993). Implementation and validation of a finite
element approach to simulate occupant crashes with airbags: Part I - airbag model. ASME
Applied Mechanics Division -Publications- AMD, 169, 269.

108
Appendix A UMAT *MAT card documentation

Appendix A UMAT *MAT card documentation


*MAT_USER_DEFINED_MATERIAL_MODELS

User defined material 48. This model is the Continuum Plain Weave Fabric Model for shell elements with one
through-thickness integration point (membrane) developed at The University of British Columbia. This model
employs a mechanistic unit cell approach and non-orthogonal formulation capable of representing the constitutive
behavior of plain weave fabrics. The unit cell formulation is based on simple yarn constituent properties and can
employ both linear and nonlinear yarn extensional behavior. Additional structural mechanisms include transverse
yarn compression, yarn bending rigidity and coating extension. A four region shear stress-shear input parameter is
capable of fitting the shear behavior of both uncoated and coated fabrics. The unit geometry can be approximated
by simple linear or sinusoidal approximation of the yarn crimp. The routine includes an instantaneous ultimate
strain based yarn failure criterion which removes failed yarns from the unit cell and invokes element erosion when
both warp and fill yarns have surpassed their capacity. The current implementation is a scalar routine.

Card 1 1 2 3 4 5 6 7 8

Variable MID RO MT LMC NHV IOTHRO IBULK IG

Type I F I I I I I I

Default none none 48* 38* 40* 0* 1* 2*

Card 2 1 2 3 4 5 6 7 8

Variable IVECT IFAIL ITHERM IHYPER IEOS

Type I I I I I

Default 0* 1 0 0 0

*Note: These values are not necessarily defaults. These variables must set as indicated to ensure proper operation
of this user material model.

Card 3 1 2 3 4 5 6 7 8

Variable K G SHTH thw0 thf0 niter TOL geoflag

Type F F F F F I F I

Default none none none 0 90 none none 0

109
Appendix A UMAT *MAT card documentation

Card 4 1 2 3 4 5 6 7 8

Variable H01 n1 cr1 ey1u EIy1 den1 yro1 lcd1

Type F F F F F F F I

Default 0 none none none none none none none

Card 5 1 2 3 4 5 6 7 8

Variable H02 n2 cr2 ey2u EIy2 den2 yro2 lcd2

Type F F F F F F F I

Default 0 none none none none none none none

Card 6 1 2 3 4 5 6 7 8

Variable gm1 G1 gm2 G2 gm3 G3 gm4 ETOL

Type F F F F F F F F

Default none none none none none none none 1.1

Card 7 1 2 3 4 5 6 7 8

Variable fda fdb stfnr Ec Nu_c tc

Type F F F F F F

Default none none none none none none

VARIABLE DESCRIPTION
MID Material Identification. A unique number must be chosen.
RO Mass density.
MT User material type. To specify the fabric model, 48 must be specified.
LMC Length of the material constant array. For the fabric model, 38 must be specified.
NHV Number of history variables. For the fabric model, 40 must be specified.

110
Appendix A UMAT *MAT card documentation

VARIABLE DESCRIPTION
IORTHO Set to 0 for the fabric model
IBULK Address of the bulk modulus in the material constants array. Set to 1.
IG Address of the shear modulus in the material constants array. Set to 2.
IVECT Vectorization flag. For the fabric code, set to 0.
IFAIL To allow element erosion, set IFAIL=1.
ITHERM For the fabric code, set to 0.
IHYPER For the fabric code, set to 0.
IEOS For the fabric code, set to 0.
K Bulk Modulus. See Remark 1.
G Shear Modulus. See Remark 2.
SHTH Shell Thickness.
thw0 Initial Warp Angle.
thf0 Initial Fill Angle.
niter Maximum number iterations for the unit cell calculation.
TOL Tolerance of the unit cell solution. See Remark3.
geoflag Unit cell geometry flag:
EQ 0.0: Linear
EQ 1.0: Sinusoidal
H01 Initial amplitude of the sine function representing the warp yarn. Set to 0.0 for linear as the
code internally determines the initial yarn height.
n1 Warp yarns per inch.
cr1 Warp yarn crimp.
ey1u Ultimate strain of warp yarn.
EIy1 Warp yarn bending rigidity. See Remark 4.
den1 Warp yarn linear density in denier.
yro1 Warp yarn material density.
lcd1 Load curve ID to specify warp yarn stress-strain curve.
H02 Initial amplitude of the sine function representing the fill yarn. Set to 0.0 for linear as the
code internally determines the initial yarn height.
n2 Fill yarns per inch.
cr2 Fill yarn crimp.
ey2u Ultimate strain of fill yarn.
EIy2 Fill yarn bending rigidity. See Remark 4.
den2 Fill yarn linear density in denier.
yro2 Fill yarn material density.
lcd2 Load curve ID to specify fill yarn stress-strain curve.
gm1 Shear strain designating a transition zone. See Remark 5.
G1 Initial Shear modulus.
gm2 Shear strain designating a transition zone. See note 5.
G2 Intermediate Shear modulus
gm3 Shear strain designating a transition zone. See Remark 5.
G3 Maximum Shear modulus.
gm4 Shear strain designating a transition zone. See Remark 5.
ETOL Value of erosion shear for highly deformed elements to maintain stability. Recommended
value of 2.00.
stnfr Factor in calculating minimum contact force. Typically set to 1.1.
fda Maximum transverse compression parameter (mm).
fdb Intra-ply contact stiffness ((N-mm)(1/3) ).
Ec Modulus of elasticity of coating

111
Appendix A UMAT *MAT card documentation

VARIABLE DESCRIPTION
Nu_c Poissons ratio of coating
tc Coating thickness

Remarks:

1. The suggested units for analysis are g , mm, ms, N, MPa. Internally, the conversion of the yarns per inch
will formulate the unit cell geometry in mm.

2. Bulk Modulus is used to determine stable time step and is required for determining transmitting boundaries,
contact interfaces, rigid body constraints. Within mechanistic unit cell empowered model, the following
equations can be used to calculate the correct bulk modulus that accounts for the crimp of the yarn.

2c*
K=
3(1 2 )

where is density of the fiber, is the Poissons ratio of the yarn material and

c
c* =
1 + cr
where c is the speed of sound in the fiber material and cr is the crimp of the yarn.

With respect to best practices, for unbalanced fabrics, the yarn crimp that results in the larger values of K
governs.

3. Shear Modulus is used to determine stable time step and is required for determining transmitting
boundaries, contact interfaces, rigid body constraints. Within mechanistic unit cell empowered model, the
following expression can be used to calculate the correct shear modulus that accounts for the crimp of the
yarn:

2c*
G=
2(1 + )
where . and c* are defined in remark 1.

4. Tolerance of the unit cell Newton-Raphson scheme to reach a solution. A smaller tolerance has a trade off
of longer computational time.

112
Appendix A UMAT *MAT card documentation

5. The theoretical bending rigidity of low-twist, non-blended yarns is bounded by two values. The lower
bound assumes each individual fiber has complete freedom of motion frictionless matter and is defined as:
EI yarn = nEI fiber

where n is the number of fiber within the yarn and EIfiber is the bending rigidity of a single fiber.

The upper bound assumes the fibers have no freedom to move and act like a complete cluster bonded by
friction. Therefore the yarn bending rigidity can computed using the number of fibers in the yarn divided
by the yarn packing factor times the rigidity of the lower bound. This yields the expression:

(EI ) yarn = E fiber I yarn


where Nfiber is the number of fibers in the yarn and Iyarn is the second moment of inertia using the cross-
sectional geometry of the yarn.

For greater discussion on the theoretical bending rigidity of yarn, determination of values within the
bounds, determination of bending rigidity of blended yarns and determination of bending rigidity of highly
twisted yarns the paper by Platt, Klein and Hamburgaer (1959) is highly suggested.

6. Due to the difficulties of evaluating fabric shear using the fabrics constituents, empirical data from the
fabric is warranted to empower the shear behavior of the proposed unit cell method. This empirical data
can be obtained either through picture frame test where the fabric under goes pure shear; rail shear test
using the Kawabata Evaluation System (KES) for fabrics; or could potentially be numerically calibrated
through a full 3D finite element model of the fabric and its constituents. Each method has advantages and
disadvantages. Picture frame requires special fixture and the design of the fixture must be carefully done to
ensure reproducible results. Kawabata Evaluation System requires specialized equipment and some
additional calculation to remove the effects of tension and to obtain pure shear behavior. The use of 3D
finite element models can be used to evaluate the pure shear behavior of virtually design fabrics but
capturing the correct boundary conditions can be challenging.

Regardless the approach to obtain the shear data, the following method can be used to represent the shear
behavior of the unit cell. A spline fit of the shear modulus as a function of shear strain illustrated in Figure
2A can be used to describe the shear stress-strain behavior based of typical behavior seen in Figure 1A.

Single-Coated Fabric
Uncoated Fabric

Coated Transition Zone


Shear Stress

Locking Transition
Zone

12 3 4

Shear Strain
Figure A1: Shear model behavior regions in the shear stress-strain curve

113
Appendix A UMAT *MAT card documentation

G3

Shear Modulus
G1coated

G2
G1 uncoated

1 2 3 4

Shear Strain
Figure A2: Secant shear modulus as a function of strain

For coated fabrics, the fabric-coating shear interaction modulus, G1 needs to be defined as well as the
extent of the transition zone which is bound by 1 and 2. For uncoated fabrics, these values can be set
equal to zero. The pre-locking shear stiffness of the fabric, G2 is the same for both uncoated and coated
fabrics of the same construction. As the fabric continues to deform in shear, the amount of rotation at the
crossovers reach a geometric limit and begin to lock. This is a gradual process controlled my friction, the
current packing state of the fibers within the yarn and geometric features of the fabric. Within the spline
model, this process is bound by 3 and 4 which are the same for both the coated and uncoated fabric.
Likewise the locking shear modulus, G3, is the same for uncoated and coated fabrics.

History Variables:

The fabric continuum model will return many more variables than normal material models. These
variables include the stretch of each yarn vector, yarn strain, unit cell geometric parameters, yarn failure
flags, number of iterations, solution error etc. To access these history variables, be sure to set the NEIPS
(shells) variables in the *DATABASE_EXTENT_ BINARY card to 40. The following table shows the
history variables available and their location in the history variable list:

History Model
Variable Variable Description
1 strchw Stretch of warp yarn vector
2 strchf Stretch of fill yarn vector
3 thw Angle of the warp yarn vector
4 thf Angle of the fill yarn vector
5 fail1 Warp yarn failure flag (0=no fail, 1=failed)
6 fail2 Fill yarn failure flag (0=no fail, 1=failed)
7 ey1 Warp yarn strain
8 ey2 Fill yarn strain

114
Appendix A UMAT *MAT card documentation

History Model
Variable Variable Description

9 sigw Un-rotated fabric stress - warp


10 sigf Un-rotated fabric stress - fill
11 gmw Shear strain - warp
12 gmf Shear strain - fill
13 R Newton-Raphson scheme residual error
14 iter Number of iterations of the unit cell N-R scheme
15 iterMXd Counter for number of times the maximum of
iterations was exceeded throughout the analysis
16 Ten1 Warp tensile force
17 Ten2 Fill tensile force
18 Fc1 Contact force from warp yarn
19 Fc2 Contact force from fill yarn
20 Fb1 Bending force from warp yarn
21 Fb2 Bending force from fill yarn
22 H1 Height of warp yarn in the unit cell
23 H2 Height of fill yarn in the unit cell
24 d1 Warp yarn displacement
25 d2 Fill yarn displacement
26 dcomp Total transverse compression within the unit cell
27-30 blank -
31 g11 Component of the displacement tensor
32 g12 Component of the displacement tensor
33 g21 Component of the displacement tensor
34 g22 Component of the displacement tensor
35 sy1 Warp yarn stress (from load curve)
36 sy2 Fill yarn stress (from load curve)
37 sig(1) Calculated stress in 1 dir of the element reported back
to DYNA
38 sig(2) Calculated stress in 2 dir of the element reported back
to DYNA
39 sig(4) Calculated stress in 4 dir of the element reported back
to DYNA
40 blank -

115
Appendix A UMAT *MAT card documentation

References:
Shahkarami, A., & Vaziri, R. (2007). A continuum shell finite element model for impact simulation of
woven fabrics. International Journal of Impact Engineering, 34(1), 104-119.

Shahkarami, A. (2006). An efficient unit cell based numerical model for continuum representation of fabric
systems. PhD thesis, The University of British Columbia

Kawabata, S., Niwa, M., & Kawai, H. (1973). The Finite-deformation Theory of Plain-Weave Fabrics Part
I: The Biaxial Deformation Theory. Journal of the Textile Institute, 64(1), 21.

Kawabata, S., Niwa, M., & Kawai, H. (1973). The Finite-deformation Theory of Plain-Weave Fabrics Part
II: The Uniaxial Deformation Theory. Journal of the Textile Institute, 64(2), 47.

Platt, M. M., Klein, W. G., & Hamburger, W. J. (1959). Mechanics of elastic performance of textile
materials. Textile Research Journal, 29(8), 611-627.

116
Appendix B UMAT pseudo code

Appendix B UMAT pseudo code


t
1. Read the current strain increments,

2. Calculate displacement gradient from


x t = x t 1 + t
3. Calculate deformation gradient using
F t = xt + I
4. Calculate the Cauchy-Green deformation tensor using
C = FT F
4. Read yarn angle orientations from input deck, Calculate yarn vectors using
q0i = [cos i0 sin i0 ]
T

5. Calculate the current yarn stretch using


ti = q0Ti Cq02

6. Calculate current angle between warp and fill yarn using


q01T Cq02
cos =
12

7. Calculate change of angle

8. Calculate new yarn orientation


1t = 10 and 2t = 20

9. Update yarn vectors to the current yarn orientation


qi = [cos( it ) sin ( it )]
T

10. Calculate the current yarn displacement using


d it = y0 i (ti 1)

11. Read yarn failure flag from history variables. If either of the warp or fill yarns have failed
go to 33

12. Set up Newton-Raphson scheme. Set initial values of Hi equal to H0i if t=0 else the initial
value Hi is equal to Hi t-1

117
Appendix B UMAT pseudo code

13. Calculate unit cell current spacing by


yi = y0i + di

14. Calculate current yarn length

15. Calculate yarn strain

16. Read current yarn stiffness from inputted curve deck

17. Calculate yarn tension using

Fy ,i = E ( ) Ai
(L
i L0 i )
L0 i

18. Calculate vertical component of yarn tension (contact force)


Hi
Fc ,i = 2 Fy ,i
Li

19. Calculate vertical component of yarn bending


8(EI ) yarn i H 0i H i
Fb ,i =
Li
2
yi L0 i

20. Calculate the continuity functional equation


f1 (H 1 , H 2 ) = (H 01 + H 02 ) (H 1 + H 2 )

21. Calculate the equilibrium functional equation


f 2 (H 1 , H 2 ) = ( Fc1 Fv1 ) ( Fc 2 Fb 2 )

22. Check convergence criteria -- If |f1 + f2| < Tol go to 29

23. Calculate derivatives and assemble Jacobean matrix

27. Calculate new Hi using


H 1new H 1guess f1
=
new guess [ J ]f
H 2 H 2 2

28. Go to 13

29. Calculate fabric end force


yi
F f ,i = Fy ,i
Li

118
Appendix B UMAT pseudo code

30. Calculate coating end force


Ec y2tc y1 y01 y2 y02 Ec y1tc y2 y02 y1 y01
Fct ,1 = and F =
1 c2 y01 y02 1 c2 y02 y01
c ct , 2 c

31. Calculate membrane stress from total end forces


N i = (Ff ,i + Fct ,i )ni

32. Check for yarn failure -- If no yarn failure go to 35

33. If yarn ,i yarn


ult
,i
then Ni = 0 , S12 = 0 and yarn is flagged in history variable -- go to 36

34. If both yarns have failed then erode element from the mesh

35. Calculate shear component from yarn rotation using

G1 0 < 1
1 G G1
G1 + 2 ( 1 )
2
1 < 2
2 2 1
1
G2 + (G2 G1 )( 2 1 ) 2 < 3
S12 = 2
1 G G2 1
G2 + 3
2 4 3
( )
( 3 )2 + (G2 G1 )( 2 1 ) 3 < 4
2
1 1
G3 + (G3 G2 )( 4 3 ) + (G2 G1 )( 2 1 ) 4
2 2

36. Transform yarn stress tensor into the element coordinate system using
N1 N S
x = cos 2 1 + 2 cos 2 2 + 12 2 cos1 cos 2
h h h
N N S
y = 1 sin 2 1 + 2 sin 2 2 + 12 2 sin 1 sin 2
h h h
N N S
xy = 1 cos1 sin 1 + 2 cos 2 sin 2 + 12 (cos1 sin 2 + cos 2 sin 1 )
h h h

37. Return updated stress tensor to LS-DYNA to continue analysis

119
Appendix C Justification of the selection of a linear unit cell over sinusoidal geometry

Appendix C Justification of the selection of a linear


unit cell over sinusoidal geometry
Early in the course of the research, the linear unit cell was compared to a more complex
sinusoidal geometry by (A. Shahkarami, 2006) that more closely resembles the geometry of the
unit cell. The linear unit cell assumption results in a misrepresentation of the yarn height at the
crossover. It should be noted that the initial lengths of unit cell yarns in both geometries are
equal to the actual yarn length and the yarn spacing is also true to the fabric structure (based on
the measured yarns/in and % crimp ). Referencing Figure 1C, the percent difference between
the approximation of the yarn height for the linear and sinusoidal unit cell during biaxial
deformation is about 9%. Referencing Figure 2C and Figure 3C, the trigonometric functions
that resolve the yarn forces into horizontal and fabric components have a smaller percent
difference. Finally, the fabric end stress computed by the linear and sinusoidal unit cell is shown
in Figure 3D. The difference between the computed stress of the two geometries is between
1.5% and 2.0% under biaxial deformation.

0.090 100.00

0.080 90.00

80.00
0.070

70.00
0.060
warp yarn height (mm)

warp yarn height - sin


warp yarn height - tri 60.00
0.050 % difference difference (%)
50.00
0.040
40.00

0.030
30.00

0.020
20.00

0.010 10.00

0.000 0.00
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
strain

Figure C1: Approximation of yarn height by linear and sinusoidal unit cell

120
Appendix C Justification of the selection of a linear unit cell over sinusoidal geometry
0.295 100.00
sin(theta) - sin
sin(theta) - tri
0.290 90.00
% difference

0.285 80.00

0.280 70.00

0.275 60.00

difference (%)
sin(theta)

0.270 50.00

0.265 40.00

0.260 30.00

0.255 20.00

0.250 10.00

0.245 0.00
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
strain

Figure C2: Approximation of sin theta by linear and sinusoidal unit cell

0.970 100
cos(theta) - sin
cos(theta) - tri 90
0.968 % difference
80
0.966
70

0.964
60

difference (%)
cos(theta)

0.962 50

40
0.960

30
0.958
20

0.956
10

0.954 0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
strain

Figure C3: Approximation of cos theta by linear and sinusoidal unit cell

121
Appendix C Justification of the selection of a linear unit cell over sinusoidal geometry

45.000 100.00
warp-stress - sin
warp-stress - tri 90.00
40.000
% difference

80.00
35.000

70.00
30.000
warp stress N/mm

60.00

difference (%)
25.000
50.00
20.000
40.00

15.000
30.00

10.000
20.00

5.000 10.00

0.000 0.00
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
strain

Figure C4: Approximation of warp fabric membrane stress by linear and sinusoidal unit
cell

Overall for the fabric used in this study, no dramatic difference on the fabric stresses was found
between geometries was found. Therefore, the linear unit cell was chosen since it is less
computationally intensive and requires less preprocessing to determine the initial unit cell
geometry,

122
Appendix D Conversion of membrane stress into specific stress

Appendix D Conversion of membrane stress into


specific stress

In this study in order to keep consistent with the thin membrane assumption, stresses are
normalized based on force per unit width. Another popular stress normalization unit within the
textile community is specific stress based on force per mass per unit length.

Hearle (Hearle, 1989) shows that

force/width in N/mm
Specific stress on fabric in N/tex =
fabric " weight" in g/m 2

A traditional engineering unit of stress (force per unit area) can then be determined from the
specific area provided the density of the fabric is known as shown below:

Specific stress of 1 N/tex = density in g cm-3 stress of 1 GPa

123

Das könnte Ihnen auch gefallen