2005-01-0289

SAE TECHNICAL
PAPER SERIES

An Integrated Testing and CAE Application

Methodology for Curtain Airbag Development
Nirmal Narayanasamy, Mohamed Hamid, Deren Ma and Victor Suarez
Delphi

Reprinted From: Side Impact and Rollover

(SP-1938)

2005 SAE World Congress

Detroit, Michigan
April 11-14, 2005

400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org
The Engineering Meetings Board has approved this paper for publication. It has successfully completed
SAE’s peer review process under the supervision of the session organizer. This process requires a
minimum of three (3) reviews by industry experts.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of SAE.

For permission and licensing requests contact:

SAE Permissions
400 Commonwealth Drive
Warrendale, PA 15096-0001-USA
Email: [email protected]
Tel: 724-772-4028
Fax: 724-772-4891

For multiple print copies contact:

SAE Customer Service

Tel: 877-606-7323 (inside USA and Canada)
Tel: 724-776-4970 (outside USA)
Fax: 724-776-1615
Email: [email protected]

ISSN 0148-7191
Copyright © 2005 SAE International

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE.
The author is solely responsible for the content of the paper. A process is available by which discussions
will be printed with the paper if it is published in SAE Transactions.

Persons wishing to submit papers to be considered for presentation or publication by SAE should send the
manuscript or a 300 word abstract to Secretary, Engineering Meetings Board, SAE.

Printed in USA
 2005-01-0289

An Integrated Testing and CAE Application Methodology for

Curtain Airbag Development
Nirmal Narayanasamy, Mohamed Hamid, Deren Ma and Victor Suarez
Delphi

Copyright © 2005 SAE International

ABSTRACT The fabric materials are typically coated with silicone to

reduce the leakage through fabrics. Usually there is
The Curtain Airbag (CAB) is used currently to provide leakage through the stitch lines where the fabrics are
head and neck protection for the front-seat and rear-seat sewn together. Newer sewing technology like the
vehicle occupants during side-impact collisions and seamless method reduces the leakage through the
vehicle rollovers. seams.

The coated fabric materials are used in CABs for In a normal driver and passenger airbag, the vents and
occupant protection in side impact and rollover events. In permeability play a vital role in influencing the airbag
this paper the design and development study of CABs is performance. It was shown in [2], that the effect of
described by using simulation and physical tests. The permeability was the largest compared to other airbag
mechanical properties for the airbag material are parameters like the fabric modulus and the tether
determined by uniaxial test in the fill and warp directions. stiffness, using a Design of Experiments (DOE) study for
Shear strength is also evaluated by using the uniaxial a driver airbag. However, in the CAB the porosity is
test, but the specimen is cut along 45º angle. These test almost negligible because the CAB has to stay inflated
values are used in the finite element (FE) simulations. for longer duration. Therefore, appropriate
characterization of the mechanical properties of the
In this paper, a methodology of the design study is fabric material is an important factor in evaluation of the
discussed. A Free Motion Headform (FMH) impacting a performance using FE simulation.
pole with a pillow shaped airbag is used in the design
study. The influences of CAB design parameters such as Airbag FE simulations are used in design and
pressure, chamber width, impact speed and hit location development for occupant protection [1,2]. In this paper,
are evaluated. The simulations were correlated with experimentally determined material constants for fabric
actual tests under various conditions such as impact are used in the FE simulation. A Free Motion Head form
speeds, hit locations and various pressures in the airbag. (FMH) is used in the design and development process
The simulation results were compared to the physical for CAB, which represents the occupant head motion in
test by Computer Aided Engineering (CAE) model an actual crash. An impact speed of 24 to 29 kph is used
correlation grading system. to replicate the field conditions similar to those of
FMVSS201 [3]. In the CAB design process, a Free
INTRODUCTION Motion Headform (FMH) with pole impact test is
performed to optimize the pressure in the airbag.
Understanding the influence of the pressure is critical in
The CAB is increasingly used as a countermeasure in
designing an effective countermeasure under various
order to help protect the occupant during the side impact
impact speeds, thicknesses of the airbag and hit
and rollover event. Unlike in the case of frontal impact,
locations.
the space between the occupant and the deforming
structure is much less. The CAB has to be deployed
quickly and needs to be in position in a very short time- Many parameters such as airbag chamber width,
typically 20-30ms [1]. It is desirable for the CAB to be pressure, speed of impact and hit locations influence the
inflated for a longer duration so that adequate protection design of a CAB. Experimentally determining the
can also be provided in a rollover scenario. All these influence of these parameters is time consuming and
requirements demand that CAB fabric material have expensive. An alternative to this is to use numerical
negligible leakage. simulation in the design process.
In this study, a numerical design study using FE is FE MODELING OF FABRIC MATERIAL
performed with a pillow shaped dual-chamber airbag as
a simplified representative of a CAB. Resultant head To generate the most accurate model to represent the
acceleration versus pressure charts for different cushion fabric material, an FE model would have to be created at
thickness under various head impact conditions was the thread level. As the individual threads are loaded, the
generated [4]. The correlation of the various head impact interaction between them changes. Even though the
conditions with the physical test is discussed and a material constitution at the molecular level does not
correlation grading system based on the procedure change, the rearrangement of the threads results in
described in [5] is used to evaluate the simulation changes in the global behavior, with a nonlinear
results. response [6]. Figure 3 shows such a FE model.

CAB MATERIAL MODEL AND TESTING

The basic material compositions of the fabrics are nylon,

polyester or other polymers. It is the manufacturing
process, which changes the way the material behaves
compared to the raw state. Fabrics are a web
arrangement of extruded plastic threads, and its
mechanical properties are different from the original
material. This presents a challenge for FE modeling.
Figure 1 shows an enlarged view of the fabric material
structure. Figure 3. Thread Level FE Model

It is especially time consuming to create such a detailed

model and will be computationally expensive in product
development applications. In addition, it is very difficult to
capture the leakage through the fabric, which is an
important parameter in airbag applications. This type of
modeling is difficult, if not impossible to be applied in a
production environment. Under this condition, a
membrane shell element is more practical in design and
development studies. Commercial FE packages like LS-
DYNA3D [7] solvers have a special fabric membrane
element that is being used in airbag applications. To use
Figure 1. Enlarged View of the Fabric Material Structure such a material model macro level mechanical properties
of the fabric have to be determined through physical test.
Yarns of fabric are weaved in a pattern shown in figure 1.
The fill direction represents the fabric that is laid in the FABRIC MATERIAL MODEL
longitudinal direction. In the warp direction, the yarns are
laid in the lateral direction. Normally the angles between LS-DYNA3D material model FABRIC (MAT_34) is used
both the directions are at 90º. Figure 2 shows the fill and to simulate the airbag material. It is a variation of the
wrap directions in a roll of fabric material. Fabric material layered orthotropic material model [7]. To model the
exhibits directional physical properties in the fill and warp airbag as an orthotropic model, three material constants
direction. One of the ways of characterizing the material have to be provided. Assuming a plane stress condition,
is using the finite element method. the material constitutive equations are given by [8]:

­
°°σ
§
½ ¨
1° ¨
Q Q 0 ·¸­° ε
11 12 1°
½
° ¸
®
°
σ 2
¾=¨
° ¨
Q Q 0 ¸¸°®° ε
12 22 2
°
¾
°
τ
°¯ 12 °
¿ ¨
© 0 0 Q ¸¹°¯γ 66
°
12 ¿

Where σ is the normal stress and τ

is the shear
stress, the subscript refers to the principal material
directions, i.e. the fill and warp directions. Also
ε γ
and are the strain components. The material elastic

Q
Figure 2. Fabric Roll
constants are given by the following equations:
ij
Q =
E 1
From the load versus elongation curve the stress versus
strain curve can be plotted. The stress versus strain
11
ν ν 1−
12 21
curves for the fill, wrap and 45º material directions, are
shown in figure 5. The Young’s modulus values can be
Q = 1ν− E 12 2
=
ν E 21 2 calculated from the stress versus strain curves.
12
ν ν 12 21
1 −ν ν
12 21
Stress Vs Strain
E
7.E+04

Q = 2

ν ν
6.E+04 Warp
22 1−
12 21 Fill
5.E+04

Q =G 45'

Stress (KPa)
66 12 4.E+04

E E
3.E+04
Where 1
and 2
are the Young’s modulus in the fill

G
2.E+04
and wrap directions and is the shear modulus of
12

ν
1.E+04
the fabric material. ij
is the Poisson ratio of the
0.E+00
material. Additionally in the LS-DYNA3D material model, 0 0.05 0.1 0.15 0.2
fabric leakage can be accounted for. In addition, when Strain
modeling thin fabrics [6], buckling can result in an
inability to support compressive stresses. A linearly
elastic liner can also be included to reduce such Figure 5. Stress versus Strain Curves for the Fabric
tendency. However, for this CAB material, the leakage is Material
almost negligible and therefore no leakage is specified.
The airbag tested is seamless and the leakage through For the shear modulus since the fabric is tested using a
the seams could be ignored. The linear option is not sample cut at 45º material directions, the shear modulus
invoked in this study because the pillow airbag is not can be calculated using the relation below [8].
folded in the initial configuration. The orthotropic material
1
axes can be defined by using AOPT option [6] in LS-
DYNA3D solver, this allows to specify the fill and wrap G 12
=
2ν
4 − 1 − 1 + 12
direction. The mechanical properties are determined
from the physical test. E 45 E 1 E 2 E 1

MATERIAL TESTING Where E 45

is the modulus from the stress strain curve
at 45º material direction.
The mechanical properties of the fabric material can be
determined from the uniaxial tensile testing. A
From the stress versus strain curves, it can be seen that
predetermined fabric coupon with a width of 76.2mm and
the Fabric material is nonlinear. Nieboer, J.J. et.al. [8]
effective length of 203mm is cut along the fill and wrap
proposed using an averaged linear material model.
direction. A sample with the above mentioned dimension
HongSoongu [2] uses optimization studies to estimate
is also cut at 45º directions from the fabric roll to
the modulus values for a given scenario. However, in this
determine the shear modulus. The coupons are then
study the modulus at 10% strain is used. This value is
clamped at the edges and pulled at a loading rate of 150
chosen because the airbag is subjected to higher-
mm per minute. The specimen is loaded until failure and
pressure levels and this assumption eliminates the
the load versus elongation are recorded. Figure 4 shows
variability of this parameter.
a typical test setup.
HEAD IMPACT DESIGN STUDY
In the FMH impact test, a headform with a given initial
velocity is made to impact an inflated airbag. Figure 6
shows a FMH physical test setup. In this case, the airbag
is inflated through compressed air and a constant
pressure is maintained. The responses such as HIC &
3ms clip acceleration of the headform are evaluated.

As a simplified representative of a curtain airbag (CAB),

a dual-chamber pillow-shaped airbag is studied to
provide general design guidelines for CAB. Resultant
head acceleration versus pressure charts for different
cushion thickness under various head impact conditions
Figure 4. Typical Tensile Testing Setup is generated to help design the minimum cushion
pressure for different cushion thickness [4].
 To provide enough representative test data for computer
model correlation with minimal physical tests, a design of
experiment (DOE) analysis was conducted to design the
physical test matrix. A fractional factorial matrix
consisting of 19 test configurations was generated from
the DOE analysis. Each test was repeated twice to
consider the test-to-test variations. Figure 8 shows the
physical tests that were conducted based on the DOE
study.

Figure 6. FMH Impact Physical Test Setup

Computer simulation models are used extensively as a

complement to the physical tests to create the CAB
design charts. This study also provides a method to
validate the fabric material model. The study matrix for
the design process is described below.

CAB PERFORMANCE EVALUATION MATRIX

A matrix of 84 runs with three different airbags with

varying chamber thickness, (152 mm, 178mm and
203mm), two hit locations (center and seam), two
speeds (24 kph and 29 kph) and seven different
pressures (20,30,40,50,60,80 and 100 Kpa) was studied.
Figure 7 shows the matrix that was used in this study.
The design criteria for this study were 1) the HIC value
should be less than 700 and 2) the bag should not
bottom out.

Figure 8. List of Physical Tests Conducted

FINITE ELEMENT MODEL OF TEST

A FE model of the airbag with the rigid pole was

used to simulate the test condition. The FTSS [10] FMH
is used to simulate the experimental setup as shown in
figure 9.

Figure 7. Evaluation Matrix

Figure 9. FE Model Setup for the FMH Head Impact

The airbag material is modeled in LS-DYNA3D using
MAT_FABRIC (type 34) as mentioned in the previous
section. The porosity of the fabric is assumed zero. Pole
is modeled as a MAT_RIGID material and the degrees of
freedom are constrained in all the directions. It is
assigned the similar properties of steel for contact
purposes. The Head form material properties are used
as given by FTSS [10].

The headform is given an initial velocity corresponding to

the test condition (24 kph or 29 kph). Optimal damping
properties are used to stabilize the inflated bag in the
numerical simulation.

Inflator Mass Flow Calculation

The airbag control volume in LS-DYNA3D is modeled as

an AIRBAG_SIMPLE_AIRBAG_MODEL to simulate the
shop air test condition. For this model, a mass flow curve Figure 10. Mass Flow Curve
needs to be estimated for the each pressure condition in
the airbag. CORRELATION GRADING PROCESS

A baseline model of the airbag is run using A baseline FE model is created using the 178 mm bag
AIRBAG_LOAD_CURVE option. This gives an estimate and this is correlated to the 40Kpa, center hit and 29kph
of the volume of the airbag. Using this estimated volume, test conditions. Many more analyses were done to
the mass needed to get this pressure is calculated using simulate the whole matrix for the 178 mm airbag. Only
the following ideal gas relationship. the pressure, speed and hit locations where changed in
the model to generate the test conditions. The responses
PV = mRT from this simulation matrix were compared to the
physical test data. The correlation method described in
P = Pressure (Known quantity) [5] was used to compare and rank the responses. As
mentioned in the previous section, a total of 19 physical
V = Volume (Estimated from the baseline analysis) test data (repeated twice) were available, and used for
this correlation study.
m = mass of the gas needed
The correlation method compares the simulation curve to
the physical test curve and assigns a grade for each
R
R= -1 -1
(R=8.314 J mol K , MW = molecular weight criterion in the dynamic response category. Each test
MW result is compared and assigned a grade [5]. Figure 11,
of the gas, air in this application) shows the grading criteria used for this study.

T = Temperature (ambient temperature is assumed to be

constant)

Using the computed mass, a mass flow curve is

generated as shown in Figure 10. The area under the
curve should be equal to the total mass calculated. The
mass is applied within first 10 milliseconds in this
analysis. The mass flow curve has to be scaled up or
down to get the exact pressure needed in the airbag and
this normally takes one or two iterations.
Figure 11. Grading Criteria

The overall grade for a specific bag (152mm, 178mm

and 203mm) is arrived by taking the average of the all
the grades.

The grading charts for the airbags 152 mm, 178 mm and
203mm airbags are shown in the Tables 1, -3. In the
table, abbreviated notations are used for the various
cases. The expansion for the notation is given in figure 8.
 Peak Statistic Analysis
Dynamic
Overall
Test# Grading system Standard 2nd Response Remark
Kinematics Avg Correlation 0 Moment 1st Moment
Magni-tude Deviation of Moment Score
Residual Coefficient Difference Difference
Residual Difference

1 152_40_C_24 Good 2.32% 4.15% 8.24% 0.984 10.00% 8.40% 10.00% Good

2 152_60_C_29 Good 12.35% 2.35% 5.79% 0.991 7.20% 4.40% 3.00% Good
Cases

3 152_100_C_29 Good 8.98% 1.19% 4.97% 0.993 4.00% 3.10% 2.50% Excellent
4 152_40_C_29 Good 31.30% 0.78% 4.16% 0.978 6.80% 8.00% 10.10% Good Bag Bottomed Out

5 152_60_S_29 Good 7% 0.31% 3.22% 0.998 1.20% 3.40% 4.50% Excellent

6 152_100_S_24 Good 6% 2.21% 3.51% 0.998 7.90% 8.90% 9.40% Good

Table 1. Correlation Grading for the 152 mm Airbag

Peak Statistic Analysis

Dynamic
Overall
Test# Grading system Standard 2nd Response Remark
Kinematics Avg Correlation 0 Moment 1st Moment
Magni-tude Deviation Moment Score
Residual Coefficient Difference Difference
of Residual Difference

7 178_40_C_29 Good 6.80% 4.39% 7.88% 0.987 9.00% 10.60% 15.40% Good

8 178_60_S_24 Good 6.25% 1.27% 4.35% 0.997 3.10% 1.20% 0.10% Excellent
Cases

9 178_80_S_29 Good 5.14% 1.10% 3.10% 0.998 3.40% 6.10% 8.20% Excellent
10 178_80_C_24 Good 6.34% 0.51% 3.89% 0.997 1.30% 1.60% 2.40% Excellent
11 178_20_C_29 Good 21% 0.54% 5.34% 0.968 5.40% 11.20% 16.50% Adequate
Bag Bottomed Out

Table 2. Correlation Grading for the 178 mm Airbag

Peak Statistic Analysis

Dynamic
Overall
Test# Grading system Standard 2nd Response Remark
Kinematics Avg Correlation 0 Moment 1st Moment
Magni-tude Deviation Moment Score
Residual Coefficient Difference Difference
of Residual Difference

12 203_40_C_24 Good 11% 1.03% 5.38% 0.994 2.60% 1.40% 1.30% Good

13 203_40_S_24 Good 2.30% 3.69% 8.75% 0.987 9.20% 4.80% 3.30% Good

14 203_40_S_29 Good 5.80% 0.21% 4.31% 0.996 0.70% 3.60% 4.20% Excellent
Cases

15 203_80_S_29 Good 8.40% 0.39% 5.30% 0.994 1.30% 5.60% 16.80% Good
16 203_30_C_24 Good 7% 6.19% 6.05% 0.99 12.50% 12.50% 15.30% Adequate

17 203_60_C_29 Good 7% 3.76% 5.44% 0.992 10.30% 9.90% 11.30% Good

18 203_20_C_24 Good 11% 4.30% 7.53% 0.985 10.90% 11.50% 13.00% Adequate

19 203_20_S_29 Good 5.45% 0.16% 4.58% 0.987 1.50% 5.10% 5.50% Good Bag Bottomed Out

Table 3. Correlation Grading for the 203 mm Airbag

COMPARSION BETWEEN THE SIMULATION AND
THE TEST RESULTS

The overlay between the tests and simulations for the

resultant head acceleration of the head form for some of
the cases are given in figures 12, 13,14 and 15. In
theses figures, there are two test curves (repeats) and
one simulation curve. In figure 14, the pillow airbag
bottoms out and therefore the head impacts the pole,
resulting in high resultant head acceleration values.

Figure 14. Comparison of Test and Simulation for a 178

mm Airbag-20Kpa Pressure-Center Hit-29kph Condition

Figure 15. Comparison of Test and Simulation for a

203mm Inch Airbag-30Kpa Pressure-Center Hit-24kph
Condition
Figure 12. Comparison of Test and Simulation for a
152mm Airbag-100Kpa Pressure-Seam Hit-24kph DISCUSSION
Condition
It can be seen from the stress versus strain curve in
Figure 5, that the fabric material is nonlinear. For this
study, a linear material model for the fabric was used.
The modulus was computed at 10% strain. The assumed
constant material properties at 10% strain level gave
good correlation with the tests, similar assumptions were
made by others [2,9]. The nonlinear behavior of the
material can be modeled in the later versions (970) of
LS-DYNA3D solver by giving directly the stress versus
strain experimental data as the input. This method would
eliminate the need for estimating the modulus or avoid
the need for averaging the modulus. Proper test methods
have to be performed to accurately capture the shear
stress versus shear strain curve for generating material
input data [7].

The maximum difference in peak resultant acceleration

of the headform between the FE simulation and the test
Figure 13. Comparison of Test and Simulation for a 178 did not exceed 10% for most of the cases as shown in
mm Airbag-60Kpa Pressure-Seam Hit-24kph Condition Tables 1-3. It can been seen from these tables that
whenever the airbag was bottoming out, the difference in
peak head acceleration between the simulation and test,
was more than 10%. The reason for this discrepancy
may be due to the material modeling of the pole as a
rigid material. However, in reality the pole had some
compliance, which was neglected in this analysis. In
addition, test-to-test variation in the peak magnitude was
also observed when the airbag bottomed out as shown in
Figure 13. These discrepancies were ignored for the
design study since the bottomed-out bag violated the 9. Nieboer, J.J.et al, “Airbag Modeling Techniques”,
chosen design criteria. SAE 902322, TNO Road-Vehicles Research
Institute, Delf, Netherlands, 1990.
CONCLUSION 10. FTSS- First Technology Safety Systems, Finite
element models of crash test dummies, MI, 1999.
Observations based on this application of this
methodology to CAE design are as follows:

• Proper material characterization and

appropriate boundary conditions are critical and
sufficient for the prediction of headform
performance.

• This study provided an understanding on the

influence of design parameters such as the
impact speed, hit location, chamber width and
pressure on the performance of CAB.

• This component methodology shows promise in

helping full system design by reducing iteration
time and minimizing the number of physical
tests.

ACKNOWLEDGEMENTS

The authors wish to thank the management of Delphi for

the permission to publish this material. The authors
express deep appreciation and thanks to their
colleagues, Srini V Raman and Lousie Zhang for their
contribution.

REFERENCES

1. Zhang, Honglu, et al, “ CAE-Based Side Curtain

Airbag Design”, SAE 2004-01-0841, SAE World
Congress, Detroit, Michigan, March 8-11, 2004.
2. Hong, Soongu, “A Study on the Modeling Technique
of Airbag Cushion Fabric”, SAE 2003-01-0512, SAE
World Congress, Detroit Michigan, March 3-6, 2003.
3. U.S. Department of Transportation, National
Highway Traffic Safety Administration, “Federal
Motor Vehicle Safety Standards;
Side Impact Protection”.
(http://www.nhtsa.dot.gov/cars/rules/rulings/sideimpa
ct/part6.html)
4. Pillow Airbag Study, Nirmal Narayansamy, et al,
Delphi Internal Report 2004.
5. Ma, Deren, et al, “Correlation Grading Methodology
for Occupant Protection System Models”, SAE 2004-
0101631, SAE World Congress, Detroit, Michigan,
March 8-11, 2004.
6. Cao, .J, et al “Numerical Determination of
Mechanical Elastic Constants of Textile
th
Composites”, 15 Annual Technical Conference of
the American Society for Composite, College
Station, Texas, Sept.25-27, 2000.
7. LS-DYNA3D User’s Manual version 970.
8. Jones, M. Robert, Mechanics of Composite
Materials, Hemisphere Publishing Corporation, 1975.