Detailed Simulation of the Airbag Inflation Process Using a

Coupled CFD/FE Method
Dr. G.E. Fairlie
*
, Dr. A.C. Steenbrink
**
*
Century Dynamics Ltd., Hurst Rd, Horsham, West Sussex, RH12 2DT, UK, Tel: +44 1403 270066
**
TNO Automotive, Schoemakerstraat 97, PO Box 6033, 2600 JA Delft, NL, Tel: +31 15 269 7427
Abstract
There is increasing pressure to reduce injuries and fatalities by effectively using airbags for side impact and out-of-position
impact loading conditions. The simulation of these scenarios is particularly difficult as the occupant is interacting with the airbag
at early times during the airbag deployment when the detailed behaviour of the airbag can have a significant effect.
A new airbag inflation simulation capability that overcomes this problem has been integrated into the crash simulation software
MADYMO. This allows accurate simulation of the gas flow within the airbag throughout the inflation and unfolding process.
The new module uses CFD techniques to numerically model the interior volume of the airbag in detail. The pressure field within
the airbag is directly coupled to the standard MADYMO FE model of the airbag. Hence, the acceleration and movement of the
airbag and any bodies in contact with it can be predicted more accurately.
Introduction
The accurate simulation of the early stages of the airbag inflation process requires the resolution of the
flow of the inflator gases within a complex unfolding airbag. This cannot be achieved using standard
isobaric inflation models that assume a uniform pressure within the airbag at each time during the
inflation process. Multi-chamber inflator models are little better as they only allow a single pressure
within each chamber. Moreover, the parameters that govern gas flow between chambers cannot easily be
related to experiments.
A new airbag inflation simulation capability that overcomes this problem has been integrated into the
crash simulation software MADYMO [ 1]. This allows accurate simulation of the gas flow within the
airbag throughout the inflation and unfolding process.
The new module uses CFD techniques to numerically model the interior of the airbag that is directly
coupled to the standard MADYMO FE model of the airbag. The CFD model divides the interior volume
of the airbag into many thousands of small regions or cells and the movement of inflator gases between
these cells is calculated from first principles. The distribution and state of the inflator gases within the
airbag can therefore be accurately resolved, and this gives accurate simulations of the pressure forces
acting on the airbag. Hence, the acceleration and movement of the airbag and any bodies in contact with
it can be predicted more accurately.
This paper will briefly describe the technical details of the new facility. An example of a typical airbag
application will be given. Comparisons between experimental results and simulations conducted using
the new module and standard isobaric or multi-chamber inflator models will be shown.
CFD/FE Coupling Methodology
The new module implemented in the MADYMO crash dynamics code uses a combined CFD/FE coupling
methodology. The standard FE capabilities of the MADYMO code are used to model the airbag
components and their interaction with the vehicle or occupants. The interior volume of the airbag is
modelled using a CFD type approach. In the coupling algorithm the FE airbag mesh forms a closed
surface that defines the volume enclosed by the airbag. This surface acts as a boundary condition applied
to the CFD mesh that constrains the motion of the inflator gases. The gasses within the airbag in turn
TP073
apply loads onto the FE airbag elements that are calculated by integrating the pressure in the CFD cells
adjacent to each airbag element over the element area.
The volume enclosed by the airbag at the start of the simulation is automatically discretised into a large
number of computational cells. For computational efficiency the CFD mesh uses hexahedral cells aligned
with a fixed local coordinate system with a uniform cell size in each direction. The motion of the gas
within the CFD mesh is calculated from first principles using the Euler conservation equations for
inviscid compressible hydrodynamic flow. These equations are solved simultaneously in all three spatial
dimensions using the high resolution Flux Corrected Transport method [ 2] [ 3]. Time integration is by a
second order accurate explicit method where the time step is controlled by the Courant stability
requirement. This condition requires that for numerical stability a disturbance should not be able to travel
across more than a single cell dimension in a single time integration step. This means that small cell
dimensions or high flow velocities in the CFD mesh will require small time integration steps. The
integration time step for each cycle is the minimum of that required by the FE and CFD parts of the
simulation.
The portions of the CFD mesh inside and outside the airbag are updated as the airbag surfaces move. An
efficient approximate geometric calculation is used to determine the fraction of each cell volume and face
area that is open to gas flow. If a cell face is completely blocked gas cannot flow through this face. If a
cell volume is completely covered then the cell does not take part in the CFD calculation. The calculation
of cell and face blockage fractions is numerically expensive so it is only performed if the airbag surface
has moved significantly since the last geometry update.
Unless treated specially CFD cells that are almost covered will need a small Courant stability time step.
A special technique is used to overcome this problem by blending nearly covered cells with a neighbour.
Inflator gases are added into the CFD mesh where the inflator outflow orifice intersects with the CFD
mesh. The state of the inflator gases is specified in the same way as for the MADYMO isobaric airbag
model using standard data that is usually derived from tank test experiments. This process adds mass,
internal energy and momentum into the CFD mesh. For efficiency the gas within the airbag is modelled
as a single ideal gas material. The ideal gas parameter, gamma, can be varied as the composition of the
gasses within the airbag changes during the inflation process, and the average pressure and temperature
within the airbag change.
In order to successfully model the inflation process within a complex folded airbag the cells used in the
CFD mesh must be small enough to resolve the features of the interior volume. For example a folded
airbag with the folds initially separated by 2mm would require an initial cell size no larger than 2mm.
Enclosing the entire volume of the inflated airbag with cells with dimensions this small would require a
prohibitively large number of numerical cells, a large amount of memory and would entail long
computation times. For this reason the number of cells within the CFD mesh is kept approximately
constant throughout the simulation but the cells are allowed to grow as the calculation proceeds and the
overall airbag dimensions change. The cell sizes are changed at discrete times and during this process the
state of the gas is conservatively mapped from the old CFD mesh to the new mesh.
A 2D example is shown in Figure 1 to illustrate the coupling methodology. In this example the airbag
folds are initially separated by 1mm and the airbag initially has overall dimensions of 40mm by 5mm, as
shown in the left hand picture. The inflator is positioned half way up the left hand vertical edge of the
airbag and injects gas in a horizontal direction towards the right hand side. The Euler cells are initially
0.5mm by 2mm but they grow as the analysis proceeds. This requires that the Euler grid be moved
approximately every 100 cycles during the calculation. The remaining three plots show pressure contours
within the bag during the inflation process at times of 0.25, 0.5 and 4ms after the inflator fires. Note that
the pressure contour scale is different for each of the plots. During the early part of the simulation there is
a strong pressure variation within the airbag. At the end of the simulation the pressure distribution is
close to isobaric.
Very little additional data is required to conduct a detailed inflation simulation compared with using the
standard isobaric inflation model available within MADYMO. The user must specify the number of
numerical cells to be used in each coordinate direction, the minimum cell size to be used and the
orientation and dimensions of the inflator orifice. All other data such as the inflator specification and the
FE airbag are identical to that required for an isobaric simulation.
Example Simulation Results
Tank Test
Figure 2 shows a comparison between an experiment, an isobaric and a detailed CFD simulation of a tank
test. The tank is initially at atmospheric pressure of approximately 1 bar rising to a pressure of 2 bars as
the inflator fires. The isobaric simulation used the standard MADYMO isobaric inflator model with a
fixed volume 60 litre tank. The trace labelled GF-Module shows the average pressure within the tank
using a 20x20x20 cell numerical mesh to simulate the same problem.
Flat Circular Airbag
A simple example of the inflation of a flat circular airbag is shown in Figure 3 to Figure 5. The initial
shape and location of one half of the airbag is shown in Figure 3. The airbag is constructed from two
700mm diameter membranes joined along their outer edge with an initial separation of 10mm between the
front and back faces of the bag. A 20x20x20 cell CFD mesh is defined at the start of the simulation with
cells 5mm thick through the thickness of the airbag and approximately 40mm in the other two co-ordinate
directions, as shown in the left hand picture in Figure 4. At the start of the simulation only some of the
numerical cells are within the airbag, and only these cells initially contain gas, as shown in the right hand
picture in Figure 4.
The inflator is located at the centre of the rear face of the bag and injects gas into the bag towards the
right hand side of the plots. Figure 5 shows the pressures within the CFD mesh at six times after the start
of the inflation process. Only half of the numerical mesh is shown in each plot to show the pressure
distribution within the mesh. At early times the airbag is within the initial numerical mesh so the cell size
remains constant but more cells are filled with gas. At later times the cells are enlarged so that the airbag
remains within the CFD mesh. At early times there is a significant pressure variation within the airbag,
and this causes more rapid expansion of the airbag opposite the inflator. At later times the pressure
within the airbag becomes close to uniform.
This simulation took about 40 minutes to complete to a time of 40ms on a 700MHz PC.
Folded Airbag
Figure 6 shows a folded airbag system. The inflator is located half way along the fold of the airbag that is
at the left hand side of this plot and injects gas horizontally towards the right hand side. This simulation
again used a 20x20x20 cell numerical mesh with 1.5mm cells through the thickness of the airbag. Half of
the initial CFD mesh within the airbag is shown in Figure 7. Figure 8 and Figure 9 show a comparison
between a coupled and an isobaric inflation of this airbag at 5ms after the start of the inflation process.
Figure 10 and Figure 11 show the airbag at a time of 8ms. The plots of the inflation process show that the
detailed pressure distribution within the airbag in the coupled simulation has a significantly affect the
airbag shape when compared with the isobaric inflation simulation.
The coupled simulation of the folded airbag took about 80 minutes to complete to a time of 10ms on a
DEC Alpha workstation.
Conclusions
A new CFD/FE coupled method for the simulation of airbag inflation phenomena has been implemented
in the MADYMO crash dynamics software. This method allows the calculation of the pressure
distribution within the airbag at all times during the inflation process and can therefore be used to
investigate out of position impact loading conditions.
The new facility is simple to use and only requires a small number of additional control parameters
compared with using the standard isobaric inflation module within MADYMO. Standard tank test data is
used to model the inflator in the CFD simulation.
Initial testing of the new module has shown that coupled simulation of both folded and un-folded airbags
has a significant effect on the inflation process when compared with a standard isobaric inflation model.
References
[ 1] MADYMO Version 5.4 Users Manual, TNO-Automotive, Delft, Holland, May 1999
[ 2] JP Boris, DL Book, ‘Solution of Continuity Equations by the Method of Flux-Corrected Transport’, Vol 16, Methods
in Computational Physics, Academic Press, 1976
[ 3] ST Zalesak, ‘Fully Multidimensional Flux-Corrected Transport Algorithms for Fluids’, Journal of Computational
Physics Vol 31, pp335-362, 1979







Figure 1 2D Example of Coupled FE/CFD Inflation Process

1
1.25
1.5
1.75
2
2.25
0 20 40 60 80
Time (ms)
P
r
e
s
s
u
r
e

(
B
a
r
)
Test
MADYMO
GF-Module


Figure 2 Isobaric and GF-Module Tank Test Pressure Time History


Figure 3 Flat circular Air Bag Initial Material Location




Figure 4 Initial CFD Numerical Mesh


1ms

5ms


10ms

15ms


20ms

30ms

Figure 5 Pressure in CFD mesh at 1, 5, 10, 15,20 and 30ms after start of Inflation Process

Figure 6 Folded Airbag Initial Airbag Surface


Figure 7 Folded Airbag Initial CFD Mesh


Figure 8 Folded Airbag Coupled Inflation at 5ms


Figure 9 Folded Airbag Isobaric Inflation at 5ms


Figure 10 Folded Airbag Coupled Inflation at 8ms


Figure 11 Folded Airbag Isobaric Inflation at 8ms