Professional Documents
Culture Documents
JWL Values For Abaqus
JWL Values For Abaqus
The mission of the JRC-IPSC is to provide research results and to support EU policy-makers in
their effort towards global security and towards protection of European citizens from accidents,
deliberate attacks, fraud and illegal actions against EU policies.
European Commission
Joint Research Centre
Institute for the Protection and Security of the Citizen
Contact information
Address: Martin Larcher, T.P. 480, Joint Research Centre, I-21027 Ispra, ITALY
E-mail: [email protected]
Tel.: +390332789004
Fax: +390332789049
http://ipsc.jrc.ec.europa.eu/
http://www.jrc.ec.europa.eu/
Legal Notice
Neither the European Commission nor any person acting on behalf of the Commission is
responsible for the use which might be made of this publication.
Europe Direct is a service to help you find answers
to your questions about the European Union
Freephone number (*):
00 800 6 7 8 9 10 11
(*) Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.
A great deal of additional information on the European Union is available on the Internet.
It can be accessed through the Europa server http://europa.eu/
JRC 57322
EUR 24288 EN
ISBN 978-92-79-15245-0
ISSN 1018-5593
DOI 10.2788/73796
Introduction
much smaller. In the case of an explosion inside a train station, for example,
the most fragile part is the glazing. The glass failure occurs depending on
the type of material used (e.g. annealed or safety glass) and results in flying
debris, but also in a relief surface for the air blast wave.
To design buildings and other structures against air blast waves, the
loading resulting from such waves needs to be known. In most cases, the
behaviour of structures under air blast load is computed by the finite element method. The loading should therefore be compatible with this method.
Several methods can be found in the literature.
The behaviour in the explosive and in the surrounding air can be calculated by the well-known Jones-Wilkins-Lee (JWL) equation (see Lee et
al. [1] and for the parameters of the equation Dobratz and Crawford [2]).
Liu et al. [3] present a work using smooth particle hydrodynamics (SPH)
to determine the pressure inside an explosive. More often used are arbitrary
Lagrangian Eulerian (ALE) meshes (Cendon et al. [4], Alia and Souli [5]).
Fluid calculations are very much influenced by the element size, as shown by
Cendon et al. [4].
The influence of air blast waves on structures is for example investigated
experimentally by Langdon and Schleyer [6] and by Gram et al. [7] with
a blast simulator. Numerical investigations are presented by Langdon and
Schleyer [8], Luccioni et al. [9]. The design of buildings against air blast
waves is discussed by Smith and Rose [10]; the influence of a protective
barrier on the blast wave is described by Zhou and Hao [11].
The present work concerns mainly the mid- and far-field where the geometry of the charge has small influence. Several blast models and tools are
available to calculate the pressure at a given distance from the charge for
spherical and hemispherical conditions (see Section 2.2).
The behaviour of air blast waves in complex geometries like cities is investigated by Smith and Rose [12]. Their paper shows that shielding and
channelling has to be considered in urban city streets. This can be done
for simple geometries using several charts. For more complex geometries numerical investigations should be used instead (see e.g. Remennikov and Rose
[13]). Experimental data for explosions between buildings are given by Smith
et al. [14] and Fairlie [15]. The interaction with the buildings in combination
with a damage criterion is shown by Luccioni et al. [16].
This work presents several simulation methods for the air affected by explosions inside large, complex structures. The first part of the work describes
the pressure behaviour of an air blast wave and compiles a set of equations
from the literature, which can describe the pressure of a spherical air blast
wave. Several simulation methods are then presented and compared on the
basis of the calculation time and of their capability to simulate the real be2
Figure 1: Pressure-time curve for a free air blast wave, approximation: dashed
line
haviour inside complex structures. This is done using two experiments with
different geometries.
Air blast waves result from a rapid release of energy. Here, the air blast
waves investigated result from the detonation of a solid explosive. Other
phenomena such as the combustion of gas/air mixtures can also result in an
air blast wave. Most of the models presented are also applicable to other
types of detonation using an equivalent of the released energy.
2.1
at the arrival time ta . This period is defined as the time from the
beginning of the explosion including the time of the detonation itself.
The pressure attains its maximum (peak overpressure pmax ) very
fast (extremely short rise time). The pressure then starts decreasing
until it reaches the reference pressure p0 , which in most cases is the
atmospheric pressure.
The duration of the positive phase td is the time taken to reach
this reference pressure. After this point the pressure drops below the
reference pressure to the maximum negative pressure pmin ; then
it rises again up to the atmospheric value. The duration of the
negative phase is denoted as tn .
The overpressure impulse (positive impulse) is the integral of the overpressure curve over the positive phase td .
The idealised (free air blast) form of the pressure wave in Figure 1 can
be greatly altered by the morphology of the medium encountered along its
propagation. Reflections depend on the geometry, the size and the angle
of incidence. The situation is much more complicated if there are several
reflection boundaries, as happens between buildings.
All parameters of the pressure-time curve are usually written in terms of
the scaled distance:
d
Z=
(1)
3
M
where M is the mass of the explosive charge in kg of equivalent TNT and d
the distance from the centre of the charge in m.
2.2
A widely used way of describing the form of an air blast wave is the socalled modified Friedlander equation (see e.g. Baker [17]), which proposes a
function for the positive phase of the air blast wave. The pressure p at time
t can be calculated with:
tbt
d
t
p(t) = p0 + pmax 1
(2)
td
The slope of the pressure-time function can be adapted with the form parameter b, which can be taken from Baker et al. [18]. All parameters of Equation
(2) can be taken from several diagrams and equations (e.g. Baker [17], Kinney and Graham [19], Kingery and Bulmash [20]). Also widely used is the
4
Numerical Investigations
3.1
Material Laws
(3)
The heat capacity ratio (assumed here as constant) is set to 1.35, whereas
the density of the air at normal atmospheric pressure (p0 = 105 Pa) is 0 =
1.3 kg/m3 , which leads to a specific internal energy at atmospheric pressure
of eint,0 = 2.1978 105 J/kg.
5
) eR1 + B(1
) eR2 + eint
R1
R2
(4)
sol
V
=
Vsol
(5)
where sol is the density of the explosive in solid state, is the current
density and V is the volume. The first and second terms in (4) describe
the behaviour during the detonation, whereas the behaviour of the air after
the energy release is represented by the last term, which corresponds to the
perfect gas law (3).
The parameters for the JWL equation can be found, for example, in
Dobratz and Crawford [2]. The values used for TNT and C4 explosives are
shown in Table 1. The release of pressure is combined with an ignition law,
which uses the detonation velocity vdet .
3.2
In this work calculations are performed using the finite element method as
well as using the finite volume method. All calculations use Eulerian meshes
since the deformation of structures is not considered (except in Section 4.3,
where ALE is used). An explicit time integration scheme is used.
The finite volumes (for the implementation used, see Galon [25]) use a
second-order approach for the flux in order to improve the calculation of
A
B
eint,0
R1
R2
sol
vdet
Unit
Pa
Pa
J/kg
kg/m3
m/s
TNT
3.738 1011
3.747 109
3.68 106
4.15
0.90
0.35
1630
6930
C4
5.98155 1011
0.13750 1011
5.4341 106
4.5
1.5
0.32
1601
8500
3.3
Simulation Methods
There are several methods of numerical modelling that can be used in order
to load a structure with an air blast wave. These methods differ in the
number of finite elements (or finite volumes) used and, accordingly, in the
calculation time.
The solid TNT model describes the mechanical behaviour of the (initially solid) explosive with a material law, e.g. the JWL equation (4).
A fine mesh is essential to obtain realistic results. The size of the elements in the region around the explosive should be much smaller than
the size of the explosive charge. These calculations are therefore very
expensive. To reduce the computation time, spatial partitioning (see
7
Casadei and Halleux [30]) can be used, which reduces the calculation
time for models with a large spectrum of element sizes.
The computational explosion and blast assessment model proposed by
Clutter and Stahl [31] is also a solid TNT model that uses only one
element for the explosive. This is possible using several special methods in combination with the Becker-Kistiakowsky-Wilson EOS for the
explosive.
1D to 3D mapping. This method is used, for example, in AUTODYN
(see remapping manual [32] and Birnbaum et al. [33]), and is also a
solid TNT model. A 1D calculation is used until the air blast wave
reaches a surface. Then the values of the density, energy, velocity, and
pressure are mapped onto a 3D mesh. Rose and Smith [34] map the
1D model to 2D when the wave arrives at the first surface and map
the 2D model to 3D when the wave arrives at a second surface with
another direction. This method is a combination of the solid TNT and
of the control volume methods. The calculation time should therefore
be shorter than for the solid TNT model. This method is provided only
by some numerical codes.
Compressed balloon method (phenomenological model). The pressure-time function resulting from a compressed balloon can approximately match the curve of an air blast wave. The amount of compression can be calibrated with the maximum pressure or the impulse. The
calculation time is smaller than for the solid TNT model. This method
is described in detail later on.
Control volume. A volume around the explosive is loaded by a
pressure-time curve. This curve can be taken, for example, from Kingery
and Bulmash [20]. An initial velocity has to be applied to the air particles. The air inside the control volume can be removed. This leads to a
reflection at the surfaces of the control volume, which can be neglected
if the control volume is small. In contrast, the air inside the control
volume can also be modelled. Then, the applied pressure wave is also
produced inside the control volume, which results in a reflection inside
the control volume. This method should be therefore validated through
comparisons with calculations using the solid TNT model. The computation time is similar to that of the compressed balloon approach.
The method is suitable for Eulerian calculations far away from the explosive (planar blast loads), where the control volume transforms into
p(t)
TNT
Pressure-time function
Solid TNT
Control
volume
Compressed
balloon
Compressed balloon
p(t)
Control volume
Method
air/structure
Solid TNT
Mapping
Compressed balloon
Control volume
Load-time function
reflections,
channelling
considered
considered
considered
(considered)
not
considered
calculation
time
very long
medium
medium
medium
short
3.4
First, the solid TNT model is tested for its capacity to simulate air blast
waves. A rectangular parallelepiped model with dimensions 2.2 x 0.6 x 0.6 m
is used to compare the experimental results of Kingery with numerical results.
The element size of the tetrahedral elements is chosen as 0.005, 0.0067, and
0.01 m respectively. Three symmetry surfaces are used to calculate only one
eighth of the model (i.e. 1.1 x 0.3 x 0.3 m). The other three surfaces of
the parallelepiped are defined as absorbing boundaries. The corresponding
mass of the explosive (TNT) in the 1/8 model is 1.6 kg. One eighth of the
charge is modelled as a sphere by setting the elements inside a sphere of
0.062 cm radius as explosive (spherical charge). The pressure is recorded
at the following measuring points: P1(0.5/0.1/0.1), distance 0.52 m from the
charge centre and P2(0.9/0.1/0.1), distance 0.91 m.
Figures 3 and 4 show that the solution using finite elements results in
smaller pressures and in an additional time lag compared with the experimental results. The time lag and the difference between the numerical and
experimental pressures increase with the distance. The finite element model
can represent the impulse for the intermediate and for the fine mesh investigated (see Table 3). For finite element calculations, this effect increases with
the distance from the charge. The influence of the element sizes is observable
but small. Also a cubical charge instead of the spherical one has, apart from
a time lag, only a small influence on the pressure time curve.
The cubic charge overestimates the impulse for longer distances due to a
non-perfectly spherical air blast wave.
To summarize, the finite element as well as the finite volume solution
can be limited used far away from the explosive. The finite volume solution shows a better description of the wave and also over a longer distance.
10
6.0E+6
0.
3
Kingery P1
solid TNT
0.25
5.0E+6
P1
0.05
Pressure [Pa]
4.0E+6
P2
0.1
0.05
0.45
[m]
0.40
0.20
3.0E+6
FE, coarse mesh, spherical charge
2.0E+6
1.0E+6
0.0E+0
0.0001
0.00015
0.0002
0.00025
0.0003
0.00035
0.0004
Time [s]
Figure 3: Solid TNT model, finite element and finite volume approach, results
at measuring point P1 (distance 0.52 m)
The calculation time for solid TNT models is long. The calculation time is
compared with other simulation methods in Section 4.2.
3.5
The computational effort required by solid TNT models is very large. In addition, the results of this model depend very often on the size of the elements.
The models which use a pressure-time function, in contrast, are sometimes
not realistic due to the fact that these models do not consider reflections,
street channelling and shadowing.
The idea of the balloon method is to use a region (balloon) with compressed air of initial size much bigger than the size of the solid explosive.
This method goes back to Brode [36]. He proposes to use a balloon filled by
gas at an initial pressure of:
pBrode =
ETNT ( 1)
+ p0
Vbal
(6)
where ETNT is the total initial energy of the charge, Vbal the volume of the
balloon, p0 the atmospheric air pressure, and the heat capacity ratio of the
11
2.5E+6
3
0.
solid TNT
0.25
Kingery P2
2.0E+6
P1
0.05
Pressure [Pa]
P2
0.1
0.05
0.45
[m]
0.40
0.20
1.5E+6
FV, coarse mesh, spherical charge
FE, coarse mesh, cubic charge
FE, fine mesh, spherical charge
1.0E+6
5.0E+5
0.0E+0
0.0003
0.0004
0.0005
0.0006
0.0007
0.0008
0.0009
0.001
Time [s]
Figure 4: Solid TNT model, finite element and finite volume approach, results
at measuring point P2 (distance 0.91 m)
gas in the balloon. In the literature (see e.g. Baker et al. [18]) it is customary
to relate the energy of the charge (ETNT , expressed in J) to the mass of the
charge MTNT expressed in equivalent TNT kilograms, by assuming that 1 kg
of TNT releases 4.52 106 J of energy, so that:
ETNT = MTNT 4.52 106
(7)
Brodes equation (6) is used by several authors, e.g. Ritzel and Matthews
[37] or Omang et al. [38], to calculate the far-field behaviour of an air blast
wave in an easy and effective way.
Trelat [39] presents a similar experimental method that can be used to
reproduce an air blast wave using balloons filled with burnable gas. The
difference in the air blast characteristics (peak pressure and impulse) between the air blast wave resulting from the ignition of the gas and from solid
explosive detonation can be fitted with a special equation (see [39]).
12
Model
Kingery
FE, coarse mesh, spherical charge
FE, intermediate mesh, spherical charge
FE, fine mesh, spherical charge
FE, coarse mesh, cubic charge
FV, coarse mesh, spherical charge
pmax,1
5.94E6
3.13E6
3.54E6
3.74E6
3.86E6
3.59E6
pmax,2
2.35E6
8.52E5
9.47E5
1.04E6
9.93E5
1.35E6
i1
132.8
116.4
131.8
137.0
120.2
137.5
i2
102.4
98.6
106.9
108.3
128.9
-
3.5.1
Several calculations using a compressed balloon are done in this work to test
how accurately the release of pressure from the balloon can represent the
air blast waves resulting from a real explosion. For our first investigation,
3D spherical balloons are used (see Figure 5). The balloon is the red zone
shown in the Figure. The surrounding air, shown in blue on the Figure, is
meshed by projecting the balloon surface radially and absorbing boundary
conditions are imposed along the external envelope of the model. Thanks to
symmetry conditions only one eighth of the model is meshed. The model uses
40 elements in the circumferential direction (for one eighth of the sphere),
and a total of 104,000 brick elements. Three different geometries are used
with different sizes of the balloon radius (Rb = 0.5, 1.0, 5.0 m) and of the
corresponding air radius (Rair = 5, 10, 50 m, respectively).
The computed results (in particular the maximum pressure and the impulse) are then compared with the experimental values resulting from a real
explosion. The impulse is the most important parameter as concerns air blast
wave effects on a structure. A comparison of the computed impulse vs. the
experimental values of Kingery and Bulmash [20] is presented in Figure 6, as
a function of the scaled distance defined by Equation (1). For the simulation
shown in this Figure a value of Rb = 0.5 m and Rair = 5 m is used. The
assumed mass of the charge is Mbal = 3.28 kg TNT equivalent, and from
Equation (6) this gives an initial balloon pressure of 107 Pa.
The comparison between the impulse using Kingerys equations and the
balloon model (Figure 6) shows that the latter is too small, but has the same
shape over a broad range of scaled distances, except in a small region immediately adjacent to the balloon. This trend is observed in all calculations done
with the geometry parameters given. Furthermore, there is no dependency
on the element size. The comparison of pressures shows that the peak pres-
13
balloon
air
air
R air
Rb
compressed balloon
symm. condition
Efit ( 1)
Vbal
(8)
where Efit = Mfit 4.52106 according to Equation (7). For the above example
14
400
Impulse [Pa s]
200
Radius of
the balloon
0
0
0.5
1.5
2.5
3.5
Figure 6: Comparison of the impulse of the balloon method against the experimental values of Kingery and Bulmash [20] for Rb = 0.5 m, Rair = 5.0 m, initial
balloon pressure 107 Pa
the best fitting charge is Mfit = 2.5 kg TNT, and this results into pfit =
7.55 106 Pa.
The equivalent TNT charge of the balloon Mbal is represented by the
upper curve in Figure 8. The two energy curves in this Figure show similar
behaviour in the logarithmic scale used for the over-pressure so that, by
computing the ratio between them, we obtain the scaling factor bal as a
function of pfit for this particular geometry (Rb = 0.5 m and Rair = 5.0 m),
see the red curve in Figure 9. The same procedure is then repeated for the
other values of the balloon radius, namely Rb = 1.0 m and Rb = 5.0 m,
showing very little dependency of results upon the balloon radius, see the
blue and yellow curves in Figure 9, respectively.
All these curves are very close to one another and approximately linear,
except perhaps for very low values of the initial balloon pressure, see comment
below. They can be fitted quite accurately by the following linear expression,
also plotted in Figure 9:
bal = 0.2205 log(pfit ) + 2.265
(9)
where pfit indicates the best-fit over-pressure in the balloon and must be
15
5E+6
Kingery
3E+6
Balloon method
Size of the balloon
Pressure [Pa]
4E+6
2E+6
1E+6
0E+0
0
0.5
1.5
2.5
3.5
Figure 7: Comparison of the peak pressure of the balloon method against the
experimental values of Kingery and Bulmash [20] for Rb = 0.5 m, Rair = 5.0 m,
initial balloon pressure 107 Pa
To summarize, the following procedure can then be used to obtain the initial
pressure of the balloon pbal for a given TNT equivalent mass M (in kg) and
a given balloon volume Vbal (in m3 ):
1. Calculation of the nominal energy of the explosive charge (ETNT , in J)
with Equation (7):
ETNT = M 4.52 106
(10)
2. Calculation of the over-pressure in the balloon in Pa with Brodes formula Equation (6):
pTNT = pBrode p0 =
ETNT
( 1)
Vbal
(11)
(12)
14
12
10
0
1E+6
1E+7
1E+8
pTNT
+ p0
bal
(13)
(15)
1.5
Rb=0.5
Rb=1.0
Rb=5.0
Proposed formula
pbal/p0=10
1.0
pbal /p0=5
pbal/p0=2.5
abal [-]
pbal/p0=5
0.5
0.0
1E+5
1E+6
1E+7
1E+8
1E+9
Figure 9: Scaling factor bal for different sizes of the balloon and proposed
linear formula
The use of the factor fbal deserves some explanation. The initial pressure
of the balloon results typically from the perfect gas Equation (3) and is
the combination of the higher internal energy (temperature) and the higher
density of the balloon gas, with respect to atmospheric values. Therefore, for
a given pressure of the balloon, only one of these two quantities is independent
while the other one results from the equation of state. So in principle one can
arbitrarily decide how to distribute the scaling factor bal between and
eint . Numerical tests have shown that in practice there is no influence of the
chosen distribution rule on results, in particular as concerns the computed
impulses. Therefore, it is proposed to adopt the square root factor as given
above in Equations (14) and (15).
With this simple procedure it is possible to calculate the initial state of the
balloon gas for a given balloon size and a given charge. While the proposed
procedure is applicable in general in any computer code, it should be noted
that the coefficients appearing in Equation (12) have been calibrated for the
specific Finite Volume model used in our code. If a different numerical model
(e.g. Finite Elements) or a different computer code is used, the calibration
18
should of course be checked. For example, when using the Finite Element
model for fluids also available in our code, slightly different values of the
coefficients in the expression of bal (12) have to be used to obtain the best
possible results with the proposed simple balloon model.
3.5.3
The proposed procedure raises the question, whether there are limitations on
the size of the balloon or on the ratio of the initial balloon pressure versus
the atmospheric pressure (pressure ratio pbal /p0 ).
The curve of bal (see Figure 9) has some limits. Clearly, values smaller
than 0.0 cannot be used, because they would result in negative pressure.
But, even using values of bal less than 0.5 represents an extrapolation of
the numerical data shown in Figure 9. These values should therefore be used
with care (a warning is issued by the code). In addition, if the value of bal
is relatively small, a small change of bal results in a large change of the
overpressure. Also in that case accuracy is relatively low.
Values of bal larger than 1.0 are also dangerous, because none of the
computed points in Figure 9 is above 1.0, at least for our own numerical
model. Therefore, also in this case a warning is issued in our implementation.
It is observable in Figure 6 that the distance to any structure should be
at least twice the radius of the balloon.
3.5.4
19
ial
r
ate
g
bin
or
s
Ab
tr
me
Sy
0.4
Symmetry
Balloon
1.0
[m]
y
0.4
Rb
Figure 10: Model to check the influence of the balloon size (resp. overpressure)
on the form of the pressure-time function
3.5.5
Rb=0.04 m, pbal/p0=1280
Rb=0.06 m, pbal/p0=280
Rb=0.1 m, pbal/p0=46
Rb=0.2 m, pbal/p0=5.2
Friedlander equation
Pressure [Pa]
4E+5
0.4 m distance
3E+5
2E+5
0.8 m distance
1E+5
0
0.0005
0.001
0.0015
0.002
Time [s]
Figure 11: Influence of the balloon size (resp. overpressure) on the form of the
pressure-time function. Distance to the charge: 40 cm resp. 80 cm
much more on the mesh size. The model of Section 3.5.4, see Figure 10 is
used with a size of the balloon of 10 cm and different finite volume sizes using
a regular mesh. The resulting pressure-time functions are shown in Figure
13. For this geometry an element size smaller than 2 cm is recommended,
which corresponds to at least 5 elements in the balloon region along each
spatial direction, which has also be shown in the first experiment.
The compressed balloon method is used to simulate several air blasts with
different geometries. The results are compared with solid TNT models and
with Kingerys experimental curves.
21
300
Kingery 2.5 kg
nseg=4
nseg=6
nseg=10
nseg=20
nseg=30
nseg=40
Impulse [Pa s]
200
nu
be
ro
fs
eg
100
en
ts
lair
lb
0
0.5
1.5
2.5
1/3
4.1
Explosion Experiment
The experiment presented by Alia and Souli [5] involves an explosion of 227
g C4 under spherical conditions, corresponding to 268 g TNT (see Cusatis
et al. [40]). The explosive is placed at a distance of 1.22 m from a rigid wall.
The pressure wave is reflected by this wall. The pressure is measured above
the explosive at a distance of 1.52 m from the charge. The geometry of the
numerical models is presented in Figure 14.
Five different numerical simulations of this experiment are performed:
two with the solid TNT model and the others with the balloon method.
Due to the fact that the wave reflected by the wall hits the pressure gauge
after a relatively long time, the wall is neglected in the numerical model with
solid TNT. With this simplification, the model can be built much smaller
(0.40.42.0 m, small model in Figure 14) than the numerical model used
e.g. by Coussin [41]. Absorbing elements replace the ignored air. All models
use symmetry conditions; only one eighth of the geometry is built.
The solid TNT model is meshed with cube finite volumes with a secondorder approach in both space and time. The centre of the explosive is built
22
l=0.01 m
4E+5
l=0.02 m
l=0.04 m
Pressure [Pa]
Friedlander equation
3E+5
0.8 m distance
2E+5
0.8 m distance
1E+5
0
0.0005
0.001
0.0015
0.002
Time [sec]
as a cube with 7 (coarse mesh) or 10 (fine mesh) finite volumes along each
direction. Around this cube, the mesh is generated in a similar way to the
model in Section 3.5.1. This leads to a spherical charge. The space between
the spherical charge and a bigger cubic box around the charge (see Figure
14) is also filled with cubic volumes and is defined as air using the ideal gas
law. The explosive is modelled by the JWL equation and the parameters
shown in Table 1.
For the calculations with the balloon method, two different balloon sizes
are used: one with a volume of 1.0 103 m3 (corresponding pressure 9.44
107 Pa), and the other one with a volume of 2.8103 m3 (corresponding pressure 2.74107 Pa). Both models are built with larger dimensions (1.20.82.0
m, full model), thus taking also into account the reflection at the wall.
The experimental result obtained by Alia is shown in Figure 15, where
it is compared with Kingerys pressure-time curve, which describes the peak
pressure and the positive impulse very well. The pressure wave in the experiment is slower than the wave in Kingerys model. A reason could be that
Kingerys experimental data is developed for TNT and not for C4. The experimental result of Alia shows also the reflection of the air blast wave at the
wall, arriving at the gauge point 5.2 ms after the detonation of the charge.
23
Absorbing material
1.52
Reflection wall
Pressure
gauge
Small model
2.00
0.40
1.20
0.60
The solid TNT model also shows a time shift which, however, is smaller
than for Kingerys equation. The positive impulse is similar to the experimental one. However, the pressure drop in the negative phase is much faster
than in Alias experiment. The difference between the two meshes is observable but small, while the difference in the calculation time is dramatic (1.75
h to 19.6 h).
The impulse and the pressure-time curve of the compressed balloon method
(full model) are similar to the experiment, if one disregards the time shift.
The compressed balloon method with higher overpressure (smaller volume of
the balloon) shows a larger time shift of the pressure wave than the model
with the smaller overpressure. The pressure-time curve is shifted in Figure
16 to allow better comparison of the different parts of the curve. Comparison with the results of Coussins finite volume calculations [41] shows that
the compressed balloon method has the same behaviour but is much faster
(about 1.3 h on a current desktop PC) in contrast with Coussins finite volume solution (about 46 h on a similar processor, but with 8 times more
memory).
24
250000
Experiment
Kingery (0.268 kg TNT)
Solid TNT model (FV), raw mesh
200000
Pressure [Pa]
150000
100000
50000
1.0E-3
2.0E-3
3.0E-3
4.0E-3
5.0E-3
6.0E-3
Time [s]
Figure 15: Results for the experimental model from Alia with different simulation methods
It can be seen, that the numerical solution follows in most parts Alias
experimental curve. If one neglects the time shift (which can also be seen
in Kingerys equation), the compressed balloon method can represent the
pressure-time curve very well. The reflection at the wall is fairly well represented, both in the arrival time and in the pressure magnitude.
It can be concluded that the compressed balloon method can be used to
calculate this type of problems. The time shift in the numerical solutions
could be partly due to the different kind of explosives. In any case, the shift
has very little influence on the behaviour of the structure.
4.2
2.5E+5
Pressure [Pa]
2.0E+5
1.5E+5
Reflection from the stand
1.0E+5
5.0E+4
Experiment
Balloon method: time shift 0.5 ms (Vbal =8e-3 m3)
FV solid TNT from Coussin
0.0E+0
1E-3
2E-3
3E-3
4E-3
5E-3
6E-3
7E-3
Time [s]
Figure 16: Results for the experimental model from Alia after shifting the
numerical results of +0.5 ms
the end of the street and at another point around the corner of the building
at the end of the street (see Figure 17).
One solid TNT model and two different compressed balloon models are
used to calculate the development of the air blast wave in this geometry. All
models use two symmetry conditions. The solid TNT model is built with
tetrahedral finite volumes (second-order approach for the flux). The element
size inside the explosive is chosen as 0.002 m, then it increases up to 0.02 m
in the rest of the model. About 130,000 elements are used.
The first compressed balloon model is built with cube finite volumes with
a constant element size of 0.01 m corresponding to about 126,000 elements.
The other model uses about 179,000 tetrahedral finite volumes with a maximum size of 0.02 m. The balloon volume of both models is set to 9.6105 m3 ,
which corresponds to a pressure in the balloon of 5.44 107 Pa.
Figure 18 shows the development of the air blast wave by using the compressed balloon method (tetrahedral finite volumes). The air blast wave is
reflected at 1.57 ms. The comparison between the numerical and the experimental results (Figure 19) at gauge location 1 shows that the pressure-time
curve resulting from the solid TNT model underestimates the impulse as
26
2.0
0.3
0.15
0.3
building
altitude 0.3
charge
0.24
building
altitude 0.3
2
gauge locations
well as the peak pressure. The finite volume model results in a fairly good
pressure-time function. Both compressed balloon models show a good description of the pressure-time curve. The time shift of the tetrahedral finite
volumes is bigger than when cube finite volumes are used.
The numerical results of the compressed balloon methods for the second
gauge location (see Figure 20 show also a good correspondence with the
experimental result, apart from the time shift. The pressure-time function
of the finite volume solid TNT model underestimates the positive impulse.
The pressure-time function proposed by Kingery cannot be used because
Kingery neglects any street channelling effect and thus predicts much smaller
pressures.
There is also a big difference in the calculation time. The solid TNT
calculation needs approximately 7.2 h on a current desktop PC. Thanks to
a much more regular mesh, the calculation time of the compressed balloon
method is much smaller: 1.2 h with tetrahedral finite volumes and 0.6 h with
cube finite volumes.
4.3
tained from on-site 3D laser scannings by the JRC 3D Reconstructor software (see Bostrom et al. [43]) and is shown in Figure 21. The length of the
model is 14.7 m, the width of the train is 2.68 m, the height is 2.14 m.
An aluminum sheet of 3 mm thickness is welded on a frame structure. The
aluminum sheet is modelled using quadrilateral shell elements. A Lagrangian
formulation for all solid material is used. An erosion criterion is used for the
aluminium sheets with a Van Mises failure limit of 245 106 Pa. The cross
section of the frame structure tentatively has the standard profile IPE80.
These profiles are placed near the doors and at the upper border of the
carriage. The floor of all trains is considered rigid since it is a very stiff
construction and since in most terrorist attacks the reported displacements
and failure of the floor were small.
Laminated safety glass is used for the windows (for the simulation method
see Larcher [44]). Since the failure of the laminated glass can not be represented with the element sizes used for the calculations, a displacement
criterion of 30 % of the span of the windows is used. Reaching this limit, the
border of the window is eroded so that the air can escape.
A charge of 10 kg TNT is located in the centre of the train. The diameter
of the balloon is chosen to 0.7 m. The mesh size of the fluid mesh is set to
28
4E+5
Experiments
Kingery analytical, hemispherical, reflected
FV solid TNT model
FV balloon method, TET 2 cm
FV balloon method, CUB 1 cm
Pressure [Pa]
3E+5
2E+5
1E+5
0E+0
0.001
0.0015
0.002
0.0025
0.003
Time [s]
0.1 m, the mesh size of the structural part is set to a maximum of 0.1 m.
Non-conforming fluid-structure interaction is used to allow calculation of the
fluid even after failure and erosion of parts of the structure.
The development of the pressure, the displacements, and the failure of the
structure are presented in Figure 22. The air blast wave resulting from the
compressed balloon destroys the structure immediately around the explosive.
At t = 13 ms all windows are failed due to reaching the failure criterion. The
extremities of the carriage are also destroyed due to the channelling effect
and the reflection of the wave at the closed ends. The reflection leads to
higher peak pressures and impulses at the ends than in the other parts of the
tube. In comparison to the terrorist attack in Madrid near Atocha station,
where a charge of 8 to 12 kg was used, the failure of the structure is similar,
even though the frame structure of the real train carriages involved is not
known in detail. Using fluid-structure interaction the risk inside the train
can also be calculated using the impulse and the pressure (see Larcher et al.
[45]).
29
1.4E+5
Experiment
FV balloon method, TET
FV balloon method, CUB
FV solid TNT model
Pressure [Pa]
1.2E+5
1.0E+5
8.0E+4
0.0015
0.0020
0.0025
0.0030
0.0035
Time [s]
Conclusion
This paper presents several methods for simulating air blast waves, which
can also be used inside complex structures. Developments are done to better fit the compressed balloon method to experimental data. This method
allows a relatively fast calculation by taking into account reflections, shadowing and channelling. The overpressure of such a balloon can be calculated
by simple equations presented in this paper. Several models with different
balloon sizes, different element sizes and different geometries have shown that
the compressed balloon method produces representative impulses and peak
pressures. The limitation of the pressure ratio inside the balloon is discussed.
The compressed balloon method is only a phenomenological model. The
method cannot therefore represent all kinds of possibilities. The pressure in
the region nearby the balloon is not very accurate. The distance between the
balloon and the structure should be large enough.
The implementation of the compressed balloon method in existing codes,
which deal with fluid-structure interactions, is quite simple. The real explosive charge and part of the surrounding air inside a structure (balloon
30
References
[1] E. L. Lee, H. C. Hornig, J. W. Kury, Adiabatic expansion of high explosive detonation products, Tech. Rep. No. UCRL-50422, Lawrence
Radiation Laboratory (1968).
[2] B. Dobratz, P. Crawford, LLNL Explosives Handbook: Properties of
Chemical Explosives and Explosive Simulants, Tech. Rep. UCRL-5299;
Rev.2, University of California; Lawrence Livermore National Laboratory (1985).
[3] M. B. Liu, G. R. Liu, Z. Zong, K. Y. Lam, Computer simulation of high
explosive explosion using smoothed particle hydrodynamics methodology, Computers and Fluids 32 (3) (2003) 305 322.
[4] D. A. Cendon, V. S. Galvez, F. Galvez, Modelling explosions using ALE
meshes: The influence of mesh refinement in pressures and in efforts induced by blast/structure interaction, Structures and Materials 15 (2004)
173180.
[5] A. Alia, M. Souli, High explosive simulation using multi-material formulations, Applied Thermal Engineering 26 (2006) 10321042.
[6] G. Langdon, G. Schleyer, Inelastic deformation and failure of profiled
stainless steel blast wall panels. Part I: Experimental investigations, International Journal of Impact Engineering 31 (2005) 341369.
31
[7] M. M. Gram, A. J. Clark, G. A. Hegemier, F. Seible, Laboratory simulation of blast loading on building and bridge structures, WIT Transactions on the Built Environment 87 (2006) 3344.
[8] G. Langdon, G. Schleyer, Inelastic deformation and failure of profiled
stainless steel blast wall panels. Part II: Analytical modelling considerations, International Journal of Impact Engineering 31 (2005) 371399.
[9] B. M. Luccioni, R. D. Ambrosini, R. F. Danesi, Analysis of building
collapse under blast loads, Engineering Structures 26 (1) (2004) 63 71.
32
[10] P. Smith, T. Rose, Blast loading and building robustness, Prog. Struct.
Engng. Mater. 4 (2002) 213223.
[11] X. Q. Zhou, H. Hao, Prediction of airblast loads on structures behind
a protective barrier, International Journal of Impact Engineering 35 (5)
(2008) 363375.
[12] P. D. Smith, T. A. Rose, Blast wave propagation in city streets - an
overview, Progress in Structural Engineering and Materials 8 (1) (2006)
1628.
[13] A. M. Remennikov, T. A. Rose, Modelling blast loads on buildings in
complex city geometries, Computers and Structures 83 (27) (2005) 2197
2205.
[14] P. D. Smith, G. C. Mays, T. A. Rose, K. G. Teo, B. J. Roberts, Small
scale models of complex geometry for blast overpressure assessment,
International Journal of Impact Engineering 12 (3) (1992) 345360.
[15] G. E. Fairlie, Efficient analysis of high explosive air blast in complex
urban geometries using the AUTODYN-2D and 3D hydrocodes, analytical and experimental methods, in: 15th Int. Symposium on the Military
Aspects of Blast and Shock, 14-19 September, Banff, Canada, 1997.
[16] B. Luccioni, D. Ambrosini, R. Danesi, Blast load assessment using hydrocodes, Engineering Structures 28 (12) (2006) 17361744.
[17] W. Baker, Explosions in the Air, University of Texas Press, Austin and
London, 1973.
[18] W. Baker, P. Cox, P. Westine, J. Kulesz, R. Strehlow, Explosion Hazards
and Evaluation, Elsevier, Amsterdam, 1983.
[19] G. F. Kinney, K. J. Graham, Explosive Shocks in Air, Springer, Berlin;
Heidelberg; New York; Tokyo, 1985.
[20] C. N. Kingery, G. Bulmash, Airblast parameters from TNT spherical
air burst and hemispherical surface burst, Tech. rep., Defence Technical
Information Center, Ballistic Research Laboratory, Aberdeen Proving
Ground, Maryland (1984).
[21] D. Hyde, ConWep, US Army Waterways Experimental Station, US
Army (1991).
33
lEnergie
Atomique (CEA), Gif-sur-Yvette, France (2008).
[42] L. Feng, Modelling blast in urban area, Tech. Rep. No. 11 MSc Weapons
Effects on Structures Project Report, Cranfield University Weapons Effects on Structures, Royal Military College of Science, Shirivenham, UK,
(1997).
[43] G. Bostrom, M. Fiocco, J. Goncalves, D. Puig Alcoriza, V. Sequeira,
B. Chartier, R. Kiesser, F. Mariotte, M. Richard, P. Zamora, 3D reconstruction in nuclear security, ESARDA Bulletin 38 (2008) 1724.
35
[44] M. Larcher, Simulation of laminated glass loaded by air blast waves, in:
DYMAT 2009 - 9th International Conferences on the Mechanical and
Physical Behaviour of Materials under Dynamic Loading, EDP Sciences
(www.dymat-proceedings.org), 2009, pp. 15531559.
[45] M. Larcher, F. Casadei, G. Solomos, Risk analysis of explosions in
trains by fluid-structure calculations, Journal of Transportation Security, preprint, DOI: 10.1007/s12198-010-0038-z.
36
European Commission
EUR 24288 EN Joint Research Centre Institute for the Protection and Security of the Citizen
Title: Explosions in complex geometries - a comparison of several approaches
Authors: Martin Larcher, Folco Casadei
Luxembourg: Office for Official Publications of the European Union
2010 41 pp. 21.0 x 29.7 cm
EUR Scientific and Technical Research series ISSN 1018-5593
ISBN 978-92-79-15245-0
DOI 10.2788/73796
Abstract
For the design and calculation of structures loaded by air blast waves, especially from inside the structure,
assumptions on the applied load are needed. This paper presents several simulation methods for the air blast
loading of structures and their ability to be used for complex geometries. Experimental-analytical pressure-time
functions of spherical load conditions applied to the structure by disregarding the air are not applicable in such
cases because they do not account for reflections, shadowing and channelling effects. Fluid calculations, which
model also the solid explosive, are very expensive due to the extremely small elements for the explosive and
the air nearby.
This paper therefore presents a review of a well-known simulation method, which uses a balloon with
compressed air instead of the explosive. A procedure is developed which makes it possible to determine the
overpressure of such a balloon for a given size of the explosive more accurately than before. The pressure-time
function and the impulse-distance function of calculations using this method show good correspondence with
experimental-analytical data. The functioning of the method is verified against experimental results.
LB-NA-24288-EN-C
The mission of the JRC is to provide customer-driven scientific and technical support
for the conception, development, implementation and monitoring of EU policies. As a
service of the European Commission, the JRC functions as a reference centre of
science and technology for the Union. Close to the policy-making process, it serves
the common interest of the Member States, while being independent of special
interests, whether private or national.