Professional Documents
Culture Documents
Fulltext01 PDF
Fulltext01 PDF
Abstract. The demand for protection of personnel in military vehicles has increased dramatically in
the past recent years. Numerical simulations may be an effective tool in order to obtain an estimate if
a military vehicle is able to withstand a specific threat, or to compare different design concepts against
each other. The success of the outcome is however dependent on several parameters, where valid input
data to chosen material models is the focus of the current paper. A detonating explosive acting on a
deformable structure is a highly transient and non-linear event. In field blast trials of military vehicles,
a standard procedure is followed in order to reduce the uncertainties and increase the quality of the
test. The explosive is buried in the ground, where the state of the soil must meet specific demands. Soil
may be considered as a three phase medium, consisting of solid particles, water and air. Variations
between the amounts in these phases affect the stiffness of the soil, and may affect the result of the
test. Laboratory experiments have been done in the present work to characterise the behaviour of a soil
material. The experimental outcome has formed input data to represent the material behaviour under
influence of the expanding detonation products, in combination with an analytical three-phase model.
The material data derived in this work along with material data from the literature is used in numerical
simulations representing field blast trials. The blast trials included explosive buried at different depths
in wet or dry soil. A dependence of water content of the soil can be shown, both in the past field trials
along with the numerical simulations. Even though some deviations exist, the simulations showed in
general good agreement with the experimental results.
1 INTRODUCTION
Soil characteristics are of interest in a number of different fields, covering both quasi-static and
dynamic loading. Quasi-static soil properties are for instance of interest in dam engineering
and construction, while dynamic soil characteristics are of concern regarding earth quake
physics and defensive protection against land mine explosions. Soil materials consist of many
single grains of different size and shape, forming a skeleton where the voids are filled with
water and air. Soil can thus be considered as a three-phase medium, consisting mainly of
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 2
solid grains with portions of water and air. When a small amount or no water at all is
included in the soil, the sample may be considered dry. If no air is included in the soil, the
sample is on the other hand said to be fully saturated. When the soil is under loading, it
undergoes a change in shape and compressibility. The volume decreases due to changes in
grain arrangements. Microscopic interlocking with frictional forces between the contacting
particles lead to bending of flat grains and rolling of rounded particles. If the load is further
increased, the grains eventually become crushed [1]. In quasi-static applications, the water
is often assumed to be incompressible. If the soil is compacted at a slow rate, the air and
portions of water are squeezed out of the soil skeleton. This assumption is however not valid
under shock loading, where water and air can undergo severe volumetric compression. In a
highly dynamic compaction such as an underground explosion, the air and water does not
have time to escape and may be considered as trapped in the soil [2].
In the NATO standard [3], recommendations for testing conditions of field blast trials with
explosive positioned in the ground are given. The soil shall be saturated with water prior to
testing, with the total wet density of the soil 2.20.1 g/cm3. The soil type in the ground shall
be sandy gravel, with 100% passing a 40mm sieve and maximum 10% passing a 0.08 mm sieve.
According to Table 1, definitions for gravel, sand, silt and clay based on particle size are given
based on three different classification systems [4]. Based on this, the soil type sandy gravel
may contain a particle size between 040 mm, with maximum 10% of the grains smaller than
0.08 mm.
Similar to rock and concrete, the strength of a soil is pressure dependent [5]. However, in
contrast to rock and concrete, soil has very low strength without some kind of confinement.
The soil material response depends on several parameters, such as for example grain size
distribution, grain density, in situ density and degree of saturation. It is common to use
a triaxial apparatus for characterisation of soil and other granular materials. A cylindrical
sample is placed into a pressurised chamber, where the stress can be individually set in the
radial and axial direction. It is thus possible to obtain both the yield function and the
compaction curve with this test. This method has been used for example in a derivation of
mechanical properties for sand with 6.57% moisture content by Laine and Sandvik [6], where
a triaxial apparatus was used up to 60 MPa confining pressure. Above 60 MPa, the data
representing the compaction curve was estimated by a 5th order polynomial function. The
material data provided by Laine and Sandvik is very often used in the industry and defence
applications involving land mine simulation due to its simplicity [7], but have also been used in
civil applications such as road side safety [8]. It is difficult to reach high hydrostatic pressures
with triaxial testing devices, but magnitudes up to 1 GPa has been reached for example by
Gabet et al. in [9]. However, a triaxial press capable of such high pressures becomes very large
and expensive. At the present, a common experimental method to reach higher pressure for
constructing a compaction curve (i.e. shock Hugoniot) is the plate impact experiment, carried
out at high strain rates. This has been done for both dry and water saturated quartz sand
by Chapman et al. in [10]. Bragov et al. [11] evaluated data using both a modified Kolsky
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 3
bar and plate impact experiments to obtain material parameters for dry quartz sand covering
a wide range of strain rates. Using the Kolsky bar, pressures close to 280 MPa was reached,
and close to 1.8 GPa for the plate impact. No significant strain rate dependence could be
found between the two experimental methods, indicating that a quasi-static characterization
would also be valid.
It is of interest to have one generic model which includes all the important parameters affecting
the soil response. Grujicic et al. [7] used an analytical and computational approach, where
a three-phase model was constructed in order to estimate sand response under blast loading
at different degree of saturations. An extension to include differences in grain size has also
been done in [12]. Regarding material data for larger particle sizes corresponding to gravel
or sandy gravel, not much data is available in the literature. Lekarp et al. provided yield
function data for sandy gravel in [1] which also has been used by Wu and Thomson in [8], but
no volumetric compression data has been found.
The aim of this work is to, by using relatively simple experimental techniques, characterize a
sandy gravel material previously used in blast trials. The blast trials are simulated with the
developed material data in addition to published data from the literature, and compared to
the experimental results. A three-phase modelling approach have been used to construct the
compaction curve at different degrees of saturation based on the characterized material.
In an earlier work by Zakrisson et al. [13], blast trials with the explosive placed in sandy
gravel at three different burial depths were carried out. The explosive charge was 0.75 kg
Swedish military plastic explosive m/46 (commercially known as NSP71) with a density of
1500 kg/m3 , consisting of 86% PETN and 14% fuel oil. The charge was formed to a cylindrical
shape resulting in a diameter to height (D/H) ratio of 3.The experimental setup is shown in
Figure 1, where the ground blast rig with the sides 3 2 m and total height 2.7 m includes a
hanging test module. The test module consists of a square target plate of steel quality Weldox
700E with dimension 600x600x8 mm, held in place with a plate holder with the most important
dimension shown in Figure 1. The target plate is able to deform into a tube with inner radius
250 mm, where the tube edge is rounded to a 15 mm radius allowing a smooth deformation
of the target plate. The maximum dynamic target plate deformation was measured with a
deformable crush gauge made of a small block of thin walled aluminium honeycomb mounted
inside the tube of the test module. Knowing the initial distance from the target plate to
the crush gauge, the distance the crush gauge has deformed is thus used to determine the
maximum dynamic deformation, max , of the target plate. Further, the residual target plate
deformation, res , was determined at the plate centre. The test module also includes ballast
weights, giving a total test module weight of 2120 kg. The maximum test module jump, Ztm ,
was then determined by using a deformable crush gauge in combination with a linear position
sensor. The momentum transfer, Itm , can then be estimated by
p
Itm = m 2gZtm , (1)
where m is the test module mass, g the gravity constant and Ztm is the test module jump.
measurement of momentum transfer.
In addition to 10 tests with explosive placed in sandy gravel, two tests with explosive posi-
tioning in a steel pot were done. This work focuses on the simulation of the explosion in sand,
while the steel pot simulations along with further information about the test procedure can
be found in [13, 14]. Three different depth of burial (DOB) were tested, 0, 50 and 150 mm,
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 4
627 mm
Test module
Soil box
Figure 1: Plate holder and test module to the left, the complete experimental setup to the right.
measured from the sand top surface to the mine top surface. For all three DOBs, the soil
was watered and compacted manually. Prior to the tests, a sample using a confined volume
of 1.218 dm3 was taken from the top of the soil box to determine the initial in situ density,
moisture content and degree of saturation. Also, two tests at DOB 50 mm were carried out
without watering the soil, and may thus be considered as dry. The square soil boxes were
made of wood with the side 950 mm while the height varied between 500 and 600 mm de-
pending on DOB of the mine. The height of the soil box was 500 mm for DOB 0 and 50 mm,
while the height was increased to 600 mm for DOB 150 mm. The stand-off to the target plate
was nominally held at 250 mm and measured from the target plate to the surface of the soil.
Since the boxes varied in height, the test module had to be adjusted vertically in the hanging
chains to keep the nominal stand-off distance, even though a slight variation was experienced.
The charges buried at 0 and 50 mm depth had a stand-off of 246 mm, while 150 mm DOB
had a 235 mm stand-off.
The complete description of all 10 tests is given by Zakrisson et al. in [13], while the average
initial states and the experimental results are given in Table 2 with deviations. The definitions
of volumes of a three-phase soil sample is shown in Figure 2, where the initial relative volume,
, for each phase is defined as
Vi0 0
= = i0 , (2)
V0 i0
where i = a, w, s represents air, water and solid grains, respectively. The saturation, S, is
determined as
S = Vw /Vp , (3)
and if the degree of saturation is 1 (i.e. 100%), the soil is fully saturated and only consisting
of solid grains and water. The moisture content is denoted w and determined as
w = mw /ms , (4)
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 5
where mw is the mass of water in the soil sample and ms is the mass of the sand particles.
The sandy gravel used in the tests is commercially distributed as concrete gravel, with solid
grain density of 2.70.03 g/cm3 and a particle size ranging between 0-8 mm.
Table 2: Initial conditions for the soil along with test results.
DOB 0 mm 50 mm 150 mm 50 mm
State Wet Wet Wet Dry
(kg/m3 ) 18401717 185448 31 18421717 177155
s0 (%) 63.80.5
0.5
1.3
64.51.1 0.5
64.20.7
w0 (%) 13.60.4
0.4
1.4
13.21.3 0.2
12.80.3
a0 (%) 22.60.9
0.9
1.4
22.32.7 0.6
23.00.8
w (%) 8.00.2
0.1 7.70.6
0.8
0.1
7.40.1
S (%) 37.61.5
1.5 37.35.4
3.9
1.2
35.60.7
max (mm) 92.21.6
1.7 102.51.82.3 0.4
72.30.5 0.1
92.20.1
res (mm) 84.60.9
0.9 91.82.0
1.1
0.8
59.51.3 0.4
82.20.4
Itm (Ns) 19902020 2623307373 28337743 220555
55
Air Va0
Vp0
Vw0 V0
Water
Vs0
Solid grains
3 MODELLING
The general numerical methods used in this work to describe the blast loading and structural
deformation are described in this section. The explicit solver of the commercial finite element
(FE) software LS-DYNA V971 R5.1.1 [15] was used for the calculations. The numerical
simulations in this work were executed on a Linux SMP cluster with 4-8 Gb of available
memory. Only one core was used on a 2.8 GHz dual core AMD Opteron 2220 processor, with
double precision. The numerical code has been used without modifications.
The LS-DYNA material model 16 together with a tabulated equation of state (EOS) is used
to describe the soil in this work. The deviatoric and volumetric response is decoupled, where
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 6
the deviatoric response is given by a yield function and the volumetric response is given by
the EOS [16]. The material model gives the yield function relation between the deviatoric
stress, q, versus pressure, P , in a piecewise tabulated form, together with a tensile cut-off
value. The initial density and the Poissons ratio are also given in the material model. The
EOS is given in piecewise tabular form of 10 points, relating pressure versus volumetric strain,
v , given by the natural logarithm of the relative volume based on the initial density given by
the material model. The EOS also relates the unloading bulk modulus, K, to the tabulated
pressure points. LS-DYNA extrapolates if needed [15]. Thus, Poissons ratio, , given in
the material model and the bulk modulus from the EOS complete the relation between the
isotropic elastic constants as
3K(1 2)
G= , (5)
2(1 + )
where G is the shear modulus. If the bulk modulus is varying, also the shear modulus will
vary, leaving Poissons ratio constant. Thus, the sound speed, c, in the material will also vary
since
p
c = K/. (6)
An example of the yield function and the volumetric response defined by the equation of
state is shown in Figure 3. The volumetric response is in the figure given in the form of
pressure-density, since it usually is more intuitive to relate to compared to volumetric strain.
Figure 3: Soil material input. Yield function defining shear stress against pressure shown to the left,
compaction equation of state representing pressure versus density shown to the right.
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 7
A soil sample can, as mentioned in the introduction and shown in Figure 2, be divided
into three phases, solid particles, water and air, with the volumes denoted Vs , Vw and Va ,
respectively. In an explosive process, it may be assumed that the air and water does not have
time to escape, i.e. all three phases exist during the compaction. The total volume is thus
the sum of the volume of each of the three individual phases. The procedure used here was
originally proposed by Henrych [17], and has previously been adapted by An, Fiserova, Wang
et al. and Grujicic et al. to mention a few [2, 4, 7, 18].
The initial relative volume for each phase is defined according to Equation (2) and can be
written as
V0 Va0 + Vw0 + Vs0
= = a0 + w0 + s0 . (7)
V0 V0
Equation (9) shows that the current density of the soil sample can be described by the sum
of the relative volumes of each individual phase. All three phases will experience the same
pressure at the same time, but each phase may experience different individual densities due
to different compressibility defined by its equations of state. For air, the equation of state can
be expressed in an adiabatic form such as
Va0
P = P0 , (10)
Va
where P0 is the atmospheric pressure and = 1.4 is the adiabatic constant. Equation (10)
can be rewritten using Equations (2) and (8) to
a a0
P = P0 = P0 . (11)
a0 a
where Cw0 = 1415 m/s, kw = 3, w0 = 1000 kg/m3 [4]. Using Equations (2) and (8) and
rearranging Equation (13) analogously as for the air EOS, we get
1/kw
P P0
w = w0 2 kw + 1 . (14)
w0 Cw0
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 8
Using Equation (9) and rearranging, the expression for the current density becomes
0
= , (15)
a + w + s
where a and w is calculated using Equations (12) and (14) for a given pressure. If a relation
for the current relative volume for the sand, s is known, Equation (15) is completed and the
pressure-density data to the equation of state can be derived.
where A, B, R1 , R2 and are material constants, is the relative volume and E is the internal
energy per unit reference volume. The JWL equation of state for the plastic explosive m/46
used in the blast trials have been calibrated and validated using cylinder tests presented in a
report by Helte et al. [19], and is used in previous applied work by Zakrisson et al. [14]. The
material- and JWL parameters for m/46 are given in Table 3.
Table 3: Material- and JWL-parameters for the plastic explosive m/46 [19].
D PCJ A B R1 R2 E0
kg/m3 m/s GPa GPa GPa kJ/cm3
1500 7680 21.15 759.9 12.56 5.1 1.5 0.29 7.05
The air is modelled with an ideal gas form of equation of state, defined as
p = (1 ) E, (17)
0
where is the current density and 0 the initial density while E is the internal energy per unit
reference volume. Also, = 1.4 is defined as the ratio between the specific heat at constant
pressure and volume, respectively. With initial density 1.169 kg/m3 , the initial pressure is 1
bar which results in an initial internal energy E0 of 250 kJ/m3 [20].
where A, B, n, C and m are material constants, eq and 0 are the equivalent plastic- and
reference strain rate, respectively. The homologous temperature, T , is defined as T =
(T Tr )/(Tm Tr ), where T is the current temperature, Tr the room- or initial temperature
and Tm the material melting temperature. In this work, the modified JC model is used for the
steel plate Weldox 700E, which undergoes large plastic deformation. The material parameters
for Weldox 700E regarding the modified JC model is given by Brvik et al. in [22], shown in
Table 4. The strain rate parameters and C has been adjusted and used in previous work by
Zakrisson et al. [14], in order to better correlate the model response with the experimental
strain rate experiments presented in [22].
Table 4: Weldox 700E material constants for the modified JC constitutive model [22].
Figure 4: Test setups for the material characterisation. In a), the confined compression is shown and
in b), the uniaxial compression is shown, c) shows the Brazilian disc test.
P
K= , (20)
v
where P is pressure defined positive in compression and v is the volumetric strain. In uniaxial
strain, Equation (20) becomes
P
K= , (21)
z
Evaluating Equation (19) for z using the assumption of uniaxial strain yields
(1 )
z = 3K z , (22)
(1 + )
Inserting Equation (21) in Equation (22), rearranged for the pressure, we have
(1 + )
P = z , (23)
3 (1 )
where z may be estimated from the surface pressure of the punch, knowing the press force
and punch area. Reported values in the literature regarding Poissons ratio for sandy gravel
is between 0.150.35 [8], where = 0.25 is chosen for this evaluation. Since Equation (23)
is only valid during elastic loading-unloading, it is only the maximum value on the pressure-
density curve and the following elastic unloading that is represented, i.e. not the non-linear
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 11
P (MPa)
k2 = 86608 700
P (MPa)
700
k3 = -95882 600
600
k4 = 52939
500 500
k5 = -14589
400 k6 = 1607 400
300 300
200 200
100 100
1.6 1.8 2 2.2 2.4 2.6 1.6 1.8 2 2.2 2.4 2.6
3
(g/cm3) (g/cm )
Figure 5: In a), measurement data is shown together with polynomial function and fully compacted
line. In b), the corresponding equation of state is shown.
compaction path. The measured maximum values are shown in Figure 5a, representing press
forces 50 kN, 140 kN and 240 kN, resulting in a maximum pressure of about 270 MPa.
Even though Equation (23) is only valid for the elastic loading-unloading, it is used during
the non-linear compaction for all measured press force points to evaluate the test with the
highest density up to 270 MPa. The curve, named Function in Figure 5a, is close to the
measured points at all evaluated pressures, and here used as a reasonable assumption also
for the compaction curve outside the measured points. A 5th order polynomial function is
constructed and used up to the full compaction is reached, given in Figure 5a with coefficients.
The full compaction is assumed linear from the theoretical maximum density of the individual
grains, using the bulk stiffness of Westerly granite with 21501 GPa/(g/cm3 ) [23] (also used
in [6]). The compaction curve is then used to form the input for the EOS model described in
Section 3.1, where 10 tabulated points at equally spaced densities are formed. Since LS-DYNA
uses extrapolation outside the data points, both the two last points is on the fully compacted
curve to ensure the fully compaction line is never exceeded. The elastic unloading is also
defined by 10 tabulated points for each of the 10 pressures. The first value is defined by the
linear unloading between the two first pairs of data points, while the two last points represent
the fully compacted line. The values at the intermediate points are interpolated between the
1st and the 9th value of the bulk modulus. Note that the input data in LS-DYNA is based
on pressure versus logarithmic relative volume, and not density as shown in Figure 5b. The
input data for the unloading is represented by the bulk modulus. The complete input data
for the virgin material tested is shown in Figure 5b is given in Table 6, together with the
corresponding density.
The initial relative volumes for the sample used to create the input data given in Table 6
is a0 = 0.392, w0 =0.017, s0 = 0.591. Using the polynomial function given in Figure 5a
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 12
Table 6: Equation of state data for sandy gravel with 0.94% average moisture content.
together with Equation (12), (14) and (15), a relation for P(s ) can be derived. For a given
pressure and desired values for the initial relative volumes representing the three phases, a
and w is determined by Equation (12) and (14). The current relative volume for sand, s , is
determined by interpolation from the previously derived relation P(s ). Equation (15) finally
gives the total density of the three-phase system, and completes the soil compaction curve.
The three-phase approach can now be used to create the EOS data corresponding to the
relative volumes representing the field blast trials, given in Table 2.
It is reasonable to assume that the theoretical maximum density for the three-phase medium
should reduce if a greater content of water exist in the medium, since the TMD for water is
lower than the solid grains. The density of air is low in comparison to water and sand and
is therefore neglected. A function for the theoretical maximum density of the three-phase
medium depending on the moisture content is introduced as
TMD = ww0 + (1 w) TMD,s . (24)
Thus, if no water is included in the medium then TMD = TMD,s , and if no solid grains is
included in the medium then TMD = w0 . The bulk stiffness of Westerly granite representing
the fully compacted material is identical for all states, but the start value at P = 0.1 MPa is
shifted to the TMD according to Equation (24).
in the vertical direction, where a uniaxial stress state is assumed. With the von Mises yield
stress defined as r
p 3
q = 3J2 = Sij Sij , (25)
2
where J2 is the second stress invariant and Sij the deviatoric stress tensor, the shear stress in
uniaxial stress state while the press pushes in the vertical direction becomes
q = |z |. (26)
ways testing, but at =2.234 g/cm3 points can be connected. As can be seen in the figure,
the inclination of the yield function would not deviate largely if any of the other points would
have been suitable. Since the deviatoric and volumetric response of the yield model used in
the simulations is decoupled, the model is thus independent of density and one yield function
for a general density is sufficient. The relation between shear strength and pressure is close to
linear for soft soils loaded up to 500 MPa [11]. In this work, the failure line is extrapolated up
to the unconfined strength of Pikes Peak granite at 226 MPa [23]] where a von Mises cut-off
is used, in analogy with previous work by Laine and Sandvik [6]. The complete yield function
is shown in Figure 3, and given in tabular form in Table 7. Further, a tensile cut-off of 0.001
Pa is used, which is reasonable since soils with larger grain size are generally considerable
cohesion less [12].
y
p
2
x
o
r
Figure 7: Yield function for sandy gravel constructed by uniaxial compressive test and Brazilian disc
test.
An attempt was made to determine the Poissons ratio during the BD test with use of digital
speckle photography (DSP) with CCD cameras shown in Figure 4c, using the commercial sys-
tem ARAMIS. The DSP measures movement of small speckle points (usually applied) on the
surface of a specimen, and is for example used to measure localized strains in characterization
of steel materials [25]. The sand has a natural speckle pattern due to the different sizes of
the sand grains, and no external pattern was applied on the surface. The Poissons ratio is
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 15
evaluated at the center of the specimen, and is according to Hondros [26] defined as
(3x + y )
= , (32)
(3y + x )
where x and y is the strains in the x- and y-directions, respectively. Since Poissons ratio is an
elastic parameter, Equation (32) is only valid in the elastic part of the diametral compression.
The Poissons ratio for two tests are shown in Figure 8, where it seems to be stabilizing
after a short time around = 0.25 0.32. Fracture occurs quickly in the specimen and the
elastic assumption fails, where the Poissons ratio becomes non-linear. This indicates however
that the assumption of = 0.25 used in section 4.1 seems reasonable, even though the DSP
measurements was sensitive where many tests resulted in unsatisfactory data.
Figure 8: Poissons ratio determined with DSP during Brazilian disc testing.
to both yield functions from the literature [1, 6]. The literature data was derived using a
triaxial apparatus. The sand & gravel data was derived with the primary aim to characterize
the development of cumulative permanent axial strain with the number of load applications
for different tests [1]. A number of 80000 cycles was done. How the multiple cyclic loading
affects the yield function compared to a single cycle test is not reported.
q (MPa)
1.5
500 1
Sand [6]
0.5
Sand & gravel [1]
Sandy gravel (present work)
0 0
1.4 1.6 1.8 2 2.2 2.4 2.6 0 0.5 1 1.5 2 2.5
3
(g/cm ) P (MPa)
Figure 9: In a), equation of state data is shown, b) shows the yield function. Literature data is
compared with the present work.
5 NUMERICAL SIMULATIONS
The finite element models and the numerical simulations corresponding to the field blast
trials are described in this section. In order to simulate the large expansion of gases (such
as explosion in sand and air), an Euler description have been used. The structural plate
deformation is described with a Lagrange description, where frictionless contact is defined
between the surrounding structures. A penalty based fluid-structure-interaction (FSI) is used
to couple the state variables from the gases in the Eulerian domain to the structural parts
in the Lagrangian domain [11]. All models and simulations are representing the geometry of
the test rig used in the field blast trials. The experimental geometry is represented in 3D,
which is modelled using quarter symmetry, see to the right in Figure 10. The gas dynamics
of the blast process can however be considered as axisymmetric until the shock wave reaches
the structure. The gas expansion is therefore simulated in a 2D axisymmetric Eulerian model
until the symmetry condition is about to be violated. A map file of the state variables is then
written from the last state of the 2D simulation. The map file is then used to initialise, or
fill, the 3D Eulerian domain with the last state of the 2D simulation, where the sequence
is shown in Figure 10 from left to right. In this way, a denser Euler mesh can be used in the
initial 2D simulation to preserve accuracy of the shock wave build-up, while a coarser Eulerian
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 17
mesh is used in the 3D domain to save computer (CPU) time without significantly reducing
the accuracy. This approach has recently been used by Zakrisson et al. in [13].
Air
Test module
Explosive Target plate
Sand
Axis of
revolution
Figure 10: To the left, the 2D model is shown prior to detonation. In the middle, the shock wave
reaches the upper boundary of the 2D model. To the right, the 3D model is shown after mapping from
the last 2D state.
The FE model of both the initial 2D Eulerian model and the 3D Eulerian and Lagrangian
model is shown in Figure 10. The 2D model regarding DOB 0 and 50 mm has dimensions
500 750 mm while DOB 150 mm has dimensions 500 850 mm, with uniform quadrilateral
element size 0.5 mm leading to a total of 1492000 and 1700000 elements respectively. In order
to preserve the accuracy, a uniform 2D mesh size of 0.5 mm was found necessary in free air
detonation using identical stand-off, charge size and geometry in [14], and is thus used here
for detonation in soil as well.
Regarding the 3D Eulerian mesh, smaller element sizes were used closer to the axis of rev-
olution and towards the target plate, with coarser elements towards the boundaries. The
same Eulerian domain is used irrespectively of DOB, and consists of 174000 solid hexagonal
elements in total. The size and mesh distribution in the XZ-view is given in Figure 11, where
the sign # is followed by the number of elements along a distance. The element length bias
ratio across a distance is defined as B = Le,max /Le,min , where Le,max and Le,min is the largest
and smallest element side length respectively. For the 3D model, outflow is prevented on the
two symmetry planes and at the bottom surface, but allowed on the top and outward lateral
boundaries. The Eulerian domain consists of air, soil and explosive.
The Lagrangian domain consists of the test module, including the target plate and plate
holder, and is identical to the model used in [14]. The Lagrangian parts consist of 12180
elements in total, where the target plate is represented by 3600 elements. Only fully inte-
grated shell elements with 5 through thickness integration points have been used for structural
calculations in 3D, where also thickness change due to membrane stretching is accounted for.
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 18
800 mm
#30, B9 550 mm
Le,min = 5.0 mm
#30, B3.25
250 mm
Le,min = 4.3 mm
#10, B1
300 mm
#15, B6
600 mm
Le,min = 14.1 mm
#40, B6
Le,min = 4.5 mm
Figure 11: The XZ-view of the mesh distribution in the Eulerian domain is shown, with a total of
174000 elements. Denser element distribution is located laterally towards the symmetry axis as well
as vertically towards the position of the target plate. Bias B indicates the ratio of the largest/smallest
element length across a distance.
Two types of soil materials, with data derived in the present work and data from literature [6],
are used in the simulations and represent the soil conditions of the experiments presented in
Section 2. The simulation cases are presented in Table 8, together with the test conditions of
the soil material. The sandy gravel characterised in section 4 is named SG, and the sand data
derived by Laine and Sandvik in [6] is named S. For the soil type SG, the three-phase model
is used for all characterisations except Case 7, where the virgin (dry) material is used from
Table 6 but adapted to the pre-compacted dry density. Simulation cases 1-6 are based on the
average input data from all wet tests presented in Table 2, while cases 7-8 are based on the dry
tests. Cases 9 and 10 represent the high and low variation of the initial density and saturation
of the wet tests at DOB 50 mm, since there was a large scatter in both the soil state and the
measured response. Case 11 represent fully saturated soil, where no measured data exist for
comparison. However, it represents the conditions suggested for field blast trials in [3]. The
simulation cases are representing the condition of the test with the stand-off distance from
the surface of the sand to the target plate as stated in Section 2. Equation of state data for
the simulation models related to cases 16 and 8 are provided in Table 9.
The mapping from 2D to 3D is used in all simulations, and since all simulations in this work
has different initial conditions, an initial 2D simulation has to be done for every simulation
case. The actual time when the shock wave reaches the boundary in the 2D model deviates
due to the different soil states and burial depths, but is about 0.06, 0.3 and 0.65 ms with
the increasing DOB tested. In 3D, the simulation continues to 2 ms for DOB 0 mm, 4 ms
for DOB 50 mm and 8 ms for DOB 150 mm. At those times, the FSI force is close to zero
and the Eulerian domain is deleted since the blast acting on the structure has past. The
calculation is thereafter restarted to allow the residual deformation of the plate to converge,
and continues until 12 ms for DOB 0 and 50 mm, and 16 ms for DOB 150 mm. The maximum
plate deformation, max , is stored in the calculation, while the residual plate deformation, res ,
is determined with the inner edge of the rigid rig as reference. From the time of the maximum
plate deformation to the simulation end time, a mean value is calculated which represents the
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 19
residual plate deformation in the calculations. The impulse is determined by integrating the
total vertical FSI force over time, IFSI, where the end value represents the transferred impulse
to the quarter symmetry model. The presented values in comparison to the experiments are
thus multiplied by four. This is an identical approach as previously used in [14].
6 RESULTS
The results from all simulation cases are shown in Table 10. A relative comparison between
the numerical results for cases 18 and the experimental results presented in Table 2 is shown
in Figure 12. The filled symbols represent sandy gravel, while blank symbols represent sand.
The symbols diamond, square and circle represent maximum plate deformation, residual plate
deformation and impulse, respectively. Also, an error bar representing the experimental scat-
ter in results is given.
In general, no great difference is shown between the two soil parameter setups, even though
the sand model results in overall larger values compared to the sandy gravel model except
for flush buried explosive. The largest deviation in results between the two soil parameter
setups is shown at 150 mm overburden, where the sand model overestimates the response
between 0.633.6%. The sandy gravel model overestimates the residual plate deformation
and underestimates the impulse transfer with about 13%.
The dry soil condition shows good agreement to the experimental results for both sandy gravel
and sand, with total deviation in results between 3.36.3%.
By using the three-phase model, it is possible to tune the input data to the specific state of
the sand at the test. Case 9 and 10 uses the lower and upper variations of density and relative
volumes of the corresponding test at 50 mm depth of burial. The numerical comparison to the
corresponding experimental result is shown in Figure 13. With the upper limit of input data,
both maximum and residual plate deformation is close to the corresponding experimental
result. The impulse is however underestimated with about 20%, even though an increase
of 164 Ns can be seen when comparing Case 3 to Case 9. With the lower limit of input
data regarding Case 10, the comparison is less accurate. Case 11 represents fully saturated
sand, where no experimental data exist for comparison. A significant increase in both plate
deformation and impulse is seen compared to the other numerical and experimental results.
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 21
Figure 12: Numerical results for Case 18, relative to the experimental results given with confidence
bounds from the experiments. Black markers represents sandy gravel, white markers represents sand.
Diamonds, squares and circles represents max plate deformation, residual plate deformation and im-
pulse, respectively.
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 22
Figure 13: Numerical results for Case 9 and 10, relative to the experimental results given with confi-
dence bounds from the experiments. Diamonds, squares and circles represents max plate deformation,
residual plate deformation and impulse, respectively.
and sandy gravel is shown against the experimental results. However, the numerical results
using the sand material slightly overshoots the results for sandy gravel. One explanation may
be that the moisture content of 6.57% is still represented in the sand data, while sandy gravel
only consists of about 0.94% moisture content. For flush buried explosive, all numerical results
underestimated the measured values with about 1012% except for the sand material, which
underestimated the impulse with 23%. No reason has been found for this large difference in
impulse between sandy gravel and sand at 0 mm overburden. Experience from earlier work
shows that an underestimation of about 10% may be expected concerning the plate response,
when the explosive gas is expanding freely in air using a similar numerical approach as in
this work [14]. However, an underestimation of only about 11.5% is expected regarding the
impulse. At 50 mm burial depth, both models of sandy gravel and sand gave satisfactory
results regarding the plate deformation, with an underestimation 1.15.4% compared to the
average experimental result. The impulse was underestimated with 1216.6%. However, a
large deviation in the experimentally measured impulse was seen, and if the numerical results
are compared against the lowest experimentally measured impulse, the numerical deviation
is only about 2.5%. The largest deviation in results between the two soil parameter setups
was experienced at 150 mm DOB. The sand model overestimated the experimental results
up to 33.4%, while the numerical results using the sandy gravel model deviate 13%. One
reason for this deviation between the models may be that accurate soil response is of more
importance at larger burial depths.
The three-phase model was used to simulate individual tests at 50 mm DOB, using the upper
and lower values of density and saturation, where results are compared to the corresponding
experimental values. The impulse was underestimated with about 2026%. For the upper
values of the input data, the plate response was within 1.5% of the experimental values, while
the results using the lower values underestimated the corresponding experimental results with
8.612.1%.
The density and water content measurements in the blast trials were done manually using
a confined volume, hence there exist an experimental uncertainty regarding the state and
uniformity of the soil. There was a large scatter in the density and water content measurements
at 50 mm depth of burial in wet soil, along with the experimental results. Even though, the
large underestimation of the impulse in the corresponding numerical simulations can perhaps
be attributed uncertain initial conditions.
Altogether, the numerical results using material data for sand taken from the literature showed
in general good agreement to the experimental results, except at 150 mm depth of burial.
The material data derived for sandy gravel along with use of the three-phase model showed
acceptable results at all explosive burial depths. Further validation of the three-phase model
should if possible be carried out at a greater variation of saturations, with more reliable
determination of the initial state of the soil.
REFERENCES
[1] Lekarp, F; Richardson, I.R.; Dawson, A.; Influences on permanent deformation behavior of un-
bound granular materials, Transportation research record 1547:68-75, 1997.
[2] Wang, Z.; Hao, H.; Lu, Y.; A three-phase soil model for simulating stress wave propagation due
to blast loading, Int. J. Numer. Anal. Meth. Geomech. 28(1):33-56, 2004.
[3] AEP-55: Procedures for evaluating the protection level of logistic and light armored vehicles, 1st
Ed., vol. 2, NATO, 2006.
[4] Fiserova, D.; Numerical analyses of buried mine explosions with emphasis on effect of soil prop-
erties on loading, PhD Thesis, Cranfield University, 2006.
[5] Hawkins, A.B.; Aspects of rock strength, Bulletin of Engineering Geology and the Environment
57(1):17-30, 1998.
B. Zakrisson and H.-A. Haggblad | New Design Concepts in Light-Weight Armour for Vehicles 24
[6] Laine, L.; Sandvik, A.; Derivation of mechanical properties for sand, Proc. of the 4th Asia-Pacific
Conference on Shock and Impact Loads on Structures, Singapore, 361-368; CI-Premier PTE LTD,
2001.
[7] Grujicic, M.; Pandurangan, B.; Qiao, R.; Cheeseman, B.A.; Roy, W.N.; Skaggs, R.R.; Gupta, R.;
Parameterization of the porous-material model for sand with different levels of water saturation,
Soil Dynamics and Earthquake Engineering 28(1):20-35, 2008.
[8] Wu, W.; Thomson, R.; A study of the interaction between a guardrail post and soil during
quasi-static and dynamic loading, International Journal of Impact Engineering 34(5):883-898,
2007.
[9] Gabet, T.; Malecot, Y.; Daudeville, L.; Triaxial behaviour of concrete under high stresses: In-
fluence of the loading path on compaction and limit states, Cement and Concrete Research
38(3):403-412, 2008.
[10] Chapman, D.J.; Tsembelis, K.; Proud, W.G.; The behaviour of water saturated sand under shock-
loading, Proc. of the Conference of the American Physical Society Topical Group on Shock Com-
pression of Condensed Matter, Baltimore, Maryland, 834-840; 2005.
[11] Bragov, A.M.; Lomunov, A.K.; Sergeichev, I.V.; Tsembelis, K.; Proud, W.G.; Determination of
physicomechanical properties of soft soils from medium to high strain rates, International Journal
of Impact Engineering 35(9):967-976, 2008.
[12] Grujicic, M.; Pandurangan, B.; Coutris, N.; Cheeseman, B.A.; Roy, W.N.; Skaggs, R.R.;
Computer-simulations based development of a high strain-rate, large-deformation, high-pressure
material model for STANAG 4569 sandy gravel, Soil Dynamics and Earthquake Engineering
28(12):1045-1062, 2008.
[13] Zakrisson, B.; Wikman, B.; Johansson, B.; Half scale experiments with rig for measuring structural
deformation and impulse transfer from land mines, Proc. of the 24th International Symposium
on Ballistics, New Orleans, 497-504; DEStech Publications, Inc., 2008.
[14] Zakrisson, B.; Wikman, B.; Haggblad, H.-A.; Numerical simulations of blast loads and struc-
tural deformation from near-field explosions in air, International Journal of Impact Engineering
38(7):597-612, 2011.
[15] Hallquist, J.O.; LS-DYNA Keyword users manual, Version 971 / Rev 5 (Beta), Livermore Soft-
ware Technology Corporation, 2010.
[16] Malvar, L.J.; Crawford, J.E.; Wesevich, J.W.; Simons, D.; A plasticity concrete material model
for DYNA3D, International Journal of Impact Engineering 19(9-10):847-873, 1997.
[17] Henrych, J.; The dynamics of explosion and its use, New York, Elsevier, 1979.
[18] An, J.; Soil behaviour under blast loading, Dissertation, University of Nebraska, 2010.
[19] Helte, A.; Lundgren, J.; Ornhed, H.; Norrefeldt, M.; Evaluation of performance of m/46, (in
Swedish), FOI-R2051SE, Swedish Defence Research Agency, 2006.
[20] Sonntag, R.E.; Borgnakke, C.; Van Wylen, G.J.; Fundamentals of thermodynamics, 6th ed., New
York, Wiley, cop., 2003.
[21] Brvik, T.; Hopperstad, O.S.; Berstad, T.; Langseth, M.; A computational model of viscoplas-
ticity and ductile damage for impact and penetration, European journal of mechanics. A, Solids
20(5):685-712, 2001.
[22] Brvik, T.; Dey, S.; Clausen, A.H.; Perforation resistance of five different high-strength steel plates
subjected to small-arms projectiles, International Journal of Impact Engineering 36(7):948-964,
2009.
[23] Lawrence Livermore Laboratories: Geoscience parameter data base handbook: granites and
basalts, UCRL-15152, 1979.
[24] Jonsen, P.; Haggblad, H.-A.; Sommer, K.; Tensile strength and fracture energy of pressed metal
powder by diametral compression test, Powder Technology 176(2-3):148-155, 2007.
[25] Kajberg, J.; Wikman, B.; Viscoplastic parameter estimation by high strain-rate experiments and
inverse modelling - Speckle measurements and high-speed photography, International journal of
solids and structures 44(1):145-164, 2007.
[26] Hondros, G.; The evaluation of Poissons ratio and the modulus of materials of a low tensile
resistance by the Brazilian (indirect tensile) test with particular reference to concrete, Australian
Journal of Applied Science 10(3):243-268, 1959.