A Review of S-ALE Solver for Blast Simulations

Abstract

Blast modeling and simulation is a very important field in the military land vehicle industry. Increasing demands for higher protection levels leads the engineers to more challenging design and simulation cases. In most situations, Arbitrary Lagrange Euler (ALE) method is the most well-known method for blast simulations and also for determining the effects of blast loads on structures. Various studies are performed for the effect of mesh size and the domain shape for traditional ALE solver of LS-DYNA. The newly implemented SALE solver is stated to give shorter simulation times and also less memory requirements using the advantage of structured mesh. In this work, the SALE solver is compared to the traditional ALE solver for mine blast in steel pod. Different mesh sizes and advection methods are used for comparison. In addition to the displacement, momentum and deformation pattern, the solution times and memory requirements are also examined. Fluid-structure interaction (FSI) performance for solid interfaces is reviewed, as well.

Full text

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
A Review of S-ALE Solver for Blast Simulations
İlker Kurtoğlu
FNSS Savunma Sistemleri A.Ş. Oğulbey Mah. Kumludere Cad. No: 11 Gölbaşı – ANKARA, TURKEY
Abstract
Blast modeling and simulation is a very important field in the military land vehicle industry. Increasing
demands for higher protection levels leads the engineers to more challenging design and simulation
cases. In most situations, Arbitrary Lagrange Euler (ALE) method is the most well-known method for
blast simulations and also for determining the effects of blast loads on structures. Various studies are
performed for the effect of mesh size and the domain shape for traditional ALE solver of LS-DYNA.
The newly implemented S-ALE solver is stated to give shorter simulation times and also less memory
requirements using the advantage of structured mesh. In this work, the S-ALE solver is compared to
the traditional ALE solver for mine blast in steel pod. Different mesh sizes and advection methods are
used for comparison. In addition to the displacement, momentum and deformation pattern, the solution
times and memory requirements are also examined. Fluid-structure interaction (FSI) performance for
solid interfaces is reviewed, as well.
1 General Introduction
Blast simulations take a great portion of the design of armored military vehicles. Determining the
deformations on the main structure and the human response are critical issues in military industry.
Previous studies are performed for comparison of simulations with the field tests and increasing the
accuracy in the simulations [1-2]. In those studies, small scale models are used both in simulations
and tests. For the full scale vehicle simulations, larger ALE domain and finer ALE mesh are needed in
order to capture the interface forces and hence estimating the impulse more accurately. Larger ALE
domain means higher computation times and higher memory requirements. This brings challenges in
supporting the projects within the schedule and also hardware investment.
The S-ALE solver implemented recently is stated to reduce the solution time and the memory
requirements [3]. It also brings simplicity in defining the ALE domain with a couple of keywords. Since
there is no external mesh generation requirement, it also reduces the keyword sizes of ALE domain
from gigabytes to a few kilobytes.
2 Blast in Steel Pod
The validation of military vehicles against mine blast is defined in relevant standard [4]. In this
document, two alternative blast testing methodologies are defined. The one that is used in this work is
the steel pod case, where the soil effects are eliminated and only the effect of explosive plays role in
the deformation of structure. The pod geometry is shown in figure below.
Fig.1: Steel pod specifications [4].
For buried charge, the reference explosive is stated as TNT [4]. However, for blast in steel pod, the
explosive is stated as PETN or C4 with their TNT equivalents. In this study, 6 kg TNT equivalent PETN
is used and mass of the explosive is given as 5.04 kg [4]. The explosive parameters are taken from
literature [5].

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
3 Model Information
For the impulse comparison of traditional ALE solver and S-ALE solver, the impulse on a rigid plate is
examined with different mesh sizes of ALE domain. For consistency of the fluid structure interaction,
the rigid plate is also meshed with the same size of elements as the ALE domain. The general view of
the model is shown in figure below.
Fig.2: General model view.
The materials in the ALE domain if filled with *INITIAL_VOLUME_FRACTION_GEOMETRY keyword
using appropriate parameters. The domain is fixed from the bottom in the Z direction with the keyword
*BOUNDARY_SPC_SET and this set is generated by using *SET_NODE_GENERAL. The coupling of
ALE domain with the structure is established with *CONSTRAINED_LAGRANGE_IN_SOLID keyword.
Before doing the final simulations, some trials are performed to optimize the coupling parameters in
order to prevent leakage.
For S-ALE solver, the new keywords, which are implemented in R9 release of LS-DYNA®,
*ALE_STRUCTURED_MESH and *ALE_STRUCTURED_MESH_CONTROL_POINTS are used [3]. It is
stated that the current implementation of S-ALE solver supports only the Donor Cell (1st order) and
Van Leer (2nd order) advection methods. For rigid plate impulse comparison and deformable plate
impulse comparison, only Van Leer advection scheme is used. For the simulations without a target
plate, hence no fluid structure interaction, both methods are examined.
The air is modeled with *MAT_NULL and *EOS_LINEAR_POLYNOMIAL with the values that can be
found easily in the literature and also from previous blast simulation studies. The explosive is modeled
with *MAT_HIGH_EXPLOSIVE_BURN and *EOS_JWL. The necessary parameters are taken from [5].
The steel pod is modeled with *MAT_SIMPLIFIED_JOHNSON_COOK but there is no deformation on
the pod as it expected by the experience from observations in field tests. The *CONTROL_ALE card
used in the simulations is shown below.
Fig.3: ALE parameters used in the simulations.
All the simulations are performed with LS-DYNA® R9.1 (SVN 113698) MPP version. Intel® MPI is
used in the studies and 40 cores are used in all simulations. In one case, which is 10 mm meshed
ALE domain, the memory requirements of traditional ALE solution made us to use double precision
version. In the rest of the simulations single precision version of LS-DYNA® is used.

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
4 Simulation Results for Rigid Plate
The interface forces comparisons of traditional ALE solver and S-ALE solver for different mesh sizes
are shown in figure below.
Fig.4: Interface pressures for traditional ALE solver and S-ALE solver.
By looking at the figure one can conclude that as the mesh size decreases, the interface pressure
converges to a peak value and the sharpness of the peak is also increasing. The total comparison of
the two methods is shown in the figure below.
Fig.5: Comparison of the two methods.
When the figure above is examined, it can be seen that as the mesh size decreases, the interface
forces become closer between to methods. When the mesh size is higher, traditional ALE gives higher
interface forces. This situation also affects the momentum transferred to the target plate. Only when
10mm mesh size is used, the momentum results are close for traditional ALE and S-ALE which is
shown below.
Fig.6: Momentum comparison for the target plate.
It is stated that the S-ALE solver reduces the completion time of the simulation when compared to the
traditional ALE method [3]. However, when the simulation times are compared, as the mesh size is
decreasing the traditional ALE gives smaller simulation time than the S-ALE.

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
Mesh Size
Traditional ALE
S-ALE
Ratio
20mm
8540s
7242s
0.848
15mm
18959s
20802s
1.097
10mm*
82582s
144149s
1.745
Table 1: Elapsed time comparison (* double precision is used).
The memory requirements (memory required to complete solution reported in d3hsp file) are also
compared and the results are presented in the table below.
Mesh Size
Traditional ALE
S-ALE
Ratio
memory
memory2
memory
memory2
memory
memory2
20mm
329M
23M
74M
21M
0.225
0.913
15mm
762M
50M
164M
44M
0.215
0.880
10mm*
2494M
149M
533M
132M
0.213
0.885
Table 2: Memory requirements for simulations (* double precision is used).
When the solution time steps are investigated in detail, it is found that the S-ALE solver time steps are
decreasing during the simulation. For all cases, the time step is defined with LCTM parameter in
*CONTROL_TIMESTEP keyword as 0.1 microseconds. The TSSFAC parameter in the same keyword
is taken as 0.5. The time step comparison is shown in figure below.
Fig.7: Time steps during the simulations.
After observing this situation, the air internal energy in all simulations is compared to check if there is
an abnormality in the trend of energy. By experience it is known that, having very low density and high
compressibility, air is a challenging material in blast simulations especially when there is a structure
coupling with air.
Fig.8: Internal energy comparison of Air material.
Although exactly the same parameters are used for traditional ALE and S-ALE, there is a significant
difference in the internal energy trend of the air. Several trials on different parameters were made to
eliminate this difference but it still remains. When the volume fraction of the explosive is examined, the
difference can be observed. The volume fraction for the explosive are shown in the table below.

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
@0ms
@0.2ms
@1.0ms
@2.5ms
Traditional
ALE
S-ALE
Fig.9: Volume fraction of explosive.
5 Simulation Results for Deformable Plate
The deformable plate simulations are only performed with 20mm mesh to see the difference between
traditional ALE and S-ALE. The behavior is very similar to the simulations performed with rigid plate.
The time step in the traditional ALE solution seems to be constant, but in S-ALE solution after some
time it starts to decrease. This situation eliminates the advantage of S-ALE solver from the point of
solution time. The momentum results are also different in deformable plate simulations.
Fig.10: Z momentum in deformable plate.
The distribution of volume fraction of explosive is very similar with the one shown for rigid plate
simulations. For this purpose, the pressure distributions are compared for the deformable plate case
and shown in figure below.
@0ms
@0.5ms
@1.0ms
@2.0ms
Traditional
ALE
S-ALE
Fig.11: Pressure distributions for simulations with deformable plate.

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
6 Simulation without Target Plate
After the findings in previous sections, two successive simulations are performed with both methods
without using the target plate, hence eliminating the effects of FSI. In these simulations, both Donor
Cell and Van Leer advection methods are tested. For all simulations, 20mm meshed model is used.
The total energy comparison of the simulations is shown in the figure below.
Fig.12: Total energy comparison
As it can be seen from the figure above, there some differences in the trend of the total energy and in
the final value, as well. The different advection methods seem to have similar trend for both traditional
ALE and S-ALE. For a detailed investigation, the explosive kinetic energy, steel pod internal energy
and air internal energy are also compared, respectively. Comparisons are shown in the figures below.
Fig.13: Explosive kinetic energy and steel pod internal energy comparison.
In the figure above, one can note that the kinetic energy of the explosive has similar trends when
traditional ALE and S-ALE are compared. Also, the trend in the difference between advection methods
are again similar. However, when the same advection method is compared for traditional ALE and S-
ALE, the peaks are different although the trends seem to be similar. A similar behavior is also
observed in the internal energy for the steel pod. The air internal energy trend has also some
differences as shown in figure below.
Fig.14: The air internal energy comparison.

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
Pressure distribution and the volume fraction of the explosive are also investigated and some
differences are observed between traditional ALE and S-ALE. The pressure distributions and volume
fractions are shown in figures below.
@0.10ms
@0.25ms
@0.50ms
@0.75ms
Traditional ALE
(METH=1)
S-ALE
(METH=1)
Fig.15: Pressure distribution for Donor Cell advection scheme.
@0.10ms
@0.25ms
@0.50ms
@0.75ms
Traditional ALE
(METH=1)
S-ALE
(METH=1)
Fig.16: Volume fraction of explosive for Donor Cell advection scheme.
@0.10ms
@0.25ms
@0.50ms
@0.75ms
Traditional ALE
(METH=2)
S-ALE
(METH=2)
Fig.17: Pressure distribution for Van Leer advection scheme.

11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
@0.10ms
@0.25ms
@0.50ms
@0.75ms
Traditional ALE
(METH=2)
S-ALE
(METH=2)
Fig.18: Volume fraction of explosive for Van Leer advection scheme.
7 Summary
In this work, a comparison between the traditional ALE method and recently implemented S-ALE
method is performed for mine blast in steel pod case. It is stated that S-ALE solver has much less
memory requirements than the traditional ALE and also have smaller simulation times in the identical
models. However, due to the differences in advection, air and explosive behavior, the time step
decreases during the simulation with S-ALE and this cause the extension of total simulation time
except 20mm mesh configuration. When the memory requirements are compared, it is definitely
obvious that the S-ALE requires less memory than the traditional ALE since there is no ALE mesh and
hence no keyword to read and process.
The reason for time step decrease cannot be determined for S-ALE. When the pressure distribution
and volume fractions are examined, differences between traditional ALE and S-ALE can be seen
easily. For convenience, the traditional ALE model is tried to be solved by LS-DYNA® R7.1.3 (SVN
107967) version but the time step suddenly decreased after the explosion and the simulation is
terminated with error. Hence, no comparison can be made between the versions R9.1 and R7.1.3 for
traditional ALE.
In general, METH=-2 is used in blast simulations and METH=3 is also tried. However, since the S-ALE
is only implemented currently for Donor Cell (METH=1) and Van Leer (METH=2) advection schemes,
no comparison can be made using the former advection methods.
8 Literature
[1] Kurtoğlu İ., Salihoğlu B., Tekin G., Taşan C., Validation of Mine Blast Simulations with Field
Tests, 9th European LS-DYNA Users Conference, Manchester, UK, 2013.
[2] Kurtoğlu İ., An Assessment of ALE Mapping Technique for Buried Charge Simulations, 10th
European LS-DYNA Users Conference, Wurzburg, GERMANY, 2015.
[3] Chen H., LS-DYNA® Structured ALE (S-ALE) Solver, 14th LS-DYNA International Users
Conference, Detroit, MI, USA, 2016
[4] AEP-55, Procedures for Evaluating the Protection Level of Armoured Vehicles Volume 2.
November 2010.
[5] Dobratz P.M., Crawford P. C., LNLL Explosives Handbook – Properties of Chemical Explosives
and Explosive Stimulants, 1985, Lawrence Livermore National Laboratory.
[6] LS-DYNA® Keyword User’s Manual Vol. I (R9.0), Livermore Software Technology Corporation,
2016.