IMPACT CODE VALIDATION & BENCHMARK PROJECT
A collective effort of the impact cratering modeling community
The objective
of this project is to provide a set of information to serve as a standard for
comparison and validation of impact cratering and fragmentation modeling
codes. The information set consists of
two parts. The first contains a
description of the initial conditions and measurements from laboratory and
field experiments that can be used to validate code calculations over a wide a
range of event sizes and geological materials.
The second part consists of several hypothetical impact events of
varying complexity that serve as benchmarks for comparison of the numerics and
physical models employed in various codes.
Different
codes work better for different problems. For example, simulations of
disruption events exercise different aspects of the material models than do
simulations of cratering events.
Similarly, the material properties that are important in studies of
impact melting and vaporization differ from those in studies of late-time
crater collapse. Thus, it is important
to characterize the best applications for various codes, and provide this
information to the scientific community to help prevent incorrect use of
codes. Toward this end, this project
includes simulations that test a gamut of physical mechanisms involved in
impact events and includes comments on the applicability of various codes
wherever possible.
At the same
time, it is recognized that codes are in a continual state of development. Details such as the version of a code and
the specific input file used in a given simulation are recorded. Additionally, the information set is
designed to be extensible to accommodate future improvements and the
development of new codes.
Comments on modeling impact events
In validation
of the codes, the dominant uncertainties lie in the type, extent and accuracy
of the material modeling. For the
purposes of choosing benchmarks, it is important to know what benchmark
exercises what part of the material model.
For example, in the early time coupling, melt and vapor are determined
by the thermodynamic models, and of
course, sufficient numerical resolution.
For just the initial flow field in a nonporous material, a Mie-Gruneisen
or a Tillotson equation of state (EOS) suffices. Porosity adds considerable complexity, in which case an explicit
crush model is necessary. (A
low-density material is not the same as a porous one. The low density does not have the "P-dV" work of the
crush in it.) A nonlinear elastic model
such as a Murnaghan model will not have the energy dissipation necessary at the
higher velocities.
For melt, one
only needs to determine which material reaches a particular pressure, then the
material is known to unload into melt states even if the code may not
explicitly model the phase change during the isentropic expansion. For vapor, one must have explicit
three-phase equations, and details such as the blow-off velocities depend on
both the front side and the backside of the vapor dome.
Gravity must
be included in all simulations of the later stages of crater growth for large
craters, and for any size of crater in cohesionless materials such as dry sand
or transient craters in water. Although
gravity is not difficult to model, it is important to get an initially
quiescent pressure and density state compatible with the EOS to avoid waves
running around in the problem having the magnitude of the initial pressure.
The various forms of strength models must
also be considered, including both yield models and fracture models. For ductile metals, a Von Mises or Tresca
yield condition is sufficient. For
granular geological materials with little cohesion, such as dry soils, there
must be a pressure component to the yield strength, a Mohr-Coloumb or a Drucker-Prager
model is generally used. There is
little evidence of significant size or rate effects in the shear strength for
non-cohesive materials such as dry sand.
For geological
materials such as rocks, the cohesive components of the yield strength become
important, so that tensions are possible.
The tensile strength is often observed to be strongly rate dependent, at
least in small-scale experiments. In large impacts the strain rates may be
sufficiently slow rate that the rate dependence is not too important. However, large-scale rocks certainly behave
differently than small lab samples usually picked for their pristine
character. For impact disruption
events, it is the tensile part of the failure envelope that is of primary
importance, while for cratering the shear part is dominant.
Finally there
is the case of highly porous materials, in which craters have been observed to
form mostly by compaction (as opposed to shear). In this case, an accurate porous compaction model is needed.
Validation Tests
Numerous
laboratory and field experiments of cratering and fragmentation have been
conducted under a variety of conditions.
The validation part of this project employs a set of experimental
results that provides stringent tests of the physical models in used in codes.
The
experimental results are drawn both from laboratory studies of impacts and from
field tests of explosion cratering.
Laboratory tests are useful because they provide direct information on
impact events. However, they are
necessarily conducted at small scales.
Field explosion tests are complementary in that they provide important
data at much larger size scales.
When
validating a code it is important to consider as many aspects of the impact
process as possible. For example, one
must assure that a simulation not only predicts the correct final size of a
crater, but also the kinematics of the crater growth, material flow, ejection,
etc. Therefore, the test results
included here were selected to encompass as many observables as possible and to
sample a wide a range of experimental conditions. Results are included for simple materials, such as water and
metal, and materials such as soil and rock, whose constitutive behavior is
quite complex.
The validation
test cases are summarized in Table 1 below.
Each case is given a unique designator of the form VXYn, where XY is one
of IC (impact crater), EC (explosion crater), IF (impact fragmentation) or EF
(explosive fragmentation), and n is an integer.
The number of
tests in the validation suite is fairly large, and will undoubtedly grow with
time. It is not expected that every
code will be applied to the full suite of tests. For example, many impact codes do not currently implement a
pressure-dependent strength model.
However, because the suite includes materials ranging from very simple
to very complex, virtually all impact codes should be applicable to at least
some of the test conditions.
The results
for impacts in water eliminate the need for a strength model. Furthermore, gravity only needs to be
included to model the later stages of crater growth, as it approaches its
maximum volume before collapsing. Hence
the water tests are relatively uncomplicated and should be the starting point
for validation testing.
The impact
event in an aluminum target is the next step in complexity. In this case, gravity is not needed, and a
simple Von Mises criterion is sufficient to model the strength as noted in the
previous section.
Results are
included for impact cratering in dry sand.
In this case, the strength model needs no cohesion, but must include the
pressure dependence. In order to
exercise the pressure dependence, results are included for dry sand, which has
a friction angle of 35 deg, and glass beads, which have a friction angle of
about 20 deg.
Several
experiments are included for alluvium targets.
This material model requires both a cohesion and pressure
dependence. Results are provided for
both small-scale impact and for large-scale explosion craters. The field tests were selected to cover a
wide range of sizes and as many diagnostics as possible. Models of the explosion tests may either
employ a model of the explosive detonation or simply an initial region of high
pressure, as outlined in Appendix A.
Simulation of
the experiments in porous targets requires a model of porous crush-up, which
must be coupled appropriately with the strength model, which is not included in
most codes at the time of this writing.
Experiments
are included for impact cratering and disruption of rock. As noted in the previous section,
simulations of these events must consider the rate dependent strength of rock.
Finally, the
test suite contains the results of explosive fragmentation tests using targets
made from weakly cemented basalt. The
experiments were conducted in a pressurized atmosphere as an analog of the
gravitational overpressure in large asteroids.
These are presently the only experimental results that address the
shattering of large, gravity-dominated asteroids.
Details of the initial conditions, material properties and test results are provided in Appendix A. This information, along with summaries of the corresponding code calculations will be made available to the public on the web at http://keith.aa.washington.edu/craterdata/.
Table 1. Summary of
Validation Tests
|
ID |
Source |
Target |
Remarks |
|
VIC1 |
1.9 km/s pyrex sphere |
Water |
Transient crater growth in simple
material |
|
VIC2 |
4.6 km/s glass sphere |
Water |
Transient crater growth in simple
material |
|
VIC3 |
1 km/s steel sphere |
Water |
Pressure histories in simple material |
|
VIC4 |
8 km/s aluminum sphere |
Aluminum |
Final crater size in simple material
w/strength |
|
VIC5 |
2 km/s polyethylene cylinder |
Dry Sand |
Crater growth, tracer particle
motions |
|
VIC6 |
2 km/s polyethylene cylinder |
Dry Sand |
Final crater profile at 500G |
|
VIC7 |
6 km/s aluminum sphere |
Glass beads |
Final crater size in
low-friction-angle material |
|
VIC8 |
2 km/s polyethylene cylinder |
Alluvium |
Final crater profile in cohesive soil |
|
VEC1 |
20 Tons TNT, 10m deep |
Alluvium |
Final profile, mound growth, tracers
(Stagecoach) |
|
VEC2 |
500 Tons TNT, 38m deep |
Alluvium |
Final profile, tracers, ground motion
(Scooter) |
|
VEC3 |
4.4 KTons ANFO, surface |
Alluvium |
Final crater size (Minor Scale) |
|
VEC4 |
100 KTons Nuc., 190m deep |
Alluvium |
Final crater profile, mound growth,
tracers (Sedan) |
|
VIC9 |
2 km/s polyethylene cylinder |
50% Porous |
Final profile (500G) in highly porous
soil |
|
VIC10 |
2 km/s polyethylene cylinder |
70% Porous |
Final profile (500G) in highly porous
soil |
|
VIC11 |
X km/s XXX |
Gabbro |
Crater profile and subsurface
fracturing |
|
VEC4 |
0.42 KTon Nuc, 34 m deep |
Basalt |
Crater/ejecta profile, est of rupture
zone (Danny Boy) |
|
VEC5 |
0.5 KTon TNT, surface |
Basalt |
Crater/ejecta profile (Sailor Hat) |
|
VIF1 |
0.6 km/s Al cylinder |
Granite |
Collisional fragmentation of 1.9 cm
target |
|
VIF2 |
0.6 km/s Al cylinder |
Granite |
Collisional fragmentation of 3.2 cm
target |
|
VIF3 |
0.6 km/s Al cylinder |
Granite |
Collisional fragmentation of 17 cm
target |
|
VIF4 |
0.6 km/s Al cylinder |
Granite |
Collisional fragmentation of 34 cm
target |
|
VEF1 |
X gm PETN |
Grout |
Explosive frag of weakly cemented
basalt |
|
VEF2 |
X gm PETN |
Grout |
Exp frag of weak cem basalt, 75 psi
overpressure |
|
VEF3 |
X gm PETN |
Grout |
Exp frag of weak cem basalt, 1000 psi
overpressure |
Benchmark Tests
A
fundamental part of this project includes a set of simple benchmark simulations
that should be carried out to test a code against other codes routinely used
for impact modeling calculations. This approach provides a standard method for
comparing codes against each other, tests a code’s capabilities and accuracy
for different types impacts, and provides a public record of performance for
the same set of test cases.
The
set of benchmark simulations includes relatively simple cases, such as impact
into water, which should be addressable by virtually all impact
codes, as well as more complicated cases that are meant to test the behavior of
material models. The main purpose here
is not to validate the accuracy of a
code, but rather to compare its results against those of other codes. Hence, the benchmark includes some events
that cannot be performed in the lab or in the field, and involves detailed
comparisons of quantities that generally cannot be measured in experiments.
The
simulations are divided into two classes.
The early-time modeling
focuses on the early stages of the impact, dealing with the propagation of the
shock wave through the target and projectile. Therefore, the modeling should
focus on maximum shock pressure, shock pressure decay, internal energy,
temperature, melting/vaporization, and tracer particle histories during crater
growth.
For shock levels and melt/vaporization estimates, gravity can be ignored in the calculations. Additionally, it is not necessary to have a sophisticated EOS, as long as the available EOS does a good job at modeling the Hugoniot. This means that EOS can be as simple as Mie-Gruneisen. All that needs to be determined is if the material reaches the threshold shock pressure associated with melting/vaporization upon unloading from the shock state. A strength model is also not needed for the early stage calculations.
The second class contains the late-time modeling, which focuses on the slowing and possible collapse of the transient crater. In this case, both gravity and a good strength model are important. As noted in a previous section, the necessary complexity of the strength models varies with the material being modeled. For ductile metals a von Mises or Tresca yield condition is sufficient, whereas granular materials with little strength, like dru soils, must include a pressure component to the yield strength, such as in the Mohr-Coulomb or Drucker-Prager models. For rocky materials the cohesive component of the yield strength, which allows tension, becomes important. For cratering it is the shear part of the failure envelope that is crucial, and for disruption it is the tensile part of the failure envelope.
The late-time model results should focus on crater depth, radius, and volume as a function of time, transient cavity characteristics, final crater shape, tracer histories, and stress/strain fields.
Table 2 summarizes the benchmark simulations. Each simulation is identified by a designator of the form BXn, where X is either E (early time) or L (late time), and n is an integer. The details of the initial conditions for the simulations are provided in Appendix B.
Table 2. Summary of Benchmark Tests
|
ID |
Source |
Target |
Remarks |
|
BE1 |
1 km Al sphere, 5 km/s |
Water |
Early-time low speed impact in water |
|
BE2 |
1 km Al sphere, 20 km/s |
Water |
Early-time high speed impact in water |
|
BE3 |
1 km Al sphere, 5 km/s |
Aluminum |
Early-time low speed impact in metal |
|
BE4 |
1 km Al sphere, 20 km/s |
Aluminum |
Early-time high speed impact in metal |
|
BE5 |
1 km Al sphere, 5 km/s |
Dry Sand |
Early-time low speed impact in dry
soil |
|
BE6 |
1 km Al sphere, 20 km/s |
Dry Sand |
Early-time high speed impact in dry soil |
|
BL1 |
1 km Al sphere, 5 km/s |
Water |
Late-time low speed impact in water |
|
BL2 |
1 km Al sphere, 20 km/s |
Water |
Late -time high speed impact in water |
|
BL3 |
1 km Al sphere, 5 km/s |
Aluminum |
Late -time low speed impact in metal |
|
BL4 |
1 km Al sphere, 20 km/s |
Aluminum |
Late -time high speed impact in metal |
|
BL5 |
1 km Al sphere, 5 km/s |
Dry Sand |
Late -time low speed impact in dry
soil |
|
BL6 |
1 km Al sphere, 20 km/s |
Dry Sand |
Late -time high speed impact in dry
soil |
AUTODYNE, CTH,
SALEB, SAGE, SOVA, SPH, ZEUSS, ALE3D, GEODYN
Natalia
Artemieva nata_art@mtu-net.ru
Erik Asphaug asphaug@emerald.ucsc.edu
Willy Benz Willy.Benz@phim.unibe.ch
James Cazamias cazamias1@llnl.gov
Rob Cocker robc@lanl.gov
Gareth Collins gareth@lpl.arizona.edu
David Crawford dacrawf@sandia.gov
Galen Gisler grg@lanl.gov
Keith
Holsapple holsapple@aa.washington.edu
Kevin Housen kevin.r.housen@boeing.com
Boris Ivanov ivanov@lpl.arizona.edu
Don Korycansky kory@es.ucsc.edu
Jay Melosh jmelosh@lpl.arizona.edu
Patrick Michel michel@obs.nice.fr
Dugan O’Keefe dinosr@aol.com
Elisabetta
Pierazzo betty@psi.edu
Sarah Stewart sstewart@eps.harvard.edu
Zibi Turtle turtle@lpl.arizona.edu
Appendix A
Description of
validation experiments
VIC1 - Initial Conditions
Description: 1.9 km/s impact of a pyrex sphere at normal incidence into a water target. This was a quarter-space experiment in which the target fixture had a thick Plexiglas window as its front face. The projectile struck the target just inside the interface between the water and the window, allowing the crater growth to be viewed against the window.
Diagnostics: Various snapshots of the profile of the transient crater.
Projectile:
Radius:
Velocity:
Angle: Normal to target surface
Material:
Density:
Mass:
Shape:
Environment:
Pressure:
Gravity:
Target:
Material: water
Density: 1.0 gm/cm3
Properties: include any relevant material property information
Comments: Simulations of quarter-space experiments should use….
VIC1 - Results
Include plots of the results. Web site will also have data files (tab-separated text).
Comments:
Appendix B
Description of
benchmark simulations
BE1 - Initial Conditions
Description: Impact of a 1-km aluminum sphere into a water target.
Diagnostics: Provide …
Projectile:
Radius: 0.5 km
Velocity: 5 km/s
Angle: Normal to target surface
Material: Aluminum
Density: 2.7
Shape: Sphere
Environment:
Gravity: not required (or 1G if required by code)
Strength model: not required
Target:
Material: water
Density: 1.0 gm/cm3
EOS model:
Simulation duration:
Tracer locations:
Mesh:
X extent: 10 km in from impact point to edge of mesh.
Z extent: -6 km to +5 km (water for z<=0, vacuum for z>0)
Spatial resolution:
Comments:
Code
Diagnostics
1)
Peak shock pressure, temperature, energy, density
2)
Shock decay: pressure, particle velocity (temperature?)
3)
Crater morphology: diameter, depth, ??
4)
Melting/vaporization (shock pressure contours)
5)
Ejecta distribution and speed
6)
Post-impact temperature distribution
7)
Stress/strain distribution