Success Stories
Behind Armor Debris
Simulation and Analysis of Behind Armor Debris Fields
Hubert W. Meyer Jr, Mechanical Engineer, WMRD, ARL
Behind armor debris is a major cause of damage in military vehicles that have been perforated by a penetrator, bullet or fragment. The ability to predict the debris field resulting from attack by such a threat is critical to assessing and improving the survivability of our tactical systems. ARL’s Survivability and Lethality Analysis Directorate (SLAD) has the mission of providing such assessments to vehicle designers. ARL’s Weapons and Materials Research Directorate (WMRD) has been working to develop the capability to model numerically the behind armor debris resulting from armor perforation, for application to the SLAD mission.
Modeling of the debris field historically has been done by statistically analyzing data from carefully controlled experiments. The difficulty of collecting this information makes it an expensive and lengthy process. Supplementing these experiments with numerical simulations is a natural synergy, but has not yet been successfully exploited because previous computer systems were unable to cope with the daunting size of the problems.
 |
| CTH “Flat Mesh” simulation of the experiment, shown at 600µs after impact of the 30mm APDS round on the armor plate. |
With the addition of the latest computers to the ARL Major Shared Resource Center (MSRC), numerical modeling of these experiments is now within reach. The Eulerian shock physics code CTH was chosen to model the experiment. The experiment modeled as a demonstration of the technique consists of a 30mm Armor Piercing Discarding Sabot (APDS) round perforating a 1-inch-thick armor steel plate. The resulting behind armor debris impacts a large (2-ft x 2-ft), thin (1/32-inch) mild steel witness plate placed two feet behind the armor. Perforations made in the witness plate by the debris are measured, and conclusions drawn about the size, mass, spatial distribution and velocity of the debris field. This is painstaking work, but it results in a reasonably accurate characterization of the debris field.
The difficulty in modeling this experiment arises primarily from two factors. First, the experiment is inherently three-dimensional in nature, due to the random distribution of failure in the plate, so any simulation of the experiment must be done in three dimensions. Second is the wide range of length scales; the 1/32-inch witness requires a fine grid resolution that, when extended over the two-foot air space and large area of the witness, requires one-half billion cells for a relatively coarse resolution (two cells through the thickness of the witness plate). Compounding the problem, the small cell size requires a small integration time-step, so a huge number of computation cycles are required to fly the debris through the two-foot airspace.
 |
| CTH Adaptive Mesh Refinement (AMR) simulation of the experiment, shown at 600µs after impact. |
The result of modeling such an experiment with CTH can be seen in the graphic above, which shows the state at 600µs after impact of the 30mm APDS round on the armor plate, when the debris field is well developed but has not yet impacted the witness plate. This 3-D simulation ran on the IBM Opteron cluster (stryker), using 2048 processors, or 9 teraflops of compute power. To run the simulation to 1200µs, when most debris has perforated the witness plate, required five days, the equivalent of over 28 cpu-years. The simulation required more than 1 terabyte of memory, and wrote over 300 gigabytes of information to disk. A few years ago, solving this problem would not have been possible.
CTH itself provides a means to reduce the problem size; the technique is called Adaptive Mesh Refinement, or AMR. Using this technique, CTH refines the grid only in areas where necessary. As a result, the large empty areas of this problem are coarsely resolved, and grid refinement follows the fragments as they fly toward the witness plate. This 3-D simulation ran for 54 hours on 512 Opteron processors using a total of 0.5 terabyte of memory, still a daunting computation, but almost one-tenth the cpu-hours of the “flat mesh” simulation. One objective of the current work was to verify that AMR CTH will provide the same result as the classic, flat mesh version of CTH for this problem. The work showed that grid resolution in CTH has an impact on the predicted fragmentation, and must be carefully controlled. A difference between the finest resolution of target material in the AMR calculation and the constant resolution of the flat mesh simulation contributed to the differences seen in the debris fields. If resolution is consistent, AMR CTH was found to be an accurate and computationally effective substitute for flat mesh CTH for this problem. Another objective of this work was to verify the ability of the large-scale cluster computers of the MSRC running CTH to efficiently conduct extremely large computations on large numbers of processors. This work provided a realistic test that demonstrated scalability.
In a second part of this work, a fracture model is being developed to improve the CTH prediction. Researchers at Lawrence Livermore National Laboratory (LLNL) (Becker et al) have demonstrated with a Lagrangian code the effectiveness of providing a statistical distribution of fracture properties in simulations. Here the technique is incorporated into the Eulerian code CTH and applied to modeling this ballistic experiment. In a conventional CTH simulation, all cells containing target material have the same set of fracture model parameters, so all fail in the same way. The new model installed in CTH provides a spatially random distribution of values for the principal fracture model parameter, although in the aggregate its population is Weibull-distributed. This causes non-uniform, stochastic failure of the armor plate. The resultant behind armor debris field is strongly dependent on the nature of the Weibull distribution of the fracture parameter, as quantified by the Weibull modulus m, which is a user-supplied input to CTH. As an analogy, think of the Weibull modulus as determining the standard deviation of the distribution of the fracture model parameter.
 |
| Bulge of the rear of the plate showing damage just prior to breakout (60µs after impact) for m=0 (left) and m=2. Blue is no damage; red is fully damaged. Note the symmetry of the damage in the bulge for the m=0 case, and the stochastic nature of the damage for the m=2 case. As can be seen by comparing figures, a more realistic fragmentation of the target is obtained with the distributed fracture parameter approach. |
In a third part of this work, Jerry Clarke of ARL CISD is developing software based on the Interdisciplinary Computing Environment (ICE) which will automatically identify and quantify all contiguous bodies in a CTH calculation. This type of automatic analysis was not previously possible with CTH calculations. Called FragFinder, this software identifies regions (i.e., fragments) where the volume fraction for each material is above a certain threshold, and determines the volume and velocity of these regions. FragFinder was used to analyze the debris field in a conventional CTH simulation (m=0) and in a CTH simulation using statistical fracture (m=2).
 |
| CTH prediction of mass distribution of fragments compared to experimental results. Here the statistical fracture techique is applied to only the target. This figure indicates that with the proper choice of Weibull modulus, and with the statistical model applied to both the target and the penetrator, a more realistic debris field can be obtained than arises from the classic method of using a constant parameter (m=0). |
This work has shown that simulating SLAD’s behind armor debris experiment is now within the ability of MSRC assets, and simulations can be successfully exploited to supplement the expensive experiments. Furthermore, the new capabilities of statistical fracture and automatic fragment quantification make the technique more useful.
The author would like to thank Jerry Clarke for his efforts in developing the FragFinder and visualization capabilities for CTH, and for doing the visualizations for this article, and Stephen Schraml, James Moody and Thomas Kendall for their assistance in running the simulations on the new MSRC Opteron cluster, stryker.
This work has been accomplished under the Capability Applications Project “Simulating the Formation and Evolution of Behind Armor Debris Fields.”