New Army Research Lab model knocks three centuries of analysis down to two weeks
June 19, 2013
- ARL software helps military analysts better choose, or build weapons to overpower future threats.
- Army weapons developers and infantry training organizations are incorporating the model in current studies.
- Supercomputers plus legacy software provide unmatched analytical muscle.
Choosing and building future weapon systems just got easier and faster for military planners, thanks to a convergence of legacy and new software and the might of supercomputers at the U.S. Army Research Laboratory.
With the Smart Weapon End-to-End Performance Model, developed by the U.S. Army Research Laboratory, Army leaders can more quickly decide a lot quicker to modify or replace existing weapon systems.
SWEEPM can model the overall effectiveness of all types of munitions throughout the entire target engagement, from target detection through damage estimation with a modular Monte Carlo simulation.
Right now, the tool is being considered for adoption by the U.S. Army Armament Research, Development and Engineering Center in Picatinny, N.J. Engineers there say the SWEEPM model has the potential to simplify performance evaluations of both guided and unguided munitions through the use of modular program design and High Performance Computing capabilities.
"Unique to SWEEPM is the adaptive nature of the program which allows analysts at ARDEC to implement various Guidance, Navigation and Control schemes in SWEEPM via modules specific to that part of the munition," said Ingrid M. Dombroski, competency manager in ARDEC's System Analysis Division.
"Previously, the process for evaluating new GNC schemes required either writing new code from scratch or modifying existing legacy codes to fit a particular need, both of which are tedious and labor intensive endeavors", she said.
SWEEPM is also being used in a study requested by the Maneuver Center of Excellence at Fort Benning, Ga., to look at performance variables for the 40mm grenade.
With SWEEPM, ARL experts can pass along performance indicators about the 40mm grenade in about two weeks, but without the model and the use of supercomputers, it would take research analysts 316 years to perform such analysis.
"The model is a set of files and software that can be easily packaged and shared across a file share network or burned to a disk", explained Mary K. Arthur, the principle investigator who is credited with developing SWEEPM after recognizing a void in modeling and analyzing smart weapon systems from target acquisition through damage estimation. She is a mathematician in the Lethality Division of ARL's Weapons and Materials Research Directorate.
SWEEPM was conceived in 2008, and completed in April, 2013.
"Users have a couple of options as to what they would get. First, they can choose between the Linux and the Windows versions, which only differ in that the Linux version supports parallel processing. They also have the option of receiving just the executable, or receiving the entire source code and necessary files and directions for compiling on their own machines."
As a platform independent code, it can be run as a single Monte Carlo simulation or spawned off as parallel processes allowing for more complex physics-based features to be employed throughout the engagement scenario.
Embedded simulations include multiple firing platforms, scouts, moving targets and collaborative smart projectiles to model both beyond line of sight and non-line of slight smart weapon systems, where targeting data is passed from the scout vehicle to remote firing platforms.
"SWEEPM is a framework of models that integrates a number of legacy and stand-alone codes", Arthur said. It incorporates the U.S. Army Natick Soldier Systems Center's Integrated Casualty Estimation Methodology model and its analyst manual published in December 2003; the U.S. Army Materiel Systems Analysis Activity's Casualty Reduction model, which traces its early development to 1956. Weapons and Materials Research Directorate in-house models Total System Error Budget Analysis Model and the TargetMotion model, both developed in 2003; the Smart Weapons Encounter Model developed 2000; the Semi-Active Laser model developed in 2011; and the yLoFM, a utility model for determining vehicle loss-of-function from MUVES damage tables developed in 2012.
The Operational Requirement-based Casualty Assessment modeling system, created by ARL's Survivability/Lethality Analysis Directorate in 1996, is also embedded in SWEEPM. ORCA was developed for tri-service to allow assessments of Soldier performance following weapon-induced injury.
It permits casualty assessments to be performed in a consistent manner across virtually all types of military platforms, jobs, and weapon-induced threats. The ORCA computer code enables the analyst to calculate anatomical damage and the effect on individual performance as a result of exposure to one or more insult types including kinetic energy (fragments), chemical, and blast overpressure.
The ORCA modeling system incorporates previously developed as well as newly developed injury criteria models, algorithms and scoring systems to characterize human bio-response to trauma from various types of battlefield insults and derives estimates of Soldier performance degradation.
"These high-fidelity physics models require longer runtimes compared to simply flipping a coin to decide what happened; our Monte Carlo performance models can require tens of thousands of iterations (model runs) to produce statistically meaningful results. Long run times for each model run takes too long to produce results since we often need to run a large number of permutations on the inputs," explained Dr. William F. Oberle, chief of the Advanced Weapons Concepts Branch.
Researchers rely on high performance computers, housed within the ARL Supercomputing Research Center managed by ARL'S Computational and Information Sciences Directorate at Aberdeen Proving Ground. SWEEPM's high-fidelity physics sub-modules can only be run on these supercomputers.
Put in perspective, a single iteration in a Monte Carlo simulation takes one second and, "Let's say that you want to perform a 10,000 iteration simulation, then that simulation would take roughly 3 hours," Arthur explained. "Not too bad. However, say that a single iteration takes 20 seconds. That same simulation now takes closer to 2.5 days to complete".
"An engagement scenario with 10 moving targets and 10 guiding projectiles can easily take 20 seconds if not much more. Finally, say you need 100,000 iterations for good convergence of the data. The simulation will now take over 7.5 months to run to completion."
For the 40mm grenade study, ARL research analysts want to perform 1,000,000 simulations of 10,000 iterations each.
"I calculate that it would take 316 years to perform this analysis in serial even with a decent processor. By taking advantage of the HPC, I am confident that we can cut the runtime down to under two weeks as a conservative estimate," said Arthur.