NOT YOUR GRANDFATHER’S COVART
by Rodney Stewart
As the Computation of Vulnerable Area Tool (COVART)—formerly called the Computation of Vulnerable Areas and Repair Time—approaches its 44th anniversary as a system-level vulnerability/lethality (V/L) evaluation tool, it is now supporting the third generation of engineers and analysts it has seen since its initial release in 1973. The specifics of the software have changed over the last four decades, but COVART’s mission has remained the same: provide decision-makers with the V/L information they need to make informed acquisition decisions.
COVART has definitely grown and evolved over the years. Since its initial creation (from the merging of the VAREA and HART codes developed in the 1960s), COVART development teams have added modeling techniques for new threat types, expanded the number of supported materials, provided additional methods for capturing system damage from threats, and improved assessments of threat penetration. And despite various quirks and idiosyncrasies that have arisen within the tool over time, the COVART we have today is more accurate and capable than any other COVART in history.
As shown in Figure 1, the Joint Aircraft Survivability Program (JASP) has been instrumental in making improvements to COVART possible. Over the last 15+ years, JASP has provided near-continual funding for model management and
sponsored numerous projects supporting a range of analysis needs. Recent examples of JASP-supported enhancements include the modularization of COVART, the integration of FASTGEN and BRL-CAD into COVART (i.e., COVART6), the verification of community-shared modules, and the development of common threat file formats supporting numerous Department of Defense (DoD) V/L models. Numerous Air Force programs and the Joint Technical Coordinating Group for Munitions Effectiveness (JTCG/ME) have also provided support for COVART in the past.
In FY16, funding under JASP project M-16-01 (the current model management project) led to the resolution of 31 software change requests (21 for COVART and 10 for FASTGEN) and resulted in an updated testing process, which will aid the verification of future releases. One of the major JASP-funded changes last year was the extraction of ProjPen-related material substitution decisions from COVART and the integration of these decisions in ProjPen. This change will improve the development team’s ability to conduct verification and validation (V&V) of ProjPen in the future by providing consistent material assumptions. Other fixes made under M-16-01 this year addressed an air blast distance calculation issue and a number of discrepancies observed in shared module verification testing. The Combat Effectiveness and Vulnerability Analysis Branch (EZJA) of the Air Force Life Cycle Management Center (AFLCMC) merged these improvements with updated FASTGEN and FLASH modules developed by the KC-46 program to create COVART 6.8, which was released in December.
Moving forward, JASP is providing the funding to realize the next step of COVART’s evolution. Under M-16-01, the COVART development team will start employing software parallelization, reorganize COVART’s damage estimation library to make it more portable, update the underlying code to a newer language standard, and remove older, deprecated features. The goal of this task is to place COVART in a position where it can be more easily maintained and aid future projects, which will standardize how the community addresses V/L questions.
Parallelization is the next step in improving the runtime performance of COVART. Efforts funded by JASP project M-12-02 and the KC-46 program investigated approaches to improve the speed at which FASTGEN traces rays. M-12-02 identified other ray tracing approaches with a potential speed improvement and experimentally implemented the Spatially Enumerated Auxiliary Data Structure (SEADS) ray tracer in FASTGEN to evaluate its effectiveness. The results of the study highlighted the need for the optimization of COVART processing to realize the benefits of modern ray tracing algorithms. KC-46-funded speed improvement efforts paved the way for future parallelization by creating a thread-safe version of FASTGEN. With this preliminary work completed, the next step is to optimize COVART to take advantage of this earlier work and to process rays in parallel. COVART 6.9, scheduled for a summer 2017 release, will include changes to improve ray tracing runtimes. Future development efforts will then investigate other areas of COVART to determine where parallelization may be used to make further speed improvements.
MODULARIZATION AND LANGUAGE STANDARD UPDATE
Apart from parallelization, another goal under M-16-01 over the next two years is to complete the modularization of COVART and update the underlying language standard to improve maintainability. COVART was initially modularized under JASP project M-04-03, which led to the creation of COVART5. Efforts under JASP project M-07-03 carried forward this modularization concept in merging FASTGEN, BRL-CAD, and COVART into COVART6. However, areas remained where the conceptual interfaces between the major functions of the software were not as clean as they should be. One of the areas needing most work is the component damage function of the program. In this case, all of the code related to the feature does not entirely lie within the boundaries of the library containing it. This fact makes COVART-specific damage estimation methodologies less portable and difficult to V&V separately. Efforts under M-16-01 in FY17 will address this functional boundary issue.
The primary programming language employed in COVART is FORTRAN; however, the software includes a mixture of subroutines written in FORTRAN77 and FORTRAN90. To support the features found in modern compilers, it is important to uplift the FORTRAN77 portions of the program to a newer standard. In FY17, the COVART development team will uplift the code in the component damage and vulnerable area calculation libraries to the FORTRAN 2008 standard, and future year efforts under M-16-01 will uplift other portions of the software until it is consistent.
Unfortunately, several features that have been added to the program over the last 20 years have not weathered the test of time. Some of them no longer work as intended, reference older programs that have disappeared over time, or simply do not provide benefit to the typical user. Under M-16-01, the COVART development team will remove deprecated features to improve maintainability and reduce the effort required to conduct a thorough V&V of the software. The features with the highest visibility on the shortlist for removal include support for the JTCG/ME fragment penetration equation library (sometimes called FRAGPEN) and the legacy mode of COVART6, in which users call FASTGEN/ RTG3 and COVART alternatively to conduct studies. Moving ahead in this regard is a necessary step to standardize how the V/L community approaches and performs V/L analyses.
While efforts under M-16-01 will address the maintainability of COVART, a second JASP project, V-17-04, will improve the fidelity possible for V/L studies. From its creation, COVART has assessed threat penetration two-dimensionally. The threat, acting as a point mass with a velocity and orientation, travels along a ray and makes contact with target components modeled as lines of sight with orientations. Figure 2 provides a generalized illustration of the current 2-D process/construct that COVART uses to estimate threat penetration and component damage.
The ray tracers commonly used by the V/L community process the target in three dimensions with three axes of rotation, but COVART requires them to simplify their outputs to fit the aforementioned 2-D construct. Recent improvements in the penetration methodologies employed by COVART (initially FATEPEN and now ProjPen with the near conclusion of JASP project M-15-03) have made it possible for them to estimate penetration using 6 degrees of freedom (6DOF).
Unfortunately, because COVART still forces the ray tracers to strip the second and third axes of rotation, the development team has had to make assumptions to provide the orientation information FATEPEN and ProjPen require. This practice has led to unintentional errors in COVART results because this assumed orientation often varies from the real orientation of the threat and target at impact. However, these errors can be eliminated by expanding COVART to handle 6DOF from the ray tracers through the penetration assessment. This capability is the goal of V-17-04, which will create the next generation of COVART, COVART7.
Without a doubt, COVART7 will be radically different from the COVART that first appeared in 1973. All ray tracing will be performed internally, and parallelization will lead to improved runtimes for both parallel-ray and diverging-ray analyses. In addition, the code itself will be more functionally organized, will be trimmed of unnecessary features, will reflect a more recent programming standard, and will have much less of the “spaghetti” processing found in previous versions. Finally, COVART7 will improve the realism and fidelity available for the modeling of threat penetration. As always, the overall mission of COVART will not change, but the upcoming versions of the tool will be, quite frankly, nothing like your grandfather’s COVART.
ABOUT THE AUTHOR
Rodney Stewart serves as a systems analysis engineer for the Air Force Life Cycle Management Center at Wright- Patterson Air Force Base in Dayton, OH. He currently serves as the Model Manager for FASTGEN and COVART and has more than 10 years of experience in the development, use, and management of these models. Mr. Stewart has a B.S. in mechanical engineering from the University of Colorado at Colorado Springs and an M.A. in leadership from Barclay College.