IDI Named Edelman Finalist!
IDI was a key member of the US Army CECOM (Communications-Electronics Command) team that was named one of six finalists in the 2016 Franz Edelman Award for Achievement in Operations Research. INFORMS, sponsor of the award, is the largest society in the world for professionals in the field of operations research and analytics. The Edelman Award is their premier award for the practice of analytics.
IDI supported CECOM Training Support Division (TSD) in developing the Army’s Virtual Logistics Assistance Representative, VLAR for short. The VLAR system was developed to answer two questions:
- How do we take the collective knowledge of CECOM system operations and maintenance and codify that in a diagnostic and training tool usable by anyone, anywhere?
- How do we extend the CECOM knowledge base to remote locations and to remote operators, and how do we make these remote operators more self-sufficient?
CEDAT VLAR has proven to reduce troubleshooting time and increase accuracy. This directly translates into higher equipment availability and significantly reduces the risk for soldiers in combat. CEDAT VLAR helps keep the systems soldiers depend on for their lives up and running.
The developer of the VLAR concept and the overall team leader was our government customer Dave Aebischer of US Army CECOM. The software was developed by Veterans Technology Group. Causal Bayesian networks serve as the expert knowledge layer of VLAR. IDI trained the team of analysts from American Trade School that built these causal Bayesian networks. The IDI team included Joe Tatman, Amanda Hepler, Darrin Whaley, Suzanne Mahoney, Gary Smith, Sean Tatman, Dave Brown, and Dennis Buede.
The video of our team’s 40 minute presentation of VLAR can be seen at: https://www.pathlms.com/informs/events/533/thumbnail_video_presentations/26169
The executive summary for the project follows.
US Army soldiers, faced with conducting combat operations in Afghanistan, have numerous logistics challenges. One major challenge is maintaining complex electronic weapon systems and equipment used to conduct operations against the enemy and provide critical life-support functions. On-site technical assistance for power grid optimization and system diagnostics is costly as representatives require transport via helicopter or vehicle convoy and their availability is subject to weather, terrain, altitude, and threat constraints. Another major challenge is reducing fuel consumption and the corresponding need for resupply convoys.
The CECOM Training Support Division (TSD) developed CEDAT VLAR to address the onsite needs of soldiers in theater by mitigating knowledge gaps in the operating environment. CEDAT VLAR has proven to reduce troubleshooting time and increase accuracy. This directly translates into higher equipment availability and significantly reduces the risk for soldiers in combat. CEDAT VLAR helps keep the systems soldiers depend on for their lives up and running.
Causal Bayesian networks serve as the expert knowledge layer of CEDAT VLAR, which reside within a self-educating and self-optimizing application. CEDAT VLAR has harnessed the power of operations research to develop a suite of expert systems – for both probabilistic diagnostics and optimization – and in so doing has breathed new life into Army sustainment processes.
The CEDAT VLAR team developed a set of networks designed to assist in troubleshooting tactical power generators. The greatest initial challenges in this application area was a lack of hard diagnostic data. CECOM does have access to a very robust cadre of experts with years of experience working with soldiers and equipment in combat. The VLAR team developed a formal expert-based knowledge engineering process. The process, characterized by a cyclical implementation of four steps: Define, Structure, Elicit, Verify (DSEV), has proved a robust and comprehensive framework for building networks from expert knowledge. CEDAT VLAR identified faults in minimum time with near 100% accuracy and it was clear to CECOM that this methodology would have utility for a number of other applications. Demonstrating knowledge capture across multiple domains with consistent results was critical to the success of future VLAR applications. The CEDAT VLAR identified faults in minimum time with near 100% accuracy. The total cost savings from personnel reductions alone since 2013 are $20M per system VLAR- with projected future annual savings of $10M per system.
The Headquarters Power Optimization pairs loads to generators and turns off currently unneeded generators, reducing fuel consumption costs and saving lives. We conservatively estimate cost savings in battalion headquarters at $3,000,000 per year across the Army, a 17% reduction in generator fuel consumption. As we implement the optimization at brigade headquarters scale, we anticipate increased savings as a percentage of fuel consumption due to an exponentially increased solution space. In combat operations, reductions in fuel consumption will reduce the requirement for fuel resupply convoys. This reduction will ultimately save soldiers’ lives as the frequency of exposure to ambushes and Improvised Explosive Devices (IEDs) decreases.
CEDAT VLAR is changing the Army’s sustainment paradigm by applying knowledge engineering to equipment diagnostics and using advanced optimization techniques to electrical power grids in combat environments via virtualized application of expert knowledge. CEDAT VLAR enables cost savings and cost avoidance by reducing the Army’s logistics footprint: leaner sustainment personnel strategies, optimized fuel consumption, and better repair and supply decisions, all made possible by reducing uncertainty at the tactical edge.