Incubator Reports

Setting Reliability Requirements for Subsystems

PUBLIC RELEASE
December 2024

COMPLETED
May 2024

AUTHORS: Dr. Azad Madni, Dr. Dan Erwin, Dr. Michael Sievers, Dr. Ayesha Madni
UNIVERSITY OF SOUTHERN CALIFORNIA

Critical defense systems are required to be always available for use. The means for achieving availability goals, meeting needs, and identifying key availability parameters vary across organizations and platforms, making identifying opportunities and areas for efficiency improvements challenging. Although data related to system downtime are captured, they are seldom well-organized and in a form suitable for performing trade studies to identify promising candidate efficiency enhancements.

During this seedling effort, an AIRC research team from the University of Southern California contacted U.S. Department of Defense (DoD) personnel responsible for system maintenance to discuss their current approach to providing and documenting availability-related issues. A key finding is that although systems can be significantly different, the means employed and availability concerns are strikingly similar:

• Systems fail more often than expected.
• Systems take longer to restore service than expected.
• Parts needed for repair may not be available and are sometimes borrowed from other systems.
• Some systems get deployed with degraded capability.
• Some systems have internal spares that enable continued use until repairs are possible, but most systems tend to be “single-string.”
• Systems occasionally tend to undergo additional repairs found during preventive maintenance cycles.
• Predictive maintenance remains a desired but elusive goal.

The study examined the potential for using semi-supervised machine learning methods that look for patterns in vast amounts of data. Using synthetic data, they identified and used such patterns to discern availability trends suitable for performing trade studies and evaluating key factors such as costs, risks, maintenance depot efficiencies, and redundancies. They also created concepts for a trade study dashboard that defines and analyzes availability scenarios. Exemplar scenarios included:

• Depot maintenance time vs. inherent reliability
• Maintenance time outliers
• System usage vs time-to-maintenance
• Redundancy vs time-to-maintenance