In an increasingly challenging and complicated national security landscape, making any sort of improvement almost always comes at a cost – spanning time, resources or budget. But investments in predictive maintenance are among the few that can actually drive costs down while improving availability rates and the operational readiness of our nation’s most critical fleet assets and weapon systems.
For example, within the United States Air Force, advanced system health monitors allowed a contractor to pinpoint the location of a shorted wire that maintenance personnel could not find in the airframe of a B-1 bomber, potentially avoiding hundreds of hours of additional effort.
After 18 months of predictive maintenance implementation, the Marine Corps reduced maintenance hours for amphibious assault vehicles by 69% and reduced downtime for those ground combat systems by 32%. It also increased weapon system availability by 6%.
For the Army, predictive maintenance helped avoid four serious aircraft accidents – accidents that would have resulted in property damage of $2 million or more, loss or destruction of aircraft, and/or personnel fatality or permanent total disability.
These examples, all documented by the Government Accountability Office‘s 2022 report, illustrate the tremendous value of predictive maintenance to Department of Defense missions.
Predictive maintenance refers to the use of hardware, software, and service components to provide predictive analytics forecasting failures and degradations of internal systems. These techniques use real-time, condition-based monitoring and artificial intelligence/machine learning inferencing to identify expected failure points and determine remaining asset life, rather than waiting for equipment to break.
As it stands, military components employ several suboptimal maintenance practices, including “scheduled maintenance,” “no fault found,” and “run to failure.” Scheduled maintenance can be wasteful, as it dictates vehicles undergo maintenance at certain points in time, often when no maintenance is required. “No fault found” maintenance refers to when a maintainer checks and tests a system that was declared as faulty by an operator, but they cannot locate or replicate any anomaly. Run to failure is exactly what it sounds like: Run a system until it fails, then fix it. These strategies may be acceptable for your car or your refrigerator, but not for large, expensive, mission-critical weapon systems. None of these strategies maximize the efficiency of labor and materials that go into sustaining a level of operational readiness required for the future fight.
Predictive maintenance has failed to launch
Despite documented benefits and DoD guidance on predictive maintenance dating back to 2002, predictive maintenance remains in the pilot stage for many military organizations. In some cases, predictive maintenance techniques aren’t used at all. A recent survey of DoD operations, maintenance, and IT leaders found that 55% have not grown their predictive maintenance programs beyond the pilot stage; of that number, 6% are learning about predictive maintenance, and 21% are planning for it.
To move beyond planning and piloting, GAO noted that each military service should develop a comprehensive implementation plan for predictive maintenance that includes action plans and milestones for current weapon systems, outcome-related goals and objectives, a process for evaluating progress, and a framework to develop and track milestones. GAO also said each military service should track specific quantifiable metrics and goals for evaluating predictive maintenance, and each service should routinely monitor and report the results from predictive maintenance for major weapon systems. These are important recommendations. However, they don’t specifically address the technology requirements for successful predictive maintenance implementation. In the same survey, nearly three-quarters of DoD operations, maintenance and IT leaders said their current tooling fails to provide the data access and observability needed for effective predictive maintenance. The biggest obstacle to implementing or expanding predictive maintenance, according to 41% of respondents, is legacy operational technology infrastructure and/or systems.
Legacy tech is no excuse
The DoD is pursuing IT modernization, of course, but legacy operational technology (OT) will continue to be an inherent part of many systems. At this year’s Air, Space & Cyber Conference, Gen. Mark D. Kelly, commander of the Air Combat Command, noted that many aircraft and munitions are decades old. While they have been modernized, at a certain point “you’ve squeezed the last ounce of combat capability out of our sensors, weapons and platforms,” he said.
However, legacy OT does not inhibit predictive maintenance. With modern tooling and technology, it is possible – today – to implement effective predictive maintenance despite the well-known challenges associated with legacy weapon systems.
First, military organizations need complete observability of sensor data across the entire landscape of components onboard fleets and weapon systems. Today, onboard data collected from fleet assets and weapon systems consists only of fault codes and conditions metadata – not providing a full picture.
To improve this, military organizations need modern capabilities that enable continuous, full-take data capture at the edge. This level of data is typically captured from legacy OT networks comprising dozens of onboard computing devices with obscure and proprietary protocols, all communicating on a serial bus network.
Full-take data capture records every frame down to the microsecond, providing a clear picture of the health and performance of a vehicle or other system in real time. Absent this innovative capability, select data elements are usually only sampled at prescribed intervals. As a result, when an anomaly pops, it often escapes detection by the system’s fault monitoring or existing data collection capabilities.
Once full-take data capture is in place, modern tooling is able to compress the data for storage, translate it into a human-readable format, tag and enrich it, and then export it to a dashboard for real-time monitoring.
Then, by applying AI- and ML-informed behavioral analytics to the collected data set, leaders, operators and maintainers can learn how a system or device functions over time, so they can detect abnormal behavior that could indicate a mechanical problem or cyber event. The insight gained is the difference between data-driven decision-making and servicing, repairing or replacing equipment or systems only when they break.
Effective predictive maintenance helps increase operational readiness of critical fleet assets and weapon systems while driving costs – including time, resources and budget – down.
Michael Weigand is the co-founder and chief growth officer at Shift5.