Military procurement has exploited commercial off-the-shelf (COTS) hardware and software for decades to cut costs and improve performance. However, the maintenance enterprise has progressed at a slower pace. Like an aircraft carrier, it has been difficult to redirect, but in today’s dire budgetary environment the need to do so is ever more pressing.
Many solutions have been tried, with varying success. Yet, the challenge is great, and the military is dealing with aging airplanes, ships, and ground vehicles worn down by higher-than-expected operational tempos in more than a decade of fighting. Perhaps it’s time for an off-the-shelf solution.
The latest concept, Condition-Based Maintenance (CBM) Plus, or CBM+, was mandated by the U.S. Department of Defense (DoD) about two years ago, and to date no resounding success stories have emerged. The idea of knowledge-based maintenance, where fixes are applied only upon evidence of need, is a sound one. It can reduce unscheduled maintenance and accelerate depot maintenance. Still, CBM+ systems that have been designed by the government or its contractors are sometimes costly, inflexible, and complex, adding size, weight, and power (SWaP) requirements to military platforms without improving their tactical performance.
What to do?
Despite the conundrum, the answer to the problem may be staring us in the face – if the military can apply tools currently being used in the commercial market. For one thing, these techniques promise to leverage the mountains of data that military systems already generate without adding yet more on-board sensors and hardware. Only a fraction of this “big data” is used by military maintainers today because its sheer bulk exceeds the capacity of traditional databases and processing systems. But algorithms exist in the commercial world that can harness huge data sets and extract actionable information from them. Successful predictive analytic programs use existing asset data streams and consume little bandwidth. They are broadly applicable across the multiplicity of original equipment manufacturers (OEMs) found in both military and commercial environments.
Second, today’s predictive analytics tools would enable the military to exploit the economies of remote processing via cloud computing, using the Industrial Internet – the “Internet of Things” (IoT). Data can be collected and analyzed with no additional maintenance manpower and minimal intrusion on the daily operation of platforms.
Employing software-as-a-service or traditional software licensing models, the military could improve readiness and reduce costs, realizing the potential of CBM+ with minimal development time and engineering costs.
Multiple companies are applying predictive analytics solutions via cloud computing to business problems today. In the maintenance arena, industries such as oil and gas, mining, and commercial air transport have lowered operational costs while increasing asset readiness (see Figure 1). Predictive analytics – spurred by cloud computing – promises greater timeliness and more precise alerts than CBM+. An example is GE Intelligent Platforms’ Proficy SmartSignal solution, which has established a powerful track record in industries that simply can’t afford idle assets.
Although CBM+ is data-driven, its alert thresholds often are based on statistical averages. This methodology reduces costs, compared to past practices, but is limited. It can’t know what it doesn’t know. Because it anticipates failures based on past experience, it is less agile, less handy at spotting unexpected anomalies. Newer algorithms build models of healthy behavior using data points that fall outside of statistical alarm thresholds, so that they can identify not only the “usual suspects,” but unusual behavior that may be the precursor of new problems. These techniques enable customers to be both proactive and reactive in their maintenance approach.
Perhaps it’s time for the military to employ maintenance early warning systems that are more granular – tailored to specific airplanes, tanks, or submarines. This makes sense, as each weapons system has a different operational history that can’t be reduced to a statistical average. These models can be built using existing hardware instrumentation with very few “pings” per day. Data can be processed in-house or remotely. Analysts can monitor a larger number of assets, flag potential problems, and alert end users long before failures occur.