Military Embedded Systems

Predictive analytics and the industrial IoT: Thinking about military adoption


September 10, 2015

Todd Stiefler

Abaco Systems

The amount of attention being paid to the "Industrial Internet" or Internet of Things (IoT) has spiked dramatically over the past year or two. While some of this is clearly natural hype over a "new new thing," it is also true that the combination of brilliant machines, ubiquitous connectivity, and powerful analytics carries truly transformative potential. The question facing individuals and organizations across the world, including militaries, is: "How do we make the vision real?"

A recent report from global consulting firm McKinsey and Company said, “The ability to monitor and manage objects in the physical world electronically makes it possible to bring data-driven decision making to new realms of human activity – to optimize the performance of systems and processes, save time for people and businesses, and improve quality of life. From monitoring machines on the factory floor to tracking the progress of ships at sea, sensors can help companies get far more out of their physical assets – improving the performance of machines, extending their lives, and learning how they could be redesigned to do even more.”

The potential of the marriage of “big iron” and “big data” has not been lost on the military, either, and it seems like there is a conference on the subject almost every week. It hasn’t just been talk, either: Visionary leaders across the services are testing and implementing solutions that use digital information to drive better decisions and outcomes for operators of physical assets. However, for every decisionmaker willing to lean forward, there are many more skeptics who see the vision of the Industrial Internet as pie-in-the-sky; a pretty picture painted by consultants and academics that will never (and perhaps should never) come to pass.

Why the skepticism?

This skepticism is most likely born of three causes: First, the natural conservatism and risk-aversion of the military, which is a big bureaucracy with a mission in which it absolutely cannot fail; second would be past experience with condition-based maintenance (CBM) and other data-driven initiatives that generated uneven results. Another reason for skepticism about the IoT could be the daunting prospect of trying to move from the status quo to the space-age vision of autonomous-integrated-distributed everything that some well-meaning Industrial Internet enthusiasts have painted.

The benefits claimed by advocates of distributed sensing and data analytics seem diffuse, speculative, and far off in the future. The costs, conversely, are much more tangible and much more present. Sensors cost money and so do the servers to store the data they generate, while analytics software and data scientists are expensive as well. Changing organizational behavior requires leadership time and effort. Then there’s the unknown cyber-vulnerabilities opened up by connecting ground vehicles, aircraft, or ships to a network.

Each of these costs is real and all of these drivers of skepticism are rational and understandable. Overcoming resistance to the Industrial Internet in the Pentagon and out in the field will require a graduated approach that establishes incremental steps towards the end vision and concrete metrics for measuring value along the way. GE, for its part, calls this a “maturity model” and looks at it as a critical tool for organizations to use when asking if, how, when, and where they should begin their Industrial Internet journey.

The maturity model

That journey encompasses five basic stages of maturity: connect, monitor, analyze, predict, and optimize.

  • Connect: Gather data from all machines using embedded computing capability (programmable logic controllers, industrial computers, and data historians) and store that data in a central repository so that it can be accessed and analyzed in subsequent steps.
  • Monitor: Use time-series data from connected machines to visualize and understand the current performance of all assets and processes. See which systems and subsystems are performing nominally and which aren’t. Enable aggregation of data at the asset level, unit level, and fleet level.
  • Analyze: Determine the root cause of failures and poor performers based on historical and real-time data so as to understand relationships, correlations, and trends, and enable effective troubleshooting of problems.
  • Predict: Utilize advanced predictive analytics to provide weeks or months of foresight into impending problems so issues can be averted in the first place, driving greater process consistency and asset uptime. Put actionable intelligence in the hands of frontline maintainers to drive readiness up and maintenance costs down.
  • Optimize: Transform operations and maximize the performance potential of all assets and processes. Use the information learned from the data to make better decisions, change business processes, and even design better physical systems for the future.

Few organizations can be described monolithically as being at a certain point on the maturity spectrum. In most real-world cases, organizations are made up of subunits that may be at radically differing levels of maturity in terms of how they collect, process, and leverage data. By the same token, different types of equipment used by the same organization may be in completely different leagues when it comes to their Industrial Internet maturity.

So where is the military today? The answer, of course, is “it depends.” The Army ground community, for example, is struggling with legacy assets that are sparsely instrumented and do not even allow, in many cases, for the assets to capture data. The first step for theses platforms and organizations is to think about getting fully connected and starting the process of monitoring their assets in real time. The good news is that leaders in the government and among the ground-vehicle OEMs have recognized that they have gaps and are working to pilot new connected solutions at this very minute.

The Navy surface fleet, by contrast, made significant investments in instrumentation and connectivity in the early part of the last decade and now has a fairly robust system for connecting, monitoring, and analyzing the data from their ships. They are now in the process of exploring options for leveraging the data they already capture to get predictive using advanced analytics. (See Figure 1.)


Figure 1: What phase is the U.S. military in today when it comes to Industrial Internet maturity? Every segment of the military is different. In this photo, Airman Tuan Hoang directs an MH-60S Sea Hawk helicopter attached to the Chargers of Helicopter Sea Combat Squadron (HSC) 14 during takeoff from the flight deck of the aircraft carrier USS John C. Stennis (CVN 74). (U.S. Navy photo/Mass Communication Specialist 3rd Class Christopher Frost.)

(Click graphic to zoom by 1.9x)




Among the aircraft at any given Air Force base, there might be fifth-generation fighters managed “predictively” and older aircraft whose sensor data is dumped into a database and never reviewed again unless there’s a major in-flight incident.

Uneven progress

The services have been making uneven progress, then, and some efforts to lean far forward have been met with bureaucratic resistance, technical challenges, or both. The good news, if it can be called that, is that the military is not that far behind the majority of commercial industry when it comes to using data. The McKinsey study referenced earlier found that, globally, less than one percent of industrial data is actually used in decisionmaking, and the vast majority of that is used for alarm and control purposes (i.e., “monitoring”). Certain parts of the military are already doing better than that.

The critical thing is that, for any organization, the steps in the model must each be taken in sequence, and must be completed before moving on to the next step. Assets that are not connected cannot be reliably monitored, just as operations cannot be optimized for a platform with no access to predictive analytics. No project or initiative that aims to go from disconnected assets to optimized operations in one fell swoop is going to succeed. That reality should be a relief to those who find the vision painted by Industrial Internet enthusiasts as daunting or unrealistic.

Moreover, it is not true that every organization should set the goal of getting to “optimize” for all of its assets. Each step in the maturity model carries incremental benefits and incremental costs. Whether the former outweigh the latter depends on how critical a given mechanical asset is to the organization’s mission, how it underperforms or fails and how often, and how costly those deviations from ideal performance are to the organization from both a mission-effectiveness and a cost perspective.

Goals can vary even among assets within an organization or among systems on a platform. It is probably worth pushing for optimized performance of the gas turbines on an Arleigh Burke-class destroyer, for example, but it may be enough to simply monitor the status of the air-conditioning system.

As with any journey, the first step is figuring out where you are and where you want to go. GE helps its customers perform self-assessments using the Maturity Model all the time; it can then suggest Industrial Internet hardware and software solutions, from embedded computer systems to data historians to SmartSignal predictive analytics, to help customers realize their particular Industrial Internet vision.

Todd Stiefler leads the Military Analytics team at GE, where data scientists, engineers, and equipment experts help defense organizations increase readiness and decrease operations and support costs. Todd went to GE after a decade in Washington as a defense and foreign-policy advisor to three U.S. senators, including members of the Armed Services and Defense Appropriations Committees. Readers may follow him on Twitter at @ToddStiefler.

GE Intelligent Platforms


Featured Companies

GE Intelligent Platforms, Inc.

5 Necco Street
Boston, MA 02210
A.I. - Big Data
Topic Tags