Data mining for predictive fleet management in the military


Why do military fleet managers drive everywhere looking in the rear-view mirror?

Can you imagine driving your car down the freeway but navigating by only using your rear-view mirror?  In essence that’s what most military fleet managers use to manage their assets, be they tanks, trucks or aircraft.  Whatever the type of asset, most military fleet managers oversee their fleets by looking in the rear-view mirror using traditional fleet management metrics.

Examples of common fleet-management metrics include:

  • Mission capability rates.
  • Equipment readiness & utilization rates.
  • Mean time between system aborts.
  • Mean time between failure rates.
  • Unscheduled maintenance rates.

One might think “Hey, these metrics are pretty good – it’s what I use.”   Metrics like these are essential for even the most basic DoD fleet management effort; however, they all have a serious limitation: all are historical, backward-looking metrics that tell us only what has already happened in the past.  Most commonly-used metrics are lagging indicators which are good at informing fleet managers where the fleet is today and how it has performed in the past, but offer little insight into where the fleet will be tomorrow or next week or next month or next year.  I don’t know about you, but as a fleet manager in a department of defense or the private-sector I’d *really* want to know how things will be in the future.

This is where state-of-the-art predictive modeling techniques, using data-mining and other machine-learning methods can dramatically enhance the effectiveness and insight gleaned from fleet management metrics. Using statistical and mathematical techniques, predictive fleet management models can be developed to provide a forward-looking capability with enhanced metrics that predict when and where critical performance metrics drivers and other factors are more likely to move in the future.

Some argue that trend-lines can be projected into the future, by connecting the dots and extending that line forward – Voila!   While trend-line projections may suggest where one’s fleet is headed, such approaches provide no analytic rigor, no measure of certainty, and lack the ability to actually predict with any confidence where the fleet’s performance will be in the future.  Lacking an analytic methodology, trend-line projections are just a guess.  At times an educated guess, but still… a guess.

Do you think Fortune 500 companies routinely use “guesses” to manage their fleets? Do you think Wall St would react kindly to a company missing their earnings-projections because they used “a guess” to manage their business?  Do you think your spouse would react well to using “a guess” to manage your retirement-account?  If guessing is unacceptable in business and at home, then why should guessing be acceptable in the DoD?

For anyone interested in specific technologies and solutions that are relevant to predictive fleet management, check out SAS® Predictive Asset Maintenance.


About Author

Allan Manning

A 17-year veteran at SAS, as a DoD Logistics and Supply Chain Consultant Allan is responsible for understanding the logistics, sustainment and supply-chain challenges in the defense/aerospace sector and aligning those requirements with SAS' advanced analytic methodologies and solutions. In addition, Allan is often invited to speak at conferences and educational programs, requested for interviews by various industry periodicals, and has participated in defense sector panel discussions. Prior to joining the Defense and Aerospace team, Allan spent over 10 years as a solution-architect supporting many of SAS’ key Fortune 500 customers in the Global Manufacturing and Supply Chain practice.

Leave A Reply

Back to Top