As the Defense Department sets a goal for “zero unplanned maintenance,” the B-1B bomber, C-5M airlifter and KC-135 tanker fleets are now scheduling maintenance actions using algorithms designed to predict failures based on its condition in real-time, Air Force officials said on April 9.
The predictive maintenance approach marks a paradigm shift in military aircraft maintenance. Instead of scheduling repairs based on unplanned failures or manufacturer-imposed intervals, the Air Force has started to adopt a condition-based maintenance approach over the last six months. As a new set of data analytical tools become widespread, Air Force officials hope such information allows them to avoid operational disruptions caused by unexpected part failures and plan maintenance actions more efficiently when they’re needed.
The predictive algorithms for the B-1B and C-5M fleets became activated on Oct. 31, with the KC-135 fleet following on April 1, said Col. Robert Jackson, chief of the Mobility Aircraft Division for Air Mobility Command. The “condition based maintenance-plus pilot initiative” is intended to spread to more aircraft fleets as the maintenance community becomes more familiar and comfortable with the new approach, said Jackson, a panelist for Aviation Week’s Military Aviation Logistics and Maintenance Symposium.
The algorithm analyzes a database of sensor data on the aircraft that track the performance of parts, looking for signs of deterioration that signal a failure may be imminent. The Air Force trained the algorithm using historical data, then applied it cautiously at first on the B-1B and C-5M last October.
“We went live and we watched the algorithm, so we didn’t act upon it. We validated that our algorithms were accurate,” said Debbie Naguy, head of the Product Support Engineering Division at the Air Force Life Cycle Management Center.
The CBM+ pilot program fits into the Defense Department’s ambitious goal to eliminate unplanned maintenance events overall, said Kenneth Watson, deputy assistant secretary of defense for Material Readiness.
“Zero unscheduled maintenance for us is plausible,” Watson told the MALMS audience in April 10 keynote address. “We can do it.”
But Watson acknowledged the challenges go beyond the maturity of machine-learning algorithms. He noted that the trend requires a “cultural shift” within the military’s sustainment community.
It’s an issue that Air Force leaders have already encountered as they roll out the CBM+ program to a sometimes skeptical maintenance staff.
“There’s a lot of consternation, some push back,” Naguy said. “Some folks are just excited to see the vision and other folks say, ‘Oh, gosh, you just don’t understand.’ This is changing everything we’ve always done.”
Ultimately, the Air Force also hopes to pursue additional benefits from the data collected for the algorithm. As data illuminates the actual condition of parts, the Air Force plans to ask manufacturers to review their maintenance intervals on certain parts.
“As we start to get good data,” said Dennis D’angelo, Director of the 448th Supply Chain Management Wing, “we may be able to go back to the OEM to expand the lifespan at that part because it didn’t break as [expected.”