We used machine learning to show the Air Force how to boost the reliability of their engines, predict part failure, and save money.
To improve Aircraft Uptime and Planning Capabilities
Last year alone, over 150 J85 engines were repaired and reinstalled on aircraft only to fail within 2 flying hours and be removed from the aircraft to be repaired again.
The inability to predict and prevent engine parts failure was becoming ever more costly for the Air Force. They needed a tool predicting the likelihood of failure and optimizing cost; a tool that would enable mechanics to proactively replace vulnerable parts while the engine is already in the shop. The tool had to be accurate, actionable, extensible, and ultra-simple to use.
We created a hands-on cloud-native tool for maintenance shop chiefs leveraging IBM’s Watson predictive machine learning capabilities. The tool predicts which parts will soon fail on an engine, and indicates when replacement is most cost-effective, maximizing the number of missions the aircraft can fly before the next repair. Detailed part replacement guidance is displayed in an extremely clear and actionable interface.
After proving the capability with a single engine, we’re extending capabilities to support and optimize between multiple engines, engine types, across bases, and incorporating complex data inputs such as weather, geography, and pilot flying style into its recommendations.