Pentagon Wants Better Data For Its Predictive Aircraft Maintenance AI
The Pentagon’s AI center is looking to break through past challenges it has faced advancing predictive maintenance among select aircraft.
The Pentagon’s Joint Artificial Intelligence Center is looking for ideas on how to improve the way it can use artificial intelligence technologies to predict when the Defense Department’s thousands of planes, helicopters and unmanned aerial vehicles need maintenance and repairs.
The JAIC issued a request for information May 7 indicating it has run into challenges in its pathfinder project using AI to predict maintenance on H-60 helicopters and their T700 engines. Those challenges include “the accuracy and completeness of historical aircraft data,” developing models trained on historical outputs and applying them to real-time data for real-time predictions and providing field maintainers trustworthy model outputs.
The JAIC “has been working to address these challenges and is seeking a partner to advance these efforts to achieve timely and trusted model outputs that accurately predict engine maintenance and servicing,” the RFI states.
The RFI seeks responses from industry, academia and other agencies by June 8. According to the RFI, the JAIC seeks a partner to:
- Assist in the data collection, curation, and connection to produce holistic and historical data on each H-60 aircraft in the Army, Navy, and Air Force.
- Develop and train AI models on that historical and holistic data to accurately predict the probability of a condition requiring a maintenance action on the engines within a certain number of flight hours.
- Assist in establishing near real-time and automated transmittal and connection of current holistic data.
- Provide a visual representation of model output to maintainers and planners, which will include the integration with existing reporting tools and dashboards.
“The partner will train models on this data to predict common engine issues, work with JAIC Test and Evaluation, assist with creating automated near real-time data inputs, provide methods for field units to access the model, supply model output to those units, provide methods for improving trust in the model, and develop interfaces in an agile manner with end users (aircraft maintainers at field units) to ensure a positive user experience,” the RFI states.