SBIR/STTR Award attributes
The DF&NN team proposes to demonstrate the feasibility of artificial intelligence (AI) deep neural net machine learning to predict required maintenance activities for Naval aircraft. We estimate the approach will automatically detect and predict anticipated maintenance activities and discover previously unknown required maintenance for aircraft. The system will continue to improve over time as new data are used to re-train and update prediction models. Operational deployment of this capability will be well-suited for at-sea limited connectivity to shore with onboard prediction and fleet-wide updating when in port. The team will deliver a prototype based on an in-place prototype that has been tested on five years on USAF C-130 aircraft engine, pilot debrief, and maintenance/repair data. DF&NN will apply their operationally-proven neural network development platform which has been successfully applied in numerous machine learning environments. The capability includes normal engine behavior learning, historical signature clustering with automated cluster labeling and abnormality class-categorization NN training. These NNs support on-line unknown abnormality detection and known abnormality categorization to enable discovery of repair correlations for predictive maintenance. The team provides an affordably extendable and automatically retrainable C-130 Goal-Driven Condition-Based Predictive Maintenance (GCPM) capability and decades of AI and Navy experience to reduce risk.
