8th September 2026 13h00-13h30
Stefano Donné (MWMW)
Predictive maintenance gives Defence earlier and more actionable insights into asset health. By combining sensor, operational and maintenance data, it supports anomaly detection and prediction of future condition. The cases presented show how this translates into decisions: whether an unmanned vehicle should continue a mission, whether a naval engine requires inspection, or how aircraft maintenance and resources should be planned before availability is affected. For complex and mission-critical systems, PdM can reduce unplanned downtime and improve readiness.
However, readiness starts with data. Defence must collect the right sensor, operational and maintenance data before a specific PdM need emerges, because years of history cannot be created retroactively. Generated data only becomes useful when it is collected consistently, stored with the necessary context and made accessible for analysis. Governance then keeps it available, understandable, trustworthy and secure through clear rules. This foundation enables PdM, while the same data can also support operations, logistics and future AI applications.
At RMA, Machine Learning for Predictive Maintenance for weapon systems (MLPM) project is building this expertise, with KU Leuven as academic partner and also the Royal Netherlands Navy’s DvO/DMI with who we share common hardware, needs and ambitions. Research focuses on anomaly detection when failures and labels are scarce, and on methods that remain reliable when real operational data are noisy, incomplete or contaminated.