AI/ML for Anomaly Detection and Health Monitoring for the NSN
Description
The project develops ML/AI methods/tools to enhance the operation and sustainment of GSFC Space Network (ACCESS-managed) assets. We analyze telemetry from TDRS and STPSat-6 to assess state of health, detect anomalies, and predict remaining useful life (RUL) of spacecraft components. The ML/AI Algorithms are implemented in Python and MATLAB and are reported and validated with engineering teams during Sustaining ML meeting. Validated models are being integrated into a GUI to provide an intuitive platform for engineers to analyze spacecraft telemetry and support real-time monitoring. Documentation and results are summarized and posted to SharePoint.
Detailed example
Anomaly scores/flags from telemetry (After thresholded), Predicted Telemetry (SA current, Battery Voltage), SOH index for selected subsystems, remaining-useful-life (RUL) predictions with confidence bands, Summary of the ML/AI results for review and action by engineers.
AI / analytics pattern
Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relationships.
Automation level / stage
a) Pre-deployment – The use case is in a development or acquisition status.
Expected benefit
Automated telemetry anomaly detection and forecasting reduce manual review and improve Space Network availability. In internal tests, the TDRS-8/10 Bus Voltage Limiter (BVL) detector achieved ~98% true-positive rate with very few false alarms and processed ~10 years of telemetry in ~5 minutes. Early, prioritized alerts (hours–days sooner) have surfaced incipient battery issues—e.g., two early-warning signals of failing cells—and the battery tool identified diverging cells on TDRS-9/10. Remaining-useful-life (RUL) and short-term forecasts enable condition-based maintenance and proactive scheduling, reducing unplanned downtime and associated costs.
Controls / human review
ATO: Not reported; PIA: Not published