We propose an artificial intelligence (AI) technology that significantly expands NASA’s real-time and offline ISHM capabilities for future deep-space exploration efforts. Our proposed system, Anomaly Detection via Topological feature Map (ADTM), will use Self-Organizing Map (SOM)-based architecture to produce high-resolution clusters of nominal system behavior. What distinguishes SOMs from more common clustering techniques (e.g., k-means) in the ISHM-space is that they map high-dimensional input vectors to a 2D grid while preserving the topology of the original dataset. ADTM will use SOMs as the building blocks for a hierarchical case-based model of a system. Using a combination of case-based reasoning (CBR), clustering, and classification techniques, ADTM will detect, predict, and explain anomalies, and guide users in implementing effective mitigations. This approach provides the critical ability to handle previously unknown anomalies and faults. An additional feature of ADTM is the ability to cross-correlate subsystems in order to capture the cascading effect of faults from one subsystem to another, as well as discover latent relationships between subsystems. Such analysis would significantly aid in the maintenance activities of NASA’s deep-space missions. ADTM will include tools to allow users to visualize the status of the system at various levels of granularity, configure and receive alerts about current or predicted future faults, and navigate the system models to trace root causes.
The Phase I effort implemented prototype versions of ADTM’s SOM-based clustering and classification techniques. The prototype was successfully demonstrated on three real-world datasets from NASA and on one simulated dataset for a CubeSat. The level of success attained during Phase I provides a sound foundation for the Phase II effort. We have assembled a strong team that collectively reflects deep expertise in AI, NASA space missions, and predictive health maintenance.
The Phase I effort has shown value to NASA. The Phase I prototype successfully detected anomalies in three different datasets provided by NASA. We will continue to apply ADTM to the NASA Sustainability Base system and the Lunar Orbital Platform Gateway System. Other application areas include the ISS as well as space habitat simulations like HI-SEAS, D-RATS, NEEMO, and HERA, and various future manned and unmanned spacecraft. We have a statement of interest from Northrop Grumman Innovation Systems as a possible transition to their space systems.
We have stated interest from Northrop Grumman which has several programs relating to predictive health maintenance for air and ground systems. There is an increasing demand for such systems in the military as well. We will leverage the funding opportunities provided by military technology acceleration programs like AFWERX and Defense Innovation Unit to transition ADTM to military uses.