This paper describes Anomaly Detection via Topological-feature Map (ADTM), a data-driven approach to Integrated System Health Management (ISHM) for monitoring the health of spacecraft and space habitats. Developed for NASA Ames Research Center, ADTM leverages proven artificial intelligence techniques for rapidly detecting and diagnosing anomalies in near real-time. ADTM combines Self-Organizing Maps (SOMs) as the basis for modeling system behavior with supervised machine learning techniques for localizing detected anomalies. A SOM is a two-layer artificial neural network (ANN) that produces a low-dimensional representation of the training samples. Once trained on normal system behavior, SOMs are adept at detecting behavior previously not encountered in the training data. Upon detecting anomalous behavior, ADTM uses a supervised classification approach to determine a subset of measurands that characterize the anomaly. This allows it to localize faults and thereby provide extra insight. We demonstrate the effectiveness of our approach on telemetry data collected from a lab-stationed CubeSat (the “LabSat”) connected to software that gave us the ability to trigger several real hardware faults. We include an analysis and discussion of ADTM's performance on several of these fault cases. We conclude with a brief discussion of future work, which contains investigation of a hierarchical SOM-architecture as well as a Case-Based Reasoning module for further assisting astronauts in diagnosis and remediation activities.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Semi-Supervised Machine Learning for Spacecraft Anomaly Detection & Diagnosis


    Beteiligte:

    Erschienen in:

    Erscheinungsdatum :

    2020-03-01


    Format / Umfang :

    691443 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Semi-Supervised Active Learning for Anomaly Detection in Aviation

    Memarzadeh, Milad / Matthews, Bryan / Templin, Thomas et al. | AIAA | 2023


    Topological Learning for Semi-Supervised Anomaly Detection in Hyperspectral Imagery

    Ramirez, Juan / Armitage, Tristan / Bihl, Trevor et al. | IEEE | 2019




    A Semi-Supervised Learning Approach for Pixel-Level Pavement Anomaly Detection

    Ren, Ruiqi / Shi, Peixin / Jia, Pengjiao et al. | IEEE | 2023