Dynamic Data-driven Avionics Systems: Inferring Failure Modes from Data Streams
Semantic Scholar · Article (International Conference on Conceptual Structures) · 0bda5c296748632806a9b84b535b3f9e9d6e855e · Published 2015-01-01 · International Conference on Conceptual Structures · 3 authors
Abstract and citation only, verbatim from Semantic Scholar; full text lives there. All credit to the authors and International Conference on Conceptual Structures.
Abstract
Abstract Dynamic Data-Driven Avionics Systems (DDDAS) embody ideas from the Dynamic Data- Driven Application Systems paradigm by creating a data-driven feedback loop that analyzes spatio-temporal data streams coming from aircraft sensors and instruments, looks for errors in the data signaling potential failure modes, and corrects for erroneous data when possible. In case of emergency, DDDAS need to provide enough information about the failure to pilots to support their decision making in real-time. We have developed the PILOTS system, which supports data-error tolerant spatio-temporal stream processing, as an initial step to realize the concept of DDDAS. In this paper, we apply the PILOTS system to actual data from the Tuninter 1153 (TU1153) flight accident in August 2005, where the installation of an incorrect fuel sensor led to a fatal accident. The underweight condition suggesting an incorrect fuel indication for TU1153 is successfully detected with 100% accuracy during cruise flight phases. Adding logical redundancy to avionics through a dynamic data-driven approach can significantly improve the safety of flight.
Authors
- Shigeru Imai
- A. Galli
- Carlos A. Varela
Keywords
- Computer Science
- Engineering
Citation
Shigeru Imai, A. Galli, Carlos A. Varela (2015). Dynamic Data-driven Avionics Systems: Inferring Failure Modes from Data Streams. International Conference on Conceptual Structures. Semantic Scholar ID 0bda5c296748632806a9b84b535b3f9e9d6e855e. https://doi.org/10.1016/j.procs.2015.05.301 ↗