Skip to main content

On-board anomaly detection from the OPS-Sat telemetry using deep learning

Closed

Closed

Prime contractor
Organisational Unit
Activity Type
Implementation progress
90%
16 February 2022

Duration: 12 months

Objective

Detecting anomalies from the satellite telemetry is critical  in its safe operation. Classical solutions for this task often utilize  the out-of-limit checks, which allow us to find point anomalies. To  detect other types of events, such as contextual or collective ones, we  can exploit data-driven deep learning-powered models – this project will  build upon the current advances in the field, and we will incorporate  recurrent neural networks into the proposed event detection pipeline  that will be deployed on-board OPS-SAT. Additionally, we will explore  the possibility of detecting anomalies from multi-modal sequential  signals, e.g., captured by different components of the satellite,  as such telemetries may highlight the events that would not be possible  to detect using a single channel exclusively. Overall, our main objective is  to demonstrate that deep learning can be effectively deployed on-board  OPS-SAT for automated event detection from its operational telemetry  data. We hope that the project will be an important step towards  data-driven Fault Detection, Isolation and Restoration systems that will  ultimately  improve the safety of emerging satellites, and reduce their operational  costs through delivering clear insights into the behavior of the entire  spacecraft and its pivotal (sub)systems.

Contract number
4000137339
OSIP Idea Id
I-2021-02625
Related OSIP Campaign
OPS-SAT
Main application area
Operations
Budget
50000€
ON-BOARD ANOMALY DETECTION FROM THE OPS-SAT TELEMETRY USING DEEP LEARNING