Automatic behavior detection and interpretation from low level data sets such as telemetry

 
In Space Operations, it is necessary to detect potential anomalies as early as possible so that operations engineers can have the chance to prevent them. ESOC has already developed initiatives internally in this direction, such the Novelty Detection project.

With this project, the idea was to further explore how to best detect unusual behavior in telemetry data that may develop in anomalies to better support space operations engineers.

Objectives
Explore knowledge extraction techniques and automatic behavior detection approaches from telemetry data. To fulfill this goal, it is envisage to devote some effort to analyze existing approaches, adapt them to the problem of monitoring telemetry data and develop a prototype that works with real telemetry data from an ESA mission.

Achievements and status
Technologies researched: static data analysis (based on statistical features) and dynamic analysis (based on time based patterns) using XMM telemetry data Software prototype developed: KETTY (Knowledge ExtracTion from TelemetrY).

The purpose of KETTY is to help flight control engineers focusing and getting greater insight into the monitored system dynamics and automatically detect and diagnose anomalous or novel conditions.

As the novelty detection project, it does not requires engineering knowledge to configure it, making it suitable to monitor a variety of system in several industries. At present, KETTY reports still too many false alarms.

Benefits
Allows to detect unusual behaviors in spacecraft telemetry, without requiring engineering knowledge a priori. Possible applications: space operations and monitoring system in other industries that monitor time series.

Next steps
This study has a follow-on with the GSTP study on: “Context aware spacecraft telemetry checking”.

The outcome of the study has good potential to support both space related applications as well as non space domains.