Data mining mission history to automatically check present spacecraft behaviour
06 November 2015
Tags:

Two parallel GSP studies on "Data mining history for telemetry checking and analysis" looked into the potential of automated data-mining of historic telemetry data of space missions. The simple but powerful idea was, that what followed a certain pattern in the past should follow the same pattern in the future.

Real-time telemetry data checks take effort to define, implement and manage. At the same time the number of telemetry parameters and commands on a spacecraft is constantly increasing. The result is that the vast majority of telemetry parameters are not thoroughly checked and simply pass through ESA's control systems for storage until something goes wrong. The objective of these studies therefore was to analyze data mining techniques normally used in context of "Big Data" to find patterns in historic telemetry data and make it more useful without increasing the workload.

KU Leuven (Belgium) and SATE (Italy) defined within this study certain telemetry checks (including telecommand verifications) given one year of historical mission data from the mission Venus Express. No engineering knowledge was provided and the golden rule was that the derivation of the checks had to be completely automatic, i.e. the checks had to be derived solely on the provided data with no human intervention.

The initial time period ended just before the aero-braking activities of Venus Express started. Once the checks were submitted, they were applied successfully and with surprising accuracy on the following three months of data from Venus Express, which included interesting activities, such as the preparation for aero-braking and the aero-braking itself.

 

Two parallel GSP studies on "Data mining history for telemetry checking and analysis" looked into the potential of automated data-mining of historic telemetry data of space missions. The simple but powerful idea was, that what followed a certain pattern in the past should follow the same pattern in the future.

Real-time telemetry data checks take effort to define, implement and manage. At the same time the number of telemetry parameters and commands on a spacecraft is constantly increasing. The result is that the vast majority of telemetry parameters are not thoroughly checked and simply pass through ESA's control systems for storage until something goes wrong. The objective of these studies therefore was to analyze data mining techniques normally used in context of "Big Data" to find patterns in historic telemetry data and make it more useful without increasing the workload.

KU Leuven (Belgium) and SATE (Italy) defined within this study certain telemetry checks (including telecommand verifications) given one year of historical mission data from the mission Venus Express. No engineering knowledge was provided and the golden rule was that the derivation of the checks had to be completely automatic, i.e. the checks had to be derived solely on the provided data with no human intervention.

The initial time period ended just before the aero-braking activities of Venus Express started. Once the checks were submitted, they were applied successfully and with surprising accuracy on the following three months of data from Venus Express, which included interesting activities, such as the preparation for aero-braking and the aero-braking itself.