On-The-Fly pattern recognition for satellite time series data
- Sponsored by: German Aerospace Center (Deutsches Zentrum für Luft- und Raumfahrt e.V.)
- Project Lead: Dr. Ricardo Acevedo Cabra
- Scientific Lead: M.Sc. Leonard Schlag, Dr. Clemens Schefels
- Term: Summer semester 2019
Results of this project are explained in detail in the final documentation and presentation.
Pattern recognition for time series data has an abundancy of applications as well as methods to approach the problem at hand. When applying an approach specifically for satellite telemetry, one is faced with a combination of constraints which do not necessarily appear elsewhere. These include the sheer amount of parameters, with a single modern satellite transmitting up to 80 000 different telemetry parameters, as well as the large variance in their behavior.
While our engineers, who are actively working on maintaining the satellites under our control as well as keeping them healthy, have all telemetry data available to them, the analysis in case of e.g. unforeseen events or interesting behaviors can further be simplified and improved by providing means of finding similar events in the past. In case of failures, this could also imply new and faster means of finding solutions.
Within this project, a pattern matching technique which fits these specific needs best will be researched and evaluated. This also includes preprocessing steps and optimizing the performance to ensure the results are suitable for an operational use case. A database with real and up to date satellite telemetry reaching back many years will be provided to allow for achieving this goal.