DrMUST

Ref-Nr: TDO0167

Technology abstract

DrMUST is a system for rapidly diagnosing the cause of unusual or anomalous events in a telemetry stream. It uses pattern matching to find occurrences similar to a specified event, and determine which parameters are involved in the observed behaviour.

 

Technology Description

MUST (Mission Utility & Support Tools) is a system at ESA?s European Space Operations Centre (ESOC) that provides high-performance access to telemetry, telecommand history, events and ancillary mission data from spacecraft. DrMUST is a MUST client that enables faster diagnosis of unusual events. DrMUST performs pattern matching to find behaviour similar to a specified event, and correlation analysis to find which other subsystems or parameters behave in a consistent way at the times of such events (anomalies), reducing the amount of labour intensive data analysis that is needed.

DrMUST characterizes time series behaviour by its shape over a time period. It detects past related behaviour by computing the similarity between past data and the anomaly under investigation. The time series similarity is computed allowing for dynamic time warping.

Innovations & Advantages

  • Performance: queries run very quickly compared to manual searches
  • Finding correlations for the engineer, even those that are not anticipated
  • Reduces the time needed for anomaly investigation
  • It can be easily adapted to work with any kind of time series data

Further Information

When investigating an anomaly, the first thing an engineer would want to know is whether such an anomaly has occurred before. The initial step is to define a time period for the parameter ? that contains the behaviour to be investigated (e.g. anomaly period). In Figure 1, users specify target time periods (T) where the event of interest (e.g. anomaly) happened, and time periods (N) where the behaviour is nominal.
DrMUST then finds other occurrences of parameter ? with a similar pattern. When that?s done, DrMUST performs a correlation analysis by finding other parameters that behave in a consistent way at the times when the anomaly occurs.
An example of where DrMUST has proved its worth in a Space application is the ESA mission Venus Express. The problem was that from time to time the attitude errors in the Venus Express? Y-axis were higher than expected. These time periods were input to the correlator as target periods and periods where the attitude error levels were as expected were input as nominal periods. The correlator found, to the surprise of the flight control engineers, that the activation of a payload ? the ASPERA scanner ? was introducing a torque that had not previously been considered. This demonstrates the power of the DrMUST correlator in finding correlations systematically; even those that engineers might think a priori have no connection.

Current and Potential Domains of Application

DrMUST could be applied to many fields including engine manufacturing, factory monitoring, analysing cardiograms or even for stock market analysis.