Two-interval overlap test for failure prediction in kalman filter


Hacızade C.

18th IMEKO World Congress 2006: Metrology for a Sustainable Development, Rio de Janeiro, Brazil, 17 - 22 September 2006, vol.1, pp.7-12 identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 1
  • City: Rio de Janeiro
  • Country: Brazil
  • Page Numbers: pp.7-12
  • Keywords: Confidence intervals, Failure prediction, Innovation sequence, Kalman filter, Tolerance intervals
  • Istanbul Technical University Affiliated: Yes

Abstract

An approach to failure prediction in Kalman filter is developed which uses confidence and tolerance intervals for a innovation sequence. The algorithm proposed is based on the criterion of overlapping these intervals and allows the detection of potential failures in Kalman filter. The problem of failure prediction in multidimensional Kalman filters is solved too. In this case a multidimensional innovation sequence is replaced by one- dimensional sequences.