Functional Mahalanobis semi-distance for measuring military aviation safety performance
Brusi, Kalle (2014)
Tässä tietueessa ei ole kokotekstiä saatavilla Treposta, ainoastaan metadata.
Brusi, Kalle
2014
Tilastotiede - Statistics
Informaatiotieteiden yksikkö - School of Information Sciences
Hyväksymispäivämäärä
2014-06-24
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:uta-201407212016
https://urn.fi/URN:NBN:fi:uta-201407212016
Tiivistelmä
This thesis introduces a recently developed method, functional Mahalanobis semi-distance. The method is applied as a safety performance indicator for military aviation. The data was recorded from routine flights by Finnish Air Force BAE Hawks and collected in the year 2012. The goal was to apply a method for finding anomalous approaches at the final approach segment to the selected runway.
Flight Data Monitoring is used for improving flight safety, training, maintenance and enhancing operational readiness. Systematic information about the performance of the aircraft fleet operations enhances tactical situation awareness. The filtered spatial flight data was transformed to a metric system, centered and rotated in a direction of the desired runway. Then it was functionalised to B-splines. Finally, a functional principal component analysis was performed and functional Mahalanobis semi-distances were calculated.
B-Splines are a flexible tool for researching flight data.With the splines, di?ent sample rates or pauses can be tackled quite ectively. It is also possible to use the definitions of flight phases straight from the regulatory basis. Functional principal component analysis is a tool to research features which characterize typical flights in the data. However, for anomaly detection it is possible to find the flights that are not typical for the data. Functionalising the data is a key to conduct more accurate analysis than in earlier applied methods.
The functional Mahalanobis semi-distance scores are quite a straightforward and robust way of applying anomaly search because the information is packaged into a single number. These scores describe the normality of the flight. The scores can be handled much like exceedance severity values in regular flight data monitoring.
Flight Data Monitoring is used for improving flight safety, training, maintenance and enhancing operational readiness. Systematic information about the performance of the aircraft fleet operations enhances tactical situation awareness. The filtered spatial flight data was transformed to a metric system, centered and rotated in a direction of the desired runway. Then it was functionalised to B-splines. Finally, a functional principal component analysis was performed and functional Mahalanobis semi-distances were calculated.
B-Splines are a flexible tool for researching flight data.With the splines, di?ent sample rates or pauses can be tackled quite ectively. It is also possible to use the definitions of flight phases straight from the regulatory basis. Functional principal component analysis is a tool to research features which characterize typical flights in the data. However, for anomaly detection it is possible to find the flights that are not typical for the data. Functionalising the data is a key to conduct more accurate analysis than in earlier applied methods.
The functional Mahalanobis semi-distance scores are quite a straightforward and robust way of applying anomaly search because the information is packaged into a single number. These scores describe the normality of the flight. The scores can be handled much like exceedance severity values in regular flight data monitoring.