Novelty detection requires models of normality to be learnt from training data known to be normal. The first model considered in this paper is a static model trained to detect novel events associated with changes in the vibration spectra recorded from a jet engine. We describe how the distribution of energy across the harmonics of a rotating shaft can be learnt by a support vector machine model of normality. The second model is a dynamic model partially learnt from data using an expectation-maximization-based method. This model uses a Kalman filter to fuse performance data in order to characterize normal engine behaviour. Deviations from normal operation are detected using the normalized innovations squared from the Kalman filter.

No abstract

Damage detection in a gearbox is reported based on analysis of vibration signals measured on a running aircraft engine on a test bed. The experiment was stopped due to material being found on the magnetic chip detectors in the oil system. Subsequent inspection found a significant damage to a bevel gear inside the drive train system. Three damage detection techniques are applied to the vibration signals: advanced demodulation, advanced residual technique, and the classical residual technique. Fault indicators are defined from each technique and trended over time. Results show that these techniques have the ability to provide early damage detection; however, with some noticeable difference in their feature evolution. The advantages of the techniques are compared and discussed.

No abstract