SUMMARY

Rolling element bearing fault identification is an important sub-topic of predictive health monitoring. Most state-of-the-art fault identification approaches utilize bearing configuration, shaft rotational speed or bearing harmonics that are only present in the spectrum if the sampling rate is sufficiently high. In industrial application, these three factors are often not available. This thesis investigates the performance of various state-of-the-art bearing fault identification approaches under unknown rotational speed and bearing configuration for a range of sampling rates. The aim of this thesis is to give recommendations for effective bearing fault identification under uncertain low-sampling rate circumstances. The recommendations are based on simulated and experimental data.