Sammendrag
This report covers the statistical and experimental methods needed to estimate the RUL of a roller bearing and how to calculate and minimize the maintenance cost for OWT with respect to the bearings. An experiment is conducted to give a better understanding of how the bearings degrade and also collect degradation data.
\newline
The first objective was to carry out run-to-failure experiments for roller bearings and collect vibration data. Nine runs were completed. It would however be ideal to have completed more. As there is a lot of noise in vibration data some signal processing is done. The signal processing was done in the time domain, and the time-domain feature used was RMS.
\newline
The second objective was to estimate the RUL for the bearings using the vibration data. This was done by fitting the HI to a Wiener Process. These results were then used to develop an Inverse Gaussian distribution for the failure time. The model did estimate the RUL to a certain degree, however, there is room for improvement. This could be done by implementing Bayesian updating and utilize more collected data.
\newline
The third objective was to develop a model for reducing the maintenance cost for OWTs due to the bearings being prone to failure. The general idea is that was lowering the rotation speed for wind turbines the maintenance cost can be reduced, as the bearings lifetime would increase. This model was dependent on the estimated RUL. The model did work as designed, however, due to uncertainty about the estimated RUL the calculated values have a high degree of uncertainty.
\newline