Using the correlation dimension for the detection of faults in rolling element bearings
During the last fifteen years, predictive maintenance has become increasingly recognised as a more cost-effective alternative to conventional run-to-failure and fixed-time-replacement maintenance strategies. One of the key
components of a predictive maintenance strategy is the art of condition monitoring, with vibration-based condition monitoring probably the most common technique. Its popularity is due to its cost-effectiveness, non-
intrusiveness, and applicability to a wide range of machinery. Vibration analysis techniques allow trending of condition, fault detection, fault diagnosis and often prognoses of residual life. The essence of vibration-based
monitoring, however, is early detection of impending failure. The earlier afault is detected, the sooner more advanced and complex techniques can be brought into play to determine the exact cause of the fault and subsequently make prescriptions for future maintenance.
This thesis examined the application of a relatively new field of study to condition monitoring problems, that of chaos. A wide variety of chaos techniques and quantification methods were examined. Finally, the correlation dimension of a time series was selected for more detailed study.
This thesis was scanned from the print manuscript for digital preservation and is copyright the author.
Author requested conversion to open access 26 Oct 2022
