Highly-loaded low-speed roller bearings form crucial connections in offshore structures, such as heavy-lifting vessels, single-point mooring systems, and wind turbines. In order to safeguard the integrity and reliability of these assets and their operations, a quantitative methodology for condition monitoring of the bearings can be of substantial value. To date, a number of assessment methods have been proposed to for this purpose, e.g. based on strain, vibration, lubrication, and acoustic emission (AE) monitoring. Despite their demonstrated potential for medium- and high-speed bearings (>600 rpm), no notable success has yet been reported in the assessment of low-speed bearings subjected to naturally-developing degradation. In this dissertation, a novel methodology for the analysis of damage-induced AE and inferring the bearing condition has been proposed. Acoustic emissions in this context are ultrasonic signals generated by the release of elastic energy in a material. In solid media, these signals propagate as stress waves and can be recorded by dedicated transducers.
A mathematical framework to describe the generation, propagation, transmission, and detection of transient ultrasonic waves in complex geometries has been presented. An assessment of inter-component stress-wave transmission has been performed utilising this framework. For a representative sheave bearing, results indicate that a transmission loss in the order of 15 dB is to be expected in the amplitude of the AE waves for a single rolling contact arrangement. In conjunction with a preliminary field trial regarding the ultrasonic background noise in representative operational conditions, this evaluation has shown that it is feasible to detect damage initiated AE signals from each of the rolling elements upon field implementations.
A waveform-similarity based clustering algorithm has been proposed for the
identification of damage-induced AE source mechanisms. Consistency in the source mechanism is theorised to indicate gradual progressive failure, such as crack growth. Through the descriptive framework, it has been shown that high similarity of the recorded signal must be the result of high similarity in the emitted source. Additional numerical verification of this assumptions on transfer path similarity has been performed, confirming the equivalence derived from the descriptive framework.
A low-speed run-to-failure test was performed with a purpose-built linear bearing segment, representative of the main bearing of a mooring turret, to assess the performance of the clustering algorithm. Intermediate and final visual inspections report the development of wear comprising erosion, surface roughening, pitting and surface initiated fatigue. In independent analysis of the recorded AE signals, several highly-consistent structures of clusters were identified over multiple measurement channels. The nose raceway could be identified as the source of these structures of clusters, which matched the observed evolution of localised damage during the inspections.
Based on the source-identified AE activity, a novel quantitative indicator has been proposed to infer bearing condition. The bearing condition index (BCI) adopts a value of 1 when the bearing is in good condition. The BCI drops in value as the bearing degrades, as represented by a more significant detection of clusters of similar AE signals within the normalised period of a load cycle over a multitude of measurement frequencies.
Run-to-failure experiments have been conducted to assess the proposed BCI. Intermediate and final inspections report the progressive erosion and surface roughening. Additional lubrication samples collected during these inspections contained high levels of particle contamination. A direct correlation between the AE hit-rate and the particle contamination of the lubricant was observed. Utilising progressive scaling based on cluster size, the excessive influence of lubrication contamination-induced AE signals on the BCI could be reduced, while still providing a timely warning.
In review, it is concluded that the proposed methodology can effectively describe the complex generation and propagation of AE due to damage evolution in highlyloaded low-speed roller bearings. The developed clustering method has shown to effectively identify patterns and trends in the AE signals at different stages of degradation, and provide the basis for filtering out noise-related signals. The formulated BCI can subsequently provide an intuitive indication of the condition of a low-speed roller bearing in an in-situ non-intrusive manner. As such, the methodology is believed to offer promising potential to contribute to the safe and continued operation of the offshore energy infrastructure.