부산대학교 도서관

With the improvement and expansion of technology, systems are becoming more expensive to operate and maintain. In an attempt to reduce these costs structural performance and condition monitoring systems have seen a significant boost in interest as companies try to reduce unplanned down time and extend component life. Whilst vibration monitoring and acoustic emission are the favoured monitoring methods at the moment, they have significant flaws when it comes to their application to rotating machinery. As an alternative monitoring technique, this thesis investigates the application of a non-contact magnetic technique for monitoring rate of change of torque to gear systems. Using both existing and newly developed novel test rigs, data was captured at realistic conditions allowing for four studies to be undertaken. These studies evaluate the performance of the rate of change of torque technique (ROC) and compare it to an already established monitoring method. The work showed that fundamentally the ROC technique was capable of successfully detecting very small levels: 2.7μm, of tooth bend damage on test gears under realistic operating conditions. Tests conducted on back-to-back gear testing rigs used sensors positioned both outside the torque loop, measuring ROC in parasitic losses, and inside the torque loop making direct measurements of ROC within the loaded gear pairs. These data have been analysed using a range of metrics and other analysis techniques, largely derived from current practice in vibration analysis. Whilst some techniques have been shown to be more suitable than others, it is clear that further development of ROC-specific analysis methods is required in order to achieve a robust methodology for repeatable and unambiguous detection and characterisation of gear tooth faults based on ROC signals. A further study examined signals generated during tests conducted whilst a bearing was failing. This bearing was positioned on the drive system of the test rig, external to the torque loop. No evidence was found in the in-loop ROC signals from this bearing failure. This was due to the relative power levels in and out of the torque loop, with the much higher amplitude ROC signals generated by the gear meshes swamping any indications of bearing failure. However, in a realistic application (i.e. a bearing in a torque loop) the work in this thesis, demonstrating the high sensitivity of the ROC technique, gives confidence that it is suitable for damage detection in bearings as well as gears. The final study compared ROC and vibration signals captured simultaneously during a set of gear tests. Whilst both techniques were found to be capable of detecting larger scale tooth bend defects, ROC was found to be more consistent in general, and more capable of detecting the 2.7 μm bend tested. The thesis concludes that ROC has the potential to be a highly sensitive and reliable condition monitoring technique for rotating machinery. Due to its novelty, further development is required of enhanced techniques for signal processing and classification.