Condition monitoring of rotating machinery and machine systems has attracted extensive researches, particularly the detection and diagnosis of machine faults in their early stages to minimise maintenance cost and avoid catastrophic breakdowns and human injuries. As an efficient mechanical system, helical gearbox has been widely used in rotating machinery such as wind turbines, helicopters, compressors and internal combustion engines and hence its vibration condition monitoring is attracting substantial research attention worldwide. However, the vibration signals from a gearbox are usually contaminated by background noise and influenced by operating conditions. It is usually difficult to obtain symptoms of faults at the early stage of a fault. This study focus on developing effective approaches to the detection of early stage faults in an industrial helical gearbox. In particular, continuous wavelet transformation (CWT) has been investigated in order to select an optimal wavelet to effectively represent the vibration signals for both noise reduction and fault signature extraction. To achieve this aim, time synchronous average (TSA) is used as a tool for preliminary noise reduction and mathematical models of a gearbox transmission system is developed for characterising fault signatures. The performance of three different wavelet families was compared and henceforth a criterion and method for the selection of the most discerning has been established. It has been found that the wavelet that gives the highest RMS value for the baseline vibration signal will show the greatest difference between baseline and gearbox vibration with a fault presence. Comparison of the three wavelets families shows that the Daubechies order 1 can give best performance for feature extraction and fault detection and fault quantification. However, there are limitations that undermine CWT application to fault detection, in particular the difficulty in selecting a suitable wavelet function. A major contribution of this research programme is to demonstrate a possible route on how to overcome this deficiency. An adaptive Morlet wavelet transform method has been proposed based on information entropy optimization for analysing the vibration signal and detecting and quantifying the faults seeded into the helical gearbox. This research has also developed a nonlinear dynamic model of the two-stage helical gearbox involving time–varying mesh stiffness and transmission error. Based on experimental data collected with different operating loads and the simulating results vibration signatures for gear faults are fully understood and hence confirms the CWT based scheme for signal enhancement. These results also indicate that the dynamic model can be used in studying gear faults and would be useful in developing gear fault monitoring techniques.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:700609 |
| Date | January 2016 |
| Creators | Elbarghathi, Fathalla |
| Contributors | Ball, Andrew ; Gu, Fengshou |
| Publisher | University of Huddersfield |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://eprints.hud.ac.uk/id/eprint/30647/ |
Page generated in 0.0089 seconds