Abstract:
To meet the ever-increasing demand for maintenance of gear systems, industrial companies have traditionally depended on the shutdown of the machines before processing the fault diagnosis. Nowadays, online monitoring has proven to be effective in terms of machine state analysis and fault prediction. However, the application of such a technique in the analysis of combined multiple nonlinear faults is still a subject of research. The vibration signature of a coexisting nonlinear crack and pit in two-stage gear system is unknown, it can be regarded as one of the most difficult problems to extract and diagnose. Additionally, incorporating both a crack and a pit into numerical models is a time-consuming process that demands a breadth of mechanical understanding. Diagnostics of faulty gears, on the other hand, can be performed in the time and frequency domain or in the Time-Frequency domain, depending on the complexity of the vibration. Non-linear and non-stationary phenomena (Features) occur when repeated pitting and cracking faults occur, reducing the reliability of standard signal processing methods (Gear displacement and Fast Fourier Transform). This thesis solves each of these shortcomings by developing an eight-degree-of-freedom (DOF) gear model with a breathing crack and multiple pitted gear teeth. The identified spur-gear model enabled the investigation of the crack and pitting signatures and their effect on the ensuing vibrations independently of the action of other system components. Additionally, when pitting and cracking coexist, the study was conducted to determine the influence of such a failure on the transmission system. Theoretical results indicated that the presence of pitting and crack in the tooth of the gear resulted in a decrease in mesh stiffness. Additionally, the influence of the breathing pitting and crack results in material fatigue, which results in the generation of a random term in the vibration signal. To corroborate the acquired results, several experimental tests on a spur-gear test rig with a defined pit and crack size range were undertaken under a variety of conditions. In comparison to the presented methodologies, theoretical and experimental results indicate that 3D Frequency-RPM analysis is the most sensitive and resilient method for the early detection and identification of pit and crack faults. Furthermore, when crack or pit faults are studied individually, the STFT analysis yields interesting results. The feature analysis revealed that, when using the Time-Frequency technique, the crack and pit combination in a two-stage gear system is a priori more efficient than the other options.
Description:
M. Tech. (Department of Metallurgical Engineering, Faculty of Engineering and Technology), Vaal University of Technology.
