De-noising Methods Based On Gabor Transform And Blind Source Separation With Applications

The gear system, considered as the necessary mechanical transmission equipment in modern industry, has many advantages such as large driving force, exact transferring motion, smooth transmission and so on. Gear system is becoming more and more important in transferring power and motion along with the development of large-scale, high efficiency and high strength equipments. However, the gears and gear boxes may be damaged and become failure easily because of its complex structure and bad running conditions. Accordingly, it is significant to research on condition monitoring and fault diagnosis for gear system by advanced technologies, which could not only change the current postmortem servicing and regular examining to repairing according to the specific state, but also bring more economical and social benefit.The thesis is focused on the typical fault mechanism and vibration characteristic of gear system. It is mainly including three feature extracting techniques for fault, including the de-noising method based on Gabor transform and the Blind Source Separation(BSS) algorithm, new BSS method based on Gabor transform and under-determined BSS based on Gabor transform. The methods mentioned above are used in extracting the fault features and the results show that three algorithms can get better de-noising performance, which could provide a new thought for faults diagnosis of the gear system.The main research contents and the key conclusions are shown as follows:(1)The fundamental principle of BSS is studied. The batch-processing algorithm such as JADE, the self-adaptive algorithm such as Infomax and the fixed- point algorithm FastICA are researched in detail. And their characteristics are also analyzed. At the same time, some new de-noising performance indexes are proposed in the paper. The insufficiencies of that of the blind source separation are overcomed by these improved indexes.(2)The typical faults of gear system are introduced in detail, and the fault mechanism and vibration characteristics are analyzed, which are the basis for fault diagnosis. According to the laboratorial condition, some works are done, including designing the experimentation, improving the former gear box faults diagnosis platform, and carrying out the experiment with different diagnosis styles.(3)According to the time-frequency analysis and the blind source separation, three techniques are proposed,which are the de-noising method based on Gabor transform, BSS based on Gabor transform and under-determined BSS based on Gabor transform. The simulation results show the feasibility and validity of these methods, which provides a new research direction for the mechanical fault diagnosis.(4)In fault diagnosis of gearbox, the separation of typical diagnosis signals are quite successful by the three methods proposed in the thesis. The characteristic of the separated signals is quite accordant with the setting faults. At the same time, the results show that three algorithms can get better de-noising performance in processing the measured signals.At last, the main results of the thesis are summarized in the last chapter, and some possibly valuable research directions are also pointed out.