الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 93 |  |  |
| | | from | 93 | | |
Abstract
Rolling Element Bearings (REBs) are essential components in every rotating machine. Early REBs faults detection and identification has been hard to achieve due to weak fault features in the vibration signal. Resonant band demodulation and Envelope Analysis (EA) have been used to tackle this problem. However, vibration signals acquired from complex systems are usually contaminated with periodic and gear mesh interferences, are subject to transmission path effect and are masked by white and impulsive noise. These contaminants make it hard to select a proper frequency band for demodulation. In this thesis, Gini Index (GI), which is a recently introduced impulsiveness measure in the field of bearing fault detection, is employed as a criterion for blind frequency band selection. After investigating the suitability of GI for bearing fault detection using a variety of simulated time series, GI is used to optimize the parameters of Morlet wavelet filter. The developed filter was designed to be used in rolling bearing fault detection in an adaptive and automatic manner and was tested using simulated and experimental signals. The results showed the effectiveness and efficiency of the designed filter. A three-step adaptive and automated filtration scheme is then proposed to enhance the bearing fault feature and remove noise and interferences. The three adopted consecutive filters are: Inverse Autoregressive (IAR) filtration, Maximum Gini Index Deconvolution (MGID) and Morlet wavelet filtration. The three filters are guided and optimized by GI. The proposed approach is tested using selected experimental signals of different fault locations (ball, inner and outer race faults). The results showed the supremacy of the proposed approach and demonstrated the usefulness of GI as a guiding criterion for proper filter design for REBs fault detection.