Development of an intelligent system for vibration-based predictive maintenance /
A machine in the best of operating condition will have some vibration because of small, minor defects. The use of the human sense of touch and feel for observation is somewhat limited, and there are many common problems that are generally out of the range of human senses. Vibration monitoring is a w...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English |
| 出版: |
Kuala Lumpur :
Kulliyyah of Engineering, International Islamic University Malaysia,
2014
|
| 主題: | |
| 在線閱讀: | http://studentrepo.iium.edu.my/handle/123456789/4580 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | A machine in the best of operating condition will have some vibration because of small, minor defects. The use of the human sense of touch and feel for observation is somewhat limited, and there are many common problems that are generally out of the range of human senses. Vibration monitoring is a widely used and cost effective monitoring technique. It detects, locates, and distinguishes faults in rotating machineries. It is an established process used in predictive maintenance as it is necessary to diagnose faults in machine at early stages to prevent failure during operation. In this research an intelligent method to detect faults in rotating machineries by analyzing vibration signals was developed. The faults that can be detected are some of the most common faults in rotating machineries. An experimental set-up was designed and fabricated to observe the signals generated when it is in normal working condition and when it is in faulty condition. The components whose vibration signatures were observed are rotor disk and motor. The faulty rotor disk, mechanical looseness, and fault motor vibration signatures were studied. Four features from vibration signals for various faults were extracted in the time domain. They are Root Mean Square (RMS), crest factor, kurtosis, and skewness. These features are mapped against the respective faults using a multilayer feed forward artificial neural network. The network was trained using Levenberg-Marquardt algorithm. The simulated faults condition signal were analyzed and compared to normal condition signals. The analysis of the fault signature shows that fault conditions in the system are detected for the various components. In this research, the developed artificial neural network is able to detect the faulty conditions. The trained neural network can classify different condition with 92.5% accuracy and the precision is 0.9. For further research, it is suggested that the artificial neural network be trained to detect more inherent faults in the system components. |
|---|---|
| Item Description: | Abstracts in English and Arabic. " A dissertation submitted in fulfilment of the requirement for the degree of Master of Science in Mechatronics Engineering."--On t.p. |
| 實物描述: | xv, 86 leaves : ill. ; 30cm. |
| 參考書目: | Includes bibliographical references (leaves 67-73). |