| Summary: | 碩士 === 正修科技大學 === 電機工程研究所 === 99 === For normal operation of the rotating machinery, preventing and detecting the incipient faults of the machinery is an important task, especially in the production lines of the high-technology industry. If the incipient fault cannot be detected in its early stage, then it will lead to the serious faults. The close coupling of the machinery leads to the cause of its vibration faults more diverse and complicated; however, the vibration status is still a vital indicator for identifying whether the machinery operates normally or not. Therefore, rotating machines’ vibration monitoring and fault diagnosis method is not only significant in productivity and economic benefits of high-technology industry, but also an important research topic in the field of electrical engineering.
Vibration faults of rotating machinery are with diverse characteristics. Although the traditional error back-propagation neural network can be used to diagnose the vibration faults, it is with a long training time and its parameters are difficult to be determined. Therefore, to solve the vibration signal measurement and fault diagnosis of rotating machinery, this study first used the triaxial accelerometer to measure vibration signals, then the Particle Swarm Optimization (PSO) tuned General Regression Neural Network (GRNN) model diagnosed the vibration faults. This study has proposed PSO to adjust the smoothing parameter of the GRNN, and model performances of the optimal diagnosis model obtained were compared with those of the traditional neural networks. Test results have shown that the proposed methods are with less model constructing time and higher diagnosis accuracy than those of the traditional methods. Therefore, this study has confirmed the feasibility of the proposed approach to practical system applications.
|
|---|
