抄録
Failure detection of gears in transmissions plays a critical role in the development of machinery health monitoring system. Although many studies of metal gears health monitoring have been carried out, plastic gear failure detection has been mostly unknown. To perform more research projects on plastic gears, we constructed an automatic data acquisition system in our laboratory. Many endurance tests were carried out from healthy until broken situations to collect vibration data of plastic gears. In the endurance test, the crack growth of gear tooth root can be captured by a high-speed camera, which then was used to label data. Besides, an intelligent diagnosis system was developed to monitor the condition of plastic gear during the endurance test. The proposed system can learn from visualized images created from vibration data. In orient toward achieving high accuracy of learning by the proposed system, an enhanced method was discovered to collect valuable information into data. In this paper, the Fourier decomposition method was employed to reconstruct data from some specific frequency bands. The evaluation of reconstructed data not only revealed an unhealthy situation of plastic gear before initial crack happened but also indicated sensitive frequency bands that can be used for efficient learning.
