| （英） |
Status checks of mechanical equipment are typically being conducted by workers (manpower) at the actual place. However, since the check timing is limited (i.e., periodic check), the delay of anomaly detection could be the check interval at the maximum. Furthermore, the financial cost and the effort could be extremely high. Since some abnormalities typically appear in the operating sound, we have proposed an IoT system that collects the operating sound automatically for anomaly detection. However, outdoor operating sounds have the problem of containing various noises in the audible range (e.g., bird call and car engines). In this paper, we avoid audible noise and focus on the ultrasonic component whose frequency is higher than that of the audible sound and propose a new system that collects the ultrasonic data effectively. Through experiments, we demonstrated that the ultrasonic components appears in the operating sounds, irrespective of the environmental noise and their data can be effectively abstracted by utilizing filters and be decreased by FLAC compression technology. Finally, we have shown that it may be possible to estimate the operating status based on the amount of the compressed data size. |