A study on anomalous sound detection in factories for early failure detection using wavelet transform
Masaya Ueda, Daisuke Tanaka, Mikiko Tanaka · 2024
When a machine makes an anomalous noise, it is often necessary to take measures such as stopping the factory line. Therefore, it is necessary to detect machine malfunction early and take measures. Studies have been conducted to find an anomalous sounds in factories using machine learning. With the method of directly inputting and analyzing sounds of inside the factory, it is difficult to capture the time changes and frequency characteristics of the waveform. Thus, methods using frequency spectrograms have been proposed in recent years. However, frequency spectrograms have a problem in that it is difficult to simultaneously improve the time resolution and frequency resolution. Therefore, we propose a method using wavelet transform, which is said to be able to simultaneously improve time resolution and frequency resolution. We confirmed that the results of experiments were good.