A machine sound monitoring for predictive maintenance focusing on very low frequency band

The monitoring of machines is one of key issues in the Industry 4.0 era. Particularly, the monitoring realized by non-contact sensors is drawing attention since it is easy to install and to avoid any problems caused by sensors accidentally dropping down into the machines. For example, sound monitoring satisfies this requirement. In this paper, we propose to apply the Accumulation for Real-time Serial-to-parallel Converter (ARS) for the monitoring of machine sounds to analyse low frequency bands which have not been sufficiently investigated so far. The machine sounds captured in a real factory are analysed so that the change of the machine sounds which varies in accordance with machine status is detected. It is verified that ARS successfully detects the difference as precise as wavelet transform (WT) with Morlet wavelet even though its computational load is significantly lower than that of WT.