Proposal of abnormally detection method using cutting sound

抄録

A recognition method of the tool wear and the tool life is proposed in this research. Abnormally detection is performed using a cross-correlation function. The cutting sounds are recorded by a non-directional microphone and Spectral subtraction method is applied to remove noise spectra. Two kinds of treatment to the recorded cutting sounds is applied: volume normalization and normalization on the basis of summation of power spectrum. Some kind of frequency band are also selected when the correlation value is calculated. As a result, the volume normalization is better and frequency bands in which signal changes significantly due to the wear should be applied for this analysis. The feasibility of this research is demonstrated through some case studies.