2018 年 26 巻 6 号 p. 16-22
To construct a signal monitoring system that detects changes in the condition ofa source of sound, vibration, or the like from a signal obtained from the source,after FFT processing of the observed signal, the MT method was applied to thecentroid of the resulting spectrum. The centroid can be defined by two-dimensionalcoordinates(frequency, power)and Mahalanobis distances can be calculatedeasily from the centroid coordinates. It was confirmed that changes inthe condition of the signal could be readily recognized by this method. When thesignal being monitored is of the same type as in the unit space, theMahalanobis distance has a χ2 distribution with two degrees of freedom, so theoccurrence counts at the 5 % and 1 % significance levels of this distributionwere measured, and the MT method was applied again to these occurrencecounts. In this way a signal monitoring system in which the MT method wasapplied in two stages was constructed. When tested, it demonstrated the possibilityof implementing a signal monitoring system that could sensitively detectslight signal changes.