ハーモニックウェーブレットによる定常振動波形の異常性検出

This paper deals with the fault prediction of machines by detecting local abnormalities of the signal by using the wavelet transform. In this paper, the signal is a steady-state vibration waveform measured by an acceleration pick-up. The harmonic wavelet was used to detect the abnormalities of the signal. Then, it is necessary for applying the harmonic wavelets to remove the end effects and the unacceptable noise for the effective diagnosis. The authors proposed the method to remove the end effects and the noise from the signals by using the usual Fourier transform. The signals were limited to steady-state waveforms and the abnormalities of the signals were periodical. As the result, it was made clear that the local abnormalities were able to be detected from the signals due to the slight collisions caused by loose bolt and the fatigue crack generated in the in-line flow-induced vibration.