To improve accuracy at an initial stage of monitoring when detecting a malfunction of a bearing by frequency analysis of vibration acceleration.

In this method, a vibration acceleration from a bearing part of a rotating apparatus is measured, the power spectrum of a frequency component is obtained by fast Fourier transform, and a vibration acceleration of a pilot frequency is extracted from the power spectrum for detecting a malfunction of a bearing attached to the bearing part based upon a degree of increase from normal times. Before the fast Fourier transform, discrete wavelet transformation is performed on the vibration acceleration at a digital filter 31 for determining wavelet coefficients, those of the wavelet coefficients exceeding a threshold set from a value measured during monitoring is removed, and inverse discrete wavelet transformation is performed on the wavelet coefficients after the removal for obtaining a corrected vibration acceleration.