A method of identifying at least one defect of a rotating component, where the defect is selected from a group of predefined defects, includes extracting frequency data related to each of the predefined defects in order to form a first spectrum signature for each of the predefined defects. Also measuring at least one vibration signal produced by the rotating component to obtain at least two second spectrums, filtering each second spectrum based on an exponential smoothing algorithm, selecting peaks in each second spectrum according to a prominence of each of the peaks, setting selected peaks to zero if the selected peaks are not present in a predefined number of consecutive second spectrums, and calculating a probability that each first spectrum signature corresponds to at least one of the second spectrums.