Abstract: The mathematical morphological analysis is a nonlinear analysis method of digital signal processing. The multiscale morphology analysis can describe the signals by layers depending on the varying structure elements. By morphological opening operation, a new method describing the rolling bearing faults characteristics is presented, which is not different from the time-frequency analysis. The mathematical morphological spectrum curves are created by multiscale morphological opening algorithm with varying flat structure elements, which could quantitatively show the different faults characteristics. The morphology spectrum entropies are gained by means of the morphology spectrum values according to the information theory, which quantitatively represented the shape characteristics. From the analysis of experimental data, it can be seen that the multiscale morphology method is a new idea for the faults extraction and classification of rotating machines.

Key words: Fault, Morphological spectrum, Morphological spectrum entropy, Rolling bearing