Use of Bispectral Measures in Machines Faults Diagnostics—Examples

Abstract

It is common knows that the power spectrum based methods cannot detect the phase relationship between different frequency components and additionally suppresses the phase information. It is therefore necessary to explore spectral measures of higher order, like the bispectral measures, to detect various forms of phase coupling between frequency components. In the paper was analyzed the impact of nonlinearity of the sub-section on the behaviour of the whole system by using of bispectral measures: diagonal bispectrum, maximum bispectrum and residual bispectrum. The results pointed to high sensitivity of bispectral measures to changes of the signal’s frequency structure and to the possibility of using these relations while constructing models of development of degradation-and-fatigue-related processes. In the paper was build effective and sensitive diagnostic measure of quality changes of fatigue crack growth at low-amplitude fatigue testing, fatigue process of toothed wheel damage or electric motors bearings faults. To do this, It was create a new measure (nobody else didn’t that way) which is able to extract the relevant diagnostic information. Integral of bispectral noise from bispectral maximum diagrams and integral of bispectral noise from bispectral residual diagrams were calculated with maximum given level (everything higher than maximum level was equalized to this maximum level) for every measurements.