A study of two bispectral features from envelope signals for bearing fault diagnosis

: To accurately detect and diagnose bearing faults, bispectral analysis has received more attention recently because of its unique property of noise reduction and nonlinearity extraction. Particularly this study investigates two typical bispectra: conventional bispectrum (CB) and modulation signal bispectrum (MSB) for suppressing noise influences in envelope signals and hence obtaining more accurate diagnostic features. The first component from the diagonal slice of CB results and that of the subdiagonal slices of MSB results are taken as the diagnostic features considering effective inclusion of information and easy of computations. Simulative and experimental studies show that both MSB and CB features result in good diagnostic performances but MSB may outperform CB slightly in that it shows smaller variance in attaining the feature and more sensitive to weak fault signatures. This merit of MSB may be due to that the MSB feature has more diagnostic information as it is the combination of first three harmonics, whereas the CB feature is combined from just the first two harmonics.