Weak fault feature extraction of gearbox's bearing based on EMD-bispectrum

An EMD-bispectrum method is proposed to extract the weak fault feature of gearbox's bearing. The vibration signal is decomposed to a series of IMFS by EMD, and bispectrums of main IMFs are added up as the true bispectrum of the vibration signal. The influence of the cross-components created by the direct bispectrum of multicomponents signal and the eliminating principle of cross-components by EMD-bispectrum are explained by simulation. Three groups of vibration signals of gearbox's bearing under different states are analyzed by EMD-bispectrum, the experimental results show that the cross-components can be eliminated and the noise can be restrained effectively by EMD-bispectrum, thus the true bispectrum is obtained and the weak fault feature of gearbox's bearing could be extracted effectively.