ROLLING ELEMENT BEARING FAULT FEATURE EXTRACTION BASED ON SPECTRAL KURTOSIS AND cICA
Yu Liang · Jixie qiangdu · 2013
Independent component analysis( ICA) can be used to separate the mixtures according to the sources. However,the requirement of known the number of sources in advance limited the application of ICA. The constrained independent component analysis( cICA) is a specific ICA method,which can be utilized to extract the desired independent components( ICs)by reference signal. Envelope extraction technology is valid to separate the sources of interference signal beyond the demodulation resonance band and reduce the dimension of signal sources. The newly developed spectral kurtosis( SK) based envelope extraction can be used to extract the envelope adaptively. By combining the SK based envelope extraction and the cICA scheme,a robust envelope extraction approach for the fault feature extraction of rolling element bearing( REB) is proposed in the paper.It extracted signal envelope adaptively by SK firstly. Then,by building the reference signal by priori knowledge on the characteristic frequencies of REB faults,the characteristic envelope ICs related to different rolling bearing faults can be extracted by cICA. Tests verified the validity of the present scheme positively