Feature Extraction for Rolling Bearing Diagnosis Based on Improved Local Mean Decomposition

The Independent component analysis (ICA) can only be applied to the over determined blind source separation, in which the number of observation sources is not less than the number of signal sources. Moreover, vibration signals of rolling bearing have the problems of nonlinear, non-stationary and fuzziness. To solve the above problems, a hybrid method with ICA, local mean decomposition (LMD) and fuzzy C-mean clustering (FCM) is proposed to diagnose rolling bearing fault in this paper. First, the PF matrix is obtained by local mean decomposition of the sampled single channel vibration signals. Then, some PF components which are highly correlated with the source signals are selected to reconstruct the signal by using cross-correlation criterion, and other PF components are used to construct the virtual noise channel. Next, the virtual noise channel and the vibration signal are used as the signal source of blind source separation, and the FastICA algorithm based on the negative entropy is used to separate the signal source and the noise, so as to achieve the purpose of noise reduction. Finally, the noise reduction signal is processed by FCM based on fuzzy mathematics, and the classification and recognition of the fault types of the vibration signals are carried out. The method is applied to the ground fault signal of rolling bearing. The results show that the method is effective and practical in the fault diagnosis.