Fault diagnosis for rolling element bearings based on independent component analysis

To overcome the difficulty in extracting the characteristic signals for fault diagnosis of rolling element bearings, the independent component analysis (ICA) was employed to separate the acoustic signals generated by the rolling element bearing system from the mixed acoustic signals collected by microphone. Envelope analysis based on morlet wavelet transform was used to denoise again and extract the fault feature signals. Then, the ratio of the fault feature frequency of the fault feature signals and the frequency of the rotation frequency of the rotor was served as the input parameter of linear neural network to identify the fault pattern of rolling elemental bearings. Experimental result shows that the fault diagnosis approach for rolling element bearings is effective.