Multi-fault Feature Separation of Rolling Element Bearing Using Singular Value Decomposition and Kurtosis

In order to separate multi-fault of rolling element bearing, and improve diagnosis accuracy, a method based on singular value decomposition and kurtosis was proposed.The singular value decomposition was applied to decompose the picked two-channel vibration signals.Differential spectrum of singular value decomposition and normalized kurtosis were exploited to filter and restructure the components processed by singular value decomposition respectively.Then, hilbert envelope spectrum was utilized to extract single fault feature.Finally, Examples from experimental tests show that the developed approach is effective for bearing multi-fault detection.Compared with hilbert envelope spectrum method, this method can separate weak fault components in practice and improve the ability of extracting transient impact signals.Moreover it can recognize the rolling bearing fault types and locations effectively.The approach can be used for fault detection of failures arising from local damage of rolling element bearing.