Amplitude Modulation Feature Extraction of Bearing Vibration Signal Using Singular Value Decomposition

In order to extract the faint fault information from complicated vibration signal of bearing, a feature extraction method based on singular value decomposition (SVD) is proposed. The principle of SVD based signal decomposition and feature extraction is analyzed, and it is indicated that the essence of signal decomposition of SVD is linear addition decomposition. On this basis, a Hankel matrix is constructed for bearing vibration signal, and by SVD processing a group of component signals are obtained, then some of front component signals are selected to be added together, and the amplitude modulation feature caused by damaged roll track is clearly extracted. In addition, the amplitude modulation feature extracted by different numbers of component signals is also investigated. The comparative result shows that the amplitude modulation extraction effect of SVD is much better than that of wavelet transform.