Early Fault Diagnosis Method for Rolling Bearing Based on Improved Singular Values Decomposition

The weak early failure signals of rolling bearings are often masked by complex shock signals and background noise, making early rolling bearing failures difficult to diagnose. Singular value decomposition (SVD) has a wide range of applications in fault diagnosis. The noise reduction ability of the method is mainly affected by the matrix structure and the selection of reconstruction components. This paper proposes an improved singular value decomposition method, which is successfully used in the early fault diagnosis of rolling bearings. The core ideas of this paper include: firstly, the Hankel matrix is selected as the structure matrix, and for the selection problem of matrix embedding dimension, an adaptive method combining IDR index and periodic signal frequency is proposed to obtain the optimal Hankel matrix embedding dimension. Secondly, the sub-signal is reconstructed by the anti-angle reconstruction method, and the CFIC method based on the peak calculation of the envelope spectrum is invoked, which can obtain the sub-signal pairs containing feature information. Finally, the method is applied to the simulation signal of bearing failure and the actual test data case of early failure to verify the practicality and feasibility of the method.