Aiming at the problem that it is difficult to extract rolling bearing weak fault features under the condition of low signal-to-noise ratio

this paper proposes a fault feature extraction method based on zero-phase component analysis (ZCA) and non-convex overlapping group shrinkage (NOGS). First of all

the ZCA standardization process is performed on the bearing vibration signal to reduce the correlation between the data and improve the important component of the fault. Secondly

a non-convex overlapping group shrinkage model is established to improve the accuracy of feature extraction by using the characteristic of strong sparsity of non-convex functions. Aiming at the problem that NOGS parameter selection depends on prior knowledge

this paper introduces a binary periodic sequence to formulate parameter solutions

and combines majorize-minimization (MM) algorithms to obtain the sparse solution of NOGS. Finally

the noise reduction effects of overlapping group shrinkage (OGS)

empirical mode decomposition (EMD)

and maximum correlated kurtosis deconvolution (MCKD) were verified by simulating the fault signal and the actual fault signal of the CNC machine feed system bearing. The feature extraction effect of ZCA-NOGS is more obvious

and it improves the problem that the energy of high-amplitude pulses is suppressed when the traditional L1 norm extracts fault features.