Feature extraction of weak faults of rolling bearings is essential for fault diagnosis. The initial faults under variable speed conditions are always weak and are covered by high background noise, making the extraction of fault features extremely difficult. It is crucial to extract the weak fault characteristics of rolling bearings correctly. A weak fault feature extraction method based on the sparse low-rank model of order-frequency spectral correlation decomposition is proposed in this paper. The angle/time cyclostationarity (AT-CS) is used to obtain the order-frequency spectral correlation (OFSC) according to the cyclic statistical characteristics of bearing signal in the angle domain under the variable speed conditions. It is found that a high degree of sparsity is expressed in the periodic pulses of OFSC. Then, the sparsity is used in the sparse and low-rank decomposition model to extract fault features. The Robust Principal Component Analysis (RPCA) algorithm is used to decompose OFSC into low-rank and sparse components. The sparse components correspond to periodic fault pulses, while the low-rank components represent interference. Finally, the Squared Envelope Spectrum (SES) is used to detect the fault characteristics of rolling bearings. The simulation results show that the method can effectively extract weak bearing fault features under low SNRs.
 

Rolling bearings,Variable speed conditions,Angle/time cyclostationarity,Order-frequency spectral correlation,Robust Principal Component Analysis (RPCA),Square envelope spectrum