Abstract: Aiming at the problem that the early weak faults of rolling bearing was difficult to be extracted completely, an algorithm for the detection of weak faults in bearings based on the fast spectral correlation(FSC) and periodic optimum singular value decomposition(O-SVD)was proposed.Firstly, through theoretical and simulation analysis, the traditional truncated singular value decomposition (T-SVD) algorithm with missing detail features was improved, and the O-SVD algorithm with Correlation Coefficient as index to judge effective singular value decomposition subspace was proposed.Then,the period optimized SVD was used as the preprocessing unit to decompose and reconstructed the fault signal of rolling bearing, then the reconstructed signal was computed by fast spectral correlation. The enhanced envelope spectra with obvious features and good preservation of local detail features were obtained. Finally, based on the simulation model, the shortcomings of the existing algorithms were analyzed, and the applicability of the proposed algorithm under low signaltonoise ratio (SNR) was illustrated with fault recognition rate as an index. The results show that, compared with the contrast algorithm, the proposed algorithm can extract the features of fault signals completely under the conditions of early weak fault, compound fault and composite fault, and has good engineering applicability.

Key words:  rotary machines; weak fault; feature extraction; optimum singular value decomposition(OSVD); fast spectral correlation (FSC); noise reduction separation