Symlectic Periodic Convolution Tracking Method: Algorithm and Applications

The periodic pulse features induced by early damage to rolling bearings under a strong noise background are weak, and the signal decomposition method based on the symplectic geometry similarity transformation theory is difficult to solve such feature extraction problems. In this regard, this paper proposes the symplectic periodic convolution tracking (SPCT) method to achieve accurate extraction of weak state features. The SPCT method achieves accurate estimation of the fault period by constructing a reweighted cyclic kurtosis spectrum (RCKS). The RCKS method makes full use of the periodicity and impact of fault features, using the periodic unit pulse sequence as a weight vector to perform convolution operations with the original vibration signal. By plotting a spectral curve of the reweighted cyclic kurtosis value as a function of the candidate period, significant peaks in the curve correspond to the true failure period. Then a truncation matrix is constructed based on the estimated period, and accurate periodic pulse extraction is achieved through SGST. Through simulation and measured signals, the superiority of the period estimation technology based on RCKS are verified, and the SPCT method is verified to be an excellent periodic pulse feature extraction method.