Jixie chuandong (Dec 2020)
Feature Frequency Extraction Algorithm based on PCA and MK-MOMEDA and Its Application
Abstract
Aiming at the problem of extracting characteristic frequency of flexible thin-walled bearings，a feature frequency extraction algorithm combining principal component analysis （PCA） and multi-point optimally adjusted minimum entropy deconvolution （MOMEDA） is proposed. In the algorithm，PCA is used to perform noise reduction processing on the original signal to obtain a reconstructed signal. The multipoint kurtosis （MKurt） is used to extract the period of the periodic shock signal in the reconstructed signal，and the theoretical period is corrected to obtain an accurate deconvolution period，enhance the reconstructed signal through MOMEDA，highlight its periodic impact，and extract the characteristic frequency more effectively. This method is applied to the fault feature frequency extraction of flexible thin-walled bearings，and compared with the maximum correlation kurtosis deconvolution （MCKD） algorithm. The results show that this method can separate bearing fault shocks from periodic shocks caused by the alternation of the bearing's long and short axes，eliminate the interference of such normal periodic shocks，and effectively extract the fault characteristic frequency in the signal. The effect is better than the maximum correlation kurtosis deconvolution algorithm.
