An Optimal Parameter Selection Method for MOMEDA Based on and Its Spectral Entropy.

As a vital component widely used in the industrial production field, rolling bearings work under complicated working conditions and are prone to failure, which will affect the normal operation of the whole mechanical system. Therefore, it is essential to conduct a health assessment of the rolling bearing. In recent years, Multipoint Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA) is applied to the fault feature extraction for rolling bearings. However, the algorithm still has the following problems: (1) The selection of fault period depends on prior knowledge. (2) The accuracy of signal denoising is affected by filter length . To solve the limitations, an improved MOMEDA (IMOMEDA) method is proposed in this paper. Firstly, the envelope harmonic-to-noise ratio () spectrum is adopted to estimate the fault period of MOMEDA. Then, the improved grid search method with spectral entropy as the objective function is constructed to calculate the optimal filter length used in the MOMEDA. Finally, a feature extraction method based on the improved MOMEDA (IMOMEDA) and Teager-Kaiser energy operator (TKEO) is applied in the field of rolling bearing fault diagnosis. The effectiveness and generalization performance of the proposed method is verified through comparison experiment with three data sets.