Underdetermined Blind Separation for Bearings Faults Based on the Improved Morphological Filter
Aiming at the condition of the strong background noise, many interference sources and unknown the number of sources in the actual industrial field, a method based on multi-scale multi-structure close-open average combination morphological filtering(C-OACMF) and sparse component analysis (SCA) was proposed to deal with the blind source separation problem of rotation machines. First, the C-OACMF was used to filter out background noise signals and extract the characteristic signal of observation signals, then using the simulated annealing genetic algorithm of fuzzy C-average clustering algorithm estimates the mixing matrix, the linear programming is finally used to estimate the source signals. Through the actual environment composite rolling bearing fault vibration signal extraction experiments verify the effectiveness and accuracy of the proposed algorithm.