基于信号子空间的新型盲解卷积方法

Deconvolution method is widely used in fault shock extraction of vibration signals. However, due to complex and changeable operating conditions of equipment, fault feature period is difficult to accurately predict, and random impact interferences make the current deconvolution method be difficult to meet needs of fault impact enhancement in complex environment of industry site. Here, aiming at the above problems, a new blind deconvolution method based on signal subspace was proposed. With this method, testing signal space was decomposed with the singular value decomposition (SVD) method, and was separated into subspaces. Then, subspace noise was suppressed with sparse coding shrinkage, and the effective subspace was selected with the pulse sparsity index. Finally, fault pulses were extracted through iteration. The results of bearing varying rotating speed simulation tests and train bearing tests showed that the proposed method can not only effectively eliminate random impact and noise, and avoid effects of energy on subspace screening, but also realize accurate extraction of fault pulses in absence of correct fault feature period information.