A fully blind and adaptive filter method to solve the inverse problem in vibration-based gear damage detection

In vibration-based gear damage detection, the fault signatures are usually corrupted by noise and convolution of vibration transmission channels. Deconvolution techniques have been widely studied to solve the inverse problem for restoring fault signatures. The mainstream methods, however, often fail due to the requirement of prior knowledge about fault signatures or the experience-based determination of filter parameters. In this paper, a fully blind and adaptive filter method is proposed. Specifically, a new criterion is defined, and the filter method is formulated as a maximization problem based on the criterion. The detailed deduction and pseudo-code for iteratively solving the problem are presented. Moreover, the filter method embeds an advanced swarm intelligence algorithm—that is, the grey wolf optimizer to achieve the adaptive determination of the optimal filter parameters according to the signals. In this context, it is a fully blind and adaptive filter method without requiring any prior knowledge and experiences about the target signatures to pre-determine filter parameters. Analysis results of a simulated signal and the comparisons with several state-of-the-art methods show the effectiveness of the proposed method for restoring the repetitive fault impulses from the mixed signal containing complex interferences. Fault-injection experiment and engineering application are carried out and the results demonstrate the performance of the proposed method for solving the inverse problem in gear damage detection.