Optimal Maximum Cyclostationary Blind Deconvolution for Bearing Fault Detection
Jiahao Li, Yi Liu, Jiawei Xiang
Abstract
Maximum cyclostationary blind deconvolution (CYCBD) is a newly presented blind deconvolution method for extracting faults in mechanical systems. The method presents two challenges, which mainly come from needing to set the cyclic frequency in advance and a defining suitable filter length. To address these issues, an optimal maximum CYCBD method is developed. First, the raw signals are processed by the noise subtraction method to remove the inference of inherent impulses and environmental noise. Furthermore, the cyclic frequency is estimated by calculating the autocorrelation of the envelope signals to reveal the fault-induced impulses. Second, to determine the filter length adaptively and automatically, the residual autocorrelation energy (RAE) ratio is employed as the objective function to optimize the filter length by the improved success–failure method. Finally, the proposed method is applied to extract fault features. The effectiveness of the proposed method is validated with data obtained from an open-access bearing dataset and experimental test rigs.