Adaptive Maximum High-Order Cyclostationarity Blind Deconvolution Method for Rolling Bearing Fault Diagnosis

The blind deconvolution (BD) algorithm is an effective tool for fault identification, which can extract excitation sources from noise observation. Among them, the maximum second-order cyclostationarity blind deconvolution method (CYCBD) can effectively extract the weak periodic pulse signals associated with bearing faults. However, due to the need to set the filter length in advance, the filter length selection may be inappropriate, and the fault signal extraction needs to be more accurate. Therefore, to determine the optimal filter length, an adaptive maximum high-order cyclostationarity blind deconvolution method is proposed in this paper. The ratio of maximum peak to the amplitude of interference is the optimal criterion for determining the filter length and constructing the γ-order power envelope signal under Gaussian conditions. This method can suppress the interference frequency and find the filter length that makes the fault frequency most obvious. The method is applied to the fault feature extraction of bearing experiments, and compared with the CYCBD envelope, the results show that the method can not only successfully extract the fault feature, but also better suppress the interference frequency.