自 适 应 多 通 道 卷 积 稀 疏 轴 承 微 弱 故 障 诊 断

Strong background noise makes it difficult to extract weak fault features from the signal. Traditional methods based on single-channel signals neither consider the difference in spatio-temporal coverage nor overcome the information limitation of a single signal, making it hard to extract early weak fault features effectively under strong background noise and interference. In order to better extract the weak bearing fault features in the signal, an adaptive multi-channel convolutional sparse(AMCCS)model is proposed for extracting early weak fault features. In this model, a channel dictionary is constructed to learn the channel information in the multichannel signal, and a convolutional dictionary is constructed to learn the fault information in the signal, so as to adaptively learn the shared weak fault feature information and channel weight information from the multi-channel fault signals. Then, an efficient iterative update algorithm is developed to handle the updating of channel dictionary and convolutional dictionary. Finally, the algorithm is validated using bearing fault simulation signals and experimental signals with low signal-to-noise ratio. The results demonstrate that the proposed method exhibits significant superiority and robustness in bearing fault feature extraction under low signal-to-noise ratio conditions.