【摘要】 主动运维技术是机械装备提高维修效率、降低维修成本、保证长时间稳定运行的必要保障。本文以滚动轴承作为研究对象,提出了一种基于双维度注意力残差收缩网络的智能故障诊断方法。该方法首先通过引入卷积注意力模块,分别对振动信号在空间与通道维度上的加权赋值来强化重要的特征,使网络的注意力集中在对故障分类更关键的信息上。其次,通过基于混合域注意力的软阈值化模块实现数据降噪以及去除特征中的冗余信息。实验结果表明,所提出的方法能够有效实现故障诊断和分类,且具有良好的鲁棒性。更多还原

【Abstract】 Active operation and maintenance technology is a necessary guarantee for mechanical equipment to improve its maintenance efficiency, reduce its maintenance cost, and ensure its long-term stable operation. In this paper, an intelligent fault diagnosis method based on dual dimensional attention residual shrinkage network is proposed, taking rolling bearing as the research object. In the proposed method, firstly, a convolution attention module is introduced to weight the processed signal in its spatial dimension and channel dimension to strengthen important features, focusing the network’s attention on information which is more critical for fault classification. Secondly, the functions of data denoising and removing redundant information in features are realized through an adopting soft thresholding module based on mixed-domain attention. Experimental results show that the proposed method can realize the effective diagnosis and classification of rolling bearing faults, and has good robustness.更多还原