一种用于主轴轴承故障诊断的深度卷积动态对抗迁移网络

A deep convolutional dynamic adversarial transfer network for spindle bearing fault diagnosis

迁移学习智能故障诊断方法已经成为了机械设备故障诊断领域的一个研究热点。然而,大多数相关方法在迁移学习过程中未能合理地评估源域样本和目标域样本的相似性,且数据分布的差异会造成迁移诊断的结果不同。针对此问题,提出深度卷积动态对抗迁移网络用于主轴轴承智能故障诊断。该网络首先利用一维卷积神经网络从处理过的振动信号中自动提取特征集,然后利用动态对抗学习策略动态地调整条件分布和边缘分布在迁移学习过程中的重要程度,有效地提高迁移诊断的精度。通过数控机床主轴轴承故障诊断实验,验证了所提方法的有效性。实验结果表明,所提方法能够有效挖掘故障特征信息,实现不同工况之间的知识迁移,具有较好的应用价值。

Transfer learning based intelligent fault diagnosis method has become an important research direction in the field of me-chanical equipment fault diagnosis.However,most of the existing fault diagnosis models cannot reasonably calculate the impor-tance of marginal and conditional distributions in the process of transfer learning,and different data distribution will lead to different diagnostic results.To solve such problem,a deep convolution dynamic adversarial transfer network is proposed for intelligent fault diagnosis of spindle bearing.One-dimension convolutional neural network is used to extract transferable features.A dynamic adver-sarial learning strategy is introduced into the proposed method.The importance of marginal and conditional distributions in transfer learning is calculated according to the similarity of data distributions,which effectively improves the diagnostic accuracy.The effec-tiveness of the proposed method is verified in spindle bearing fault diagnosis of industrial machine tools.The experimental results show that the proposed method can powerfully explore fault features and realize knowledge transfer between different working con-ditions,which has important significance for the practical application industry.