一种残差注意力迁移学习方法及其在滚动轴承故障诊断中的应用

A New Transfer Learning Method with Residual Attention and Its Applications on Rolling Bearing Fault Diagnosis

提出了一种基于残差注意力卷积神经网络(CSRA-CNN)的迁移学习算法,用于提高滚动轴承的故障诊断精度。在卷积神经网络模型中加入残差注意力机制,使模型在训练过程中更加注重故障特征的提取,从而有效提高迁移准确率。为了测评基于残差注意力卷积神经网络的性能,将其与传统卷积神经网络在不同迁移学习策略下的结果进行对比。用动力传动故障诊断综合实验台和高速列车综合实验台对所提算法进行了验证,该方法可以完成变转速以及变转速变载荷下轴承不同健康状态的迁移学习,且迁移效果均优于传统的卷积神经网络。

A transfer learning algorithm was proposed based on class-specific residual attention convolutional neural networks(CSRA-CNN)to improve the fault diagnosis accuracy of rolling bearings.Residual attention mechanism was added to the convolutional neural network model,which made the model pay more attention to fault feature extraction in the training,and also improves the migration accuracy effectively.To evaluate the performance of the proposed method,the results were compared with traditional convolutional neural network under different transfer learning strategies.The proposed algorithm was verified by fault diagnosis integrated test bench of power transmission system and high-speed train comprehensive test bench.The results show that the proposed method may complete the transfer learning of different health states of bearings under variable speed and variable speed and load,and the transfer effect is superior to that of the traditional convolutional neural network.