基于深度残差收缩网络和迁移学习的变工况轴承故障诊断

Fault Diagnosis of Bearings Under Variable Working Conditions Based on Deep Residual Shrinkage Network and Transfer Learning

为了更快速、准确地提取轴承的故障特征,本文在卷积神经网络的基础上,引入残差项并添加软阈值和注意力机制,构建深度残差收缩网络,提取轴承的故障特征信息;并且为了避免出现神经元坏死现象,使用LeakReLU代替ReLU作为激活函数。由于轴承在实际应用中所处的工况并不固定,因此本文通过迁移学习方法,将训练的网络模型应用到不同工况中,并且对本文模型与传统的卷积神经网络模型在不同工况下轴承故障诊断的效果进行对比,验证本文所提方法的有效性。

In order to extract the fault features of bearings more quickly and accurately,based on the convolution neural network,in this article we introduce the residual term and add the soft threshold and attention mechanism to build a deep residual shrinkage network to extract the fault feature information of bearings.Moreover,in order to avoid the phenomenon of neuronal necrosis,LeakReLU is used to replace ReLU as the activation function.Because the working condition of the bearing is not fixed in practical application,we apply the trained network model to different working conditions through the transfer learning method,and compare the effectiveness of the model and the traditional convolution neural network model in bearing fault diagnosis under different working conditions,which verifies the effectiveness of the method proposed in this article.