摘要
针对现有轴承故障诊断方法应对噪声以及变负载条件下诊断能力不足问题,提出一种基于卷积神经网络(Convolutional Neural Networks, CNN)和有效通道注意力模块(Efficient Channel Attention,ECA)的滚动轴承故障诊断方法。该方法首先通过卷积神经网络对原始信号进行自适应故障特征提取;然后使用ECA模块生成通道注意力权重,实现对通道全局特征信息的掌握,据此增强模型在噪声及变负载条件下特征提取能力;最后将所提取的特征信息输入Softmax分类器并输出结果,实现滚动轴承故障诊断。通过对比实验证明,相比于传统深度学习方法,该方法拥有优良的轴承故障诊断性能,并在噪声干扰以及变负载条件下仍能保持出色的故障诊断准确率。
In order to solve the problem of insufficient fault diagnostic capability of current diagnosis methods for rolling bearings under noise and variable working conditions,a new fault diagnosis method based on convolutional neural network(CNN)and efficient channel attention(ECA)module was proposed.At first,the original signal features were extracted by convolutional neural network.Then,ECA module wais used to learn the channel attention of the convolution layer,and the feature weight was generated from each channel to improve the diagnostic capability under noise and different loading conditions.Finally,the extracted feature signals were input to the Softmax classifier,and the output of the diagnostic result was obtained.The fault diagnosis of the rolling bearings was realized.The proposed method was applied to analysing the fault experiment data of rolling bearings,and the results were compared with those of the traditional deep learning method.It is shown that this method has better diagnostic accuracy than the traditional deep learning method even under noise and different loading conditions.
作者
谢天雨
董绍江
XIE Tianyu;DONG Shaojiang(School of Mechatronics and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,China)
出处
《噪声与振动控制》
CSCD
北大核心
2021年第2期111-117,共7页
Noise and Vibration Control
基金
国家自然科学基金资助项目(51775072)。
关键词
故障诊断
卷积神经网络
有效通道注意力模块
滚动轴承
fault diagnosis
convolutional neural network(CNN)
efficient channel attention(ECA)
rolling bearing