用户名 密 码 登 录 忘记密码 免费注册

网站会员中心 | 设为首页 | 加入收藏

 您好,欢迎光临中国轴承行业网         退出登陆

会员中心 | 设为首页 | 加入收藏

国外文献 > 1.轴承 > 7. 智能方法与计算 > 神经网络


关键词搜索:


标 题: A Channel Attention based Large Kernel Convolutional Neural Network for Bearing Fault Diagnosis
基于通道注意力的大核卷积神经网络用于轴承故障诊断
作 者: Fan Li
Decheng Wang
Ying Wang
Liping Wang
Jun Wu
Jiayu Sun
关键词: Fault diagnosis  [故障诊断]
Deep learning  [ 深度学习]
Convolutional neural network  [ 卷积神经网络]
Channel attention  [ 频道关注]
出版年: 2023
文集名: 3rd International Conference on Neural Networks, Information and Communication Engineering (NNICE 2023)
第三届国际神经网络、信息与通信工程会议(NNICE 2023)
会议名: 3rd International Conference on Neural Networks, Information and Communication Engineering (NNICE 2023)
第三届国际神经网络、信息与通信工程会议(NNICE 2023)
届 次: 3rd
摘 要: We propose an improved end-to-end Channel Attention based Large Kernel Convolutional Neural Network (CALKCNN) in this paper. The CALKCNN can automatically extract features from the original vibration signal and accurately diagnose bearing fault without any manual feature selection operations. We break the stereotype and innovatively use a large kernel in the first layer of CALKCNN to extract the long-time scale characteristics and suppress high-frequency noise, which improves the anti-noise and generalization ability. The CALKCNN can adaptively adjust the weights of feature maps generated by large kernel convolutions based on the channel attention mechanism, selectively focus on essential features and ignore unimportant features, which improves the adaptability and robustness. Compared with five excellent baseline models, the CALKCNN achieves state-of-the-art performance in bearing fault diagnosis by experiments using CWRU bearing fault dataset.

本文提出了一种改进的端到端通道注意力大核卷积神经网络(CALKCNN)。该网络能够自动从原始振动信号中提取特征,并在无需人工特征选择的情况下准确诊断轴承故障。我们打破了传统做法,在CALKCNN的第一层创新性地采用大核,以提取长时尺度特征并抑制高频噪声,从而提升抗噪能力和泛化性能。CALKCNN通过通道注意力机制,可自适应调整由大核卷积生成的特征图权重,有选择地聚焦关键特征,忽略无关特征,显著增强了模型的适应性和鲁棒性。与五种优秀的基线模型相比,在使用CWRU轴承故障数据集进行的实验中,CALKCNN在轴承故障诊断方面取得了最先进的性能。
中文内容为网络AI翻译,专业术语未必准确,仅供参考。

VIP网员

联系我们 | 登录注册 | 会员管理 | 网员服务 | 服务信息 | 广告服务 | 高薪聘请 | 法律声明 | 关于我们

京公网安备 11010202009788    京ICP备2021015424号