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 > 机器学习


关键词搜索:


标 题: A cross domain deep learning method for rolling bearing fault diagnosis
用于滚动轴承故障诊断的跨域深度学习方法
作 者: Yue Zhao
Guorong Fan
Yuxing Cao
Yong Yang
Wenhua Gao
Zengshou Dong
关键词: Deep learning  [深度学习]
Fault diagnosis  [ 故障诊断]
Cross domain  [ 跨域]
出版年: 2024
文集名: 43rd Chinese Control Conference (CCC 2024)
第43届中国控制会议(CCC 2024)
会议名: 43rd Chinese Control Conference (CCC 2024)
第43届中国控制会议(CCC 2024)
届 次: 43rd
摘 要: Deep learning has widely used in fault diagnosis field. However, a vast of deep learning models for diagnosis focus on single perspective such as time domain and frequency domain. Therefore, the diagnostic accuracy is not satisfactory when the experiment under noise environment especially high noise. Cross Domain feature Fusion Network (CDFN) is proposed to seattle this problem with time domain and frequency domain simultaneously. Firstly, fast fourier transform (FFT) is used to transform raw vibration signal into frequency domain. Then, a Conv block is designed as main feature extractor for CDFN. Next, Fusion block composes time domain feature, frequency domain feature and previous cross domain feature together to generate new domain feature. At last, with severe Conv block and fusion block, the proposed CDFN realize cross domain feature extraction for classification. The CWRU dataset is used to validate the proposed method, the accuracy and T-SNE visulization reasults show that the proposed CDFN has strong ability for fault diagnosis problem and can retain effectiveness under hith noise environment.

深度学习在故障诊断领域已得到广泛应用。然而,现有的大量深度学习模型大多仅从时域或频域等单一视角进行分析,因此在噪声环境下(尤其是高噪声环境)的诊断准确性并不理想。为解决这一问题,本文提出了一种跨域特征融合网络(CDFN),可同时结合时域和频域信息。首先,采用快速傅里叶变换(FFT)将原始振动信号转换至频域; 其次,设计一个卷积块作为CDFN的主要特征提取器; 接着,融合模块将时域特征、频域特征以及先前的跨域特征组合,生成新的跨域特征; 最后,通过深层卷积块与融合模块,实现对分类任务的跨域特征提取。实验使用CWRU数据集验证了该方法,结果表明,所提出的CDFN在故障诊断任务中具有较强性能,并且在高噪声环境下仍能保持良好的有效性。
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