报告详情
  1. 活动首页
  2. 报告清单
  3. 报告详情
Information Fusion-Based Model-Agnostic Meta-Learning with Residual Efficient Channel Attention for Cross-Condition Few-Shot Fault Diagnosis
编号：69 访问权限：仅限参会人 更新：2026-09-22 15:59:17 浏览：51次 口头报告

报告开始：暂无开始时间（Asia/Shanghai）

报告时间：暂无持续时间

所在会场：[暂无会议] [暂无会议段]

暂无文件

摘要
To address the degradation in diagnostic performance caused by limited gearbox fault samples and variations across variable-speed operating conditions, information fusion-based model-agnostic meta-learning (IFMAML) enables cross-condition few-shot diagnosis by integrating multisensor information fusion with meta-learning, but it insufficiently exploits channel dependencies in deep fused features. IFMAML-ECA, which directly incorporates conventional efficient channel attention (ECA), scales the features using attention weights within the range of (0,1), which may excessively suppress transferable channel responses. To address this issue, this paper proposes an information fusion-based meta-learning method with residual efficient channel attention (ResECA), termed IF-REMAML, which enhances deep fused features through residual channel modulation and adopts an optimization strategy in which the ResECA parameters are fixed during inner-loop adaptation and updated during outer-loop optimization. Multiple cross-condition few-shot diagnostic scenarios are constructed on the Tsinghua University planetary gearbox variable-speed dataset. Experimental results demonstrate that the proposed method achieves favorable cross-condition diagnostic performance and outperforms several representative meta-learning methods.
关键词
Meta-learning,cross-condition diagnosis,few-shot learning,multisensor information fusion,residual channel attention
报告人
Leijun Shi
Master Student Beijing Institute of Technology

稿件作者
Leijun Shi Beijing Institute of Technology
Cuiying Lin Beijing Institute of Technology
YUN KONG Beijing Institute of Technology
发表评论
验证码 看不清楚，更换一张
全部评论
重要日期

  • 会议日期

    11月06日

    2026

    至

    11月08日

    2026

  • 10月15日 2026

    初稿截稿日期

  • 11月10日 2026

    注册截止日期

主办单位
IEEE Instrumentation and Measurement Society
承办单位
Sichuan University
联系方式
登录查看完整联系方式
历届会议
移动端
在手机上打开
小程序
打开微信小程序
客服
扫码或点此咨询