报告详情
  1. 活动首页
  2. 报告清单
  3. 报告详情
Lightweight Bearing Fault Diagnosis via Cross-Domain Knowledge Distillation for Edge Deployment
编号：175 访问权限：仅限参会人 更新：2026-10-19 18:07:08 浏览：1次 张贴报告

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
Edge-based bearing fault diagnosis must cope with cross-domain distribution shifts under tight computational budgets. This paper proposes a semi-supervised cross-domain knowledge-distillation framework using dual-view short-time Fourier transform and envelope-spectrum inputs. A DenseNet-lite teacher is pretrained on labeled source data and adapted with a small labeled target set; a depthwise separable convolutional neural network student is trained by output and feature distillation, cross-domain class-center alignment, and confidence-filtered pseudo-label supervision. Across representative single- and multi-fault-size cross-domain bearing fault diagnosis benchmarks, the student achieves average accuracies of 97.83% and 93.61%, with standard deviations of 1.31% and 3.51%, respectively, while using 18.83% to 21.48% of the teacher parameters and 19.35% to 24.65% of its per-sample multiply-accumulate operations (MACs).
 
关键词
Bearing fault diagnosis, edge computing, knowledge distillation, semi-supervised learning, transfer learning
报告人
HaoYi He
student AFEU

稿件作者
HaoYi He AFEU
Xinyu Yang AFEU
Siyu Shao AFEU
Shulin Cai AFEU
Chenyang Wang AFEU
Rui Mo AFEU
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

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