报告详情
  1. 活动首页
  2. 报告清单
  3. 报告详情
Multimodal Generalized Zero-Shot Fault Diagnosis: A New Paradigm via Relation Graph-Guided Semantic Refinement
编号：164 访问权限：仅限参会人 更新：2026-10-19 18:04:13 浏览：19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
Generalized zero-shot learning  has shown great promise in fault diagnosis by enabling the simultaneous identification of known and un known faults using only known category data. However, existing methods remain constrained by two fundamental challenges: the information silo resulting from underutilized multimodal signals, and the semantic shift between static predefined prototypes and dynamic deep features. To overcome these bottlenecks, this paper establishes a novel diagnostic paradigm, multimodal generalized zero-shot fault diagnosis, and proposes a relation graph-guided semantic refinement  framework. Specifically, a modal relation-induced fusion network models inter-sensor dependencies via a learnable graph structure to achieve effective multimodal fusion while preserving modality-specific characteristics. In addition, a multimodal semantic reconstruction and refinement mechanism is developed to resolve semantic shift. It reconstructs semantic prototypes using known multimodal features guided by mutual information, and subsequently performs dynamic alignment via optimal matching during inference to closely approximate the true distribution of unseen classes. Extensive experiments on rotating machinery experimental platform demonstrate that the proposed model significantly outperforms six state-of-the-art zero-shot and multimodal baseline methods.
关键词
Fault diagnosis; Graph neural network; Multimodal Learning; Open-set learning; Zero-shot learning
报告人
Lv Wang
Doctor Chongqing university

稿件作者
Lv Wang Chongqing university
Yi Qin Chongqing university
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