Multimodal MOE for fault diagnosis of external gear pumps
ID:21 View Protection:ATTENDEE Updated Time:2025-11-10 10:55:06 Hits:193 Oral Presentation

Start Time:2025-11-22 17:20(Asia/Shanghai)

Duration:20min

Session:S2 Parallel Session 2 » S2-1Parallel Session 2-22 PM

No files

Abstract
The external gear pump, as the core component of the fuel system in aircraft engines, is crucial for its reliable operation. Therefore, conducting effective fault diagnosis is of great significance. In view of the limitations of methods based on single-modality data, and the problem that existing methods for integrating multimodal data fail to fully exploit the interaction between modalities, this paper proposes a new fault diagnosis method. Firstly, masking is applied to the pressure and vibration signals based on the physical effect of the gear pump, and features are extracted using a fully convolutional network(FCN). Secondly, based on the theory of partial information decomposition(PID), a Mixture of Experts(MOE) is constructed to depict the interaction relationship between different modalities. Finally, the outputs of each expert are fused through a re-weighting mechanism to obtain the final prediction result. Experiments on the gear pump dataset show that the proposed method has superior diagnostic performance.
Keywords
Fault Diagnosis
Speaker
Jinyang Wang
Mr. Hefei University of Technology

Submission Author
Jinyang Wang Hefei University of Technology
Xiaochuan Li Hefei University of Technology
Juan Xu Hefei University of Technology
Chuan Li Chongqing Technology And Business University
David Mba Birmingham City University
Submit Comment
Verify Code Change Another
All Comments
Important Date
  • Conference Date

    Nov 21

    2025

    to

    Nov 23

    2025

  • Oct 20 2025

    Draft paper submission deadline

  • Dec 08 2025

    Registration deadline

Sponsored By
IEEE Instrumentation and Measurement Society
South China University of Technology
Organized By
South China University of Technology