Presentation Details
  1. Home
  2. Presentation List
  3. Presentation Details
A Dual-Stage Progressive Feature Disentanglement Method for Cross-Condition Gearbox Bearing Fault Diagnosis
ID：35 View Protection：ATTENDEE Updated Time：2026-09-18 14:01:01 Hits：6 Poster Presentation

Start Time：Pending（Asia/Shanghai）

Duration：Pending

Session：No Session »

No files

Abstract
Rolling bearing fault diagnosis for subway train transmission systems faces severe feature distribution shift when operating conditions (rotational speed and lateral load) vary between training and deployment. A single feature extraction paradigm can hardly guarantee global cross-condition robustness and high resolution under extreme low-speed heavy-load conditions at the same time. To address this problem, this paper proposes a physics-driven dual-stage progressive diagnosis strategy. First, the rotational frequency is accurately estimated from the interference-free motor current signal. Then, in Stage I, wavelet packet band energy ratios are used to construct a globally robust baseline that is inherently immune to amplitude fluctuations caused by load variation. In Stage II, aiming at the feature aliasing of wavelet features under low-speed heavy-load conditions, envelope spectrum order energy ratios are introduced for microscopic refined diagnosis. Experiments on the PHM Data Challenge dataset (gearbox bearing faults TYPE9-TYPE12) show that the proposed method effectively overcomes cross-condition interference and improves the diagnostic accuracy from 83.33% of the wavelet-only baseline to 100%, providing a highly interpretable engineering paradigm for small-sample, cross-condition diagnosis of complex mechanical systems.
 
Keywords
Fault diagnosis,gearbox bearing,wavelet packet energy ratio,envelope spectrum,order analysis,cross-condition,PHM Data Challenge
Speaker
Chao Liang
Senior Engineer The Fifth Electronics Research Institute of Ministry of Industry and Information Technology

Submission Author
Yuan Li China Wuzhou Engineering Design Group Co., Ltd.
Chao Liang The Fifth Electronics Research Institute of Ministry of Industry and Information Technology
Zhiwei Ma China Wuzhou Engineering Design Group Co., Ltd.
Panpan Han China Wuzhou Engineering Design Group Co., Ltd.
Mou Sun China Wuzhou Engineering Design Group Co., Ltd.
Kai Zhou China Wuzhou Engineering Design Group Co., Ltd.
Shiyu Jia China Wuzhou Engineering Design Group Co., Ltd.
Submit Comment
Verify Code Change Another
All Comments
Important Date

  • Conference Date

    Nov 06

    2026

    to

    Nov 08

    2026

  • Oct 15 2026

    Draft paper submission deadline

Sponsored By
IEEE Instrumentation and Measurement Society
Organized By
Sichuan University
Contact Information
Login to view full contact details
Previous Conferences