Sinc-BAPSI: A Fault Diagnosis Network with Band-Aligned Interaction Across Measurement Points for Aero-Engine Inter-Shaft Bearings Under Unseen Operating Conditions
Changes in low-pressure/high-pressure (LP/HP) rotor speed combinations cause aero-engine inter-shaft bearings to exhibit different casing vibration distributions under different operating conditions. These changes make recognition under unseen conditions more difficult. To establish a common input-band reference for interactions across casing measurement points, this paper proposes Sinc-BAPSI. Parameter-shared SincConv filters four casing acceleration signals separately. Filters with the same index at different measurement points share cutoff-frequency parameters to provide a common input-band reference. Band-Aligned Pairwise Sensor Interaction (BAPSI) then encodes the responses at measurement points and their difference and product relationships within corresponding bands. It introduces interaction information through shared encoding, aggregation across measurement points, and residual enhancement. A CNN and a BiGRU subsequently perform local feature extraction and bidirectional temporal modeling. Evaluation with mutually exclusive operating conditions was conducted on 17 LP/HP operating conditions in the HIT aero-engine inter-shaft bearing dataset. Sinc-BAPSI achieved a mean Macro-F1 of 0.9924 and a mean Accuracy of 99.46%, with a mean Macro-F1 improvement of 0.0060 over CNN-BiGRU. Additional speed-range extrapolation tests showed that diagnostic performance varied with the extrapolation direction.
Authors
- Chong Dong (ORCID: https://orcid.org/0000-0002-3542-305X)
- Shuqi Liu (ORCID: https://orcid.org/0000-0003-0092-3529)
- Ting Fang (ORCID: https://orcid.org/0000-0002-8057-0649)
- Jiaming Han (ORCID: https://orcid.org/0000-0001-5513-5031)
- Hua Guo
- Shaosen Han
- Ming Wang
- Hongyu Liu
- Long Ma
Institutions
- Chinese Academy of Geological Sciences (CN)
- Anhui University of Technology (CN)
Publication Details
- Journal
- Sensors
- Published
- 2026-11-18
- DOI
- https://doi.org/10.3390/s26227296
- Primary Topic
- Machine Fault Diagnosis Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00