Feature decoupling and cross domain alignment with transfer learning for cross working condition mechanical fault diagnosis
Honglian Xiao, Jianhua Xiao · 2026
Read Feature decoupling and cross domain alignment with transfer learning for cross working condition mechanical fault diagnosis by Honglian Xiao, Jianhua Xiao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13473462.txt.
Description
Honglian Xiao and Jianhua Xiao presented in 2026 a neural network for spotting machine faults when operating conditions change. Fault diagnosis from vibration and current signals is hard because sensors behave differently. A machine's signal pattern also shifts with load and speed, so models trained in one setting often fail in another. Their model, called WDH-Net, combines three ideas: wavelet-based enhancement, a dual-stream encoder separating fault signs from condition signs, and a mapping aligning features across domains.
Wavelet packet decomposition with cross-attention picks out fault details from mixed signals. The dual streams strip condition interference, and a multi-kernel discrepancy measure pulls feature spaces of different conditions together. Tests on two public bearing datasets, CWRU and PU, showed strong performance across working conditions, though no single accuracy figure is reported. For plant engineers, the value lies in a model that keeps diagnosing correctly when regimes shift.
Catalog record
| CREATOR | Honglian Xiao, Jianhua Xiao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 104 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |