Szczegóły publikacji
Opis bibliograficzny
ID CNN-based domain adaptation for robust bearing fault diagnosis / Paweł KNAP, Urszula JACHYMCZYK // W: IEEE MetroXRAINE 2025 [Dokument elektroniczny] : 2025 IEEE International Conference on Metrology for eXtended Reality, Artificial Intelligence and Neural Engineering : 22-24 October 2025, Ancona, Italy : conference proceedings. — Wersja do Windows. — Dane tekstowe. — Piscataway : IEEE, cop. 2025. — Dod. ISBN: 979-8-3315-0280-5 (print on demand). — e-ISBN: 979-8-3315-0279-9. — S. 860-865. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 865, Abstr.
Autorzy (2)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 170312 |
|---|---|
| Data dodania do BaDAP | 2026-10-30 |
| Tekst źródłowy | URL |
| DOI | 10.1109/MetroXRAINE66377.2025.11340399 |
| Rok publikacji | 2025 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
Abstract
Predictive maintenance relies on vibration-based fault diagnosis, yet deep modEIS suffer from severe domain shift when transferred between different operating conditions, faults, machines, or environmental conditions. We investigate five unsupervised domain-adaptation (UDA) techniques-CORAL, MMD, DAN, JAN and CDAN-built on a compact ID-CNN that processes raw vibration waveforms. Experiments on the public CWRU benchmark and a proprietary vibration dataset show that DAN achieves the best overall trade-off: it boosts target-domain accuracy from 79.2 % to 87.6 % on CWRU and from 30.5 % to 94.3 % on the laboratory set while adding only marginal latency and no extra parameters at inference time. CDAN attains similar accuracy but triples the model size and relies a lot on hyperparameter tuning. We discuss memory-latency-accuracy trade-offs and provide deployment guidelines for edge devices.
