Szczegóły publikacji
Opis bibliograficzny
Domain adaptation-based convolutional neural networks for robust cross-condition fault identification / Paweł KNAP, Urszula Jachymczyk // W: 2025 26th International Carpathian Control Conference (ICCC) [Dokument elektroniczny] : 19-21 May 2025, Starý Smokovec, Slovakia : proceedings. — Wersja do Windows. — Dane tekstowe. — Piscataway : IEEE, cop. 2025. — Dod. ISBN: 979-8-3315-0126-6, 979-8-3315-0128-0. — e-ISBN: 979-8-3315-0127-3. — S. [1-6]. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. [5–6], Abstr. — Publikacja dostępna online od: 2025-06-10
Autorzy (2)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 161213 |
|---|---|
| Data dodania do BaDAP | 2025-07-25 |
| Tekst źródłowy | URL |
| DOI | 10.1109/ICCC65605.2025.11022924 |
| Rok publikacji | 2025 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
Abstract
Accurate and timely fault detection is critical for ensuring the reliability and safety of industrial machinery. While convolutional neural networks (CNNs) have shown promise in identifying faults from vibration signals, their generalization often suffers when moving from well-controlled laboratory scenarios to more diverse, real-world conditions. To address this challenge, our work integrates domain adaptation techniques into a CNN-based fault detection framework. By transferring knowledge across differing fault types and operating conditions, we aim to mitigate the adverse effects of distribution shifts that arise from changes in load, speed, and mechanical properties. Our approach employs unsupervised domain adaptation to align feature representations between source and target domains, reducing the need for extensive manual labeling in new conditions. Preliminary experimental results on a representative dataset indicate that our adapted model improves accuracy and robustness over baseline CNN methods, even when confronted with previously unseen conditions. While further evaluation on a wider range of data and scenarios is needed, these initial findings suggest that domain adaptation can play a pivotal role in developing more adaptable and resilient predictive maintenance solutions.
