Szczegóły publikacji
Opis bibliograficzny
Condition-invariant feature selection for operating-condition-robust fault diagnosis / Urszula JACHYMCZYK, Paweł KNAP // W: MMAR 2026 [Dokument elektroniczny] : 30th international conference on Methods and Models in Automation and Robotics : 18-21 August 2026, Międzyzdroje, Poland : technical papers : on line proceedings. — Wersja do Windows. — Dane tekstowe. — [Piscataway] : IEEE, cop. 2026. — ( International Conference on Methods and Models in Automation and Robotics ; ISSN 2835-2815 ). — USB ISBN:979-8-3195-1920-7. — Print on Demand(PoD) ISBN:979-8-3195-1922-1. — e-ISBN: 979-8-3195-1921-4. — S. 191-196. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 195-196, Abstr. — Publikacja dostępna online od: 2026-09-02
Autorzy (2)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 169799 |
|---|---|
| Data dodania do BaDAP | 2026-10-07 |
| Tekst źródłowy | URL |
| DOI | 10.1109/MMAR70562.2026.11667933 |
| Rok publikacji | 2026 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
| Konferencja | International Conference on Methods and Models in Automation and Robotics 2026 |
| Czasopismo/seria | International Conference on Methods and Models in Automation and Robotics |
Abstract
Fault-diagnosis models often degrade under changed operating conditions. This work investigates whether crosscondition fault classification can be improved using a simple and interpretable source-only feature-selection strategy based on vibration features. To identify features that remain informative under changing regimes, a ranking score was proposed that combines fault discriminability, measured by the ANOVA Fscore, with a penalty for operating-condition sensitivity computed across source regimes. Based on this score, reduced subsets containing the top 25%, 50%, and 75% of ranked features were used to train machine-learning classifiers and evaluated on a previously unseen operating condition following a leave-onecondition-out protocol. The results show that selecting conditioninvariant features substantially improved generalization under domain shift on all 4 tested regimes. For test regime 2000 rpm 50% load, the best performance was obtained by XGBoost using the top 50% feature subset, achieving an accuracy of 0.861, compared with 0.684 on all features. For The proposed approach preserves interpretability, reduces dimensionality, and offers a computationally efficient alternative to more complex adaptationbased solutions.
