Khalilian, Niousha, Sehri, Mert, Hua, Zehui et Dumond, Patrick.
2025.
« Bearing fault diagnosis using traditional machine learning via ChatGPT ».
In Proceedings of the CSME-CFDSC-CSR 2025 International Congress (Montreal, QC, Canada, May 25-28, 2025)
Coll. « Progress in Canadian Mechanical Engineering », vol. 8.
Prévisualisation |
PDF
399 - Bearing fault diagnosis using trad.pdf - Version publiée Licence d'utilisation : Tous les droits réservés aux détenteurs du droit d'auteur. Télécharger (275kB) | Prévisualisation |
Résumé
The increased use of large language models (LLMs) is facilitating a new era for machine fault diagnosis, where LLMs are now capable of conducting simple machine learning (ML) analysis for condition monitoring. This paper explores the potential of LLMs, specifically ChatGPT, for use in diagnosing rolling element bearing faults. Through structured prompts and automated model execution, ChatGPT was tested on the Case Western Reserve University (CWRU) bearing dataset using traditional ML algorithms—Support Vector Machine (SVM), Random Forest (RF), and k-Nearest Neighbors (KNN). The results demonstrate that while ChatGPT can effectively apply feature extraction techniques and execute ML models, its performance is highly dependent on structured guidance, dataset preprocessing, and feature selection. The findings highlight the strengths of ChatGPT in facilitating traditional ML-based fault diagnosis but also reveal its limitations in handling raw data and optimizing deep learning models. These insights pave the way for future research in integrating LLMs with industrial diagnostic frameworks.
| Type de document: | Compte rendu de conférence |
|---|---|
| Éditeurs: | Éditeurs ORCID Hof, Lucas A. NON SPÉCIFIÉ Di Labbio, Giuseppe NON SPÉCIFIÉ Tahan, Antoine NON SPÉCIFIÉ Sanjosé, Marlène NON SPÉCIFIÉ Lalonde, Sébastien NON SPÉCIFIÉ Demarquette, Nicole R. NON SPÉCIFIÉ |
| Date de dépôt: | 18 déc. 2025 15:15 |
| Dernière modification: | 18 déc. 2025 15:15 |
| URI: | https://espace2.etsmtl.ca/id/eprint/32435 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |