EUVIMEDEUVIMED entdecken
Crossref

Decision support system for explainable fault diagnosis through retrievalaugmented generation on electrical and thermal variables

LUIS CABELLO DE LOS COBOS, PAULA LAMO ANUARBE, Javier Rainer Granados

DYNA · 01.11.2026 · 101 (6) · 548-554

DOI: 10.52152/d11639 ↗

Kurz gefasst

Fachartikel in der Zeitschrift DYNA, online erschienen am 01.11.2026. Verfasst von LUIS CABELLO DE LOS COBOS und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper proposes a decision support system for the explainable fault diagnosis in Permanent Magnet Synchronous Motors (PMSM), based on a Retrieval-Augmented Generation (RAG) architecture. The approach integrates temporal measurements (currents and temperatures) with documented phenomenological knowledge. Unlike traditional eXplainable Artificial Intelligence (XAI) tools, which provide numerical importance attribution, the proposed methodology generates semantic reasoning grounded in physical evidence. To ensure operational viability and prevent temporal data leakage, the evaluation is conducted using a 5-fold Time Series Cross-Validation (Walk-Forward) on a public dataset of 10,892 samples, comparing the approach against conventional Machine Learning (ML) algorithms. The results demonstrate a clear trade-off between accuracy and explainability: while models such as Random Forest or XGBoost achieve an overall accuracy of 96.47%, the RAG-LLM approach obtains 82.36%. Furthermore, a conservative bias is observed in the RAG architecture regarding incipient faults (Early Recall of 84.3% compared to 99.0% for ML), as well as higher latency. It is concluded that, although traditional ML is optimal for real-time detection, the proposed RAG-LLM model offers significant value as a secondary diagnostic assistant, providing explanations grounded in the retrieved context. Keywords: fault diagnosis; decision support system; predictive maintenance; PMSM; three-phase inverter; electrical and thermal variables; explainable artificial intelligence; Retrieval-Augmented Generation (RAG).

Zitieren

APA

COBOS, L. C. D. L., ANUARBE, P. L., & Granados, J. R. (2026). Decision support system for explainable fault diagnosis through retrievalaugmented generation on electrical and thermal variables. DYNA, 101(6), 548-554. https://doi.org/10.52152/d11639

Vancouver

COBOS LCDL, ANUARBE PL, Granados JR. Decision support system for explainable fault diagnosis through retrievalaugmented generation on electrical and thermal variables. DYNA. 2026;101(6):548-554. doi:10.52152/d11639

BibTeX
@article{COBOS2026,
  author = {COBOS, LUIS CABELLO DE LOS and ANUARBE, PAULA LAMO and Granados, Javier Rainer},
  title = {Decision support system for explainable fault diagnosis through retrievalaugmented generation on electrical and thermal variables},
  journal = {DYNA},
  year = {2026},
  volume = {101},
  number = {6},
  pages = {548--554},
  doi = {10.52152/d11639}
}
Kostenlos starten →