EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

LightGBM-Based Nineteen Class Demagnetization Fault Diagnosis of an IPMSM Using Finite Element Analysis

Akhila Mohanan Nair Sreekala Devi, Himavarsha Dhulipati

IEEE Access · 2026 · 14 · 168276-168294

DOI: 10.1109/access.2026.3748414 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Akhila Mohanan Nair Sreekala Devi und 1 weiteren Autor:innen. Die Arbeit nennt 16 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Natural Sciences and Engineering Research Council of Canada.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Permanent magnet demagnetization is one of the important faults in interior permanent magnet synchronous motors, as it can reduce torque capability, increase current demand, and affect drive reliability. This paper presents a physics-informed demagnetization diagnosis framework for a 36-slot, 6-pole interior permanent magnet synchronous motor (IPMSM) using finite element analysis (FEA) and machine learning (ML). Healthy, uniform-demagnetization, and partial pole-demagnetization conditions are modeled in ANSYS Maxwell 2D by scaling the permanent-magnet material characteristics. A 19-class classification problem is formulated to identify demagnetization type, severity level, and affected pole location within a single diagnostic framework. Two stator-current feature configurations are compared. The first uses 19 time-domain and frequency-domain features extracted from the Phase-A current, while the second uses 57 features extracted from all three phase currents. Feature standardization is applied within a pipeline, and 10 supervised classifiers are evaluated using 5-fold stratified cross-validation. The reported results therefore represent a simulation-based proof of concept obtained from a 2-D FEA model at a fixed speed of 1800 rpm and fixed load under single-fault conditions; because each class originates from a single simulation, the cross-validation accuracies characterize within-simulation performance and do not establish generalization to independent simulations, machines, or experimental measurements.

Förderung

  • Natural Sciences and Engineering Research Council of Canada

Zitieren

APA

Devi, A. M. N. S., & Dhulipati, H. (2026). LightGBM-Based Nineteen Class Demagnetization Fault Diagnosis of an IPMSM Using Finite Element Analysis. IEEE Access, 14, 168276-168294. https://doi.org/10.1109/access.2026.3748414

Vancouver

Devi AMNS, Dhulipati H. LightGBM-Based Nineteen Class Demagnetization Fault Diagnosis of an IPMSM Using Finite Element Analysis. IEEE Access. 2026;14:168276-168294. doi:10.1109/access.2026.3748414

BibTeX
@article{Devi2026,
  author = {Devi, Akhila Mohanan Nair Sreekala and Dhulipati, Himavarsha},
  title = {LightGBM-Based Nineteen Class Demagnetization Fault Diagnosis of an IPMSM Using Finite Element Analysis},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {168276--168294},
  doi = {10.1109/access.2026.3748414}
}
Kostenlos starten →