EUVIMEDEUVIMED entdecken
Crossref

Long-Term Health Prediction for Power Diodes Using Nonuniform Kalman Filter

Chao Guo, Zhonghai Lu

IEEE Transactions on Reliability · 2026 · 75 · 2835-2849

DOI: 10.1109/tr.2026.3719753 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Reliability, 2026. Verfasst von Chao Guo und 1 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Long-term health prediction is frequently compromised by the accumulation of errors and uncertainty, resulting in unreliable results and degraded performance. The capability of the Kalman filter (KF) to adjust states based on current measurements can mitigate error accumulation and provide guaranteed results with quantifiable uncertainties. Conventional KF employs a uniform predict–update loop in each time step, which is optimal for short-term prediction yet suboptimal for long-term prediction. To address this, we propose a nonuniform Kalman filter (NU-KF), in which the predict step occurs at every step but the update step is performed only when the measurement is available. Theoretically, we demonstrate that, under the assumption of linear time-invariant systems with Gaussian noise, NU-KF can minimize the mean-square error, thus making it an optimal filter in long-term prediction, whereas conventional uniform KF frameworks are shown to be suboptimal in this context. Experimentally, we validate this approach for the long-term health prediction using the accelerated aging dataset of power diodes, utilizing forward voltage drop signal as the health indicator. Evaluated against three benchmark methods, including the gold standard one-step ahead KF (OSA-KF), the conventional recursive multistep KF (RMS-KF), and the state-of-the-art multistep ahead extended Kalman filter (MSA-EKF), our method achieves a reduction of up to 65.78% in mean-squared error compared with MSA-EKF, while it achieves comparable performance to OSA-KF. Unlike the exponential error accumulation patterns of RMS-KF and MSA-EKF, our method exhibits linearly bounded error accumulation along with the prediction length, ensuring stable forecasting accuracy for long prediction lengths.

Zitieren

APA

Guo, C., & Lu, Z. (2026). Long-Term Health Prediction for Power Diodes Using Nonuniform Kalman Filter. IEEE Transactions on Reliability, 75, 2835-2849. https://doi.org/10.1109/tr.2026.3719753

Vancouver

Guo C, Lu Z. Long-Term Health Prediction for Power Diodes Using Nonuniform Kalman Filter. IEEE Transactions on Reliability. 2026;75:2835-2849. doi:10.1109/tr.2026.3719753

BibTeX
@article{Guo2026,
  author = {Guo, Chao and Lu, Zhonghai},
  title = {Long-Term Health Prediction for Power Diodes Using Nonuniform Kalman Filter},
  journal = {IEEE Transactions on Reliability},
  year = {2026},
  volume = {75},
  pages = {2835--2849},
  doi = {10.1109/tr.2026.3719753}
}
Kostenlos starten →