This paper proposes a data-driven prognostics framework based on partial least squares (PLS) regression for predicting the remaining useful life (RUL) of power semicon-ductor devices. The method is developed using experimental degradation data from TO-247 IGBT devices subjected to accelerated power-cycling tests under constant stress conditions. Device degradation is monitored through the on-state collector-emitter voltage, VCE, on , and end-of-life is defined as a 5% increase in VCE, on relative to its initial value. To support early prediction, only an initial fraction of each degradation trajectory is considered. These truncated degradation trajectories are transformed into fixed-length feature vectors describing their level, variability, trend, and accumulated degradation, and these features are then used as inputs to the PLS regression model to estimate the failure cycle, from which the RUL is obtained. The framework is validated on datasets acquired under two junction-temperature swings, ΔTj=120∘C and ΔTj=140∘C. Results show that the proposed approach provides accurate number of cycles to failure predictions from partial degradation observations, with mean relative errors below 4% for monitoring fractions above approximately 30%–40%. In the external test, the relative failure-cycle prediction errors for the two held-out DUTs are 1.31% and 0.44%, corresponding to relative RUL errors of 1.88% and 0.63%, respectively.

Afif, O., Tartagni, M., Achaq, M., Magnone, P., Romani, A. (2026). Partial Least Squares-Based Prognostics for Power Semiconductor Devices under Power Cycling Tests. Piscataway : Institute of Electrical and Electronics Engineers Inc. [10.1109/iciea69691.2026.11760490].

Partial Least Squares-Based Prognostics for Power Semiconductor Devices under Power Cycling Tests

Afif, Oumaima;Tartagni, Marco;Romani, Aldo
2026

Abstract

This paper proposes a data-driven prognostics framework based on partial least squares (PLS) regression for predicting the remaining useful life (RUL) of power semicon-ductor devices. The method is developed using experimental degradation data from TO-247 IGBT devices subjected to accelerated power-cycling tests under constant stress conditions. Device degradation is monitored through the on-state collector-emitter voltage, VCE, on , and end-of-life is defined as a 5% increase in VCE, on relative to its initial value. To support early prediction, only an initial fraction of each degradation trajectory is considered. These truncated degradation trajectories are transformed into fixed-length feature vectors describing their level, variability, trend, and accumulated degradation, and these features are then used as inputs to the PLS regression model to estimate the failure cycle, from which the RUL is obtained. The framework is validated on datasets acquired under two junction-temperature swings, ΔTj=120∘C and ΔTj=140∘C. Results show that the proposed approach provides accurate number of cycles to failure predictions from partial degradation observations, with mean relative errors below 4% for monitoring fractions above approximately 30%–40%. In the external test, the relative failure-cycle prediction errors for the two held-out DUTs are 1.31% and 0.44%, corresponding to relative RUL errors of 1.88% and 0.63%, respectively.
2026
2026 IEEE 21st Conference on Industrial Electronics and Applications (ICIEA)
1
6
Afif, O., Tartagni, M., Achaq, M., Magnone, P., Romani, A. (2026). Partial Least Squares-Based Prognostics for Power Semiconductor Devices under Power Cycling Tests. Piscataway : Institute of Electrical and Electronics Engineers Inc. [10.1109/iciea69691.2026.11760490].
Afif, Oumaima; Tartagni, Marco; Achaq, Mariam; Magnone, Paolo; Romani, Aldo
File in questo prodotto:
Eventuali allegati, non sono esposti

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11585/1094650
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
  • OpenAlex ND
social impact