Abstract
DC-DC power converters are ubiquitously employed to produce an efficiently regulated voltage to a load that may be either constant or varying, from a source that may or may not be well controlled. DC-DC converters are power conversion circuits that use high-frequency switches and inductors, transformers, and capacitors to filter switching noise into regulated DC voltages. It is necessary to estimate the remaining useful life (RUL) of a power converter during operation to ensure the reliable and safe operation in aerospace, automotive, space and other mission critical applications and to provide early warning of failure for taking a pro-active action(s). This paper considers the effect of multiple components degradation on performance parameters of power converter. This study proposes a RUL prediction model by utilizing a multivariate-LSTM model to relate deviations in several performance parameters to the RUL. The superbuck power converter is used as a case study. This study follows the k-fold cross technique to validate the proposed RUL prediction model. The findings and comparison show that the multivariate-LSTM model is a better RUL predictive model with high prediction accuracy than other similar deep learning models.
Original language | English |
---|---|
Article number | 114958 |
Journal | Microelectronics Reliability |
Volume | 144 |
DOIs | |
State | Published - May 2023 |
Keywords
- Deep learning
- Multi-variate LSTM model
- Piecewise-linear degradation model
- Power converter
- Prognostics and health management
- RUL prediction
All Science Journal Classification (ASJC) codes
- Electronic, Optical and Magnetic Materials
- Atomic and Molecular Physics, and Optics
- Condensed Matter Physics
- Safety, Risk, Reliability and Quality
- Surfaces, Coatings and Films
- Electrical and Electronic Engineering