Recent studies proved that Ferrite Core (FC) power inductors working in sustainable saturation conditions can enable the achievement of switch-mode power supplies with high power density levels. Since the saturation characteristic of these magnetic components is strongly non-linear, mathematical models capable of representing FC inductors non-linear behavior are extremely valuable. This modelling problem can be of considerable complexity, especially in case of sharp saturation profiles. Neural networks are structures of extreme topological flexibility, intrinsically non-linear and able to operate on multidimensional data both in input and in output. Their ability to be universal approximators has also been proved, since a neural structure of adequate topological characteristics has been shown to have the potential to well approximate any multi-dimensional non-linear function. In this paper, we propose the use of a feedforward neural model to represent the behavior of FC power inductors up to deep saturation current levels. The technique has been verified on commercial FC power inductors via numerical simulations, thus allowing the validation of the proposed neural model. The results obtained open up the possibility of including the problem of modelling the non-linear characteristic of inductors in the great field of deep learning research.

Neural models of ferrite inductors non-linear behavior

Laureti S.;Ricci M.
2019-01-01

Abstract

Recent studies proved that Ferrite Core (FC) power inductors working in sustainable saturation conditions can enable the achievement of switch-mode power supplies with high power density levels. Since the saturation characteristic of these magnetic components is strongly non-linear, mathematical models capable of representing FC inductors non-linear behavior are extremely valuable. This modelling problem can be of considerable complexity, especially in case of sharp saturation profiles. Neural networks are structures of extreme topological flexibility, intrinsically non-linear and able to operate on multidimensional data both in input and in output. Their ability to be universal approximators has also been proved, since a neural structure of adequate topological characteristics has been shown to have the potential to well approximate any multi-dimensional non-linear function. In this paper, we propose the use of a feedforward neural model to represent the behavior of FC power inductors up to deep saturation current levels. The technique has been verified on commercial FC power inductors via numerical simulations, thus allowing the validation of the proposed neural model. The results obtained open up the possibility of including the problem of modelling the non-linear characteristic of inductors in the great field of deep learning research.
2019
978-1-7281-0397-6
Deep Learning
Ferrite Cores
Neural Models
NonLinear Systems
Power Inductors
Saturation
File in questo prodotto:
Non ci sono file associati a questo prodotto.

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/20.500.11770/307385
 Attenzione

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

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 10
  • ???jsp.display-item.citation.isi??? 7
social impact