Naslov
Adaptive Algorithm for Sampling Nonlinear Circuit Behaviour in Time-Domain

Autori
Magerl, Marko ; Stockreiter, Christian ; Eisenberger, Oliver ; Barić, Adrijan

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, neobjavljeni rad, ostalo

Skup
The 1st Croatia - Japan Electromagnetic Compatibility Workshop

Mjesto i datum
Zagreb, Hrvatska, 21.05.2018. - 24.05.2018

Vrsta sudjelovanja
Poster

Vrsta recenzije
Nije recenziran

Ključne riječi
black-box modelling ; adaptive sampling ; direct power injection ; echo state networks

Sažetak
An algorithm for sampling the time-domain behaviour of circuits for building nonlinear black-box models is presented. The two- dimensional modelling domain (frequency, forward power level), for which the time-domain behaviour of a circuit is modelled, is sampled adaptively to achieve two goals: (i) apply more focus on the nonlinear regions, (ii) avoid undersampling of the linear regions. The proposed algorithm is used to build an accurate and fast model of a strongly nonlinear circuit in broad frequency range using echo state networks. The model reduces simulation time in the nonlinear regions of operation, and the model accuracy is higher compared to the model built using uniform design space sampling for the same number of samples.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika

POVEZANOST RADA