Moving vector hysteron model identification based on neural network inversion

CARCANGIU, SARA;Cardelli, E.;Faba, A.;FANNI, ALESSANDRA;MONTISCI, AUGUSTO;Quondam, S.

2016-01-01

---

Abstract

A neural network inverse procedure has been proposed to solve the Moving Vector Hysteron Model identification. Using the analytic modified Preisach model, a multilayer perceptron neural network is firstly trained to associate the direct relationship among the model parameters and the hysteretic behavior of the material to be modeled both in case of scalar and rotational magnetization. The neural model is then inverted using as input the magnetization identifying the corresponding model parameters values. Model validations with experimental tests and simulations will be performed.