This paper aims the application of a novel synergy between a neural network (NN) and the finite element method (FEM) in the solution of electromagnetic problem involving hysteretic material in unbounded domain.
The hysteretic nature of the material is taken into account by an original NN able to perform the modelling of any kind of quasi-static loop (saturated and non-saturated, symmetric or asymmetric). An appositely developed iterative FEM procedure is presented for the solution of this kind of problems in unbounded domains.
By starting from a small set of measured loops, the NN manages the values of the magnetic field, H, and the flux density, B, as inputs while the differential permeability is the output. In particular, the proposed NN is capable to perform the modelling of saturated and non-saturated, symmetric or asymmetric hysteresis loops.
The development of an efficient method for the solution of a complicated electromagnetic problem in unbounded domain by using an iterative approach and NNs, which can be implemented also in existing FEM code.
The paper shows that the combination of FEM, iterative procedure and NNs allows us to produce effective solutions of electromagnetic problems in unbounded domains involving also nonlinear hysteretic magnetic materials with an acceptable computational cost.
Coco, S., Laudani, A., Riganti Fulginei, F. and Salvini, A. (2013), "Neural-FEM approach for the analysis of hysteretic materials in unbounded domain", COMPEL - The international journal for computation and mathematics in electrical and electronic engineering, Vol. 32 No. 6, pp. 1964-1973. https://doi.org/10.1108/COMPEL-10-2012-0205Download as .RIS
Emerald Group Publishing Limited
Copyright © 2013, Emerald Group Publishing Limited