Automated 2D field computation in nonlinear magnetic media using Hopfield neural networks

Summary form only given. It is well known that computation of magnetic fields in nonlinear media may be carried out using different analytical and numerical approaches, which become more appealing for problems involving complex geometries. Irrespective of the adopted numerical approach, geometrical domain subdivision is usually involved and local magnetic quantities are considered. We present an automated approach using 2D field computations which may be carried out in nonlinear magnetic media. In this approach, vectorial behavior within every sub-volume is represented by a superposition of finite scalar M-H functions f oriented along all possible directions /spl phi/. For isotropic media such functions are taken to be identical. The approach has been implemented for a no load transformer case and has resulted in good agreement with finite element computations. Details of the approach, computations and finite element results are given.