Neural network using genetic algorithm for magnetic performance prediction of toroidal wound cores at 50 Hz

Abstract

Geometrical and building parameters have a strong influence on magnetic performance of toroidal wound cores made from grain oriented 3% SiFe electrical steel. From a sample of 40 cores with dimensions ranging from 35 to 160 mm outer diameter, 25 to 100 mm inner diameter and 10 to.70 mm strip width and a flux density range of 0.1 to 1.7 T have been obtained and used as training data to a generalised feedforward neural network.