Prediction of magnetic properties of strip wound toroidal cores up to 2 kHz using artificial neural network

Abstract

Although magnetic wound cores have simple geometries, their magnetic properties vary in a complex manner depending on core geometry and dimensions etc. These parameters have a strong influence on magnetic performance of wound toroidal cores made from electrical steels or similar strip products. Through theoretical evaluation and experimental measurements carried out over a few years, magnetic performance of a range of strip wound cores have been quantified at low and high frequency. Using this information a neural network model has been developed for prediction core magnetic field strength, power loss and permeability a wide range of flux density and frequency. Input parameters include variables such as core geometry; dimensions strip width and thickness, induction frequency and flux density. The developed network provides flexibility in the choice of training parameters, transfer functions and training algorithm thereby enhancing accuracy.